 Page 2
Is the FORPLAN Approach of a Large Single Model Desirable in British Columbia?
Although FORPLAN would not be used in British Columbia as a land allocation model, could it be used for timber planning? Because a major strength of FORPLAN version 2 is its flexibility to model many different problems, a British Columbia formulation probably could be developed using FORPLAN's framework. But before formulating a FORPLAN model, we would first have to determine how we could use it, where it would fit into our planning system and whether it was consistent with our modeling approach. A model , by desinition, is an abstraction of reality. Its usefulness as a decision making tool comes from its ability lo simplify a complex process so that a manager may be beller able to understand that process. In building a model, the balance between complexity and simplicily must be considered. If a model is too complex it may not be understandable; if it is too simple, it may not be relevant. The balance can be difficult to achieve as managers lend to want the model to make decisions and push for it to be as real as possible. This usually means a large, complex model with much detail. The modeling exercise can become more important than the decision. In British Columbia there is continual pressure from the managers to include more and more detail and many interactions in one model. This provides a challenge for the model builders 10 develop a model that the managers will accepl, which is relevant but still understandable. Can a model be developed which incorporates all the detail and complexity requested and still be understandable? FORPLAN is an example of a model which was built to meet managers' requirements. The FORPLAN approach of one large model which brings together many forest resources and attempts to model their interactions at once has tipped the balance between complexity and simplicity. In its attempt to model every interaction--to be very realistic-- comprehension has been Corleiled. Model results must be understandable 10 the manager, who is the decision maker. The FORPLAN approach to model a very complex problem is admirable; but are FORPLAN results understandable? Are they interpretable? Are they used as support sor decision making, or do they make the decision? Most modelers understand that a model only supports decision making; but is this understood by the managers? If so, how can they use FORPLAN results as an aid to decision making when the model is very complex and the solutions dillicult to interpret? To truly use a model as a support for decision making, it must be well understood. If the model is large and complex, as FORPLAN can be, the model can easily be mystified and the solution either taken as given or discounted. In one case there is danger that the linear programming solution is taken as the "optimal" solution which must be adhered to as it is perceived to be the best. How many times have you heard, "but the model said..."? Conversely, in the other case is the results of an analysis are not in line with perception or local intuition, and if the modeling process is viewed as a large black box, the model solution will not be accepted willingly. In both of these cases the model provides little help to managers. This was expressed as an "information and
understanding gap" and identified as a "barrier to planning" in the Ottawa National Forest (Voytas 1986). This problem of comprehension could be partially overcome through education, but the education task becomes formidable because FORPLAN is a complex model, and the public as well as the forest managers must be involved. Even if the decision makers and the public understood how FORPLAN worked, could they or anybody else intelligently interpret a complex FORPLAN solution? Simple problems can be formulated, because FORPLAN is not a really single model, but a framework within which many models can be formulated. Experience shows that an analysis usually reaches the limits of a model, and that is the capacity and capability of the model are expanded the analysis grows lo that new limit. It is not the user of the tool or the manager who will usually volunteer to keep the analysis simple but the model which, through its limits, forces simplicity. The model-developers have a responsibility to build understandable models. The need for comprehension becomes very important because resource planning is not limited to biophysical and economic criteria, which can be modeled, but includes much socio-economic concerns which require a human element. Analytical techniques such as FORPLAN cannot provide the means lo resolve resource management issues, but simply provide sharper focus for decision makers. A planning process is required that will provide a framework which permits the use of models as aids and includes a large degree of human input and decision making throughout the process. As a District Ranger from While River National Forest stated, "there is always an ultimate limit to the value of a FORPLAN solution in multiple use management no matter how sophisticated we chose to make our model" (Troyer 1986). Analyses must be understood by the decision-makers in order that the results can be used intelligently and non-quantifiable values can be incorporated into decision making. British Columbia has decided to pursue a "component approach" 10 modeling. Within a planning framework, several models, processes or systems, provide support for decision making. The components span several levels of planning and provide the ability to assess many resource use and timber issues. The focus is on the planning process and not on the planning model. An example of this component approach is illustrated in TSA Planning. Timber supply analyses for TSAs have two components: long-term harvest projections leading to the determination of an Allowable Annual Cut (AAC) and selection of an overall management direction; and the geographic location and scheduling of 20-year harvesting areas for each licensee within the TSA. A component is envisioned which will be geographically specific and provide for the allocation of harvest for a 20-year period to various licensees within the TSA. The short-term strategies developed then would be evaluated for their sustainability by linking to a separate long-term harvest projection model or models. The two components provide for a 20-year time frame, geographic resolution, and long-term assessment resulting in a plan which is relevant and implementable without foregoing detail or a long-term view. Working against the component approach are the difficulties in achieving linkages between the components. and train the analysts and managers. Until recently, the Forest Service's planning systems development team consisted of one forester. This team soon will be expanded to include one other forester and a systems analyst; however, without additional resources, the implementation of FORPLAN would use 100% o our development resources for at least 1 year. Considering other priorities, the Forest Service could not afford to expend all its resources on FORPLAN. and a long time to generate much of the information required.
Staff and Expertise to Use FORPLAN
Linkages between the short-term harvest allocation and long-term harvest projections and between land allocation and timber planning, for example, have been disficult to achieve and pose challenges to modelers. Innovation is required in obtaining these linkages; the modeling effort should be directed here. When we first looked at FORPLAN we were excited about the potential to link the short and long-term aspects of timber supply planning in one model formulation rather than using separate models. The two features of FORPLAN that were of most interest were the ability to specisy variable period lengths and the use of allocation zones lo provide geographic resolution. These features could provide the capability to include a detailed, geographic specific 20 year "front end" linked directly to a less detailed long-term "tail end." This would provide the resolution required for 20 year planning and the linkage required to ensure long-term objectives of sustained yield. With closer review, we realized ihat FORPLAN would not only need variable period lengths, but also would require the specification of varying geographic detail. The overhead of maintaining the geographic resolution required (500 to 2,000 planning cells) throughout the whole planning horizon would be too high. We retrealed from the concept of one model to do both short and long-term planning and are now looking for two separate but linked approaches. FORPLAN could be used to carry out long-term harvest projections, but is it the best choice? As the long-term aspects of the plan are de-emphasized, a simpler model , such as MUSYC, may be more appropriate and is already in use. FORPLAN also could be used to carry out the 20-year allocation; but does it allow for sufficient geographic resolution? Can it be linked with a geographic information system and can it be linked to a long-term projection model? These questions for the most part have been left unanswered, because FORPLAN has not been used in this manner and would require much more detailed investigation and research. Perhaps FORPLAN could change its focus and provide a basic framework with which other components, such as a fire and pest model, geographic information system, or a timber allocation model could be linked. British Columbia would be very interested in any future efforts to simplify FORPLAN and to link il with other models, systems and levels of planning.
The planning resources in the Forest Service consist of 12 staff in Victoria, 3 in each of 6 Regions and 1 in some of 46 Districts for a possible total of 76. The total operating budget is less than $1 million. That works out to about i planner and $10,000 for every 1 million m (350,000 cunits) harvested annually. Hands-on experience using forest planning models is limited to six staff in Victoria. Knowledge of FORPLAN in British Columbia is limited to myself, one consultant firm and a graduate student at the University of British Columbia. Our understanding of FORPLAN is general and includes very little hands-on use. Because of FORPLAN's complexity, a major commitment would have to be made to increase our knowledge pool and to educate and train both the analysts and the managers to ensure successful implementation. Limited resources, other priorities and lack of expertise severely hamper our ability to do this.
The British Columbia government's computing facility is IBM. Although several IBM versions of FORPLAN have been released, IBM compatibility does not appear to be a high priority of the USDA Forest Service. We would require full support for IBM compatibility before implementing FORPLAN to ensure full support of the model.
Considering all of these factors, along with the complexity of FORPLAN, the implementation of FORPLAN in British Columbia would be a formidable task.
IBM system, it may provide a useful planning tool for British Columbia. Although British Columbia has not implemented FORPLAN, we have learned much from it which will help us to develop and use resource planning models. I hope that my comments have provided you with some insight and will help you in future modeling endeavors. We will be following FORPLAN's evolution with interest.
The Forest Service has chosen not to implement FORPLAN; but we are continuing to assess FORPLAN for its suitability in light of projected changes in British Columbia's modeling requirements and in the evolution of FORPLAN. British Columbia's requirements for resource planning models are changing. Resource planning is evolving in British Columbia as public awareness of and demand for other resource values and beller management increases. However, given the differences outlined in this paper, resource planning in British Columbia may or may not evolve in the same manner as the in the USA. As our requirements change, we will continue to evaluate American resource planning models, including FORPLAN, for their suitability. We are curious about the future of FORPLAN. If it can be simplified, easily linked with other planning models including a geographic information system to become part of an overall planning system, and can be supported on an
British Columbia Ministry of Forests. 1984. Forest and Range Resource Analysis 1984. ISBN 0-7726 - 0243-3. Iverson, David C. and Alston, Richard M. 1986. The Genesis of FORPLAN: A Historical and Analytical General Technical Report INT-214, 31 p. Troyer, Jack. 1986. FORPLAN · The District Ranger's Experience. In Proceedings of the Workshop on Lessons from Using FORPLAN. USDA Forest Service, Land Management Planning Systems Section, Washington, D.C., 5 p. and Decisionmaking. In Proceedings of the Workshop on Lessons from Using FORPLAN, USDA Forest Service, Land Management Planning Systems Section, Washington, D.C. 4 p.
The North Carolina State University Experience with FORPLAN: Software Transfer and Example Of Use
Joseph P. Roise and John Welker
Abstract.--NCSU required the use of FORPLAN software for several management research projects. FORPLAN software was obtained through the University of California, Berkeley. Several steps are discussed on how FORPLAN was transferred for use at NCSU. The first applications of FORPLAN at NCSU, a national planning model for Jamaica, is also discussed.
(2) we could move FORPLAN to the TSO system, and do all the calculations on one system. Because of the size of the problems we were working with, we felt it would be best to do all work on one system. Notice that one possible option was not included: buying or leasing MPSX. The cost of MPSX was around $17,000 and our initial budget to get FORPLAN running was less than $10,000. In hind sight, it may have been less expensive to buy MPSX and load it onlo CMS. During the summer and fall of 1986, we installed FORPLAN on the TSO system for the North Carolina State University School of Forest Resources. This document contains two major sections the first describes the strategy and specific steps taken while installing FORPLAN and the second discusses an application in planning Jamaica's forest resources.
The print out of the FORPLAN code requires 0.2 cubic feel of paper, weighing in at 9.7 lbs. It would take a loblolly pine seed in average soil about 16 years to make the wood for this print out. No big deal when compared with the amount of fiber in the national forests. However..... In the summer of 1985 the Department of Forestry at North Carolina State University acquired a lape containing FORPLAN software from the University of California, Berkeley (UCB). At UCB FORPLAN is running on the IBM-Conversational Monitor System (CMS), and because one of our campus computer systems used IBM-CMS, our hopes were high that we would soon be running FORPLAN. On the first day the program was loaded into CMS, FORPLAN could be started using the EXEC (execution) language program supplied by UCB with the rest of the FORPLAN files. On the same day our problems began. On the NCSU CMS system the standard secondary memory allocation is less than one cylinder. FORPLAN required 35 cylinders of space. This results in quite an expense just to store the model. The EXEC program was sailing on some system software differences between UCB's and NCSU's version of CMS. These needed to be reconciled. The most serious problem was that the solution software required for this version of FORPLAN is MPSX. Not only was this software not on the CMS system, but the campus had discontinued its use on all systems in favor of another large scale LP solution package. However, there was a version of MPS kept on another IBM computer system using the Time Sharing Option (TSO) operating system. The original install tape for MPS was located but because of age (original installation around 1975) The magnetic fields on the tape had deteriorated beyond the background noise level. The tape could not be read. At this point we considered two options. Either we could: (1) generate the malrix on CMS and send the matrix sile to TSO, through a filter to remove discrepancies between MPS and MPSX, solve the problem on MPS, send the results through an output filter, and back to CMS to the report generator; or
This section has two parts. The first outlines a general strategy for installing any program like Forplan and the second section details the specific steps taken while installing FORPLAN at NCSU.
General Strategy for Installing FORPLAN
There are at least four major steps required when installing a program like FORPLAN. The first step is determining whether available computers have the hardware needed to execute the program. After l'inding a computer system, the second step is moving the files with the source code onto that computer system. Once all the files with the source code are on the computer system, the third step is translating the source code into executable code. After supplying the computer system with executable code, the fourth step is supplying the executing progntun with data in a usable format. Each remaining paragraph, in this section, provides succinct elaboration on each of these
"Roise is Assistant Professor of Forestry and Operations Pesearch, and Welker is Ph.D. candidate and research assistant, at North Carolina State University, Raleigh.
Before acquiring a computer program, determine whether the available computers have enough
consist of a certain number of lines and that each line be presented in a particular columnar format. When an adequate computer system has been supplied with an executable program and the data expected by that computer program, the process of installation is finished. When installation is finished, the process of use and maintenance begins. Using a program can be a formidable task simply because the data required may be expensive to supply, alter and maintain. The costs associated with using a program almost always overshadow the costs incurred while installing that program.
Adapting the General Strategy To Install FORPLAN
hardware to allow the program to execute. The hardware of a computer system consists of a central processing unil, an arithmetic/logic unit, random access memory and peripheral input/output devices. Distributors of programs usually provide a list of the hardware required to allow the program lo execute. For example, a program requiring 5 megabytes of random access memory will not execute on a computer which has only 640 kilobytes of random access memory. Similarly, a program requiring 2 tape drives, will not execute on a computer that has no lape drives. A computer system's capacity to store and access information, using devices like fixed disks and tape drives, may not only preclude execution of a program, but may not be sufficient lo slore the program itself. Alter finding a computer system capable of storing and executing the program, the source code files are then placed into the computer system. The source code files are ASCII files which hold only information which is readable by humans. In other words, these are the liles that computer programmers have entered using a keyboard. Once again, these files are written in English. These files are transported using media that the computer system supports. For example, a tape may be sufficient, given that the computer system, not only has a compatible tape drive, but also is able to read the format in which information is stored on the tape. A tape may not be the most convenient medium for moving files to a computer system. For example, the capacity to transfer files over a telephone line is sometimes more convenient. Aller the source code has been stored on the computer system, the next step is to translate the source code into executable code. Files containing executable code cannot be read by people, but they can be read and used by computers. Requirements for consolidating source code into executable code vary depending on the complexity and structure of the computer program. For example, a simple program may only require one compile step, one link siep and one go step. Where the compiler translates source code into an object module, the linker translates an object module into a load module and the go step loads and executes the load module. However, complex programs, such as FORPLAN, may require several compile and link steps before the go step. Each lile holding source code is first translated into an object module, and second, each object module is cross referenced, by the linker, and stored, as a load module, in a library of load modules (this library is usually stored as a partitioned data sel). After all source liles have been stored as load modules, the computer program is ready for execution. Aster supplying the computer system with executable code, data must be made available in the format expected by the program. Because data can be accessed using any one or more of several devices, it is necessary to know which devices the computer program expects louise. Additionally, it is necessary to know the precise method for presenting data to the computer program. For example, it may be necessary to provide data which establishes initial conditions before presenting that data pertinent to a particular problem. Also, the physical arrangement of data may vary from one computer program to the next. For example, the computer program may require that the data
Having established the events required to install a program like FORPLAN, the following outline reconciles ihat general strategy with the steps taken during installation. Each of the numbered steps, in the following outline, corresponds to an event which occurred while installing FORPLAN. As each objective in the general strategy is accomplished, it is noted with commentary. 1. Reconcile hardware requirements with available The lape with the source code needed to install FORPLAN were written using the UCB equivalent of the IBM CMS TAPE DUMP command. Therefore, a computer, at NCSU, capable of reading this format was required. Additionally, FORPLAN produces files that are used as input to the IBM Mathematical Programming System, then Forplan reads and processes the vulput of the Mathematical Programming System. So a computer capable of operating MPS with Forplan another requirement. Finally, Forplan requires a minimum of 2 megabyles of random access memory, a lape drive and a considerable amount of space on a disk pack. Having established the requirements, the next Having sound a computer system to match the Berkeley 3. Move files from tape onto disk on CMS 4. Move subroutines from MACLIB to individual Page 3
This step is unique to installing Forplan. The Forplan program is actually a set of separate programs, the execution of which, must be coordinated. Results are passed, using files, from one module to the next. A command procedure coordinates the execution of modules depending on the information stored in the files that are passed from one module to the next.
FOREST PLANNING MODEL FOR JAMAICA
The source code had been stored, on tape, in a file called FORPLAN MACLIB. Alter loading the tape on the tape drive, the command, "TAPE LOAD FORPLAN MACLIB DI," moved the file from the tape to the D-disk. A list of files in the Maclib was displayed using the command, "MACLIB MAP FORPLAN MACLIB D1." Files stored in a CMS Maclib can be moved from the Maclib to the Punch machine using the command, "PUNCH FORPLAN MACLIB D1 MEMBER (membername)", where membername is the name of one of the liles in the Maclib. The Punch command takes the liles out of Maclib format and converts it into ASCII format. Aster moving the member from the Punch machine to disk, the file can be further processed. In our case, further processing meant moving the files from CMS at the NCSU Computing Center to TSO at the Triangle Universities' Computing Center (TUCC). The files were transserred using the electronic mail system. Having supplied the computer system with the files containing the source code, the next step was to translate the source code into executable code. 6. Produce hard copy of the source code 7. Prepare list of all subroutines and calls to subroutines 8. Prepare files for compilation and linkage 9. Compile, link and store files in a PDS at TUCC 10. Back up all files and the PDS on tape at TUCC Because the Forplan program consists of more than 230 separate subroutines, it required as many compilation and linkage steps. Subroutines which referred to no other subroutines were the first procedures compiled and stored in a partitioned Jata set. Remaining procedures were compiled and cross referenced with routines already in the PDS. Having supplied the computer system with executable code, the last objective was to supply Jala to the executing program. 11. Prepare list of all read, write and format statements 12. Match the instructions in the user's manual with the read, write and format statements 13. Establish files required to execute Forplan Files containing the source code were processed to produce a list of all Read, Write and Formal statements. These statements were cross referenced with the user's manual to determine which files are required by each module of Forplan. 14. Write command procedure to coordinate the execution of Forplan's modules and MPS
The initial use of Version 2 FORPLAN at North Carolina State University is to examine the current pine lumber import substitution strategy of Jamaica. In the early 1960's the Forestry Department of Jamaica began establishment of Pinus caribaea var. hondurensis (PCH) plantations in the eastern and generally mountainous portion of the island. The objectives were to provide watershed protection and to substitute for lumber imports, primarily from the southeastern United States and Central America. In 1986 there are about 10,000 hectares in pine plantations on government lands (FIDCO 1986). A smaller area of pine plantations also exist on private lands. About 60,000 additional hectares in the eastern portion of the island have soils suitable for PCH (FIDCO 1986). In 1978 a state-owned company, Forest Industries Development Corporation (FIDCO), was formed with the "objective of developing the potential for commercial forestry and forest industries in keeping with the nation's strategy or import substitution, job creation, and optimum land utilization" (FIDCO) 1979). Harvesting and sawmill operations for lumber production began in 1979. In 1985 FIDCO produced 4.8 million board feel of import grade lumber, representing 19% of total consumption of this grade. Inferior grades of lumber and treated posts are also produced. Soil and rainsall conditions permit high growth rates, between 15 and 27 cubic meters per hectare per year on a 20-year rotation. Also, the FIDCO sawmill is running on a single shift basis so that production could be expanded with an increase in log supply. However, about 75% of the suitable forest land requires skyline harvesting. In addition, both coffee and mixed agriculture compete for land use over much of the area, particularly following forest road development. The compatibility of forest plantation wood production and watershed management is also a significant issue for those lands near Kingston. Jamaica's location, and favorable credil arrangements in recent years, account for the current role of the United States as the major supplier of coniferous lumber. However, the long-term trend of total consumption of import grade coniferous lumber has declined over the past 20 years, -0.8% per year (External Trade Bulletins, 19661985). Constant real income per capita, increasing import supply prices (1.2% per year), and increases in duties contribute to the decline. In April 1986 the elective duty on lumber imports was increased from 9% to 27% which should further reduce consumption and imports.
and to test the sensitivity of the results to alternative price and policy expectations.
The principal policy issue in this context is that of determining the appropriate land use strategy given current knowledge of production possibilities and factor and output price expectations. The three graphs in figure 1 illustrate the conceptual model for determining the lumber production equilibrium in a static context. The graphs are identical except for the domestic supply schedules, SD, S, and De represent the perfectly elastic import supply and export demand schedules respectively. In the upper graph Os should be produced domestically, and O.-Oj should be imported. In ihe bollom graph the equilibrium is one of autarky so that domestic production equals consumption. The primary objective of this study is to test the hypothesis that the appropriate long-term land use strategy is one in which a combination of imports and domestic production provides the quantity demanded at the import supply price adjusted for duties and distribution costs. Additional objectives of the study are to show the import and domestic supply quantily schedules in the short-term
The harvest scheduling-linear programming (LP) model is convenient for examining this problem, particularly given the flexibility of Version 2 FORPLAN for representing alternative investment decision variables. Unlike the static representation shown in figure 1, the model permits the specilication of domestic wood supply curves for alternative time horizons. The base case computer run of this problem will assume that the optimal solution includes at least some importation of lumber in each period of the planning horizon (top graph sig. 1). However, if all demand is met by domestic production in at least one period for the base case, another case may be run to portray the other equilbria possibilities using the price-quantity options in FORPLAN. In this case the appropriate price-quantity locus is Pçif to the point of intersection with DD, DD to the point of intersection with De, and PFOB l'or greater quantities.
Table 2 gives a schematic representation of the prescription and timing choices to be considered for each analysis area over the chosen planning horizon of thirtyeight years. The first part of the table gives the prescription-liming choices for the land use decision variables. There are four vegetation categories: PCH, nalive vegetation, coffee, and mixed agriculture. Based on land classification by the Ministry of Agriculture, there are three long-term land capability classes: forest, collee, and agriculture (Vernon 1959). The rate of conversion from the current land use to the long-term land capability class can be modified by specifying constraints or limiting the prescription-timing choices for a given current land use. The land use capability classilication will be used to determine what alternative land uses are permilled for a given current land use. For pine management the principal identifiers for differentiating analysis areas are: age class, site index class, stocking, harvesting system, transport distance, roading status, and ownership class. Pine management treatments are based on even-aged management with no thinnings. The timing choices permit clearcuts between seventeen and thirty years of age (table 2).
Figure 1.--Conceptual Import substitution • export promotion
Table 1.--Linear programming lumber Import substitution model.
existing hectares of pine age class i, yield class j, alternative use class k, managed with prescription-timing choice | existing hectares of native hardwood vegetation of yield class ), alternative land use class k, managed with prescription-timing choice ! existing hectares of coffee (same subscript definitions as above) existing hectares of mixed agriculture land use (same subscript definitions as above) volume of lumber in use class g, imported in period m operating units of standard capacity of processing function f, depreciation status I, Installed and depreciating with timing choice |
Maximize NPV = 0, U + D'WW +DxX + DyY-0,2-D'QQ у Du. Dw. Dx Dx, Dy stand or crop level net present values per hectare at the start of the у planning horizon (excluding capital costs in Do discounted CIF price plus domestic handling cost per unit volume of imported lumber discounted capital costs per unit of capacity (purchase, installation, maintenance, and salvage)
Import Grade Lumber Demand (for each period) Imported lumber of grade g is less than or equal to lumber demand at the duty inclusive import supply price b. Domestic lumber production of grade g is less than or equal to lumber demand at the import supply price, exclusive of duties c. Domestic lumber plus imported lumber supply is greater than or equal to total lumber demand at the duty inclusive import supply price
Coffee Supply equal to the Coffee Marketing Board quotas.
Lumber and Kiln-drying Capacity Installed capacity in each period is greater than or equal to the amount required for domestic lumber production
Increases or decreases in installed capacity between periods are within levels consistent with capital costs per unit of capacity.
Prescription nos. 1-4 etc.; 5; 6-7 etc. plus : CC RC CC RC etc. Prescription nos. 1-4 etc.; 5; 6-7 etc. plus: CC RA CC RA etc. Prescription nos. 1-7 etc. except no restriction on maximum clearcut age Prescription nos. 1-4 etc.; 5; 6-7 etc.; 8-9 etc. except no restriction on maximum clearcut age Prescription nos. 1-4 etc.; 5; 6-7 etc.; 10-11 etc. except no restriction on maximum clearcut age Prescription no. 5 plus: RP CC RP AC RP CC/RP etc. AC NV AC NV etc. Prescription nos. 5; 12-15 etc. plus: AC Prescription nos. 5; 12-15 etc. plus: RA AC RA etc. Prescription nos. 5 & 12 plus: AA RP CC/RP AA RP CC RP etc. AA NV AANV etc. Prescription nos. 3. 5,12,19-20 etc., 21-22 etc. plus RC AA RC etc. Prescription nos. 5,12,19-20 etc., 21-22 etc. plus: AA
Page 4
km./ha. km./ha. km./ha. km./ha. km./ha./yr.
$/ha./yr. cd./cu.m. logs mh./cd. $/mh. $/mh. $/cd. $/cd. $/cd.$/ha. harvested km./ha. harvested mh./km. km./cu.m. logs km./post$/cu.m. - km. $/post-km. sh./cu.m. logs $/sh. $/sh.
Developed land management Construction, access roads Construction, harvest roads Upgrading, access roads Upgrading, harvest roads Road maintenance Machine hours (mh.) Road construction Upgrading Maintenance Pine Management Silviculture, labor input (by year in the rotation) Silviculture, management overhead Logging, crew-days (cd.) Logging, machine hours Logging, machine operating expenses Logging, machine capital costs Logging, labor expenses Logging, miscellaneous expenses Logging, management overhead Logging, post production expenses Construction, planned skid tralls Machine hours, planned skid trails Log transport, cubic meter-km. Post transport, # post-km. Log transport costs Post transport costs Sawmill, shift operating hours (sh.) Sawmill, operating expenses Sawmill, management overhead Head office administration, Plantation development Logging and sawmilling Transfer payments, Plantation development Logging and sawmilling Non-structural, e.g. pattern stock Structural Other, e.g. dimension Other land management Cottee land rent Cottoe production (by yr. of rotation) Mixed agriculture land rent Native vegetation protection Processing capacity Capital costs (by year from installation) Sawmili Kiln-drying Standard units of capacity (by year from installation) Sawmill Kiln-drying Lumber Importation Non-Structural grades Structural grades Other grades
$/ha. mgd. # / ha.
nominal cu.m./ha. nominal cu.m./ha. nominal cu.m./ha. nominal cu.m./ha. nominal cu.m./ha. cu.m.,ha. cu.m./ha. cu.m./ha.
$/ha/yr.
$/std. hour of capacity $/cu.m. of lumber
std. hrs./std. hr. installed cu.m./cu.m. installed
$/nominal cu.m. Page 5
A forest planning system such as FORPLAN can be evaluated in at least two ways. An evaluation of interest inainly lo model builders and operations research theorists l'ocuses on a generic matrix generation package that can be linked to a solution algorithm designed to solve a specific mathematical programming problem. A second type of evaluation focuses on the results from the application of the system and is more of interest 10 forest managers, the Congress, and groups with special interests in forest planning. The following comments address both types of evaluation, though it is loo early lo thoroughly evaluate the the results of applying FORPLAN lo forest planning. The development of FORPLAN rellects a "maximizing" strategy rather than an "optimizing" one. FORPLAN is more a lool kit than a tool. Much like a PAC MAN, FORPLAN has gobbled up several systems designcd to solve specific problems in output scheduling, land and resource allocation, transportation planning project planning, and combinations thereof. Unlike PAC MAN, however, FORPLAN has retained intact the systems it gobbled up, thereby allowing uscrs 10 analyze problems Irom varicd perspectives. The multiple model framework of Version 2 contains elements of Timber RAM, ADVENT, IRPM, RAA and others. From a researcher's standpoint, it provides a means to evaluate various paths of analysis for a variety of natural resource management problems. Whether it's of use from a manager's standpoint depends on whether the results have any practical application. Can they be implemented on the ground? Can they be used to justily budget requests? Successsul use of FORPLAN to solve planning problems and meet the needs of practitioners in forest management requires investment in scveral key areas. System sollware support, training of analysts, and consultation with users are all needed 10 assure thought'ul and efficient use of the system. Management analysts must be carefully selected for their ability to shist between the complexity of FORPLAN and the technical, social and political complexities of day-to-day forest management. FORPLAN sort out mathematically inseasible solutions, but is incapable of recognizing ones that aren't feasible in real life. It takes interaction with real forest managers with names such as MARVIN to avoid those kinds of infeasibility. Stressing the human aspects of planning should not be construcd as rejection of any analytical system, whether large mathematical model or otherwise. The point is: the system is not the solution; pcople are. Long-range planning is nothing more or less than risk-taking decisionmaking. If we don't give any thought to where we are going, any road will get us there--and without planning, there's a good chance we won't like where we've arrived. The role of planning analysts is to aid the thought process. There must be intcraction between analysts and managers. True optimization will occur if there is agreement on the problems to be analyzed and it turns out the results of the planning process are meaningful and useful
Alston, Richard M. and David C. Iverson. 1987. The road from TIMBER RAM to FORPLAN: How far have we traveled. Journal of Forestry. In press.] Barber, Klaus H. 1986. Large FORPLAN models: an exercise in folly. p. 89a-890. In: Proceedings of the Workshop on Lessons from Using FORPLAN. (Denver, Colorado, April 29-May 1, 1986] USDA Forest Service Inot numbered], 268 p. Land Management Planning Systems Section, Washington, D.C. Bchan, R.W. 1981. RPA/NFMA - time to punt. Journal of Forestry 79(12):802-805. Beuter, John H. 1985. Federal timber sales. Congressional Research Service CRS-64, 140 p. Library of Congress, Washington, D.C. Bowes, Michael D. and John V. Krutilla. 1986. The economics of multiple use forestry. Manuscript in review, 411 p. Resources sor the Future, Washington, D.C. Chappelle, Daniel E. 1977. Linear programming for forestry planning. p. 129-163. In: Convery F. J. and Ralston, C. W., cditors. Forestry and long range planning. Duke University School of Forestry, Durham, NC. Chappelle, D. E., M. Mang and R. C. Miley. 1976. Evaluation of TIMBER RAM as a forest management planning model. Journal of Forestry 74(5):283-293. Gass, Saul I. 1964. Linear programming: methods and application. 280 p. Mc-Graw-Hill Book Company, New York. Iverson, David C. 1985. The simple analytics of forest planning. Unpublished paper presented at Western Forest Economist Conference, Wemme, OR. May 1985. 31 p. Available from David C. Iverson, USDA Forest Service, R-4, Ogden, Ulah. Iverson, David C. and Richard M. Alston. 1986. The genesis of FORPLAN: A historical and analytical review of Forest Service planning models. USDA Forest Service General Technical Report INT-214, Intermountain Research Station, Ogden, Utah. Johnson, K. Norman and Brian Greber. 1986. Timber harvest levels on the national forests of (regon: causes, impacts and alternatives. (Outline of a study). College of Forestry, Oregon-State University, Corvallis. Kaul'man, Herbert. 1967. The forest ranger. 259 p. Resources for the Future, Washington, D.C. O'Toole, Randal. 1986. When both sides can win. Forest Watch 7(3): 20. II FORPLAN in project analysis. p. 130-142. In: Proceedings of the Workshop on Lessons from Using FORPLAN. (Denver, Colorado, April 29-May 1, 1986) USDA Forest Service (unnumbered publication), 268 p. Land Management Planning Systems Section, Washington, D.C. decision (revised edition), 175 p. Prentice-Hall Inc., Englewood Cliffs, New Jersey. Page 6
My final comments on LP as an economic tool emanate from the very commonly expressed concerns that LP does not seem to be very effective in dealing with phenomena such as lire, insect and disease infestations, or other dramatic ecological changes. I believe that it is not uncommon for this shortcoming to be attributed to the fact that LP does not incorporale stochastic variation in the response variables with grcal ease. I also have heard it suggested that LP fails because it does not easily incorporale risk averse perspectives or that it does not account for the desirability of llexibility the undesirability of irreversible decisions. All these points are well taken; however, all these weaknesses could be remedied and the LP still would not perform well in terms of modeling things such as sire, insect and disease infestation, etc. My reasoning is as follows. An LP is based on a calculus-oriented formulation of the optimization problem at hand, where constraints and objective l'unctions are smooth, many times dillerentiable functions. That is, the basic mathematical foundation of an LP includes an assumption that small changes in choice variables will result in small changes in response variables. Likewise, symmetrical reactions to small changes in choice variables are implicity assumed. Phenomena such as fire and insect and disease infestation simply don't behave that way. They behave in lits and starts and at times very small changes in choice variables can cause an immense impact on the ecosystem because of such a discontinuity. I would assert that managing these unstable "catastrophes" is generally more important than managing the ecosystem when it is "wellbehaved." I would also assert that we are more likely to learn fundamentally new things about the ccosystem by concentrating on the discontinuitics than we are from studying the smooth, slow changing behavior that an LP can handle. Refocusing our analytical energy on the discontinuities should not be taken lightly or undertaken prematurely. It will require moving into an entirely different area of mathematical analysis. A new branch of applied mathematics called "Catastrophe Theory" has developed out of topology and may show promise in this area. Many questions remain unanswered. A very large part of my research focus over the next 5 years is planned io be in this pursuit. Many researchers around the country are working on Catastrophe Theory, and the potential is very exiting. Al this point, however, this work still belongs on the desk of the researcher and is not ready for a planning analysis such as that in forest planning.
I doubt that there will be very much disagreement that this linkage is logical. Also, I believe it is the intent of the agency to accomplish as much of a linkage between forest planning and RPA as is possible in the development of the 1990 Program. Part of the analysis of resource interactions for the 1989 Assessment will be based on the alternatives generaled with the FORPLAN models (Hos and Pickens 1986). The approach to be taken focuses the regional/national analysis on the selection of discrete management alternatives provided (perhaps over a period of years) by the forests (see Wong, 1980; Bartlell, 1974). These discrete alternatives will include scheduled outputs and costs. They, in essence, will be zero-one choice variables in higher level models that could be linear programs or integer programs. A test of this approach is reported in Hot and Pickens (1986, 1987). This test shows that the Bartlett-Wong approach performs very well in terms of overall optimality--it is able lo locale "points" on the global production possibilities frontier with a high degree of accuracy. The Bartlett-Wong approach, however, does not closely emulate a global optimization in terms of forest level output solution values or budget allocations across forests. It is thus much more useful for Assessment analyses than for Program analyses. Given the above-quoted viewpoint that the forest plans and the Program need to be linked, however, this multilevel model may be useful in the Program as well. It should be emphasized that if forest planning alternatives are to be used in building the Program alternatives (regardless of the method), then systematically varied set of forest alternatives would be most useful. Beuter and Iverson state: "...FORPLAN runs are expensive and, probably too many are being made..." If forest planning information is to be used at a higher level of planning, then I would take serious issue with this statement. In this context, the purpose of forest planning is not just the development of a "preferred alternative" but the development of a set of alternatives that effectively describe the range of options available. My reaction would be that we probably need more runs (alternatives) not fewer, and if FORPLAN is too expensive to do this then it is a very serious shortcoming. One advantage of modeling a planning problem should be that sensitivity analysis and "what if" exercises are possible. If we can't experiment with the model, ils utility is greatly diminished. As an example, the performance of the Bartletl-Wong model is directly affected by the number of and systematic variation in the alternatives included. Given these conclusions, the Forest Service might eventually consider developing the capability to pursue an approach such as that discussed by Kornai and Liptak (1965). This approach involves a game theoretic model between a higher level planning authority (the "center") and a set of sectoral planning units. The center makes an initial provisional distribution of the "available resources, material, manpower, etc. among the sectors, and at the same time also indicates their output targets." The sectors then rigorously analyze this set of quotas and report back "one type of economic elliciency index--the shadow prices derived from programming." A model such as FORPLAN
Linkage Between FORPLAN and RPA
Beuter and Iverson state: "It is logical that there should be some link between the forest plans, the RPA program and budgets, but there isn'l--at least not any direct link that is apparent at the present time" Page 7
The Costs and Benefits of a Forest Planning Model: Discussant's Comments
Abstract.--This paper focuses on the external economics of FORPLAN. What are the costs and benefits of this particular planning model and the process based around it? It discusses the costs and benefits of the planning system, and concludes by examining alternatives to FORPLAN
differ from past ones. Beuter and Iverson's question can then be partially answered by determining il FORPLAN has resulted in decisions which differ from the ones made under earlier planning systems. Little evidence has been presented on this point. Schweitzer et al. (1986) argue that current decisions probably differ little from past oncs because "there are powerful forces in the forest planning process that influence most forest plans to be similar to traditional types of forest management." (p. 10). If the use of FORPLAN is not altering forest management decisions, then this costly planning procedure does not achieve Simon's desidcralum for procedural rationality. Suppose that FORPLAN represents overinvestment in analysis. What, then, is the optimal amount of analysis? The "Max-Loss" procedures developed by Navon el al. (1986) offer one approach for addressing this question. Rows and columns of the FORPLAN matrix can be aggregated, and a bound on the loss of efficiency resulting from the aggregation can be calculated. In their sample problems, the loss in efficiency was very small even with large reductions in problem size. Thus, in large FORPLAN models the marginal costs of increased complexity probably exceed the marginal benelits. Economic efficiency would be served by reducing model size. My own, limited experience with FORPLAN suggests that the nel present value surface may very flat near the optimum; many different management plans have very similar economic performance. These different management plans may have quite different political implications. Rather than simply solving every FORPLAN problem to optimality, the analytical clfort would more usefully be directed towards identifying all the plans which are "nearly" optimal.
National forest planning is an unmitigated example of what Simon(1979) has called procedural rationality. He argues that the objective os procedural rationality is to find optimal decisions net of analytical costs. The costs of developing the FORPLAN system and the FORPLAN models for each national forest apparently are not known with any degree of certainty, but are thought to be several hundred million dollars. To evaluate the planning system, these costs must he calculated. Such a summary has been requested by Congress for the national Resources Planning Act Assessment and Program (RPA). A similar accounting should be made for the planning activities promulgated under the auspices of the National Forest Management Act (NFMA). Beuter and Iverson ask if FORPLAN will lead to betler decisions. Because all national forest planning relies on FORPLAN, there is no way of wholly answering this question. However, better decisions must, by delinition,
The use of FORPLAN excludes important people from the national Forest planning process. That the general public cannot understand FORPLAN is by now dbvious. A public excluded is a public prone to litigation. 'This fact results in a central paradox of applying the rational planning paradigm to the national forests: is the public
1 Professor of Forestry, School of Forestry and Environmental Studies, and School of Organization and Management, Yale University, New Haven, CT 06511
"research" in the normal "research and development" process for creating new methods, the Forest Service would have been served well by the kind of exploratory work which would have occurred on individual national forests under a less centralized planning system.
participates effectively in the planning process, the results are necessarily politically determined; if the public is systematically excluded from the planning process, the process will be subverted by legal challenge. In either case, politics, not rational planning, determines the outcome. FORPLAN apparently also excludes people within the Forest Service from participating in forest planning. Volyas (1986), a Deputy Forest Supervisor on the Ollawa National Forest, commented on that Forest's plan: Because of the speed necessary to develop and or motivation or knowledge as well. Maybe some wereturned off by the complexity and sophistication of the FORPLAN model...some Forest and District people were left in the dust, so to speak. They were unwilling, or unable or not permilled to keep up with the planning process. (p. 119-120) Through this tendency for selective exclusion, the use of FORPLAN in national forest management planning incvitably alfects the internal organizational structure of the Forest Service. For example, one might hypothesize that FORPLAN is a tool for the timber management interests within the Forest Service to regain some of the ground they lost during the past decade. Bils of evidence support this hypothesis. For example, in their review of the history of FORPLAN, Iverson and Alston (1986) note that FORPLAN was acceptable to the agency only because of its roots in harvest scheduling optimization. As another example, I once visited a national forest in the midst of a forest planning crunch. Their FORPLAN model was not functioning as expected, and they needed help. Who was contacted? The Timber Management Start in ihe regional ollicel The longer term changes in the structure of the Forest Service which have been induced by FORPLAN are likely to be rather subtle. For the researcher in organizational behavior, it might be useful to follow the careers of the planners, operations research analysts and economists hired during the first round of NFMA planning. Will they rise through the ranks to become district rangers and forest supervisors? Or will the agency's immune system isolate them?
Traditionally the management plans for individual national forests have been negotiated locally among local interest groups. The exceptions have been rare and notably so: the Bitterroot and Monongahela National Forests, for example. The shift to a procedurally complicated, technically complex, standardized planning model has made the management practices on individual national forests easier targets for the powerful centralized critics--the Office of Management and Budgel, and the national interest groups such as the National Forest Products Association and the Wilderness Society. The Wilderness Society has been particularly effective in using the increased centralization to ils own advantage. Once the complexities of FORPLAN have been mastered, the cost of analyzing another forest plan is low. The scale economies of knowledge associated with a standardized planning process inevitably savor the national interest groups over the local ones. Elsewhere in this symposium others have argued that FORPLAN has served to shield the Forest Service from ils critics. Where the critics do not have the technical capability, time or money to master FORPLAN, perhaps this is so. But once the key to the central defense is discovered, the shield no longer serves well. A diversily of planning methods would be a beller, more permanent defense, particularly from the national interest groups to whom the Forest Service is most vulnerable.
Mathematical models exclude some considerations so that other aspects of a problem can be treated with great clarity and precision. This systematic exclusion of information is precisely the value of quantitative approaches to management and policy analysis. Yet this characteristic also produces the risk of Beuter and Iverson's "mathematical aphasia," or what is less eloquently known as a Type III error--solving the wrong problem. During the past three decades of large-scale social science modeling, the analytical landscape has become littered with modeling efforts wrecked by failure to include some important phenomenon. What important information does FORPLAN systematically exclude?
The 1979 directive requiring all national forests to use FORPLAN shisted power upward in the classically decentralized Forest Service. Increasing the degree of central control, while tending to standardize plans, limited on-the-ground testing of alternative analytical approaches to the forest planning problem. Neither FORPLAN nor its immediate predecessors were velled through the normal research procedures of peer review and publication. Because FORPLAN had not received the benefit of
Systematic exclusion of information in planning results in a more limited capacity to respond to future change. Because the future is inherently unknowable, some attention should be paid to "surprise rich" scenarios. By their very nature, such scenarios may be difficult to quantify: there is no historical record on which to base the conventional analyses. Data are history, but planning is about the future. "Too great a reliance on quantifying the past increases, rather than reduces ignorance aboul the future.
Beuter and Iverson cmphasize my sisth point: the current forest management planning process focuses on individual national forests without consideration of adjacent lands. This inevitably results in suboptimal social solutions to sorest sector problems. The best solutions may lie outside the national forests. For example, the Pacilic Northwest Forest Policy Project (Bruner and Hagenstein 1983) showed convincingly that a coordinalcd public-privale action was more successful in achieving a stable regional timbcr economy than was federal action alone. As another example, to serve the demands for outdoor recreation in New England, it may be more ellcctive lo improve slale and local parks near the population centers than to develop more capacity on the White and Green Mountain National Forests. Yet national forest management planning explicitly excludes consideration of such alternatives. The State and Private Forestry (S&PF) branch of the USDA Forest Service has suffered hard times recently. A conscrvative administration abandoned S&PF's historical mission of subsidizing timber production on private lands. An alternative, useful role would seek out those occasions where the best solutions to national forest problems lic on private or nonfederal public lands.
So, the costs of NFMA planning in general and FORPLAN in particular seem large. What are the benefits? Beuler and Iverson point out that the theoretical objective of FORPLAN is economic efficiency: to maximize net present benefits subject to constraints describing forest production possibilities, environmental regulations, and legal requirements. Haigh and Krutilla(1980) go further to describe economic efficiency as a requirement of the National Forest Management Act. Does FORPLAN achieve this objective? Partially. FORPLAN is apparently effective in eliminating prescriptions which are not cost-effective. This is no small task. But the larger objective of overall economic efficiency is not always served. Even where economic optimization is given free rein in FORPLAN, the resulting plan is unlikely to be selected as the preferred alternative. Others in this symposium have stressed the utility of FORPLAN as a simulation tool. If simulation is the objective, then linear programming is a poor analytical approach, and the principal benefit of linear programming. - identification of economically elficient plans--is lost. While economic elliciency might underlie NFMA, reconciliation of political discord was an overriding objcctive of the new planning procedures. The Gillmeir Index--the number of lawsuits filed against the Forest Service on mallers of policy and management--is a useful mcasure of how well the Forest Service has achieved this objective. The definitive statistics have not been compiled, but Gillmeir (personal communication) reports that the Index is probably down as administrative appeals have supplanted formal legal challenges. Perhaps the Gillmeir Index will rebound as these appeals mature into court cascs, or perhaps it will remain quiescent as controversies are reconciled through administrative channels. The long-lerm success of NFMA planning in reconciling discordant visions of the national forests remains unproved. Earlier in this symposium, two Forest Supervisors--Orville Daniels and Steve Mealy--gave optimistic views. Because FORPLAN is essentially a nonincremental, "zero-based" planning model, all the tough decisions surface at once. As a consequence, political discord peaks during the period when the plans are being prepared. Optimistic views of the planning process presume that the decisions forged during This contentious period will hold together throughout the planning cycle, and thereby the discord will be reduced over the longer run. Is this optimistic view warranted? The next decade's performance of the Gillmeir Index will provide part of the answer.
Plan implementation carries many dilliculties. Beuter and Iverson note the unrealistic budgets coming out of the current round of forest planning. Plans are implemented through budgets. What is the use of planning for a budget which is unlikely to be funded? At least some analytical essort should be devoted to exploring alternatives where the budget is constrained to realistic levels. This problem also occurs in the national RPA plans. Optimal sorcst plans do not simply "scale down" in response to budget constraints. That is, the optimal mix of program elements will shist as increasingly strict budget constraints are imposed. At a high budget level, the optimal plan might call for much recreation. As the budget is lowered, the optimal program mix might well be skewed towards a larger emphasis on timber. The gulf between assumed and actual budgets is not the only impediment to implementation. Other model assumptions may be at variance with reality. For example, the marginal willingness to pay for recreation on national forests is, by desinition zero unless a user lee is levied. In a linear programiming framework, a positive value for recreation must be used to insure that, on average, the land allocations are correct. But any prescriptions which call for additional recreation investment will be correct only if a see equal to the assumed marginal willingness to pay is actually levied. Otherwise, what guarantees that the recreationists who value the new facility arc thc ones who use it? The same problem would arise if a FORPLAN simulation valued timber at $150/mbl; but management practice gave timber away al no cost.
information be brought back into the planning via the political system? Iverson usefully distinguish model efficacy from typical of FORPLAN problems? Formal, quantitative modeling has a role in planning, managing, and developing policy on the national forests. Much work remains to exploit the full power of formal analysis. Given the size and importance of land management planning, research on the technical and instilutional aspects of the problem should be expanded.
This begs Beuter and Iverson's final question: "If not FORPLAN, what?" They argue that FORPLAN, or something very much like il, will occupy a central place in forest planning for quite some time. I agree. Sunk costs, so capriciously dismissed by economists, weigh heavy on administrators. Long bureaucratic memories and sheer institutional momentum guarantee happy pasture for formal planning models such as FORPLAN. A relevant question is how to improve the utility and performance or FORPLAN. Here, I have lwo suggestions. First, because FORPLAN is a market-like model, make greater use of market-like incentives. In the management of the national forests, for example, levy secs for recreation and wilderness use (Binkley and Mendelsohn 1987 discuss the advantages of using fees in outdoor recreation management). Imposing recreation user fees would bring management reality into concert with the economic assumptions behind the forest plans. Imposing fees would generale believable estimates of the value of recreation. Most important, imposing fees could create new revenues for the Forest Service. These revenues could fund the many, high economic relurn recreation investments which apparently exist on the national forests (Binkley and Hagenstein 1986). Similarly, use market-like incentives in the forest management organization. In a 1981 Resources for the Future conference on management of the national forests, I proposed that each national forest be organized as a profit center in the accounting sense (Binkley 1983). Where they exist, use market prices lo value the inputs and outputs associated with management of the forest; where market prices do not exist, impule surrogale prices. Pay the forest supervisor, district rangers and others of the forest according to the annual net surplus. To do so would require a set of accounts which would separate legitimate capital investments from operating expenses. By themselves, these accounts would help in analyzing some of the current national forest management controversies such as the socalled below-cost timber sales problem. If economic efficiency is an objective of using FORPLAN in NFMA planning, then market-like incentives in management and organization would surely help achieve it. Second, do needed research on land management planning. The usual research and development processes were not followed in developing FORPLAN. There apparently was little peer review, and no peer-reviewed publication of the model as it was developed. Those outside ihe inmediate circle of developers had little access even to the grey literature on the model. Lack of participation by the wider research community, particularly the research division of the USDA Forest Service, is surprising. Three areas seem to warrant particular altention in fulure research on land management planning: i. Analysis of the analytical process itself. What were the costs and benefits of the first round of to land management planning? change in response to FORPLAN? What kinds of information were systematically ignored because of the modeling efforts? How can this
Binkley, C. S. 1983. Comments. p. 237-244. Governmental interventions, social needs and the management of U.S. national forests. Roger Sedjo, ed. Johns Hopkins University Press, Baltimore, MD. Binkley, C. S. and R. O. Mendelsohn. 1987. Economic advantages of recreation user fees. J. Forestry (In press). Binkley, C. S. and P. R. Hagenstein. 1986. Economic analysis of the 1985 RPA program. (Unpublished manuscript.) 12 p. Bruner, W. and P. R. Hagenstein. 1983. Alternative forest policies for the Pacific Northwest. Prepared for the Pacific Northwest Regional Commission Forest Policy project. Washington State University, Pullman, WA.
Haigh, J.A. and J. V. Krutilla. 1980. Clarifying policy directives: The case of national forest management. Policy Analysis 6:409-439. FORPLAN: a historical and analytical review of Forest mountain Research Station, Ogden, Utah. Navon, D. I., G. Veiga, R. J. Hrubes, and A. F. Weintraub. 1986. Mar- Loss: An automated procedure for calculating the maximum loss of optimality in aggregating FORPLAN problems. p. 90-96. In Lessons from using FORPLAN: Proceedings of the workshop. (Denver, CO, 29 April-1 May, 1986] USDA Forest Service, Land Management Planning Systems Section, Washington, D.C. 268 p.
Schweitzer, D., F. Norbury and G. Alward. 1986. Economic elliciency and national forest planning. (Unpublished manuscript.) 13 p.
Votyas, F. J. 1986. Managing FORPLAN for analysis and decision making. p. 119-122. In Lessons from using FORPLAN: Proceedings of the workshop. (Denver, CO, 29 April-1 May, 1986) USDA Forest Service, Land Management Planning Systems Section, Washington, D.C. 268 p.
Simon, H. A. 1978. On how to decide what to do. Bell J. Econ. 9:494-507. Page 8
effects of data error structures on model performance, or even allow a model user to become more familiar with the model behavior, it follows that there will be more computer time spent on the models to accomplish these goals.
TECHNICAL CONSIDERATIONS WHEN MODELING ECOLOGICAL SYSTEMS USING FORPLAN
A point that is related directly to the size of models and lo pragmatic considerations involves the propagation of errors in model parameters and the reliability of the data in Jarge sels of input data. The experience in most ecological modeling exercises has been that models of intermediate complexity have normally had the greatest forecasting ability (O'Neill 1975). This can be altributed, in part, lo the fact ihal almost all parameters in ecological models have error in their estimates. It is a nontrivial problem to identify whether it is a worse error to leave out a given feature in a model rather than include it when there is error in estimating the parameters associated with its incorporation in the model. In most ecological models, this problem cannot be solved by a rule of thumb. Most sensitivily analyses of large models focus on a few parameters that appear to be the dominant factors for a given model application. One might presume (as has been done regularly) that is the error in the more sensitive model parameters is under control then the model predictions will be reliable. In cases in which the model is being used to interpolate system dynamics this may well be true, but the potentially dramatic impact of model structure sensitivity analysis must be kept in mind.
THE ECOLOGICAL IMPLICATIONS OF THE NATIONAL FOREST MANAGEMENT ACT
DIFFICULTY IN COMPARING INTUITION AND MODEL RESULTS
Somewhat related to the sorts of considerations just mentioned concerning large models is the difficulty in being able to develop an intuitive understanding about why a model produces a given result. This criticism places the model developer in a rather difficult position in that large models are often developed specifically lo investigate problems for which there is no intuitive solution. Even if This criticism allacks the model for doing exactly what it was designed to do sometimes,a real problem is that users distrust models that produce answers which are hard to inler from the causes. The solution of the problem probably lies in exercising the model to the extent that the user is more familiar with the inner workings of the algorithms, but this approach involves the user learning a lot about the model but not necessarily a lot about reality (Barber 1986).
Before one can evaluate any model there must be a basis upon which to judge the value of the tool. In this instance, il seems mosi l'air lo judge the FORPLAN model against Section 6 of the National Forest Management Act (NFMA). One might argue that the regulations also provide à basis for judgment, but they are one step removed from the act itself and embody extensive interpretation. Also, the regulations may have been written with a particular lool in mind, thereby imparting some bias in their requirements. Therefore, we will use the act alone as the anchor for the evaluation. The principal points made in Section 6 of the National Forest Management Act concerning ecology can be summarized succinctly. NFMA Sec. 6 (e) "(1) provide for multiple use and sustained yield of the products and services obtained therefrom in wilderness; and harvesting levels, and procedures in the light of management. NFMA Sec. 6 (g)(3) "(A) insure consideration of the economic and environmental aspects of various systems of , watershed, wildlife, and fish; communities based on the suitability and
COMPUTER COSTS INVOLVED IN RUNNING MODELS
Large models are often expensive to run on computers or at least appear expensive to the non-computer aficionado. This is the case with implementations of the FORPLAN model on more complicated problems. The cost of computer time is something of a red herring in that the cost of conducting research under lield conditions (at least in the ecological sciences) is usually larger than these computer costs by a great margin. This is the case even when the clata collection is justified by a model development goal. The criticisms relating to model complexity just discussed have as their solution more frequent implementation of the models on computers. If one would like to do sensitivity tests, to determine the Page 9
The objective is to provide information which allows the exploration of resource production potentials using ecologically sound management. Allempts have been made to accomplish these objectives by creating decision variables for the LP model which are designed for independent, homogeneous analysis areas and using constraints to achieve spatial feasibility. An alternative approach can also be used when one realizes that activities in a geographic arca are not independent either in terms of the activities themselves or the consequences of the activitics. Decision variables developed based on this philosophy include an implicit land allocation objective and a resource activity schedule besides the environmental essects. The activity schedule is designed spatially and temporally to meet the management objectives implied by the prescription and to comply with the attendant standards and guidelines for such things as maximum opening size and wildlife habitat dispersion requirements. This general approach, whether it is purely area based or a mixed area based, strata based approach will be termed coordinated scheduling to simplily the remainder of the discussion (Iverson and Alston 1986). Several ecological research areas to lend themselves to consideration in this regard. In this linal section we will discuss some possible considerations that could improve the FORPLAN Version 2 system in its capability as an ecosystem analysis tool used for the maintenance of the diversity of species (which we earlier identified as one of the principal ecological soci in the National Forest Management Act). The approach proposed also can be generalized to include other resources such as water, range and timber.
of a plant or animal species can survive, grow and reproduce. A species' habitat is often quantified in terms of the ranges of physical and biotic factors associated with the presence of individuals of the species. For many animals and particularly for vertebrates (Morse 1968, Shugart and Patten 1971, Dueser and Shugart 1979, Franzreb 1983) the structure of the vegetation is a major element in determining the suitability of the habital (Hilden 1965). Because the vegetation can be allered by both natural (succession, disturbances) and human-controlled processes (timber harvest), the ability to project vegetation change over time is an essential element both of managing certain species' habitat for productivity and recreational uses on a continuing basis and of minimizing the potential extinction of certain other species. During the early 1970s the USDA Forest Service was a lcader in working with the problem of managing public lands to insure the maintenance of diversity. The agency sponsored open symposia (e.g. Smith 1975, Capen 1981) on scientific problems and applications and helped to catalyze an innovative period for scientists interested in exploring habitat management techniques. Two different management approaches were developed under direct influence of the economic and political constraints attendant to the land management problems in their respective regions of the United States. These lwo concepts were the "scalured-spccies plan" (Zecdyck and Hazel 1974) used in the productive forests of the southeastern United States and the "area-diversity plan" (Evans 1974) used in the hardwood forests of Missouri. In the seatured-species plan (Zeedyck and Hazel 1974), species of birds or other animals (usually traditional game animals) were studied to better understand their ecological requirements and to develop a plan for forest management. The resulting plan would optimize the availability of habitat for the seatured species considering other resource values in the area. In contrast, the area-diversity concept (Evans 1974) was designed to create an optimal diversity of animals by managing land to insure the presence over time of all the recognizable habitat types in a region. The plan evolved in Missouri in a hardwood forest region in which all but the most productive forests were used for recreation and watershed protection. The area-diversity plan uses ecological diversity as a central principle and uses the treatment of blocks of land (ca. 40 ha) to maintain a landscape mosaic of different habitat types. Tracts of land are cut and even-aged forests are regenerated on the cutover sites. The amount of land cleared in any given year is designed to initiate a sequence of tree regeneration and regrowth leading to mature forest. Such a system is very conservative in terms of how much land might be harvestcd in a given year. These approaches to developing management plans, and others (Gill et al. 1974, Holbrook 1974), have evolved considerably since the early 1970's. For example, a management plan that was an area-diversity plan might use a leatured-species plan on areas that harbored a rare and endangered species or an important game species (Zeedyck and Evans 1975, Harris et al. 1984). The Forest Service now uses as a matter of policy a synthesized version of the two approaches called management indicator species (Mealey
HABITAT MANAGEMENT: CONCEPT DEVELOPMENT AND EXAMPLE CASES FROM THE USDA FOREST SERVICE
Continued presence of suitable habitat for the species over time is an ecological analog to the problem of managing a forest for a sustained yield of forest products. Habitat is defined as the kind of place(s) where individuals Page 10
Sincere gratitude is extended to all those who contributed their suggestions in the preparation and review of this paper: Bob Bailey, Sarah Crim, Bill Gast, Richard Holthausen, Don Jameson, Norm Johnson, Brian Kent, Mit Parsons, Hal Salwasser, Jim Sisler, and Reuben Weisz.
Aubreville, A. 1933. La foret de la Cote d'Ivoire. Bull. Comm. Alr. (ccid. Franc. 15:205-261. Aubreville, A. 1938. La foret colonial: Les forets de l'Afrique occidentale francais. Ann. Acad. Sci. Colon., Paris 9:1-245. Bailey, R. G., Icchnical coordinator. 1986. Proceedings of the workshop on lessons from using FORPLAN; April 29-May 1, 1986; Denver, Colo. Washington, D. C.: USDA Forest Service, Land Management Planning Systems Section. Bailey, R. G. 1984. Land delineation survey: Background and status. In: G. Lund, editor. Proceedings of the Forest Land Inventory Workshop. March 26-30, Denver, Colo. USDA Forest Service, Washington. Barber, K. H. 1986. Large FORPLAN models: An exercise in folly. In: R. (j. Bailey, technical coordinator. Proceedings of the workshop on lessons from using FORPLAN; April 29-May 1, 1986; Denver, Colo. Washington, D. C.: USDA Forest Service, Land Management Planning Systems; 1986. systems analysis in wildland management: An overview. Paper presented at ORSATIMS Joint National Meeling, Philadelphia, PA. March 31 - April 2, 15 p. Benson, G. L. and W. F. Laudenslayer. 1986. Simulating willlise responses to forest management strategies. In: Madison, Wisc. p. 351-356. Process in a Forested Ecosystem. Springer-Verlag, New York. 253 p. Bormann, F. H. and G. E. Likens. 19796. Catastrophic disturbance and the steady state in northern hardwood forests. Am. Sci. 67:660-669. Botkin, D. B., J. F. Janak and J. R. Wallis. 1972. Some ecological consequences of a computer model of forest growth. J. Ecol. 60:849-873. Boyce, S. E. 1980. Management of forests for optimal bcncfits (DYNAST-OB). USDA Forest Service Research Paper SE204. Southeast Forest Experiment Station, Asheville, NC.
Brand, G. J., S. R. Shilly and L. F. Ohmann. 1986. Linking wildlife and vegetation models to forecast the effects of management. In: J. Verner, M. L. Morrison and C. J. Ralph, editors. Wildlife 2000: Modeling Habitat Relations of Terrestrial Vertebrales, University of Wisconsin Press, Madison, Wisc. p. 383-388. Capen, D. E. 1981. The use of multivariate statistics in studies of wildlifc habitat. USDA Forest Service General Technical Report RM-87, 249 p. Rocky Mountain Forest and Range Experiment Slation, Fort Collins, Colorado. Casti, J. 1983. Forest Monitoring and harvest policies. Applied Mathematics and Computing. 12:19-48. Dress, P. E. 1982. Quantitative methods for national forest land management planning. 266 p. USDA Forest Service. Southern Region. Atlanta, Georgia. Dueser, R. D. and H. H. Shugart. 1979. Niche pattern in a forest-floor small mammal fauna. Ecology 59:89-98. Evans, R. D. 1974. Wildlife habitat management program: A concept of diversity for the public forests of Missouri. p. 73-83. In: J. P. Slusher and T. M. Hinckley, editors. Timber-Wildlife Symposium. Missouri Acad. of Science, Occasional Paper No.3. Franzreb, K. E. 1983. A comparison of foliage use and tree height selection by birds in unlogged and logged mixed coniferous forest. Biol. Consv. 27:259-275. Fretwell, S. D. 1969. On territorial behavior and other factors influencing habitat distribution in birds. III. Breeding success in a local population of lield sparrows а (Spizella pusilla Wils.). Acta Biotheor. 19:45-52. Gill, J. D., R. M. Desiraal and J. W. Thomas. 1974. Forest habital management for non-game birds in Central Appalachia. USDA Forest Service Res. Note NE-192. 6 . p. Green, R. H. 1980. Multivariate approaches in ecology: The assessment of ecologic similarity. Annual Reviews of Ecology and Systematics 11:1-14. Biogeography Theory and the Preservation of Biotic Diversity. The University of Chicago Press, Chicago, Illinois. 211 p. Zool. Fenn. 2:53-75. Holbrook, H. L. 1974. A system for wildlife habitat management on southern National Forests. Wildl. Soc. Bull. 2:119-123. Holling C. S., editor. 1978. Adaptive environmental assessment and management. John Wiley and Sons. New York. 377 p. for coordinated resource management. In: J. Verner, 376. Holthausen, R. S. and N. L. Dobbs. 1985. Computer assisted tools for habitat capability evaluations. A CO, July 30, 1985.4 p. Horn, H. S. 1975. Forest succession. Sci. Am. 232:90-98. Page 11
Iverson, D.C. 1986. Later development of FORPLAN: FORPLAN Version 2. p. 23-32. In: R. G. Bailey, technical coordinator. Proceedings of the workshop on lessons from using FORPLAN; April 29-May 1, 1986; Denver, Colo. Washington, D.C.: USDA Forest Service, Land Management Planning Systems; 1986. Iverson, D. and R. Alston. 1986. The genesis of FORPLAN II: A historical and analytical review of USDA Forest Intermountain Research Station. Ogden UT. optimal management strategies based on stochastic dynamic ecological models. (In Press). Johnson, K. N. 1983. Assessing the environmental feasibility of proposed timber schedules for the national forcsts: App. Management Planning Systems Section, USDA Forest Service, Fl. Collins, Colo. Johnson, K. N. and T. Stuart. 1984. Representing multiple outcomes in timber management planning. Mimeo. Land Management Planning Systems Section, USDA Forest Service, Fort Collins, Colo. 28 p + App. Structures and options guide. Land Management Collins, Colo. FORPLAN Version 2: An Overview. Land Management Planning Systems Section, USDA Forest Service, Fl. Collins, Colo. Jones, D. B. 1986. Early development of FORPLAN. p. 11- 22. In: R. G. Bailey, technical coordinator. Proceedings of the workshop on lessons from using FORPLAN; April 29-May 1, 1986; Denver, Colo. Washington, D. C.: USDA Forest Service, Land Management Planning Systems; 1986. Kessell, S. R. and M. W. Potter. 1980. A quantitative succession model sor nine Montana forest communities. Environ. Manage. 4:227-240. Kessell, S. R., R. B. Good and M. W. Poller. 1982. Computer modeling in natural area management. and Wildlife Service, Canberra. Kock, L. L. de, D. M. Stoddart and H. Kacher. 1969. Notes on the behavior and food supply of lemmings (Lemmus lemmus L.) during a peak density in southern Norway, 1966/67. Zeils. sur Tierpsychol. 26:609-622. Mealey, S. P. and J. R. Horn. 1981. Integrating wildlife habitat objectives into the forest plan. Trans. N. Amer. Wildl. and Natur. Resour. Conf. 46:488-500. Mealey, S. P., J. F. Lipscomb and K. N. Johnson. 1982. Solving the habitat dispersion problem in forest planning. Trans. N. Amer. Wildl. and Natur. Resour. Cons. 47:142-153.
Mitchell, T. R. 1986. Use of FORPLAN V2 to meet the analysis requirements on the Shushone National Forest. In: R. G. Bailey, technical coordinator. Proceedings of the workshop on lessons from using FORPLAN; April 29-May 1, 1986; Denver, Colo. Washington, D. C.: USDA Forest Service, Land Management Planning Systems; 1986. Moeur, M. 1983. Representing natural variability in multi resource simulation models using nearest neighbor sampling procedures. In: R. H. Lamberson, editor. Proceedings of the Second Pacific Coast Conference on Mathematical Modeling of renewable resources, Univ. Of Victoria, Victoria, B. C. Morse, D. H. 1968. A quantitative study of male and female spruce-woods warblers. Ecology 49:779-784. O'Connor, R. J. 1981. Habitat correlates of bird distribution in British census plots. p.553-537. In: C. J. Ralph and J. M. Scoll (cds.) Estimating the Numbers of Terrestrial Birds. Cooper Ornithol. Soc. Stud. Avian Biology. O'Neill, R. V. 1975. Modeling in the Eastern Deciduous Forest Biome. p. 49-72. In: B. C. Pallen, editor. Systems Press, New York. on a data-rich model: 'The Colorado River experience. Madison, Wisc. p. 79-86. FORPLAN in project analysis. p. 130-142. In: R. G. Bailey, technical coordinator. Proceedings of the workshop on lessons from using FORPLAN; April 29. May 1, 1986; Denver, Colo. Washington, D. C.: USDA Forest Service, Land Management Planning Systems Section; 1986. and forest management. Trans. N. Amer. Wildl. and Natur. Resour. Conf. 46:473-437 Salwasser, H. and S. P. Mealey. 1983. A perspective on national forest planning for wildlife and lish resources. Proc. Symp. on RPA/NFMA Forest Forest Resource Planning Center for National Resource Sluclies. Berkeley, CA. 90 p. + App. Salwasser, H., S. P. Mealey and K. Johnson. 1984. Wildlife population viability: A question of risk. Trans. N. Amer. Wildl. and Natur. Resour. Conf. 49:421-439. Schantz, T. von. 1981. Female cooperation, male . competition and dispersal in the red lox Vulpes vulpes. Oikos 37:63-68. Shugarl, H. H. 1984. A Theory of Forest Dynamics. Springer-Verlag, New York. 278 p. Shugart, H. H. and B. C. Palten. 1972. Niche quantification and the concept of niche pattern. p. 283-327. In: B. C. Patten, editor. Systems Analysis and Simulation in Ecology, Vol. II. Academic Press, New York. Shugart, Herman H. and Deon L. Urban. 1986. Modeling habitat relationships of terrestrial vertebrales - The researcher's viewpoint. In: J. Verner, M. L. Morrison and C. J. Ralph, editors. Wilellise 2000: Modeling Habitat Relations of Terrestrial Vertebrates, University of Wisconsin Press, Madison, Wisc. p. 425-432. Page 12
and animal populations. As the authors point out, the difficulty is in realistically representing the consequences of uncertain events. The ecosystem can be seen moving from one equilibrium to another equilibrium. Risk analysis prior to the LP model has been used to reconcile uncertainty in minimum viable populations by determining the constraint necessary to keep the system from moving out of the desired equilibrium (Salwasser et al. 1984). This approach works when the consequences of the uncertain event to be avoided can be described, extinction of a population or spccies, and the LP model contains the necessary variables which must be constrained, habitat variables. This approach does not work for events, such as (loods or pest outbreaks. As Hor (1987) has discussed, the LP framework may not be the appropriate tool to model these events. Once ihe FORPLAN model is constructed, all aspects of the structure of system appear to be known with certainly. That they are accounted for with certainly is another problem. What is meant here is the uncertainty that cannot be explained, the ignorance of processes that are reflected in the variability and that cannot be linked to a casual explanation. This type of uncertainty, called "residual unpredictability" by Walters and Hilborn (1978), assects the ability of the model to project. A long line of modelers, including some ecological modelers such as Karplus (1975) and Wallers and Hilborn (1978), have long urged acceptance of the uncertainly in ecological systems, or rather, acceptance of our inability to precisely model ecological systems. Walters and Hilborn (1978) described current ecological management as a passively adaptive discipline: "We observe disturbed systems, synthesize sometimes elaborale models, and conduct optimization exercises which pretend that our actions will not affect the way we observe and learn in the future. Eventually, with luck, we are able to devise optimal l'eedback policies. These specify the best behavior in the face of a residual unpredictability whose probability characteristics we have discovered through experience. Along the way we waste a lot of time trying to understand how this residual variability arises, forgetting thal we usually cannot do anything about it." Our understanding of ecological responses and our ability to quantify some of the components of that residual unpredictability have increased over time. To say that we usually can never do anything about that uncertainty might be a strong statement. However, it is equally absurd to conclude that the residual unpredictability does not exist, as we do in models which have a lixed structure. The model represents the best quantification of the system. The impact of the known variability on this projection estimate must be ascertained to further reduce this component of the projection variability. Dixon and Howili's (1979) work in California showed possible approaches to model refinement using control theory. It is more difficult to assess the impact of the residual unpredictability of the projection estimate. The LP model structure makes it difficult to incorporate this uncertainty. It is difficult to link the movement of ccosystems from an cquilibrium characterized by a continuous smooth surface
Shugart and Gilbert (1987) propose an interactive relationship between FORPLAN and Forest Simulators shown in ligure 1. This approach has much utility. It addresses the problems of scale, incorporates some measure of uncertainly, and offers the potential of reducing the size of the final forest-level model. Perhaps, the most important point of this proposed improvement is the recognition of scale in quantifying ecosystem dynamics. What Shugart and Gilbert have proposed is a modeling system tailored to describe ecosystem dynamics al more appropriate levels of scale by appropriate modeling techniques. What will this approach require? First, it will require that species/habitat relationship models be constructed for selected species (or indicator species, etc). This requires multivariate data linking the species with the habital. Further, the multivariate dala describing the habitat must be of the same type of the data available in the vegetation/patch simulators. Second, it will require that vegetation/patch simulators be constructed for forest types on a forest. Again, this requires dala on forest management and tree dynamics. This requirement may not be too difficult to accomplish for the individual plant vegetation simulators, because these use much of the forest managers' intuitive experience along with available data. This approach will rely heavily on the choice of patch size. This becomes the indivisible unit, like analysis areas, and all forest dynamics must be described in terms of this size. It also will require inventory data to construct the species/habitat relationship models. Ultimately, a link between the needs of the analytical tools and the inventory would improve models.
Shugart and Gilbert proposed three criticisms of using large models: a distrust arising from model complexity-small is beautiful; the difficulty of comparing intuition and model results; and the computer costs involved in running models. These criticisms can be restated in a broader framework: the role of model abstraction in land management planning. Ecosystem behavior is known only in part, and never with certainly. The ability of the analysi lo abstract the environmental system using FORPLAN is crucial to the success of the actual forest model in clescribing ecosystem behavior.
While I agree with Shugart and Gilbert that the objective of model abstraction in land management planning is to allow the exploration of resource production potentials using ecologically sound management, must this entire process be quantified in the forest modeling system? What aspects of the ecological system must be abstracted in the modeling part of the planning process? What did NFMA imply about the objective of the analytical tools? Were the models to allow resource managers to reach conscnsus in the public involvement process, or were resource managers to use the models to gain insights into the production space of the system. What has FORPLAN done? I believe that FORPLAN altempled a dillicult goal: to be both types of models.
Innis, George. 1978. Grassland Grassland simulation model. Ecological Studies 26. Springer-Verlag, New York. Karplus, Walter J. 1977. The place of systems ecology models in the spectrum of mathematical models. p. 225-228. In: George S. Innis, editor. New Directions in the Analysis of Ecological Systems. Simulation Councils Proceedings Series. Vol. 5, No. 2. The Society for Computer Simulation. La Jolla, Calif. Milne, Bruce T. 1987. Hierarchical landscape structure and the forest planning model: Discussant's comments on a paper by H. H. Shugart and B. Gilbert. In: Proceedings of FORPLAN: An Evaluation of a Forest Planning Tool. [Nov. 4-6, 1986. Denver, Colo.) Wildlife population viability: A question of risk. Trans.
Allen, T. F. H., and Thomas B. Starr. 1982. Hierarchy. Perspectives for Ecological Complexity. The University of Chicago Press. Chicago. 310 p. Dixon, B. L. and R. E. Howill. 1979. Uncertainty and the intertemporal management of natural resources: An empirical application to the Stanislaus National Forest. Giannini Found. Monog. 38. California Agric. Exp. Sta., Berkeley. 95 p. Fairfax, Sally K. 1981. RPA and the Forest Service. Unpublished papers prepared for the Conservation Foundation's Institutes on RPA, Washington, D.C. 21) p. Hor, John G. 1987. FORPLAN: An Economic Perspective. In: Proceedings of FORPLAN: An Evaluation of a Forest Planning Tool. (Nov 4-6, 1986, Denver, Colo.) Hollings, C. S. 1978. Adaptive Environmental Assessment and Management. Wiley International Series on Applied Systems Analysis, Vol. 3. Wiley Press. Chechester, United Kingdom.
Shugarl, H. H. and Brad Gilbert. 1987. An Ecological Evaluation of FORPLAN in the context of National Denver, Colo.) observation monitoring systems. A practical approach. p. 235-240. In: Lavenroth, W. K., G. V. Kogerboe, and M. Flug, editors. Analysis of ecological systems: Stateof-the-are in ecological modeling. Developments in Environmental Modeling 5. Elsevier Sci. Publ. Co. Wallers, C. 1986. Adaptive Management of Renewable Resources. MacMillan Publishing Co. New York, 374 p. Walters, C. and R. Hilborn. 1978. Ecological optimization and adaptive management. Ann. Rev. Ecol. Sept. 9:157188.
Hierarchical Landscape Structure and the Forest Planning Model: Discussant's Comments
Abstract.--Special strategies are needed for building models sensitive to the hierarchical structure of landscapes. Such hierarchical models may be effective for extrapolating from fineto broad-scale predictions. In some applications, landscape dynamics may be simulated using enhanced Markov models, especially where the effects of neighboring patches are relevant. Combinations of geographic information systems and Bayesian models of wildlife habitat offer a powerful means of linking FORPLAN results to the ground.
Shugart and and Gilbert (This volume) raise many considerations regarding the application of FORPLAN 10 resource management in the National Forests. Their review covers two general topics: (1) philosophical considerations stemming from the application of large, complex models, and (2) ecological ramifications of the assumptions and algorithms of FORPLAN. These topics are linked by the constraints that a particular paradigm, either philosophical or ecological, impuse on the design and inplementation of a model. Here, my comments emphasize the structural aspects of landscapes. First, I comment on the complex hierarchical spatial structure of landscapes, thereby highlighting an alternative paradigm which may prove useful in landscape planning. Second, I suggest several modeling strategies reflecting the hierarchical structure of landscapes. These comments were prompted by Shugart and Gilbert's awareness of scale effects implicit in the "area-diversity" approach (Evans 1974) and the "featured-species plan" (Zeedyck and Hazel 1974). The application of FORPLAN in a spatial context may require an understanding of landscape heterogeneity and structure which change predictably with scale.
(1983), but anticipated by Shugart and West (1981). Specifically, small landscape entities, such as canopy gaps created by tree falls, have a very short lise span, while landscape features spanning hundreds of hectares persist for much longer periods of time. Therefore, the landscape is viewed as a hierarchy (Allen and Starr 1982, Allen et al. 1984), with small, dynamic entities layered over large, slowly changing entities. The static nature of large encompassing entities forms a constraint on the behavior of small entities. Foresters recognize this hierarchical pallern at all levels. Each Forest is unique at the continental scale, most likely because major environmental factors such as geological features, precipitation, temperature, and insolation vary complexly throughout the United States, providing a unique combination of conditions at each Forest. Within Forests, topographical, geological, and historical patterns create unique conditions among walersheds, thus constraining the types of forests possible within these smaller regions. Environmental constraints the slowly changing conditions of large entities within which small landscape elements exist. Consistent changes occur in the structure of landscapes as a sunction of scale for areas about the size of a lypical National Forest. Here, landscape "structure" is the "spatial relationship among the distinctive ecosystems or 'elements' present" (Forman and Godron 1986), where elements include common entities such as meadows, streams, ponds, and wooded patches. The scale-dependency of landscape structure means that landscape characteristics (e.g., the diversity of landscape elements) vary with the spatial scale used to make the measurements (fig. 1). The slope of the relationship is constant for a given landscape (within a specilic range of length scales). Thus, the slope of the line is a "scale-independent" (Mandelbrot 1983) characteristic of a landscape. This relationship rellects the hierarchical arrangement of small, dynamic patches within large, slowly changing patches, such as entire watersheds. Scale-dependent landscape structure suggests a possible limitation of FORPLAN when applied in a spatial context (c.g., Armel 1986, Ryberg and Gilbert 1986). A scale
Hierarchical Landscape Structure and Model Complexity
A growing interest in landscape ecology has prompted ecologists to view ecosystems in a spatial context (Naveh and Lieberman 1984, Risser et al. 1984, Forman and Godron 1986). Landscape studies emphasize the effects of heterogeneily on the flow of organisms and resources between ecosystems. Of primary interest are the relationships between the spatial and temporal scales at which ecosystem processes vary. Urban et al. (1987) develop a l'undamental relationship between the spatial and temporal extent of landscape entities; a relationship first described by Delcourt et al.
1 Assistant Professor, Department of Biology, University of New Mexico, Albuquerque, NM 87131
dependent relationship may be anticipated for both the parameters and yield streams of FORPLAN, just as the diversity of landscape elements may change with the scale al which measurements are made (fig. 1). A major discrepancy between the predictions of the model and the intended prescription (specilied by the National Forest Management Act) may occur when the results are applied to the Forest as a whole. For example, if species, or landscape, diversity in an approximately 40 ha area is used as a constraint to determine yield or prosils, application of the model's predictions may be inappropriate when applied to a different spatial scale, such as the entire forest (sig. 2). At the very least, prorating diversity values to broad areas without consideration of scaling relationships may degrade the integrity of the landscape as a whole. Effects of scale-dependent structure need to be considered when data are aggregated at broad scales lo reduce the complexity of the matrix entered into the linear programming module. To see why, consider that spatially complex landscapes may be represented as "percolating networks" (sensu Orbach 1986) exhibiting truly remarkable properties. The simplest percolating network is a set of points on a lattice, with the probability of a point being present determined by an arbitrary threshold, 0 <P 3 1. If vacant points represent clearcuts, and occupied points are viewed as forest, then a dramatic elfect on the number of potential interactions between forested and cleared land is associated with the critical probability of 0.5928. Surprisingly, when the probability of clearing is > 0.5928, there will appear a cluster of cleared areas large that it spans the width of an infinite plane (Orbach 1986). Also at
Figure 2.--Scale-dependent landscape structure may preclude Page 13
that threshold, the relative number of interfaces between forest and clearculs begins lo diminish compared with the interfaces between clearcut patches (fig. 3). Granted, the average density of clearcuts is rarely, if ever, equal to 59%. However, the aggregation of data al different scales, to reduce the complexity of the matrix input to the linear programming routine in FORPLAN (Barber 1986, Navon et al. 1986), may alter the conscqucnces of percolating network geometry for estimations of yield at broad spatial scales. This occurs because a landscape is finite in area, and aggregation at coarse scales may result in finite clusters because of "sampling" effects which occur at coarse grid sizes. Therefore, the interfaces between large landscape entities (some of which result from the management regime) may obey the relationship in figure 3 based on a different probability value than the original fine-scale percolating network. Aggregation of data at different scales is an essective means of reducing model complexity, although it may drastically alter the interactions occurring among extensive landscape entities, such as watersheds. In summary, landscapes have a hierarchical scaledependent structure. Small entities exhibit rapid changes compared with large entities. As a result, broad-scale features (e.g., mountain ranges, watersheds) form the longterm context for smaller enlilies, such as clearculs and canopy gaps. A growing interest in spatial applications of FORPLAN (Armel 1986, Ryberg and Gilbert 1986) suggests that special strategies are needed for building models sensitive to the hierarchical structure of landscapes. Next, I suggest several possible technical solutions to problems inherent in adopting a hierarchical paradigm.
g(x) = b, the Lagrangian function, L, is desined as L(x,x) = f(x) +2 19(x)-b] = g where is a Lagrange multiplier. Specification of the stationary points of L yield two equations for the unknowns, X and X: dL = df +, dg = 0 dx dx dx
L = g-b = 0 da Solving for the stationary points of the equations gives ** and, which can be equaled to min f(x) (Haimes 1977). Application of this approach to landscapes requires a series of models for individual analysis areas (considered here to be precise geographic locations), with knowledge of the couplings between adjacent areas. Couplings might include the amount of energy, money, silt, water, and nutrients flowing from one analysis area to another. Then, as described by Haimes (1977), a hierarchy of Lagrange multipliers is developed io coordinate the analysis area models into a complete, optimized Forest system. Most important, the solution to such a model dilfers from the solution obtained by prorating a single Foresl-wide model to smaller landscape units.
Potentinl Modeling Strategies Reflecting the Hierarchical Structure of Landscapes
The scale-dependency of landscape structure implies that simple, linear extrapolation of model results obtained alone scale lo other scales may not ensure that the necessary yields or amenities are obtained. Granted, the unique nature of each Forest demands that the application of FORPLAN be tailored to each location. However, a set of independent models and prescriptions may not combine to satisfy the national-level mandate of Congress put forth by the National Forest Management Act (Haimes 1977). Individual Forests face the same problem is optimization is done for individual analysis arcas (e.g., Armel 1986). The optimal solution for a local management unit may be incompatible with the long-term goals of the Forest. Similar problems occur in water resources management, and the hierarchical approach developed by Haimes (1977), and provided below in slightly modified form, seems appropriate for the extension of FORPLAN to a series of specific, interacting, management units. The key to this approach is the Lagrange multiplier, which helps link several small models hierarchically. It has the added bencfit of accommodating nonlinear constraint cquations, which
Landscape dynamics are relevant to discussions of wildlife habitat and resource management. The appropriate habitat for a species shills spatially through time as successional plant communities age, so a dynamic model would be helpful. Two aspects of landscape structure and dynamics suggest limitations of simple Markov models besides those shown by Shugart and Gilbert. These limitations may be reduced or climinated. First, the "rules" applied to a landscape change through time because of Congressional mandales, public opinion, or Forest Service policy. Thus, the transition probabilities describing the likelihood of change from one cover type to another also change. Simple Markov models have constant transition probabilities, making them insensitive to changes in the rules controlling landscape change. Second, neighboring patches may affect the transition probabilities of particular patches. For example, the staggered cutting method precludes the clearing of uncut timber if it is next to an existing clearcul. Damage from windthrow or fire may occur if a stand is next to a clearcut (Jerry Franklin, personal communication). Thus, a
"neighborhood" effect should be incorporated into Markov models. This is done using "conditional" transition matrices, where the probability of changing from one cover type lo another is conditional, or depends, on the identity of the adjacent patch. Dr. Jcan Hartman, at Harvard University, and I are developing such a landscape model incorporating both dynamic and conditional transition matrices. The greatest assets of Markov models are that few parameters and small amounts of data are required for simulation (compared with forest gap models). Thus, Markov models are a practicable way of reducing the number of variables required.
predictability if only very coarse resolution habitat information were available or if species sensitive to linescalc habitat variation were of interest (MacArthur 1972). In summary, recognition of complex spatial structure in landscapes requires special approaches to landscape modeling. Interactions between adjacent (and perhaps distant) patches precludes simple prorating of estimates over wide areas, or among locations with vastly different environments. Hierarchical models reflecting the scaledependent structure of landscapes may be effective means ol' extrapolating from sine- lo broad-scale predictions. In some applications, landscape dynamics may be simulated using enhanced Markov models, with the advantage of using many lewer parameters than are necessary in models based on gap dynamics. Twenty years ago sew people would have envisioned the present capabilities of FORPLAN. There is every reason that future versions will be sensitive to the hierarchical structure of landscapes, and incorporate knowledge of the surrounding palches lo predict changes within particular patches.
Wildlife Habitut, Bayesian Models, and Geographic Information Systems (GIS)
Managers often require knowledge of the precise location of animal populations or resources (Connelly 1986). Correlational studies of wildlife habitat (discussed by Shugart and Gilbert) may not provide such information, or it may be difficult to translate correlational relationships to maps. Bayesian classification provides a simple, alternative method for determining the most likely habitat locations for wildlife. Once developed, this technique could be readily applied to GIS-based versions of FORPLAN (Stephan 1986), and 10 FORPLAN coupled with a simulator of landscape dynamics, such as the Markov model described above. Precise predictions of wildlife habitat locations could be made using such versions of FORPLAN. A Bayesian model requires maps of cover types and Other landscape features such as slope and elevation. Each point on the landscape is described by a vector, x, containing the list of features present. Then, the "state" of wildlife populations, w, of each point on the landscape has a probability of occurring, conditional on the vector of landscape features. Examples of states include "wildlife species present" and "wildlife absent." According to Baycs' formula,
Allen, T. F. H. and T. B. Starr. 1982. Hierarchy: Perspectives for Ecological Complexily. Univ. Chicago Press. Intcrlevel Relations in Ecological Research and Management: Some Working Principles from Hierarchy Theory. USDA Forest Service General Technical Report RM-110, 10 p. Rocky Mountain Forest and Range Experiment Station, Fort Collins, Colo. FORPLAN. p. 143-152 In Bailey, R. G. (ed.). Proceedings of the workshop on lessons from using FORPLAN. April 29-May 1, 1986. Denver, CO. Washington, D. C., U. S. Dept. of Agric. Forest Serv., Land Management Planning Syst. 286 p. in lolly. p. 89a-890 In Bailey, R. G. (ed.). Proceedings of the workshop on lessons from using FORPLAN. April 29-May 1, 1986. Denver, CO. Washington, D. C., U. S. Dept. of Agric. Forest Serv., Land Management Planning Syst. 286 p. Connelly, W. J. 1986. Integrating harvest schedules with land management options. p. 97-109 In Bailey, R. G. (ed.). Proceedings of the workshop on lessons from using FORPLAN. April 29-May 1, 1986. Denver, CO. Washington, D. C., U. S. Dept. of Agric. Forest Serv., Land Management Planning Syst. 286 p. Delcourt, H. R., P. A. Delcourt, and T. Webb III. 1983. Dynamic plant ecology: the spectrum of vegetational change in space and time. Qual. Sci. Rev. 1:153-175. Evans, R. D. 1974. Wildlife habitat management program: A concept of diversity for the public forests of
Here, p(x/w;) is the state-conditional probability density function for & P(W) is the a priori probability of state Wj and P(W;1x) is the a posteriori probability of each staté, ; given the set of landscape features present. Neighborhood effects could be included in the vector of landscape features, if such elfects were shown to be important. This model of wildlife habitat is easily applied using a GIS, and may be tuned to incorporate consequences of landscape structure. The sensitivity of this approach to the hierarchical structure of landscapes is not known, except thal the model would probably provide less
Forman, R. T. T., and M. Godron. 1986. Landscape Ecology. J. Wiley and Sons. New York. Haimes, Y. Y. 1977. Hierarchical Analyses of Water Resources Systems. McGraw Hill. New York. MacArthur, R. H. 1972. Coexistence of species. p. 253-259 In J. Behnke, ed. Challenging Biological Problems. Oxford Univ. Press, Oxford, England. Mandelbrot, B. 1983. The Fractal Geometry of Nature. W. H. Frecman and Co. New York. Naveh, Z. and A. S. Lieberman. 1984. Landscape Ecology Theory and Application. Springer-Verlag. New York.
Ryberg, S. M. and B. Gilbert. 1986. Use of version II FORPLAN in project analysis. p. 130-142 In Bailey, R. G. (ed.). Proceedings of the workshop on lessons from using FORPLAN. April 29-May 1, 1986. Denver, CO. Washington, D. C., U. S. Dept. of Agric. Forest Serv., Land Management Planning Syst. 286 p.
Shugart, H. H. and D. C. West. 1981. Long-term dynamics of forest ecosystems. Amer. Scientist 69:647-652.
Stephan, J. W. 1986. Geographic information systems: A tool for forest plan implementation. p. 192-194 In Bailey, R. G. (ed.). Proceedings of the workshop on lessons from using FORPLAN. April 29-May 1, 1986. Denver, CO. Washington, D. C., U. S. Dept. of Agric. Forest Serv., Land Management Planning Syst. 286 p.
Navon, D. I., G. Veiga, R. J. Hrubes, and A. F. Weintraub. 1986. Max-loss: an automated procedure for calculating the maximum loss in optimality in aggregating FORPLAN problems. p. 90-96 In Bailey, R. G. (ed.). Proceedings of the workshop on lessons from using FORPLAN. April 29-May 1, 1986. Denver, CO. Washington, D. C., U. S. Dept. of Agric. Forest Serv., Land Management Planning Syst. 286 p. Orbach, R. 1986. Dynamics of fractal networks. Science 231:814-819. Risser, P. G., J. R. Karr, and R. T. T. Forman. 1984. Landscape Ecology: Directions and Approaches. II. Nat. Hist. Surv. Spec. Publ. No. 2. Champaign, III. 18 p.
Urban, D. L., R. V. O'Neill, and H. H. Shugarl, Jr. 1987. Landscape ecology: A hierarchical perspective. : A BioScience 37:119-127.
Zeedyck, W. D. and R. B. Hazel. 1974. The southeastern featured species plan. p. 58-62 In J. P. Slusher and 'T. M. Hinckley (eds.). Timber-Wildlife Symposium. Missouri Acad. of Science, Occasional paper No. 3.
An Evaluation of FORPLAN from an Operations Research Perspective
B. Bruce Bare and Richard C. Field
Abstract.--FORPLAN is a computer system for developing linear programming models in support of National Forest planning. Technically, most deficiencies of LP can be overcome to correctly produce large-scale comprehensive models. However, the validity, reliability and usefulness of these models are in doubt because their opaqueness may exceed the analysts' ability to adequately interpret them for use in supporting politically-charged decisions. Better definition of the planning problem and closer tailoring of the analytical support is suggested.
The following evaluation of FORPLAN is offered in an effort to explain some of the characteristics of National Forest planning and to make future improvements in the planning process. While at times critical, our comments are not directed at any region, forest or individual. To the contrary, we recognize the many line accomplishments of a large cadre of men and women who worked tirelessly lo develop, improve, and implement FORPLAN during the past few years. The many accomplishments altributable to FORPLAN are exclusively the result of their efforts. We also recognize that our evaluation is blessed with the added advantage of hindsight. Thus, the faults we lind must be interpreted in this light. Finally, we acknowledge the direct and indirect help of the many people who have been involved in the development, refinement, maintenance, and implementation of FORPLAN.
embracing the above listed lenels, an O.R. perspective also implies the adoption of a systems approach to problem solving. Coupled with a decision orientation and reliance on a quantitative assessment of alternatives, O.R. practitioners presume that a rational-comprehensive approach will lead lo an acceptable solution to any given problem. The past few years of forest planning have raised doubts about this last assumption. Nevertheless, with such a broad charter our evaluation could easily encompass almost everything in this symposium, and, we might add parcnthetically, would be in keeping with the comprehensive nature of O.R. Further, we recognize that the analyses that natural resource economists (Krutilla and Fisher 1985) and systems ecologists (Shugart 1984) have described fall within this framework. Partly, this is because many scientific disciplines have embraced some of the lenels enumeraled above, and partly it is because ().R. itself follows the scientific method. Because we do not wish to duplicate the evaluations from these points of view we must necessarily limit our evaluation somewhat. Thus, our task is not to address the "problem" of National Forest management, or the "decisions" to be made by National Forest managers. While essential lo comprehensive evaluation, we will assume (perhaps incorrectly) that these evaluations will occur elsewhere. Instead, our objective is to evaluate the previously chosen "mathematical apparatus" -- FORPLAN: Versions 1 and 2 (Johnson 1986, Johnson et al. 1986). We will consider its lcchnical basis and how well it has provided the information needed by decision makers. Simply put, we raise three questions: (1) Does FORPLAN work? (2) Is il the right technique and is it used correctly within the decision environment of National Forest planning?, and (3) Are the results useful? Simply put, the answers are: (1) Yes, but (2) Possibly, but probably not, and (3) Occasionally. We begin by describing the mathematical basis of FORPLAN, its inherent assumptions, and their implications. We note the "problem solving" and "decision making" difliculties that have been generated as a result of these implications, and we speculate about why FORPLAN evolved as it did and note some possible alternatives.
Perhaps the most difficult part of this paper will not be the evaluation of FORPLAN, but determining what is an "Operations Research" (O.R.) perspective. The president of the Operations Research Society of America, Stephen M. Pollock (1986), speaking to the diversity and broad charter of operations research recently said that, "... the great trinity of O.R.: problem solving (agenda setting, lixing goals and objectives); decision making (evaluating and choosing); and the necessary mathematical apparatus needed to achieve this name your favorite technique, algorithm or heuristic) [leads to ... the almost hubristic willingness of most opcrations researchers to attack normative or descriptive problems of the widest conceivable scope ..." Besides
1B. Bruce Bare is Professor, College of Forest Resources, University of Washington, Seattle, WA 98195. Richard C. Field was Operations Research Analyst, USDA Forest Service, Southeastern Forest Experiment Station when the paper was prepared. His current address is 150 Cloverhurst Terrace, Athens, GA 30605.
Finally, we make some recommendations about addressing the "real problem," which heretofore has gone unanswered.
FORPLAN Is A Linear Programming Approach
Max Z c'x, subject to Ax < b, x > 0 where: A is an m by n matrix of constraint coefficients (n > m), where m is the number of constraints and n the number of decision variables.
FORPLAN (Versions 1 and 2) is a linear programming (LP) model. As such, it seeks to optimize a single linear objective function subject to a set of linear equality or inequality constraints. As described by Iverson and Alston (1986), FORPLAN is the outgrowth of a series of LP models developed and used by the Forest Service during the past 20-25 years. Without repeating this history, let il suffice to say that forest-level planning has undergone revolutionary changes during this time. Chiel among these has been the legislated need for comprehensive and integrated multi-resource planning; public involvement in all phases of planning; the introduction of compulcrized planning models and data bases; and increased pressures on all the resources of the National Forest System. FORPLAN is the mandated planning model in use on the National Forests and is an outgrowth of several earlier computerized planning systems: (1) RCS (Resource Capability System), (2) RAA (Resource Allocation Analysis), (3) Timber RAM (Timber Resource Allocation Method), (4) MUSYC (Multiple Use Sustained Yield Calculation Technique) (5) ADVENT (A Model for Program Budgeting), and (6) IRPM (Integrated Resource Planning Model). All these LP models, plus others, influenced the development of FORPLAN Version 1, and subscquently, Version 2. A full accounting of this history is available in Iverson and Alston (1980), Johnson et al. (1986), Jones (1986), and Iverson (1986). For completeness, and to betler understand FORPLAN and how it is used in forest-level planning, we begin our evaluation with a clelinition of LP along with its inherent assumptions and limitations.
b is a vector of m constraint levels, O is a vector of zeros. An LP problem is likely to have an insinite number of infeasible solutions (do not satisfy the constraints) and, it is hoped, at least one and possibly an inlinile number of feasible solutions (within the decision space). The simplex algorithm (Dantzig 1963) first finds a feasible solution, then makes systematic moves to other feasible solutions if it can improve the value of the objective lunction. If il cannot, it assumes it has found the optimal solution and stops. There may be other feasible solutions known as alternative optima which have the same optimal value of the objective function, but differ in decision space. Solution time is a function of the number of decision variables and, more important, the number of constraints which deline the decision space. Theoretically, the solution time for the simplex increases exponentially with problem size (combinations of n, m and non-zero constraint coefficients), but it rarely does so in practice. A radically different algorithm developed by Narendra Karmarkar of Bell Labs promises to do even better because its solution time appears to be polynomially, rather than exponentially, bounded. However, its performance has not been as spectacular as hoped and it has not been tested on a full array of LP problems (Hooker 1986). Whatever the outcome of the battle of the algorithms, very large problems are not likely to be solved by either method in a reasonable amount of time.
What is Linear Programming?
LP Assumptions and Limitations
Linear programming (LP) is a special case of the more general form of optimization techniques called mathematical programming. Mathematical programming is simply the representation of a problem in mathematical terms coupled with a formalized technique to find the optimal solution, usually without enumerating and examining all possible solutions. It is assumed that there is not a unique solution to the problem. Otherwise, there would be a mathematically straightforward -- but not necessarily easy -- way to solve the problem. Generally, a different solution technique (algorithm) is needed for different forms of mathematical expressions. Thus, there are many forms of mathematical programming and, as might be guessed, the clifficulty of the algorithm increases with the complexity of the mathematical expression. In an LP model, all the expressions have the form of linear equations or inequations and the accepted -- and until recently, the only algorithm for solving such problems has been the simplex. The general form of the LP problem may be expressed in matrix form as:
Although the assumptions of LP are well known (e.g., see Dykstra 1984 or Kent 1980), the implications of these assumptions are not usually consciously considered by (.R. analysts or the general users of LP models. However, the interpretation of results and their implementation by onthe-ground managers depend heavily on these assumptions and their relationship to the "real worldl." Recently, Wilson (1986) revicwed the implications of these assumptions within the context of FORPLAN. We wish to pursue this topic in more detail. First, because of its linearity assumption, LP is often criticized as being too simplistic to model many real world phenomena. "The world is not lincar!" 'This is an oft heard battle cry, and is generally a valid observation. But, the implication of lincarity that is most often challenged is Page 14
it also may be horribly misleading about the considence that should be placed in the results. Is that the case with FORPLAN? It could be. Let us look more specisically at its assumptions, structure and characteristics and compare il lo other classes of LP models.
FORPLAN as an LP: Does it Work?
proportionally. This deficiency can be effectively overcome with proper formulation. For example, FORPLAN handles nonlinear timber yields with Dealing with nonlinearities in the constraints and the objective function are admittedly more dillicult. However, more complex algorithms can be used (c.g., quadratic programming), or the nonlinear constraints can be handled in a piece-wise linear fashion (Hrubes and Navon 1976). Such is the case with FORPLAN demand constraints. However, the more insidious implications of linearity osten are overlooked. These are the additivity and divisibility assumptions. Additivity means combination of feasible values for the decision variables produces a consequence which is the sum of the individual values. No interactions are permitted to cause variations from this total. The second implication divisibility allows a decision variable to take on a continuous range of values. In many LP models, however, integer-valued decision variables are required (e.g., FORPLAN Version 2 with coordinated allocation choices). To permit noninteger value implies infeasibility in the real world, making such a solution dillicult to implement. LP does not allow interactions and it does not guarantee integer solutions. More oficn than not, these are important considerations in forest-wide modeling as found on the National Forests. Carclul formulation of the model and cautious interpretation of the solutions may allow one to avoid incorrect results without resorting to more complex and costly alternatives such as nonlinear and mixed integer programming. However, this is doubt'ul with large, complex models such as FORPLAN, especially Version 2 with coordinated allocation choices. Another commonly cited criticism of LP is that it is deterministic. This is because probability functions, which are often nonlinear, are impossible to incorporate into LP moclels. Hence, all coellicicnts in an LP model are assumed to be known with certainty. Il we attempt to incorporate uncertainty into an LP model, both model size and solution time increase dramatically. Thus, probabilistic elements are usually omitted from LP models. A final weakness of LP is that it explicitly assumes there is a single decision criterion to be optimized and implicitly assumes there is a single decision maker. Neither of these generally hold for the National Forests in practice or for public decision making in general. There are many algorithms, procedures, and alternative modeling lechniques to overcome these deliciencies. National Forest planning has used few of these, and FORPLAN even lewer. As serious as these implications are, they are not a raison d'etre for dismissing LP out of hand. Invariably, no approach will produce a perfect model of the real world, because all models are abstractions which necessarily are simplifications of reality. However, some models perform better than others, given: first, the objectives of the modeler; second, the modcler's operational constraints; and, last, the sorm of the situation being modeled. The greatest danger, then, with using LP, or any mocleling technique, is not that it doesn't work perfectly, but that it is working less perfectly than the modeler (or client) thinks it is. Having a million-dollar LP model that gives results no beller than could be worked out on the back of an envelope is not only a glorious waste of money,
From the previous discussion, we can see that LP can be deceptively simple. But the assumptions and the potential solution difficulties can not only lead to misinterpretations (as will be discussed later); they can also prevent it from working at all. It is well known that in the early days some FORPLAN moclels were never solved because of their size. Some were probably low large to even lit in the machine. See, for example, Barber's (1986) excellent discussion on large FORPLAN models, and the critique of Version 1 by Iverson and Alston (1986). Given that the Forest Service was using one of the largest mainframe computers and a state-of-the-art solution code, this implies that FORPLAN models were exceeding the limits of experience in LP (Kent et al. 1986). Furthermore, FORPLAN models are generally of the most complex form of LP, known as "blending problems." Such models typically have several choices for their size and many non-zero constraint coefficients. These severely tax any solution algorithm by increasing the computations necessary to invert the matrix (an integral part of the simplex algorithm) and by requiring more iterations lo lind an optimum solution, which may be only marginally different from a solution arrived at thousands of iterations earlier. To be more specific, FORPLAN Version 1, Model II forms typically have very sparse matrices (less than 1% non-zero constraint cocslicicnts) because of the many Model || transfer rows. The Functional Mathematical Programming System (FMPS) SPRINT algorithm (Sperry Univac 1984) is designed to solve such problems very efficiently and generally does so. However, few forests use Model Il except those in Regions 5, 8 and 9, and at least one forest in Region 8 exceeded FMPS's unadvertised row limit of 8,192 when il generated its matrix. The alternative Model i form was much smaller but also Jenser, about 810%. Most Model I matrices are not so dense but are still likcly to be an order of magnitude denser than a nominally equivalent Model II. FORPLAN Version 2 has the ability to produce more complex models than Version 1, and thus has the potential to produce even denser matrices. We have seen them with as many as 20% non-zero constraint coerficients. This seldom occurs, however, because most Version 2 models use coordinated allocation which introduces many sparse transfer rows with a corresponding overall reduction in density. The alternative FMPS solution algorithm, OPTIMIZE, was considerably less efficient in solving FORPLAN models. However, it was believed to produce more accurate marginal information than SPRINT, and beller bases for post-optimal analyses. However, little use was made of FMPS post-optimal techniques because the time and funds that would have been necessary to perform such analyses were already committed to meeting other requirements. That limited experience suggested that such procedures were not very well documented and proved to be incslicicnt, the reasons for which are uncertain. Page 15
Iverson, David C., and Richard M. Alston. 1986. The genesis of FORPLAN: A historical and analytical review of USDA Forest Service planning models. USDA Forest Service General Technical Report INT214, 31 p. Intermountain Forest and Range Experiment Station, Ogden, Utah. Johnson, K. Norman. 1986. FORPLAN version 1: An overview. 85 p. Land Management Planning Systems Section, USDA Forest Service, Washington, D. C. Johnson, K. Norman, Thomas W. Stuart, and Sarah A. Crim. 1986. FORPLAN version 2: An overview. 110 p. Land Management Planning Systems Section, USDA Forest Service, Washington, D. C. Techniques for prescribing optimal limber harvest and investment under different objectives - discussion and synthesis. Forest Science Monograph 18,31 p. Johnson, K. Norman, and Philip L. Tedder. 1983. Linear programming vs. binary search in periodic harvest level calculation. Forest Science 29(3):569-581. Jones, Daniel B. 1986. Early development of FORPLAN. p. 11-22. In: Proceedings of the workshop on lessons from using FORPLAN. (Denver, CO, April 29-May 1, 1986) 268 p. Land Management Planning System Section, USDA Forest Service, Washington, D. C. Kallio, M., A. E. Anderson, R. Seppala, and A. Morgan, Editors. 1986. Systems analysis in forestry and forest industries, Studies in the Management Sciences 21. 487 p. Elsevier Science Publishers, Amsterdam. Kent, B. M. 1980. Forest Service Land Planners Introduction lo Lincar Programming. 63 p. Land Management Planning Systems Section, USDA Forest Service, Fort Collins, CO. Jr. 1986. Experience with the solution of USDA Forest MD. economics of natural environments. 300 p. Resources for the Future, Washington, D.C. Lee, D. B. Jr. 1973. Requiem for large-scale models. Journal of American Institute of Planners 39:163-178. Liittschwager, J. M., and T. H. Tcheng. 1967. Solution of a large-scale forest scheduling problem by linear programming decomposition. Journal of Forestry 65:644-046. Mar, Brian W. 1974. Problems encountered in multidisciplinary resources and environmental simulation models development. Journal of Environmental Management 2:83-100. Martin, A. Jell, and Paul E. Sendak. 1973. Operations research in forestry: a bibliography. USDA Forest Service General Technical Report NE-8, 90 p. Northeastern Forest Experiment Station, Upper Darby, PA. McLean, D. W. 1983. Planning for forest resource management: Some thoughts on method. New Zealand Journal of Forestry 28(3):339-355.
Meising, Paul. 1984. Integrating planning with management. Long Range Planning 17(5):118-124. Michael, Stephen R. 1980. Tailor-made planning: Making planning fit the firm. Long Range Planning 13(6):74-70. Mitchell, Thomas R. 1986. Use of FORPLAN V2 to meet analysis requirements on the Shoshone National Forest. p. 45-56. In: Proceedings of the workshop on lessons from using FORPLAN. (Denver, CO, April 29-May 1, 1986] 268 p. Land Managing Planning Systems Section, USDA Forest Service, Washington, D. C. Nautiyal, J. C., H. S. Ngo, and H. K. 'Thadaney. 1975. Land use for planning: A practical application of mixed integer programming. INFOR 13(1):19-35. Navon, Daniel I. 1971. Timber RAM...a long-range planning method for commercial timber lands under multiple-use management. USDA Forest Service Research Paper PSW-70, 22 p. Pacific Southwest Forest and Range Experiment Station, Berkeley, CA. Navon, Daniel I., and Andres Weintraub. 1986. Operational model of supply for wildland enterprises. p. 353-370. In: Systems analysis in forestry and forest industries, Studies in the Management Sciences 21, M. Kallio, et al, editors. 487 p. Elsevier Science Publishers, Amsterdam. Navon, Daniel I., Geraldo Viega, Robert J. Hrubes, and Andres F. Weintraub. 1986. Max-Loss: An automated . Forest Service, Washington, D. C. models for strategic planning. Studies in Management Science and Systems 10. 184 p. Elsevier Science Publishers, Amsterdam. Nazareth, L. 1980). A land management model using Dantzig-Wolfe decomposition. Management Science 26(5):510-523. Pelz, Dieler R. 1977. Determination of optimal growing slock levels by inventory theory. Forest Science 23(2):183-189. Pollock, Stephen M. 1986. The better of diverse. ORMS Today 13(4):16-17. Romesburg, H. C. 1974. Scheduling models for wilderness research. Journal of Environmental Management 2:159 177. Sassaman, R. W., E. Holt, and K. Bergsvik. 1972. User's manual for a computer program for simulating intensively managed allowable cul. USDA Forest Service General Technical Report PNW-1, 50 p. Pacific . Northwest Forest and Range Experiment Station, Portland, OR. Shugart, Herman H. 1984. A theory of forest dynamics. 278 p. Springer-Verlag, New York. Sperry Univac. 1984. Function mathematical programming system, FMPS level 9R2: Programmer reserence. Sperry Corporation, St. Paul, MN. Steuer, Ralph E., and Albert T. Schuler. 1978. An interactive multiple-objective linear programming approach to a problem in forest management. Operations Research 26(2):254-269. Page 16
Evaluation of FORPLAN from an Operations Research Perspective: Discussant's Comments
The paper by Bare and Field (1987) provides a useful overview of the technical considerations relevant to an evaluation of any large-scale, applied programming model. I will not comment on their opinions about the motivations that led to the adoption of FORPLAN because such speculations do not seem to me to be particularly relevant to the evaluation of FORPLAN as a technical entily. For the most part their discussion of the technical aspects of planning with FORPLAN agrees largely with my own preconceptions, so my remarks are confined to a few points that I think require additional clarilication. These points Call into three categories: data errors, the delinition of analysis areas, and hierarchical, or multi-level, planning.
most existing measurement errors related to timber resources are at the very least unbiased. For non-limber resources, though, the situation is not so clear. Conventions relating to the assessment of recreation potential, aesthetic resources, soil erodibility, wildlife habitat, water yield potential, range condition, and other non-timber resources have not been clearly established. These assessments are subject to large measurement errors, if we are even willing to reser to the kinds of judgments commonly used in making these assessments "measurements." Although work underway at the Rocky Mountain Forest Forest and Range Experiment Station's Multiresource Inventory Project should help deline appropriate measurement procedures for non-timber resources, the current round of NFMA planning lacks any such comprehensive guidelines. As a result, the data used in making FORPLAN runs is inevitably flawed because of substantial (and unknown) differences in the variance associated with estimated resource levels (and projections over time) for limber and non-limber resources. Because the planning essort is specifically directed at multiresource analysis, this appears to me to be a very important difficulty. It is not inherently a FORPLAN problem; any multiresource planning method that allempts to do planning al the level of detail prescribed for the FORPLAN analyses would face the same dilliculty. Classification errors, as contrasted with measurement errors, may result directly from the FORPLAN analysis itsell. The FORPLAN analysis requires that certain decisions be taken by the planning team, and these decisions will have a profound effect on the results. One such decision is the identification of analysis areas.
As Bare and Field have stated, no large-scale programming system is ever likely to be completely free of logical and programming errors. The important thing is that the code should be free of errors ihal might have a significant impact on the problems being solved. I must admit that my experience with FORPLAN has only been second-hand, and, therefore, I am more than happy to accept the judgment of Bare and Field that, with a lew possible exceptions, FORPLAN seems to be free of significant errors of this type. However, broadly as they have interpreted their "operations research perspective," Bare and Field have lillle to say about duta errors, which I consider to be a potentially much devastating problem than programming and logical errors. Dala errors lend to be serious because of the likelihood that they will escape detection. In a model such as FORPLAN it is reasonably likely that logical and programming errors will be caught at some stage, especially because there is a large community of users, unless the errors themselves are sufficiently innocuous that they don't maller very much. The kind of data errors that are most insidious are not data-entry errors, which can usually be discovered by carefully auditing data files. Rather, the most serious data errors are those of nieasurement and classification. Measurement errors are the most fundamental, and affect all resource decisions. There is little that FORPLAN analysts can do about such crrors, because the analysts are not the ones charged with making the measurements. For certain kinds of resources, such as timber, the Forest Service has developed reasonably good procedures for avoiding such errors and we can probably be consident that
Recently, together with some of my colleagues at Northern Arizona University, I've been reviewing the plans of several of the National Forests in Arizona. One thing we've noticed is that different forests have used completely different approaches, based on an entirely different set of assumptions, in the delinition of analysis areas. On one Arizona national forest, for instance, the planning team made a considerable effort to recognize the spatial distribution of resources as one element in the delinition of analysis areas. For the most part, their analysis areas are contiguous because the planning team lelt that the plan should reflect the way the forest is managed. The planning team of another Arizona national forest, in contrast, used a more conventional forest strata approach in identifying analysis arcas, so that a single analysis area is made up of small pieces scattered all over the forest. I suspect that this is typical of the way most national forests have identified analysis areas.
Associate Professor, School of Forestry, Northern Arizona University, Flagstaff, AZ 86011.
The reason I believe this point is important is that it will influence, perhaps markedly, the results of the FORPLAN runs, particularly the calculation of present net worth. If you require that analysis areas be contiguous, you construct an implied constraint that will restrict the feasible region and possibly reduce the calculated present value of certain or all the alternatives being tested. This fact has often been given as a reason for permitting analysis areas to be noncontiguous. That's well enough if you can manage the forest in small, scattered pieces, bui il operating plans concentrate on larger, contiguous blocks, then the implicd constraints ought to be in place. Otherwise FORPLAN, or any other planning method, will overestimale nel present value. A final point about the definition of analysis areas, based on research in multiresource management we have underway at Northern Arizona University, is that different resources may require different definitions of analysis areas. Wildlife, for instance, do not respect the vegetative boundaries that have commonly been used in defining analysis areas for FORPLAN runs. Proper consideration of overlapping analysis areas for different resources will require implementation of overlay methods similar to those used in geographic information systems.
I agree with the assessment of Bare and Field that FORPLAN, as used by must national forests, does not adequately address the different but related problems of forest planning: strategic planning (allocation), tactical planning (scheduling), and operational planning (implementation). It seems to me, however, that their recommendation does not go sar enough. I believe rather strongly that to support ils future planning efforts, the Forest Service should adopt a planning system that specilically recognizes these distinct but interrelated levels of planning. Stralogic planning ought to be done at a regional or national level, and it should use an appropriate level of resolution, recognizing that the kind of detail needed to manage individual forests is unnecessary for strategic decisionmaking, and merely gets in the way. Tactical planning ought to be done at the forest level, with a higher level of resolution, and operational planning at the district level. This sinal level of planning should be very site-specisic and should rigorously maintain spatial integrity. It is at this level that planning and management should be fully integrated, and not at the higher levels. Hierarchical planning of this sort is similar to the multilevel optimization concept being developed by Tom Hoekstra, John Hol, and their colleagues in the Land and Resource Planning Project at the Rocky Mountain Forest and Range Experimeni Station. This type of approach 10 forest planning and multiresource management seems to me to be far superior to the single-level approach typified by most current applications of FORPLAN.
Hierarchical (Multi-Level) Planning
From discussions I've had with several forest planners, my impression is that one of the worst features of the current round of NFMA planning is that many of these planners and their colleagues on the planning teams have been "burned out" by the experience. I suspect that this is largely because of the level of detail required in the planning effort. This requirement has led to the necessity to develop and process vast quantities of data, and has resulted in FORPLAN models with an enormous appetile for computer time and disk space.
An Evaluation of FORPLAN from an Operations Research Perspective:Discussion Paper
Abstract.--A framework for evaluating computerized systems for public land management planning is used to review and extend Bare and Field's analysis. The linear programming (LP) system FORPLAN is a powerful and appropriate tool for public land management planning. However, thorough validation of the FORPLAN models and extensive sensitivity analysis of the LP solutions are essential to determine the reliability of the computerized analysis. Adequate and cost-effective sensitivity analysis and validation require FORPLAN LP models which are smaller and simpler than those typically used in the current cycle of planning. A hierarchical planning strategy for using FORPLAN, which is consistent with simpler and smaller FORPLAN models, is sketched and evaluated.
Bare and Field observe that in 1979, when FORPLAN was chosen as the system for claborating land management plans for all National Forests, no other system met the Forest Service needs: to satisfy the optimization requircments of the National Forest Management Act; among the National Forests; been successfully used in forestry. They argue that using FORPLAN 10 allocate land and to schedule management simultaneously may have "raised LP to its level of incompetence," and that today the FORPLAN system produces: models which are too large, 100 poorly understood, and too costly to be of much value; linkages among strategic, tactical and operational planning, and models which do not "recognize or comfortably sit the real process of making decisions on the National Forests." Based on an extensive review of recent developments in operations research, Bare and Field conclude that separate models tailored to the strategic, tactical, and operational phases of planning hold the greatest promise for effective planning. They speculate that a misunderstanding of the planning problem may be the root problem and recommend that before any changes in analytical techniques are considered, "the relationship between planning and decision making in the Forest Service must be completely and satisfactorily resolved." They conclude that only then can a task force of decision makers, planners and analysts specify
the "analytical support necessary," and can "issue bids for the purchase and installation of a "planning support system." The questions posed and answered by Bare and Field are relevant. An operations research perspective requires an examination of both the structure of a system, and of the way it will be or was used. Bare and Field quickly abandon their self imposed restraint "merely to examine the mathematical apparatus" and they review the way FORPLAN was used by Forest Service land management planners. Their assessment of the implications of LP lor forest management planning is incisive, their discussion of the questions posed is full of insights and wisdom, and their review of current developments in operations research applications to forest planning is a valuable guide to available tools. Their answers are plausible, and their conclusions, although severely critical, mostly reasonable and constructive; but their arguments are not entirely convincing. There are several weaknesses in their analysis: 1. They do not fully trace the consequences of violating the fundamental assumptions of LP; 2. Some key questions about implementation have gone unanswered; 3. Throughout their discussion they do not clearly distinguish between the problems and failures caused by the structure of FORPLAN and those altributable to the way FORPLAN was used. We shall attempt to correct these minor deficiencies, and we shall do it in way which may be useful to those who in the future will be responsible for selecting, developing, and guiding the use of operations research techniques to plan the management of National Forests. We shall present a simple but comprehensive framework for evaluating mathematical programming and simulation systems, use it to review and complete the evaluation of FORPLAN made by Bare and Field, and linally we shall show how this framework can also be used
Management Sciences Staff, Pacific Southwest Forest and Range Experiment Station, Berkeley, CA 94708.
The "framework" is simply three sets of critcria which address the questions: is the computerized system SOUND-free of logical and programming errors? Is it OPERATIONAL--can it be, or has it heen, used correctly and cost-effectively in planning? Is it RELEVANT--does it, or can il, meet institutional requirements and information needs? Note that this evaluation framework can be used either before, or alter, a system has been implemented. The inspiration for this framework is clearly traceable to Bare and Ficle; however, it will become quickly apparent that it is sulliciently different from the three questions they posed to absolve them of all responsibility for its shortcomings.
realism. The limit on achieving the most desirable level of realism in a simulation model is the cost of writing, and running, the simulation programs. This cost can be very high. Mathematical programming systems have additional restrictions which reflect their structure, and the algorithms available to solve them. For example, LP requires using only production processes which "linear and independent." Linearity here means that outputs must be directly proportional to inpuls, and that both inputs and outputs must be divisible into small increments, e.g. acres, dollars, recreation visitor days, acre feet of water, thousand board feet of timber. Note that a bridge or visitors' center is not divisible into small increments, and therefore cannot be represented in an LP. Independence requires that each production process be unaffected by the selection of any Other process, even on adjacent areas. The implications of the restrictions imposed by mathematical programming must be worked out and evaluated 10 establish whether an acceptable level of realism can be reached. Finally, the data requirements of the system must be established, and assessed against the data which is likely to be available to intended users. The range of reliability of the of data is also very important. If some of the data required to run the system are accurate and others are less so, the reliability of the LP solution may be indeterminate, and the solution may confuse rather than insorm.
Is the Computerized System SOUND?
Three principal criteria must be satisfied: is the system free of logical errors, are computer programs reliable, and can the system be used to generale models which are an acceptable representation of reality?
Is the Systeni Operational?
For simulation models of the level of complexity required in forest planning, checking the logic rigorously is likely to be impossible or at the very least impractical. But for mathematical programming systems, the logic can be checked rigorously is adequate documentation is provided. It is essential this be done, and that crrors be corrected, before the system can pass muster.
Can the System be run correctly and cosl-ellectively in operational situations? System designers tend to be optimistic about the capabilities of the hardware and soliware required by their brain children. They also lend to assume thai users are, or will become, their equal in sophistication. So, we must now add: given the institutional selling, is the system likely to be used correctly and economically? The implications of the system's inherent characteristics must be re-examined in the operational context.
Program Quality and Management
Regardless of the level of competence of the programmers, some errors will remain in the large and complex programs typically used in land management planning (LMP). Programs must be designed to ease debugging, and to limit the consequences or errors caused by bugs discovered only allier the analysis is under way, or is completed. Updates and extensions of the program rapially compound the probability of serious errors, eventually requiring rewriting the code from scratch if reliability is critical. How the computer programs have been, and are, managed is a key factor in assessing the soundness of the system.
Does the system address the information needs, and is it consistent with institutional requirements? Information nccds and institutional requirements are seldom well delined and, in the process of planning, they will be revised--perhaps radically. The system must have the flexibility to accommodate changes in in needs and requirements. The capabilities of the system must be matched against initial needs and requirements, and inconsistencies identified (e.g. a system may be incapable of producing solutions which are both site specific, and globally optimum). The system must be usable in a way which will resolve such conflicts, or which will lead to acceptable compromises. This may require the elaboration of "planning strategies" for implementing the system. The validity of the analysis conducted with a technically sound system may well depend on the design of such planning
Computerized systems used in LMP result necessarily in simplified representations of reality. Simulation systems allow the designers great freedom in choosing the level of
and which decision makers and analysts must understand to interpret the printout.
strategies. Several planning strategies for implementing FORPLAN are discussed below. The three sets of criteria are closely related. For example the assumptions underlying a system determine ils soundness, but they will also all'cct ils ease and cost of operation, and the extent to which the system addresses information needs, and perhaps even institutional requirements. In FORPLAN, linearity and certainly assumptions lead to acceptable soundness only is extensive sensitivity analysis is conducted when the system is used. This is both costly and difficult, thereby affecting the operational rating. If the sensitivity analysis shows that the solution is vulnerable to departures from linearily and certainly, some information needs may not be adequately mel. The evaluation of a system cannot proceed simply by applying each set of criteria serially. Interactions must also be considered. Grouping the criteria is simply a way of organizing the immensely complex task of assessing the potential of computerized systems. I also brings into sharper focus the source of the problem: Is it the structure of the computerized system? Is it the way the system is--or will have to be--implemented? Is it the way information needs and organizational requirements are delincu? Focusing on the source of the problem will help change the design or the implementation strategy of the system, and may help in sinding a better system.
Bare and Field infer from some "illogical results" obtained by Region 8 National Forests, that there may be something wrong with the model. They may be right. Yel, because they acknowledge that the complexity of the LPs rendered them opaque to analysis, we will never know unless the mathematical structure is systematically verified. This has been done by the developers, and a Mathematical Programmer's Guide for Version 1 has been published (Johnson 1986). A similar Guide for Version 2 is in preparation. Based on heroic--and only occasionally successful--efforts to work through the Version 1 Guide, we can only speculate that l'ew will be able to gain useful insights into the mathematical structure of FORPLAN by referring to these Guides. The notation is cruelly complex, requiring inlinite patience and stamina of those not blessed with a photographic memory. This suggests that even experienced analysts may not be able to deduce logically the relationships among the growing number of options available. Assuming there are no errors in logic, il will become increasingly unlikely that logical errors can be detected analytically if proposed extensions to the system are made. Detecting errors in logic by running sample problems is not an adequate substitute for rigorous analysis of the mathematical model. It is not likely to catch all the errors, and it is not cost effective. There are other imperative reasons for presenting the model logic rigorously and in a way accessible to analysts. First, available options--falling demand and Aggregate Emphasis/Coordinated Allocation Scheduling among others--require a clear understanding of the mathematical structure of the model to avoid potentially serious errors in formulating models and interpreting the solutions. Second, to permit an effective public review of National Forest land management plans, interested parties must be able to gain a thorough understanding of the analytical procedures used without unreasonable effort. The current Mathematical Programmer's Guide is comprehensive but so difficult to follow that its usefulness is questionable.
Let us now use this framework to review and complete the evaluation of FORPLAN presented by Bare and Field.
Is the Computerized System Sound?
Bare and Field limit their assessment of the soundness of the FORPLAN system to a discussion of the implications of the assumptions of LP for modeling management on National Forests. They propose that linearity and independence can be dealt with by "careful formulation of the model and cautious interpretation of the solution..." They note that LP is "deterministic"--that random variation in the data cannot be represented--and they suggest piecewise lincar representation as a a possible albeit expensive fix. Bare and Field conclude that these characteristics of LP do not pose insurmountable difficulties, and they proceed to evaluate FORPLAN, strictly based on the results of operational runs.
Computer Program Quality and Management
Bare and Field take for granted the conceptual soundness of the system. Because there has been no refereed publication on FORPLAN, an examination of the structure of the system is in order. Il esfort, talent, and professional integrily could assure the soundness of computerized systems, we could assume that FORPLAN were sound, and could dispense with further discussion on this subject. Because this is seldom thc case, an examination is required, if only to identify those peculiarities of the system which users must know to run FORPLAN correctly,
The probability of undetected programıning errors is widely believed to increase exponentially as the number of modifications made to large computer programs increases. FORPLAN Version 2 has undergone twelve major updates at this writing. The implications for the FORPLAN matrix generator and report writer would be chilling were il not for the care and energy and expense which has been devoted to hunting and exterminating "bugs." Some legitimate concern about the existence of bugs must remain, requiring users to validate the FORPLAN solutions with great care. The implication is that FORPLAN models must be kept simple enough to make this validation practical and Page 17
permit and timber sales administration, project planning, etc., for the District Ranger. The stakes of Supervisors and District Rangers in land management planning would be increased, and, if so inclined, these line officers could l'ully understand and gain control of the FORPLAN models. A hierarchical strategy is not without costs. Line officers, already burdened with heavy responsibilities, must participate actively in a hicrarchical planning process. Incentives must be found to induce planners to keep the FORPLAN models small and simple. Hierarchical planning only makes such models possible. Some loss in optimality will be incurred. The losses in present net value are not likely to be statistically significant given the softness of the economic data available for forest-wide planning. Losses in the timber allowable sales quantity will occur and may pose a more serious problem. Statistical analyses will be needed to determine their significance. Preliminary results suggest that very substantial reductions in model size may be achievable for reductions in allowable sales quantity which are probably not statistically significant (Navon cl al. 1986). Long before the next cycle of foresl-wide planning begins, some Forest Plans will have to be revised. Some alternative to forest-wide, comprehensive, site specific allocation and scheduling is nceded to make FORPLAN a more reliable and cost-effective planning system. A promising strategy is to use separate FORPLAN models to address the information needs of the Forest Supervisor and District Rangers. The proposed hierarchical strategy is very sketchy, and its evaluation is admittedly somewhat superficial. Additional work is needed to establish ils usefulness. In particular, the linkage between the forestwide and the District models requires further elaboration and scrutiny, and a rigorous statistical analysis of the consequences of reducing data resolution in the forest-wide model is necded. The objective here is only to start a discussion of ways and means to improve the cost-effective use of FORPLAN. A complete evaluation of FORPLAN must await a survey of the experiences of planners and decision makers on National Forests. Nevertheless, the structure of the computerized system, its documentation, and the fragmentary reports on its field use, do point to problems and to opportunities for improving the way FORPLAN can be used to elaborate implementable forest-wide alternatives. A hierarchical planning strategy holds great promise for grasping these opportunities.
Barber, Klaus H. 1986. Large FORPLAN models: An exercise in folly. In Proceedings of the Workshop on Lessons from using FORPLAN (Denver, Colo., April 29-May 1, 1986). 268 p. Land Management Planning Systems Section, USDA Forest Service, Washington, D.C. Bare, B. B. and Richard C. Field. 1986. An Evaluation of FORPLAN from an Operations Research Perspective. This volume.) Determining Optimal Policies for Ecosystems, in al. ed. North Holland. Johnson, Norman K, Thomas W. Stuart, and Sarah A. Crim. 1986. FORPLAN Version 2: An Overview. Land Management Planning Systems Section, USDA Forest Service, Washington, D.C. FORPLAN Version 1: Mathematical Programmer's Guicle. Land Management Planning Systems Section, USDA Forest Service, Washington, D.C. Jr. 1986. Experience with the Solution of USDA Forest Pub 86. Bethesda, MD. Meet Analysis Requirements of the Shoshone National Forest. In Proceedings of the Workshop on Lessons from using FORPLAN (Denver, Colo., April 29-May 1, 1986). 268 p. Land Management Planning Systems Section, USDA Forest Service, Washington, D.C. Wildlands. Proceedings of Division IV, International Union of Forest Research Organizations, XVI IUFRO World Congress, Oslo, Norway. Enterprises, in Studies in the Management Sciences, Industries, M. Kallio et al. ed. North Holland. Andres F. Weintraub. 1986. MAX-LOSS: An Automated Procedure for Calculating the Maximum Loss in Optimality in Aggregating FORPLAN Problems. In Proceedings of the Workshop on Lessons from using FORPLAN (Denver, Colo., April 29-May 1, 1986). 268 p. Land Management Planning Systems Section, USDA Forest Service, Washington, D.C. Weintraub, Andres, S. Guitart, and V. Kohn. 1986. Strategic Planning in Forest Industries. European Journal of Operational Research 24, No. 1 (Jan 1986) p. 152-162.
Adler, Ilan, Mauricio G.C. Resende, and Geraldo Veiga. 1986. An Implementation of Karmakar's Algorithm for Linear Programming. Dept. of Industrial Engineering and Operations Research. University of California, Berkeley, CA 94720. C 86-8. Page 18
to replace neutral concepts of multiple use with a This concept is likely to shape public debate in this field over the course of our careers. Conceptualizations of policy have always mallered, whether it was manifest destiny, all the way through the sagebrush rebellion, including multiple use. I expect that public use could become in that category. It's important, though, to recognize what Babbitt did not say. He recognized and would affirmatively advocate that cxtractive uses such as timber harvesting, mining, and grazing must continue. But they should be as he put it, "honestly subordinated to public uses." Implicit in his formulation are two ideas about multiple use. First, that multiple use, perhaps unlike sustained yield, has little substantive content. Ii doesn't tell you which choices should be made. Second, multiple use has tended to produce, depending upon the region, domination by timber grazing, mining, and water development interests. Public use would not and should not erase those uses. But it would produce them. The reduction in allowable cul that I think will come is an example of public use. A draw down of 25% of the total cut would greatly ameliorate the key stresses that we now see on public uses. How is the Forest Service reacting to these pressures already in place that may lead toward future recognition of public use? The returns are mixed. The first is a decision handed down by the Ninth Circuit Court of Appeals in San Francisco earlier this year. The Forest Service manages Lake Shasta in northern California. It's a major recreational resource and has had houseboats on it in the past. The Great American Houseboal Company began to sell timeshare units on houseboats, making them kind of floating condos. The Forest Service, realizing that this would greatly increase the user days on the lake, combated the sure overcrowding by adopting a permit system outlawing most time-share ownerships in houseboats on Lake Shasta. The Forest Service had no specific authority to do that in the statutes. The Appellate Court upheld the Forest Service's permitting system. Ruling that the agency's power to regulate "occupancy and use" within the forests, tracing back to the 1897 Organic Act, was broad enough to encompass the houseboal regulation for which again there was no express support in any statute. In a second example, the agency has acted in an opposite way. The Hard Rock Mining Law of 1872, perhaps the most outmoded of any public lands statule enforced toclay, allows hard rock miners the right to enter most public lands, and is they make a strike, lo receive lille not just to the minerals but to the overlaying 20 acres of land also. What that Act did was to zone the public lands for mining. The previously unrestricted autonomy of miners has been cut back in many ways partly because of the many environmental impacts froni poor mining practices. However, the Forest Service, in many cases, is refusing 10 take the tough steps. Il you talk to Forest Service mining officials in the field, you find that their marching orders are that they may "regulate" miners, but they cannot deny to them the righi to mine, even is miners refuse to comply with all necessary Forest Service requirements.
As Professor John Leshy of Arizona State University Law School concludes in his upcoming book on the Hard Rock Mining Act, the Forest Service is far too tender toward industry on this issue. There is no absolute right to mine. The Forest Service ought to assert maximum management authority while exercising it l'airly and reasonably over this major class of forest users. Third, the Forest Service has made what seems to be a truly unfortunate sel of decisions on key issues relating to water pollution. The so called "Go-Road case" was handed down last year, and then a slightly amended opinion was reissued this last summer. The Ninth Circuit Court held that proposed logging and the Go-Road itself on national forest lands failed to meet California water quality standards lo protect against erosion. The Forest Service argued that it required loggers to follow so called best management practices (BMPs), such as requiring busser strips and selling soil types in gradients where logging may not occur. But the Ninth Circuit Court said that BMPs are only a means to an end, and that loggers must meet California's substantive standards such as its requirement that turbidily cannot rise more than 20% above natural turbidity when the additional turbidity is caused by development activities. The stale turbidity standard, the courts said, is an end, not a means, and it must be complied with. A great conservation agency, especially one aware of the fragility of anadromous fish runs, and especially an agency recently chastened at Mapleton, should have applauded that decision as a key element in our continuing and frustrating struggle to find reasonable ways to reduce erosion from non-point source pollution. Instead, the Agency recently went to EPA and sought to have the Go-Road ruling administratively overturned. In 5 to 10 years, maybe sooner, the kinds of restrictions that California, and Idaho, and Montana, and others are enforcing for erosion will be accepted as a matter of course, in my view. Our rivers require those kinds of standards, and so do the animals in them. But instead of being in the vanguard of an irresistible movement, the Forest Service went into the bunker on that issue. Another example is very different. In Region 6, the Pacific Northwest, the Forest Service has gone to great lengths to follow the scientific evidence to protect old growth habitat, by using the spotted owl as in indicator species. Tremendous political pressure has been exerted on the agency to alter its proposals. But the agency has shown a great deal of courage al several points in this process, and more often than not, has followed expert scientific conclusions. Let me linish then by drawing out some very broad directions that the Forest Service might take il' I am correct about our already being into the beginning point of an evolution into public use. The Forest Service ought to do two basic things. It ought to consolidate it's authority, and it ought to be a leading advocate for good resource policies when public uses are threatened. We will learn a lot about the Forest Services willingness to consolidate it's authority, whether it will act as it did with houseboats on the one hand or with hard rock mining recently on the other. And we will learn about it's willingness to act as an advocate within the next sew years in regard to a proposed water project that is just about 40 miles southwest of here - the proposed Two-Forks Dam. Page 19
FORPLAN: An Evaluation of a Forest Planning Tool A Summary
Abstract.--FORPLAN has potential to be useful as a decision making tool in public forest management, but this potential has not yet been realized. The model needs to be used for more limited purposes in simpler and smaller forms. Also, FORPLAN should be supplemented with more specific analytical tools which have tactical and on-the-ground applications.
FORPLAN was not applied to certain regions, for example, British Columbia. The major focus of the meeting, however, was on assessing the usefulness of the FORPLAN to help in public forest planning in a context such as in the U.S. where multiple-use considerations dominate and the forest is used to produce a variety of outputs, both priced and nonpriced.
In this "summary" I recount briefly some of the major points raised by the papers of the conference that make up this volume--thcir similarities and their dillerences--and then identify the consensus that I perceived as emerging over the three days of the symposium. Some persistent differences will also be noted. While the symposium was designed to focus on thc FORPLAN model, the discussion showed that it is often dillicult to untangle the model from the planning process, and the planning process from the plan. This summary will range beyond the model, thereby reflecting the nature of the discussion of the symposium. The symposium evaluation of FORPLAN reached much consensus during the three days of the meeting. This paper will focus on that consensus.
The first one-and-one-half days presented a useful review of the variety of political, legislative, interest group, bureaucratic, technical and other background considerations that influenced the development of FORPLAN and its use in Forest Planning. The relationship of the predecessor models such as Timber RAM and MUSYC and the tendency of FORPLAN to incorporate new types of models into its broad perspective was noted. Also, the role of the Committee of Scientists was discussed. While many of the individual groups and events were familiar to most who allended, a discussion of the role of these events in the development of the FORPLAN model provided a useful backdrop for the remainder of the symposium. Some presentations were used to show the broad applicability of FORPLAN to non-Forest Service users. In the process, the FORPLAN model was described as being a software package which has broad usefulness because of its flexibility for many applications. Examples of its use included applications by industrial firms and applications to Third World countrics. We also learned why
The discussion of the first one-and-one-half days was descriptive and largely positive. The application of the FORPLAN to the Shoshone National Forest was used as an illustration of the model's ability to help in the development of a plan and in decision making. In the Shoshone, FORPLAN provided input into the decision to move ahead with certain activities that were controversial but sensible. More generally, some of the early presentations characterized Forest Planning, Forest Planning, and by inl'erence FORPLAN, as a success. This raised the question of what constitutes a success. Several criteria emerged. Out of the legislative activities that surrounded the Resources Planning Act (RPA) and the National Forest Management Act (NFMA) came two objectives that the legislation was designed to achieve. These were (1) the reduction of conflict, or more broadly what I'll term a political criterion, and (2) the improvement of decision making regarding the national forests. These objectives can be viewed as criteria with which to evaluate the success of Corest planning and also by inference the FORPLAN. Other goals and criteria also might be considered. For example, a former Associate Chief of the Forest Service has cited an important effect of the planning process as that of helping the Forest Service in the budget process. Norman Johnson, in his luncheon address, suggested that thc FORPLAN has sunctioned to protect the Forest Service from its critics by providing a shield of technical jargon with which many critics cannot effectively deal. (This view, however, was vigorously contested.) Similarly, planning may have had the positive effect of forcing the Forest Service to take an integrated view of all of its resources, not just
1 Senior Fellow and Director, Forest Economics and Policy Program, Resources for the Future, Washington, D.C. Comments presented to the FORPLAN Evaluation Symposium, November 3-5, 1986, Denver, Colorado.
timber, and to force the explicit consideration of economic and biological considerations. Others noted that the FORPLAN has led to stali training and increased technical expertise. While all might not agree that the above outcomes of the FORPLAN and the planning process were among the goals, stated or unstated, these types of effects have occurred, and have resulted in labeling the process as either a success or a l'ailure.
Other specific criticisms included that of the Bcuter/Iverson paper, in which they stated that they did not know if il improved decision making, and Binkely's paper where he contrasted seven costs (unweighted) of the FORPLAN with only two benefits. Another oftenexpressed concern was the lack of a budget in the planning process. The absence of a budget to give a scale of the relevant range of a plan could lead to serious errors in the mix and level of planned forest outputs. Several specific technical criticisms also were directed at the model. These included technical problems related to its size, problems related to linearity, lack of validation, data problems--especially when dealing with data with large but unknown variance, etc.
Reduce Political Conflict
Reducing conflict was one of the stated goals of the planning process. How well have we donc? Conflict persists, but probably no more (and possibly less) than without planning. The Beuter/Iverson paper seems to see FORPLAN and the planning process as a success largely in political conflict terms. Similarly, Binkley's comments suggest that conflict related to the National Forests may have been reduced since the advent of planning. While not eliminating conflict, the planning process may have provided a point upon which the conllict may focus. It may have allowed the conflicts to procecd in a more orderly, less chaotic fashion. It is too early to know whether conllict has been reduced or merely rechanncled or perhaps postponed. While the question still awaits the future for a delinitive answer, the preliminary assessment is that conflict has been reduced.
Despite the criticism, there was also considerable support for the FORPLAN model, in concept, as an analytical tool. There seemed to be general agreement that: 1. FORPLAN could be useful as a "strategic" model to provide a broad overview. complexity. tactical and implementation levels to complement and supplement FORPLAN. The model should be kept small, manageable, transparent and understandable. In essence, the "white box" approach is preferred in which the projections are either intuitive or, if not, the counterintuitive results can adequately be explained be tracing through the model. Currently, FORPLAN is typically a "black box," the output of which oficn cannot be explained and therefore must be accepted on faith. When dealing with counterintuitive results, the decision maker has no way of knowing whether the result is an artifact of the model or if it represents a complex but sensible reality. In this context, a good decision maker would tend to discount or wholly ignore such counterintuitive results. Such a response seriously reduces the model's value as an analytical tool.
Has forest planning and the FORPLAN model in particular led to improved decision making? If the answer thus l'ar is negative, does it give promise of leading to improved decision making in the future? As noted above, FORPLAN received high marks for its usefulness on some forests, e.g., the Shoshone. However, over the course of this symposium the FORPLAN has also received a host of criticism. A recurring criticism has been that the model, as typically applied in forest planning, is too large. The criticisms of large models are of two types, general and technical. The general criticisms include: 1. The models as developed are beyond the capability of most of the stalls; elfort of the staffs and, therefore, preclude use for analysis and stille creativity; becoming the plan rather than the tool; actions beyond its design or capability. In summary, the criticism was that in trying to be comprehensive and include everything, the model lost its ability to be a useful, flexible and perceptive analytical tool.
Concluding Remarks and Symposium Summary
Abstract.--The major points of discussion are summarized in terms of FORPLAN's ability to (1) improve the planning process, (2) improve communication, and (3) improve decision making. There was general agreement that FORPLAN met the analytical requirements of the National Forest Management Act. The sharpest criticism was the inability to model nonlinear relationships and operational difficulties with large models. Several variations of hierarchical planning were suggested. Additional analysis beyond the forest-wide FORPLAN analysis appears to be needed.
The objective of this symposium wasto document and evaluate the implementation of FORPLAN. I will not altempt to summarize each speaker's paper as this has been excellently done by the various discussants and my co-pancl member, Rodger Scego. I will try to identily the major areas in which the speakers and audicnce participants agreed or disagreed and suggest where we might go from here. Using the framework suggested by John Beuter, we can evaluate the success that FORPLAN has had in (1) improving the planning process, (2) improving communication, and (3) improving decision making. Concerning improving the planning process, there was general agreement that FORPLAN met the analytical requirements of the National Forest Management Act (NFMA). Teeguarden suggested that some type of mathematical tool would have been required to meet the analytical requirements of NFMA. Wilkinson commented that the level of analysis in FORPLAN exceeded the requirements of NFMA. There was general agreement that certain issues needed to be dealt with outside of the model. Teeguarden rated the ability of FORPLAN to achieve the planning requirements into 16 categories and concluded that the weakest categories were in linkages to issues outside of the forest such as the demand curves facing the forest which involved regional linkages and regional cumulative effects. Shugart agreed that FORPLAN adequately could represent the ecological linkages is adequate estimates of the yield cocl'licients were provided using small scale ecological models and large scale ecological models used to verify the total FORPLAN "solution." All speakers recognized the limited ability of a linear programming model such as FORPLAN to model spatial relationships but agreed that it was not a fatal llaw that could not be achieved by some form of later analysis outside of FORPLAN. There were some concerns among participants, however, about how watershed, recreation and mineral resources were being
addressed within FORPLAN and these questions went unanswered. There was considerable discussion about whether FORPLAN improved communication. Communication takes place at three levels; within the planning team, belween the planning team and the Forest Supervisor, and between the Forest Service and public. There was general agreement that FORPLAN could enhance communication between the planning team and the Forest Supervisor, but the Forest Supervisor would need to remain in close contact with the team to understand the complex modeling relationships, implicit and explicit assumptions. Within the planning team, FORPLAN provided a common framework for specialists to communicate and lo see that their concerns were integrated into the plan. There was disagreement about whether communication belween the public and Forest Service was improved using FORPLAN. It was suggested that because of its complexity FORPLAN might be used as a shield between the Forest Service and the public to justify agency positions. Others, such as Binkley, reasoned that the standard approach to analysis provided by FORPLAN increased the ability of large centralized interest groups to follow the analysis more easily after some initial investment in model study. Mealy cited examples where the public participation increased with the ability of the public to work together with the planning team in putting together prescriptions and seeing The total results in a forest context. There was general agreement that FORPLAN was not the Forest Plan, but a tool to aid decision making. Daniels and Mealy felt that FORPLAN aided decision making. Davis proposed using FORPLAN in a "will" and "without" application to measure the improvement in decisions. He suggested doing this on a smaller scale in industrial applications. Greg Jones, using the Integrated Resource Planning Model (IRPM), recently completed a similar experiment to measure the efficiency of alternative planning methods in the northern Rocky Mountains. Jones found large increases in efficiency could be gained when spatial relationships were important.
College of Forestry, Oregon State University, Corvallis.
The sharpest criticism of FORPLAN came from the operations research panelists including Bare, Fields, Dykstra, and Navon. The operations research group agreed that the major shortcomings in FORPLAN were the same as in any large linear programming model. These limitations were assumptions of linearily in modeling reality, the Jillicully or identifying measurement errors, and the cost of running the very large models. All four operations researchers proposed a hicrarchical approach to forest planning to reduce model complexity and increase the ability to analyze the problems specific to that level of planning. There was agreement that one or more additional levels of analysis would be needed besides a forest-wide application of FORPLAN. These added levels of analysis might take place above and below the forest level to meet specific needs. Al a higher level, analyses might establish guidelines for FORPLAN-forest applications. This has been proposed before in terms of a regional plan. Al a lower level, more site-specific planning could be done to insure the plans are implementable and to provide detailed information for shorl-lern budgeting. This necd for a tactical plan is becoming more widely recognized throughout the Forest Service and is being icrmed "area analysis." I concur with the previous speakers' concerns about hierarchical planning. FORPLAN must be considered a strategic planning model. Many of the current problems facing public forestry cannot effectively be deali with by strategic planning models. Problems that require spatial, sile-specific analysis can only be cursorily considered in forest-wide planning models. Typical issues include "sales below cost" and "cumulative ellects". It seems that the focus will shift from strategic planning which concentrates on estimating average clfects over the Torest for the long term (100-200 years) to tactical forest planning that concentrates on estimating site-specilic elsects (or logical geographic areas for the short term (5-50 years). This tactical planning involves a comprehensive, indepth analysis of the projects that should be selected for an area over the next lew years or decades given certain objectives and constraints. Spatial analysis, estimation of sile-specific effects and detailed consideration of logging and iransportation choices are important l'acets of this approach. Very little rescarch, development, training and extension is being conducted on tactical forest planning. Much of the recent research and development has centered with the
Management Sciences Staff under the direction of Mal Kirby. One notable output has been the Integrated Resource Planning Model. Work on increasing the efficiency of this model is being done by Greg Jones at the Intermountain Station. Outside the Forest Service, universities occasionally have projects that touch on tactical forest planning, including Teeguarden's work on the cconomical suitability of timberland and my own work on designing road systems to cfficiently access a specified set of culling units. Johnson and I have previously proposed research on tactical forest planning to the Forest Service. The needed work in tactical forest planning includes: (a) What questions a tactical planning model should answer. answer these questions. new models to do these lasks. of linkages between these planning models and geographic information systems. smoother and more informative linkage. of linkages between strategic and lactical planning applied in tactical planning, such as constraints environmental effects. In closing, I must conclude there is general agreement that FORPLAN has made a significant contribution lo forest planning. The economists, the ecologists, and the managers here have largely been satisfied by the ability of FORPLAN 10 adequately represent their concerns and address issues at the forest level. Over the last several days I have played the devil's advocate to draw out criticisms of FORPLAN and the criticism is surprisingly mild. Perhaps we are exhausted after 10 years of planning? Thank you for this opportunity to participate. Page 20
United States Department of Agriculture
Genetic Variation in Douglas-fir:
Rocky Mountain
Fort Collins, Colorado 80526
THE UNIVERSITY OF MICHIGAN
General Technical Report RM-141
Genetic Variation in Douglas-fir: in Eastern Nebraska
David F. Van Haverbeke, Research Forester
Twenty-year-old Douglas-fir trees in provenances from Arizona, New Mexico, and southern Colorado survived better and grew taller; but incurred more winter injury in eastern Nebraska than trees from provenances from northern Colorado, southern and western Montana, northern Idaho, Canada, and eastern Washington. However, surviving trees from Pacific Coast, and northern and central Rocky Mountain provenances increased in percent of plantation mean height during the past 9 years, whereas trees from southern Rocky Mountain provenances decreased. Agelage correlations indicate provenances expressing superior height growth can be identified at age 6.
The diversity of tree planting materials under study at this and other locations in the Great Plains was made possible through cooperation with the Regional Tree Improvement Project (NC-99) of the North Central States Agricultural Experiment Stations. Credits are due Jonathan W. Wright, Professor of Forestry, Michigan State University (deceased) for initiating this regional study and providing the planting stock; and to Ralph A. Read, Silviculturist (retired), and John A. Sprackling, USDA Forest Service; and Walter T. Bagley, Associate Professor of Forestry (retired), University of Nebraska, for planting, maintaining, and evaluating the early performance of this species.
Genetic Variation in Douglas-fir: A 20-Year Test of Provenances in Eastern Nebraska
Douglas-fir is a conifer species that is not indigenous to the Great Plains. However, it has been planted sparingly and in selected locations, mainly as an ornamental, throughout the central Great Plains for many years. Interest in its use for Christmas trees has increased in recent years. Identification of seed sources of Douglas-fir that are adapted to the central Great Plains environment could increase its use, reduce planting failures, and add to the number and variety of conifer species available to forestry agencies and commercial nurseries for windbreaks, Christmas trees, and environmental and esthetic plantings within the Great Plains-a region in which few conifer species are native.
Coast Douglas-fir (P. m. var. menziesii), of which two provenances are represented in these data, occurs along the Pacific Coast eastward into the Cascade Mountains from southwestern British Columbia, through western Washington and western Oregon to central California; and in the Sierra Nevada to central California and western Nevada (Little 1971, 1979) (fig. 1). Rocky Mountain Douglas-fir grows at elevations from 1,200 to 8,000 feet in the Northwest on a wide variety of soils and parent materials including granitic, volcanic, sedimentary, and metamorphic (Pfister et al. 1977). As Douglas-fir approaches its warm, dry limits (below 6,000 feet) towards the Great Plains, it becomes more restricted to basic soil parent materials such as andesite, basalt, and limestone (Ryker and Steele 1980); the latter type commonly occurs throughout the Great Plains region. The plantation reported on here at 20 years of age, is part of a larger test of Douglas-fir provenances for which 1-year-old nursery and 3- to 8-year-old field data in Michigan and Nebraska plantations were reported (Wright et al. 1971). Eleven-year field performances of the provenances in this Nebraska plantation were reported by Read and Sprackling (1976). The objective of this study was to identify adapted sources of Douglasfir for planting in the central Great Plains.
Heavy mortality in the nursery and during the first year of field establishment imposed severe limitations on the analysis and interpretation of 20-year-old Douglasfir provenance data. Numbers of individuals in some provenances, particularly those of northern origin, were reduced drastically. Also, individual trees in some provenances have declined in vigor during recent years. Despite these limitations, the overall performances of the surviving individuals in the majority of provenances have been consistent over the past 20 years. This is the only test of Douglas-fir in the central Great Plains and, thus, the sole source of data relating to the adaptability of Douglas-fir to this region. Therefore, it is deemed appropriate to report these data for their use in improving initial care and establishment procedures, and in the selection of seed sources in future tests.
Seedling stock for the Nebraska plantation originated from 55 of 128 bulked sources of Douglas-fir seed assembled from native stands throughout the species range in the United States and Canada (Wright et al. 1971). The seeds were sown in an East Lansing, Mich., nursery in 1962; the seedlings were distributed to cooperators in 1963. The Nebraska seedlings were linedout for 2 years before field-planting. Small seedling size, a heavy-textured soil, and lack of protection from sun in the summer and protection from wind in the winter, in the Nebraska line-out beds, resulted in a 90% loss of seedlings over all provenances (97% within the Pacific Coast origins, 95% within the northern Rocky Mountain origins, and 71% within the central and southern Rocky Mountain origins) (Read and Sprackling 1976). Of the original 14 Pacific Coast provenances, only 22 seedlings of 6 provenances survived; of the 26 northern Rocky Mountain provenances, 56 seedlings of 19 sources survived; and of the 15 central and southern Rocky Mountain provenances, 187 seedlings of 13 provenances survived for field planting (Read and Sprackling 1976). Surviving seedlings were field-planted in the spring of 1965 on a ridgetop of silt-loam, in single-tree, complete
Rocky Mountain Douglas-fir (Pseudotsuga menziesii var. glauca (Beissn.) Franco), the variety of primary concern in this study, occurs in most of the western mountain ranges east of the Coast Range in British Columbia and east of the Cascades and Sierra Nevada Ranges in the United States (Ryker and Steele 1980). Specifically, it is widely distributed in central British Columbia, southwestern Alberta, eastern Washington, Oregon, Idaho, Montana, Wyoming, southeast Arizona, and southern New Mexico; and locally in the mountains of northern and central Mexico (Little 1971, 1979) (fig. 1). Page 21
Figure 3.—Height growth curves for Douglas-fir origins (some grouped) after 20 years in an eastern Nebraska plantation (extension of fig. 2, Read and Sprackling 1976).
Pacific Coast 1645 WA 1646 WA Northern Rocky Mountain 1615 ID 1588 1562 ID 1507 MT 1600 MT 1616 MT 1649 MT 1520 MT 1539 MT 1595 ALB 1596 ALB 1648 MT Central Rocky Mountain 1636 CO 1529 1532 1630 1525 1611 UT Southern Rocky Mountain 1610 NM 1594 NM 1602 NM 1625 AZ 1647 AZ 1545 AZ 1593 AZ Page 22
level. However, phenotypic agelage correlations com- by shielding seedlings individually with "cedar puted at the provenance level, where the mean was ade- shingles,” or by interplanting a faster growing, but temquately described with 3 or more trees, showed steadily porary "nurse" tree species such as eastern redcedar improving values with decreasing time intervals; the in- (Juniperus virginiana L.). Avoidance of frost pockets and dication was that mean provenance height could be ade- exposed sites is also advised. quately predicted at age 6 (1970) (table 3). This study found that trees of southern Rocky Mountain origin survive better and are taller after 20 years than trees of most central Rocky Mountain and northern Conclusions and Recommendations origins. Once established, however, the surviving trees of northern sources persist and, because of lack of winter Twenty-year performance results are in close accord injury, appear to be slowly narrowing the height advanwith those reported at age 11 by Read and Sprackling tage gained earlier by the trees of southern sources. Cen(1976). Survival data suggest that, in eastern Nebraska, tral Rocky Mountain sources are relatively winter hardy Douglas-fir seedlings will incur heavy mortality in the and grow quite well; they are recommended for ornursery and during early years in the field unless seed- namental planting in the eastern part of the central Great lings are of good quality and are protected from solariza- Plains. They also may be suitable for establishing windtion, drought, and winter exposure. These results breaks around farmsteads and in urban plantings where indicate that well-developed planting stock is necessary protection and water can be provided. As in the 11-year to insure field survival; stock of 2 + 1 or 2+2 age class, evaluation, the southwest Colorado provenances 1525 possessing a well-balanced shoot-root ratio, should be (Durango), and especially the high-elevation provenance planted. Protection can be provided by sheltering seed- 1630 (Ouray), because of its trees of good vigor, continue lings in lath- or shadehouses prior to field planting; and to be recommended. Table 3.-Phenotypic agelage correlations computed at the provenance level among ages 2 to 20 for 14 provenances of Douglas-fir. Page 23
Nero, Robert W.; Clark, Richard J.; Knapton, Richard J.; Hamre, R. H., eds.
Proceedings of this first international symposium consist of 47 presented papers covering 15 owl species, and 4 workshops dealing with capture, telemetry, census, and management techniques. Basic information on habitat preferences, home range size, detecting lesser known owls, etc. will be invaluable to managers of wildlife and of forested lands; techniques information will be invaluable to researchers.
Robert W. Nero, Manitoba Department of Natural Resources Richard J. Clark, York College of Pennsylvanla Richard J. Knapton, University of Manitoba R. H. Hamre, Rocky Mountain Forest & Range Experiment Station
Sponsors: Manitoba Department of Natural Resources USDA Forest Service Natural Resources Institute, University of Manitoba World Wildlife Fund Canada World Wildlife Fund US The Wildlife Society
IN 1983, Bob Nero began to talk about the need for a gathering of "owlologists" to compare notes on Great Gray Owl research and conservation. At that time, he contacted a few colleagues who also expressed a great need to review their work and exchange ideas on research techniques. Little did any of us realize at the time that the need was international, that the forum would be 3 1/2 days of technical presentations and workshops, and that "owlologists" would be discussing all northern forest owl species at the first-of-its-kind Symposium held in Winnipeg, Manitoba, Canada, February 3-7, 1987.
veteran in producing a quality publication. He was aptly assisted by the editorial committee consisting of Bob Nero, Dick Clark and Richard Knapton. Only the authors are responsible for the material contained in their papers; their views are not necessarily those of the sponsors, i.e., the USDA Forest Service, the Wildlife Society, World Wildlife Fund, University of Manitoba, and the Manitoba Department of Natural Resources.
The timing was right on. The organizers worked hard. Sponsors' interest was high. Eventually a program was developed that included a field trip, exhibits, musical and social events, all planned to provide a restful change from intense discussions and to effect international camaraderie. It worked. How well it worked can only be known from those who were there!
How have northern forest owls benefitted from this Symposium? There was an initial, very positive response from the public and local news media to the holding of such an event. However, the full effect of this meeting will not be realized until the technical knowledge exchanged during the Symposium and recorded in this document is incorporated into action programs by management agencies and pushed by conservation groups. This challenge was presented by Monte Hummel in his opening address and recognized by Dick Clark in his summary remarks.
As the coordinating chairman for the Symposium, I had the pleasure of working with a central committee composed of Bill Koonz (Arrangements), Bob Nero (Program) and Ken De Smet (Finances). Excellent support for this committee came from office staff, university professors, volunteers, students and biologists. They are: Lori Bartley, Lynn Bergeron, Don Campbell, Brendan Carruthers, Maureen Collins, Herb Copland, Dr. Jon Gerrard, Chris Hofer, Kathryn Johnston, Rudolf Koes, Dr. Erkki Korpimäki, Dr. Emil Kucera, John Morgan, Ted Muir, Dr. Ronald Ryder, Ulrike Schneider, Renate Scriven, Dr. Spencer Sealy, Don Sexton, Dan Soprovich, Linda Tardiff, Rick Wishart and Rosemarie Young.
Canada celebrates 100 years in wildlife conservation in 1987 under the theme Wildlife '87: Gaining Momentum. During a time when nongame wildlife programs are fledging and taking wing in Canada, it is appropriate that we would begin the second 100 years with a Northern Forest Owl Symposium as the first event of this celebration. By the time the next Northern Forest Owl Symposium is held, I hope that action programs will have been implemented to protect owls where needed and to ensure that the public will have a greater understanding and appreciation of the role of owls in the natural environment.
The production of a printed Proceedings was accomplished in record fashion due to the outstanding cooperation of the many contributors who submitted camera-ready manuscripts and illustrations. Bob Hamre is obviously a seasoned
Merlin W. Shoesmith, Chief, Biological Services, Wildlife Branch, Department of Natural Resources, Winnipeg, Manitoba.
Special Aspects of Northern Forest Owls Ronald A. Ryder and W. Bruce McGillivray, Chairpersons Richard J. Clark, Dwight G. Smith, and Leon Kelso C. Stuart Houston Reversed Size Dimorphism in 10 Species of Northern Owls W. Bruce McGillivray Disease Susceptibility in Owls D. Bruce Hunter, Kay McKeever, Larry McKeever, and Graham Crawshaw The Role of the Whitefish Point Bird Observatory in Studying Spring Movements of Northern Forest Owls Thomas W. Carpenter
Strix Owls of Northern Forests Heimo Mikkola, Richard R. Howie, and Robert W. Nero, Chairpersons Reintroduction of the Ural Owl in the Bavarian National Park, Germany Wolfgang T. Scherzinger Pertti Saurola Evelyn L. Bull, Mark G. Henjum, and Ralph G. Anderson Biology of the Great Gray Owl in Interior Alaska Timothy 0. Osborne Mark F. Spreyer James R. Duncan Summer Habitat Use by Great Gray Owls in Southeastern Manitoba Maria C. Servos Status of the Great Gray Owl in Finland . Olavi Hilden and Tapio Solonen
Territorial Aspects of Barred Owl Home Range and Behavior in Minnesota Thomas H. Nicholls and Mark R. Fuller -. Thomas Bosakowski, Robert Speiser, and John Benzinger Ecology of the Three Species of Strix Owls in Finland Heimo Mikkola
Surnia Owls of Northern Forests Wolfgang Scherzinger, Chairperson Home Range Size of Hawk Owls: Dependence on Calculation Method, Number of Tracking Days, and Number of Plotted Perchings Bjorn T. Baekken, Jan O. Nybo, and Geir A. Sonerud Observations of the Northern Hawk Owl in Alberta Edgar T. Jones Kauko Huhtala, Erkki Korpimaki, and Erkki Pulliainen
Aegolius Owls of Northern Forests R. Ake Norberg and Ann Swengel, Chairpersons .. 157 Erkki Korpimaki Annual, Seasonal, and Nightly Variation in Calling Activity of Boreal and Northern Saw-Whet Owls ... 162 David A. Palmer 169 Ronald A. Ryder, David A. Palmer, and John J. Rawinski Movements and Home Range Use by Boreal Owls in Central Idaho 175 Gregory D. Hayward, Patricia H. Hayward, and Edward O. Garton Occurrence of the Boreal Owl in Northeastern Washington 185 M. W. O'Connell Home Range of Tengmalm's Owl: A Comparison Between Nocturnal Hunting and Diurnal Roosting ... 189 Bjorn V. Jacobsen and Geir A. Sonerud The Breeding Biology of Northern Saw-Whet Owls in Southern British Columbia ... 193 Richard J. Cannings Study of a Northern Saw-Whet Owl Population in Sauk County, Wisconsin . 199 Scott R. Swengel and Ann B. Swengel Remigial Molt in Fall Migrant Long-Eared and Northern Saw-Whet Owls 209 David L. Evans and Robert N. Rosenfield
Bubo, Asio, and Otus Owls of Northern Forests Geir A. Sonerud and C. Stuart Houston, Chairpersons Runar S. Larsen, Geir A. Sonerud, and Ole H. Stensrud Geographic Variations in the Diet of Eagle Owls in Western Mediterranean Europe ... Jose A. Donazar C. Stuart Houston, Roy D. Crawford, and Donald S. Houston Page 24
Some Features of Long-Eared Owl Ecology and Behavior: Mechanisms Maintaining Territoriality ..... 229 Vladimir I. Voronetsky 231 Josef Kren Fidelity to Territory and Mate in Flammulated Owls ... 234 Richard T. Reynolds and Brian D. Linkhart The Nesting Biology of Flammulated Owls in Colorado ..... .. 239 Richard T. Reynolds and Brian D. Linkhart Distribution, Habitat Selection, and Densities of Flammulated Owls in British Columbia 249 R. Richard Howie and Ralph Ritcey Censusing Screech Owls in Southern Connecticut ... 255 Dwight G. Smith, Arnold Devine, and Dan Walsh Status of the Eastern Screech Owl in Saskatchewan with Reference to Adjacent Areas 268 Christopher I. G. Adam Effects of Environmental Variables on Responses of Eastern Screech Owls to Playback 277 Thomas W. Carpenter Current Status and Habitat Associations of Forest Owls in Western Montana ..... . ... 281 Denver W. Holt and J. Michael Hillis People Power: Help for the Owl Bander 289 C. Stuart Houston Page 25
Symposium Summary and Concluding Remarks'
Abstract.--To summarize the geographic location of the researchers: of the 150 registrants, 22 (15%) were from eight European countries (Norway, Finland, and Sweden topped the list), 83 (53%) were from five Canadian provinces and one Territory, and 45 (30%) were from 17 States of the United States. Of the 52 papers presented, 39 dealt with research on a single species, four dealt with two species, and seven dealt with more than two species. Of those, three dealt with community studies of owls. Eighteen papers dealt with aspects of the basic behavior of species and 12 papers dealt with the habitat of owl species in some detail. The conference brought fromobscurity some of the basic biology of Otus flammeolus, the Flammulated Screech Owl, and its distribution on the periphery of its range in British Columbia, and the latter can also be said about the population of Spotted Owls, Strix occidentalis, in that same province. Much basic information that will be invaluable to land and wildlife managers - such as habitat preference, home range size, detecting lesser known owls, etc. presented.
SUMMARY AND CONCLUDING REMARKS
Someone has said that to summarize a conference such as this has been, is an impossible task but I would like to thank Dr. Robert Nero for providing the opportunity to try. I would also like to thank, on behalf of the participants if I might be so presumptuous, Dr. Merlin Shoesmith and all of the other Manitobans for the splendid job they have done in organizing and executing this Symposium. Having lived in Manitoba for a couple of summers I found the people of this province to be memorably hospitable and this trip has reinforced that feeling of warmth in spite of the outside temperatures.
I shall start out by admitting up front that I was unable to hear all of the papers presented. That arises from the fact that on Tuesday evening I was conducting an auto census of the owls of the forests of northern Minnesota, eastern North Dakota, and southern Manitoba. I selected a strip transect to sample the area and the strip consisted of a band starting at the Minneapolis/St. Paul Airport and ending at the Viscount Gort Hotel in Winnipeg, Manitoba, Canada. I choose 100 meters from the center of the motor car route on either side of routes 494, 94, and 49 as the specific sampling area. The dimensions of the sampling plot are actually 300 kilometers by 200 meters. Admittedly only the shelterbelts and riparian woodland were suitable habitat and I must subtract 12 kilometers of the strip where dense fog, associated with sugar beet refineries, prevented my seeing any owls. Unfortunately I did not see a single owl within the study area. The only good aspect of that fact is I do not have to ponder which statistic is most appropriate to apply to my results. All of this is offered as the reason underlying my sleeping Wednesday morning when I should have been listening to papers.
even though our colleagues from the latter three countries were unable to travel to the symposium. I trust the readers are aware of the solid contributions from Scandinavia and West Germany and my comments citing specifically representation from these countries will not offend those from other countries. While humans recognize political borders, owls do not; hence, it is important to hear from researchers from all geographic locales within owl species distributional ranges.
Secondly, I would like to define some of the technical terms that have been used at the conference for the benefit of those readers of the proceedings. Some of these terms are similar to terms used in everyday language, but they have special meaning here, thus I shall gloss them. I shall take the terms in alphabetical order. The first term is Bastard and this has to do with mixed ancestry. Now this was not the actual term used by the presenter and when I talked with him about my using the term he suggested that it perhaps had a negative connotation. So to avoid that possibility I shall use that term Complication. We saw how Strix aluco and Strix uralensis were equally implicated in complicating the ancestry of certain generations in Bavaria. Next we have Divorce which is used to refer to the dissolution of pair bonding between mates. This was used to define bonding between individuals of the same two species earlier mentioned. Then we have Secondary Females. In the human condition this might be thought of as being analogous to playing second violin in an orchestra. We saw how "playing second fiddle" has inherent risks within Surnia ulula populations. Finally we have the term Topless and when applied to the human condition this may mean that the upper portion of the torso is unadorned of garments or is naked. Here specifically it refers to the torso of a nest cavity box being naked of a roof. Enough of that--let me now try to be serious for a few minutes.
First allow me to summarize the geographic location of the researchers. This information was taken from the official list of registrants. I have deviated from that list only insofar as I have recorded Dr. Heimo Mikkola as a resident of Finland rather than Indonesia as suggested by the list. I will play the numbers game for just a moment by saying that of the 150 registrants 22 (15%) were from eight European countries, 83 (53%) were from five Canadian provinces and one Territory and 45 (30%) were from 17 states of the United States. Norway, Finland and Sweden topped the list for numbers of participants from Europe and not surprisingly Manitoba and Saskatchewan provided the largest numbers from Canada while Minnesota, Wisconsin, Colorado and Oregon were the home states providing the largest numbers from the United States.
I will now shift my emphasis to where it most appropriately belongs--to the owls themselves. I have, from the abstracts, compiled the following data (see Table 1) on a species by species basis and would caution that this compilation was done while watching slides and listening to presenters, hence must be considered a preliminary to the final report that will appear in the Proceedings. For emphasis, I will start by pointing out that seven of the 22 species targeted (perhaps a bad choice of words) selected to be the subject of this conference were not reported on at all. It is not at all surprising that five of the seven are species of Otus for 32 of the 136 species of owls commonly recognized are of the Genus Otus. They are, to enumerate, Otus kennicotti the Western Screech-Owl, Otus bakkamoena the Collared Scops-Owl of Asia, Otus brucei the Striated Scops-Owl [also of Asia] Otus scops the Common Scops-Owl of Africa, Eurasia and Indonesia and Otus sunia the Oriental Scops-Owl. In addition, we have heard nothing about Blakiston's Fish Owl Ketupa blakistoni of Japan and Korea nor Ninox scutulata the Oriental Hawk Owl which is widespread in Asia and Indonesia. Lest one think I am totally negative I would hasten to add that this conference has brought from obscurity some of the basic biology of Otus flammeolus the Flammulated Screech Owl, and its distribution on the periphery of its range in British Columbia and the same can be said about the population of Spotted Owls Strix occidentalis in that same province. Eighteen papers dealt with aspects of the basic behavior of species and we saw how techniques of hybridization, which can be an essential tool for isolating details on the genetic component of speciesspecific behaviors can be utilized with owl species. Food habits are always going to be an important aspect of predator studies; however they have reached the point where they are now well enough known on some species that they are now a means to the end of elucidating ecological relationships rather than being an end in themselves. Twelve
This says nothing about the quality of the presentations which were overall splendid from all countries. It was especially heartening to hear from Spain, Hungary, Czechoslovakia, and the USSR
1 See Northern Forest Owl subject species list for scientific binomials.
Location Legend: the following "abbreviation" scheme was used for reporting the location for each respective study--Canada [a three letter abbreviation for the Province or Territory), International (a five letter abbreviation for the country), United States (the standard two letter U.S. postal abbreviation).
papers dealt with the habitat of owl species in some detail and hopefully this area of research will expand from here, for a suitable place to live is no doubt even more critical to the survival of owl species than it is to man. I did not tally man-owl aspects and they were not here emphasized, however, we did see that there are some areas that are the cause for concern particularly with regard to species of Bubo. With regard to Bubo there is some good news and some bad. The good news is that through the dedicated efforts of an individual and his wife a cadre of volunteers was developed which changed the image of the Great Horned Owl in central Canada, while in Europe man continues to be a threat to the survival of the Eagle Owl, either directly through his activities or indirectly through his anthropogenic structures. The basic biology of nesting and population dynamics have been reported but there is certainly room for more research in these areas. The latter aspect is particularly crucial if we are to insure the survival of existing species that are rare, threatened or endangered and also if we are to manage species that are common, in a manner that will minimize conflict with man in relations with those species.
BASIC RESEARCH Figure 1.--The pyramid of sound wildlife management. Basic research must form the foundation for management of either species or communities.
I will now turn to some broader aspects of the research that has been reported here as well as some points that have been made in discussion. I view basic research, applied research, and conservation as seen in this triangle (Figure 1). You will note that I have represented basic research as the foundation for this triangle, i.e., it must necessarily form the basis for sound applied research and/or effective conservation and management of owl species. As one can see in the triangle basic research forms a connection to both applied research and conservation. Thus the material from these proceedings contributes either directly or indirectly to all aspects of owls. Put another way, even though a particular contribution may deal only with basic research it can potentially impact work of land managers, wildlife specialists, conservationists, and others if they will make use of it.
must qualify that, with regard to their surviving in darkness, having worked mostly on Asio flammeus a species that can be seen active either day or night, by pointing out that while there are some owl species that are very diurnal these are the exceptions rather than the rule. Owls represent only about 1.5% of all bird species thus reinforcing the idea that owls are unique and special. Because they have invaded a realm that is foreign to the diurnal humans they have been neglected with respect to being subject for study. Although they currently enjoy popularity amongst humans with their images being collected as statues, photographs, paintings, etc., they have been both dammed and deified in the millennia that they and man have coexisted. Because they operate in a world where man is in the dark, special techniques, apparatus, etc., are required to study them. We have seen techniques using light from the infrared portion of the EM spectrum. Perhaps light in the red segment of the visible light spectrum could also be used at least for some species. Also perhaps instruments that intensify available light, the so-called Starscopes, could be used, e.g., to minimize the risk of conducting direct observations from close range on the less timid species. Certainly radio telemetry, as we have seen at this conference, has played a key role in revealing some of the secrets that we have heard about here. No doubt it will play an even greater role in the future as the telemetry technology develops further, e.g., smaller species may be studied as smaller, lighter radios are developed. Lighter radios will also allow
One cannot only find much information in the content of the individual contributions but in the literature cited at the ends of the articles as well.
I would remind you that owls are a unique group of birds that are without equals in the specializations that they have evolved, enabling them to survive in a world of darkness. Our plenary speaker elaborated on that most thoroughly. I
tracing the migration routes, times, etc. of those species whose movements appear to be somewhat erratic in nature as well as the regular migrant species and perhaps satellite tracking would be most appropriate for some of these studies.
We have seen a number of different methods used in trapping owls (see the workshop presentations for details of numerous methodologies appropriate for owl research), some of them variations of techniques used on the diurnal raptors, i.e., hawks, falcons, etc., and some unique to owls. Successful trapping of owls is critical to many types of studies and I could not overemphasize the necessity of having known individuals while studying the basic ethology of the species in the field. At this point I will site a quote from Larry McKeever's new book "A dowry of owls"
to go, e.g., in the case of endangered species [even in those cases the species does not exist in a biological vacuum but is interrelated with other species), but a more balanced approach is that of managing an ecosystem or segment of it. As we get a better picture of the detailed habitat needs of species we are learning that absolute minimum area dimensions for species are not the only thing required for management. We must know the quality of the habitat and, in many cases the configuration of the habitat is also crucial. This poses some interesting challenges for applied research, e. g., will a habitat segment with corridors leading from it to other tracts suffice with equal satisfaction to that of a larger intact area? Such points of view and questions are going to require manager-researcher teams for they require the expertise of specialists. The list of participants of this symposium identifies a good number of the owl experts [both professional biologists and serious amateurs doing professional calibre work] and hopefully these Proceedings will carry the challenge to those in a management position that deal with owls within the domain of areas that they manage!
Better one bird in hand than ten in the wood. Better for birders, but for birds not so good"
I am sure those of you who have tried to trap owls can relate to this and would suggest that for the latter portion that depends on the professionalism of the biologist and the use that information gleaned as a result of the trapping is put to.
Management techniques have necessarily brought in habitat management. We are, I think, observing a shift in emphasis in management from the species to the community or even to the ecosystem. However, that shift in emphasis has yet to reach owl biologists (if the biologists are not looking at wildlife from that point of view how can conservationists and wildlife managers, as well as land managers, be expected to adopt that point of view?] for of 52 papers, 39 dealt with research on a single species, four dealt with two species and while there were seven papers dealing with more than two species only three of those dealt with community studies of owls. There may be occasions when the species approach is the only way
We do not have reason to be complacent about our knowledge for any species of owl. This conference will however, I think, be viewed as a landmark in the history of owl biology for it (along with the symposium on owls held in Sacramento in the fall of 1985, and the paper session on rare owls at the World Conference on Birds of Prey to be held in Eilat, Israel on 22-27 March 1987) will go a long way toward identifying owls as a unique group of wildlife and owl researchers as being unique in their own "light." There has been considerable discussion about following up this symposium with another in two or three years with suggestions that it deal potentially with any of the owl species and that it be held in a locale that would attract biologists from parts of the world that have been much underrepresented at this symposia, e. g., eastern Europe and Asia. Page 26
Evolution, Structure, and Ecology of Northern Forest Owls'
Abstract — In this introductory survey of northern forest owls I explore what distinguishes them structurally, ecologically, and energetically; what particular ecolological conditions they are subjected to; and what selection pressures govern their evolution. Comparisons are made between communities of northern forest owls in the Old World and the New World; and between northern forest owl communities and more southern ones. Forest owls, like most forest birds - and forest bats as well - have relatively short and broad wings, which are adapted for flight among vegetation. Their wing loading is low, which facilitates transportation of prey and also reduces the wings' aerodynamic noise. Reversed sexual size dimorphism is very pronounced in some species of northern forest owls. But theories of this phenomenon must also explain the same dimorphism in tropical owls and in diurnal birds of prey, and must also be compatible with some notable exceptions from the general rule. These problems have often been ignored. Forest owls are primarily "searchers" in the sense that they spend most of their hunting time searching for prey and little time pursuing and capturing them. They are "perch-and-pounce" hunters, but perch" height, giving-up time, and flight length vary with sensory, capacities, prey density, vegetation structure, and weather - aspects treated by optimal foraging theory Particular attention is given to the evolution of asymmetry of the external ears in some owls. Habitat choice, vegetation structure, and hunting technique dictate to what extent vision and hearing can be used for detection and localization of prey. Hearing is particularly useful in dense forest and for detection and localization of prey moving in dense ground vegetation or under snow. When an owl depends heavily on hearing for prey finding, demands on accurate vertical localízation causé selection for vertical asymmetry of the external ears. But ear asymmetry results in conflicting auditory information at the two ears. This may require a "training period", with extensive head tiltings, in young owls before they can fully benefit from the ear asymmetry. Interactions between owl populations and populations of small mammals are considered both in the ecological and evolutionary time scale. Owls specialized on small rodents tend to destabilize rodent population cycles, while generalized owls have a stabilizing effect, suppressing prey fluctuations. Both types of owls tend to synchronize population fluctuations of small rodents and other prey animals, both locally and over larger geographic areas. Rodent cycles give rise to different behavioral strategies in owls depending on their habitat choice, dietary specialization, hunting mode, sensory capabilities, and nesting habit.
This symposium on the biology of northern forest owls was restricted from the outset to include only forest owls occurring partly or entirely north of latitude 35° North. As a brief remembrancer of
geography, 35°N is 11.5° N of the Tropic of Cancer which is at 23.5° N. The 35° N latitude crosses USA through southern California, central Arkansas, and the southern part of North Carolina. In the Old World it passes through the northernmost corner of Africa, through the Mediterranian Sea, just south of the Caspian Sea, through northern Tibet, and across central Japan. Any species occurring wholly below this 35° N latitude has not been considered a "northern forest owl". By this criterion 22 owl species will be included (table 1). But apart from a brief mention below in a survey of owl distribution, some of these species are not treated further in any of the symposium contributions.
1Paper presented at the symposium, Biology and Conservation of Northern Forest Owls, Feb. 3-7, 1987, Winnipeg, Manitoba. USDA Forest Service General Technical Report RM-142. 2R. Åke Norberg, Department of Zoology, University of Göteborg, Box 250 59, S-400 31 Göteborg, Sweden. |
An object code must be translated into source code for a computer to read and execute it.
Pos Terkait
Copyright © 2024 kemunculan Inc.
