By Craig M. Olson, Scott P. Holmen, and Dean P. Angelides
Abstract: This paper describes the methodology used to produce four 120-year sustained yield plans for approximately 250,000 hectares of private forestland in California. The plans were created and constrained using desired future habitat conditions and other, more traditional forest planning constraints. The desired habitat mix was specified by planning sub-units, i.e. watersheds, and by decade. Current and future wildlife habitats were predicted from tree list information in a subroutine of a distance-independent individual-tree forest growth simulator. This created a time-series of habitat predictions for each forest stand managed under each silvicultural regime considered. Individual plans modeled 125 to 180 silvicultural regimes (including timing choices) for each vegetation/site combination. All resource yields, including habitats can be linked to spatially explicit locations using a GIS. Yields are combined to create a resource capability database, which is a set of biologically feasible management options. Policies are applied to this model to create policy alternative models that define a set of desired future conditions. Linear programming software is used to solve an alternative that can be mapped and evaluated. Plans were developed using Ep(x) (Ecosystem Planning eXpress) software developed by VESTRA Resources Inc., ArcInfo and ArcView software from Esri, CWHIZ linear programming software from Ketron Inc., and the FREIGHTS growth and yield simulator from LandRing Inc.
Environmental and California forest practice rules and regulations have given large industrial forest landowners in the state strong incentives to create long-term forest management plans which incorporate habitat models as integral parts of the problem formulation and planning process. Three types of legal documents are being created and filed with government agencies: Sustained Yield Plans (SYPs) for California state agencies; Programmatic Timber Environmental Impact Reports (PTEIR) for California state agencies; and Multi-species Habitat Conservation Plans (HCPs) for federal agencies. Several forest landowners, including the Louisiana-Pacific Corporation (L-P), the Pacific Lumber Company (PALCO), and the California Department of Forestry (CDF) are in the process of creating combined SYP and HCP plans to fulfill both state and federal requirements.
These documents all require long-term forest plans that must include assessments and evaluations of forest habitat changes over time. The SYP process mandates a minimum of a 100-year planning horizon that must include estimates of current and future habitats characterized using the California Wildlife Habitat Relationships System (WHR) (Airola 1988, Mayer and Laudenslayer 1988, Zeiner et al. 1990) or other vegetation classification systems. The SYP regulations also prohibit landowners from ever (over a rolling ten-year period) harvesting more that the long-term sustainable yield of their property.
The intent of the approach presented here to target the desired future condition for wildlife habitat and other public resources while ensuring sustainable production of both commodity and non-commodity outputs including high quality timber products.
The California WHR system was developed cooperatively by the University of California, Berkeley, and the California Department of Fish and Game. It contains information relating the habitat preferences of 643 terrestrial vertebrate species found in California. It allows a user to predict the occurrence and habitat quality for any of these species based upon the presence of specific habitat types and habitat elements. It includes species notes for each species including life history, range maps, legal status, habitat requirements, et cetera. In addition, it contains ArcInfo GRID habitat suitability models for 30+ species, a dBase compatible database and data-query system, and a series of books describing the system.
The WHR habitat system, like many other vegetation classification systems, uses the combination of plant species, size, and density to classify habitats. Forest habitats use a combination of tree species, average tree size, and tree crown density. Broad habitat types are based upon plant life form: tree, shrub, forb/graminoid, or aquatic. The WHR system then uses this habitat classification to identify habitat relationships between the vegetation found in an area and wildlife which is likely to be found in that area.
If WHR habitat classification predictions were incorporated into a conventional forest growth and yield model, it would be relatively simple to predict the habitat changes over time associated with different silvicultural regimes. Tying these habitat predictions back to the wildlife species database would provide a scientific basis for determining how planned forest management activities are likely to influence wildlife populations in the future. This information could then be used to evaluate whether or not a given mix of management activities is likely to meet a specific set of wildlife management objectives.
A more sophisticated planning effort could use the WHR system to help establish habitat targets for specific areas and then create a plan that meets or exceeds these habitat goals in all planning periods. The habitat targets, in part, would be used to help drive the planning process. These habitat targets would determine the location and timing of all management activities.
The WHR system is quite useful for comparative purposes and is a good general habitat classification tool. However, for any vegetation or habitat classification system there are field situations that defy the classification system. When we define rules by which definable areas on a landscape are classified, we create a system where every area is assigned a class. But natural variation produces situations where the rules don't work well. For example, we can define size classes of trees based upon the average size or quadratic mean diameter (qmd) at breast height of the larger trees (lets say trees greater than 10 inches in diameter at breast height (dbh)). So, we define the small class as stands whose average dbh is 10"-20", the medium class as 20"-30", and the large class as greater than 30". And this works well for even-aged forest stands, but when a stand has two distinct size classes, for instance numerous smaller trees and several very large trees, the qmd is computed as somewhere in between, resulting in not characterizing the stand well.
We can devise habitat and vegetation classification systems that allow for two or more size class determinations or calls per stand, but no matter how many we allow we can imagine situations where they will not adequately define the situation. In addition, highly divided, complex vegetation classification systems are difficult to interpret with respect to wildlife and wildlife habitat use because little data exists correlating habitat use to detailed vegetation descriptions. And what vegetation classification system might correlate well with one species' habitat use often does not correlate well with other species' use.
A partial solution to the problem of adequately characterizing and modeling wildlife habitats is to use existing systems such as the California WHR System and use supplemental information (if it's available) to further characterize the habitat for key wildlife species such as threatened and endangered species. If we couple the vegetation class determination with data on occurrence of large trees or proximity of other habitat elements we can tailor a system to meet the needs of HCP planning.
Our experience has shown that critical or important wildlife habitats should be modeled with a specific habitat model rather than trying to crosswalk a generalized model, such as WHR, to predict those habitats. For example, we have developed a northern spotted owl (NSO) habitat classification system that is based upon documented NSO habitat use relationships and local observations instead of trying to use WHR to understand forest management implications on NSOs. Where possible, a "chain of evidence" approach leads to better habitat models. This method uses several different ecosystem characteristics to identify habitat. We have often used trees per acre, basal area, crown cover, and volume/biomass by size class together to identify NSO habitats. For example, high quality NSO nesting habitat is found where average dbh is greater than 24", basal area is greater than 300 square feet per acre, volume is greater than 40,000 net board feet per acre, and crown closure is greater than 60 percent. If any one of these criteria is not met it is likely that the habitat is less than high quality nesting habitat. And if only one or two of the criteria are met, the habitat is unlikely to be useful for NSO nesting at all.
VESTRA Resources, a Redding, California, geographic information systems (GIS) consulting firm, was hired by L-P, PALCO, and the CDF to help them develop spatially referenced 120-year plans for the redwood region of California. These plans are based upon spatially referenced habitat targets that help to drive the planning process. The WHR system was used to estimate the quantity and quality of habitat in any planning period for a species or species guild since the models used predict the WHR habitat for each vegetation type for each silvicultural regime for each planning period.
VESTRA’s responsibilities have included: a) most GIS data capture, modeling, and analysis; b) satellite image analysis to produce vegetation mapping; c) inventory design; d) growth and yield modeling including data management, silvicultural/forest management regime script development (scripts for growth and yield modeling), and forest growth and yield simulation; e) creation of resource capability model databases ranging from 150MB to 1GB in size; f) inventory design and data processing; g) algorithm development and coding of habitat classification from tree list information and incorporation into growth and yield modeling; h) linear programming problem formulation and matrix generation in MPS data format; i) processing linear program matrix with CWHIZ; j) linkage of linear programming answers back to yield information and GIS data; k) creation of tabular summary tables, charts and graphs for documentation; l) mapping for problem alternative analysis and reporting with more than 1,000 unique maps produced for the two Louisiana-Pacific Mendocino County SYPs; m) development of a database framework and software utilities to help with all of the tasks mentioned above.
As a result of these planning efforts VESTRA has developed Ep(x)TM (Ecosystem Planning eXpress), a flexible, general purpose integrated forest-planning system which facilitates the creation and analysis of spatially explicit, sustainable forest management plans. Close integration of GIS data with traditional yield modeling is essential to this process. One of the most important data requirements is a spatially referenced vegetation inventory dataset. Without this it is impossible to create spatially referenced habitat predictions for different management alternatives.
Ep(x) is a new conceptual and technological approach to the development of long-term ecosystem management plans. It allows landowners and decision-makers to:
· define spatially explicit management units using GIS technology,
· link these management units to forest inventory information,
· simulate and evaluate hundreds of thousands of possible management activities, tracking 60 to 80 resource variables through time,
· create a resource capability model (RCM) using a relational database to store the range of biologically feasible management activities,
· link any of this RCM data to GIS coverages to create maps,
· use menu-driven data exploration tools to evaluate the RCM and help ensure that the ecosystem is being modeled correctly,
· create policy model alternatives (PAMs) which describe a set of desired future conditions to meet management goals and objectives,
· select an optimal mix of activities to achieve a balance among a broad range of management goals and desired future conditions,
· and evaluate alternative management strategies using sophisticated mapping, reporting, and data visualization tools.
Ep(x) provides a framework and software tools to seamlessly link GIS-based resource inventories, forest simulation models, operations research decision analysis, and mapping and data visualization tools to support decision-making for ecosystem management planning.
The advanced capabilities of Ep(x) are a result of a thoughtful and complete integration of sophisticated decision analysis tools with ArcInfo and ArcView, using the latest in database and software development technology. Ep(x) has been designed to take full advantage of this new computing technology as well as the latest thinking in ecosystem management. The capabilities of Ep(x) enable ecosystem management planning that:
· can be applied at the landscape or regional planning scales, on relatively short time schedules, and in a cost-effective manner,
· links various resource models into an integrated system, enabling analysis of both terrestrial and aquatic ecosystems,
· allows spatially explicit delineation of areas where land management activities and resource protection measures can be meaningfully specified,
· can be used to evaluate trade-offs among diverse stakeholder goals and evaluate long-term resource implications from planned activities,
· is based on clearly defined and easily reviewed data, models, and policy formulation,
· creates attractive maps and reports to facilitate communication of alternative outcomes,
· and encourages adaptive management through its ability to quickly incorporate changes in data, models, or socioeconomic factors.
Figure 1. VESTRA Ep(x)TM Analysis Process.
The current set of Ep(x) software tools is based industry-leading commercial products and runs on either the Microsoft Windows 95 or NT operating systems. Data sets maintained in the UNIX environment can also be used by Ep(x). GIS processing can be done either on UNIX or NT workstations.
Project Flow
The Ep(x) decision analysis framework is built on a foundation of GIS and related resource inventory data sets. The process begins with the creation of strata types from a GIS overlay of vegetation and site productivity. Forest inventory data is then linked to each strata type.
This creates the biological unit used for growth and yield modeling. Data associated with a strata type includes a tree list (species, dbh, height, live crown ratio, and tree weight per acre), a site index list for all tree species found in the strata, and strata acreage.
Through both spatial and more traditional resource simulation models applied to these data, Ep(x) creates a Resource Capability Model (RCM) that contains summaries of all important management inputs and resource outputs associated with a wide range of possible management activities on thousands of land units defined within the GIS.
The RCM defines the range of biologically feasible activities that can be considered as management options throughout the planning process. It is very important to have a wide range of management options in the RCM so that the realistic range of activities can be considered and evaluated later in the planning process. All of these plans have included some silvicultural regimes whose primary focus is to create and maintain late seral habitats with timber production as a secondary concern.
The process continues with the creation of land types from a GIS overlay of strata types with a special management concerns coverage and a watershed and wildlife analysis area (WWAA) coverage. Special concerns are geographically distinct areas (polygons) where the full range of biologically possible management actions may not be appropriate. Land types are then linked back to the RCM by joining them to their associated strata type yields.
Typical special concerns include stream and roadside buffers, areas near domestic water supplies, and areas already scheduled for harvest which will be harvested in the first period of the plan using previously determined silvicultural systems. The Louisiana-Pacific plans have also included wildlife corridors that are designed to maintain connectivity of mid and late seral habitats through time.
Ep(x) then allows the decision makers to formulate and re-formulate alternative sets of management goals and desired resource conditions by creating Policy Alternative Models or PAMs.
Each PAM contains a subset of the RCM that filters the biologically possible actions by applying policy constraints about what type of activity may or may not be appropriate in specific geographic locations (land types). For example, the RCM may include clearcutting areas adjacent to fish-bearing streams, but a PAM could restrict the range of Silviculture in these areas to regimes that never reduce the canopy closure below 70%. This is done by creating a database query to link RCM tables and fields of interest together into a virtual table and then using SQL statements to restrict the range of data rows in that virtual table.
A PAM also specifies the desired future conditions that should be created decade-by-decade by the combined set of planned activities. These typically include timber inventory targets, cash-flow requirements, late seral habitat minimums by watershed, Northern spotted owl foraging/roosting/nesting habitat minimums by watershed, and harvest flow constraints. Most alternatives also impose a harvest cap in any decade set by the estimation of long term sustained yield (LTSY). The desired future conditions are specified using linear programming constraints. Constraints are normally applied at both the project and WWAA level.
Using mathematical programming, an operations research decision analysis technique, Ep(x) optimizes the mix of management activities that should be done now and into the future on each land type. It is important to note that land types are the geographic decision unit that is modeled in the plan. Strata types are not directly modeled in the linear programming formulation.
By iteratively refining the PAMs and evaluating the decision model results, a desirable balance among the various management goals and resource conditions can be achieved and sustained through time. Ep(x) contains an analysis module that can generate a variety of report tables, charts, and graphs. Common reports include stocking, growth and harvest by decade, seral stage by decade, and Northern spotted owl roosting, nesting, and foraging habitat by decade. Reports can be made for the entire planning area or for individual WWAAs.
Figure 2. A sample graph of owl habitat acres
for a planning alternative.
Figure 3. A sample graph of owl habitat acres
for watershed in a planning alternative.
One of the key components of Ep(x) is the ability to predict the location of wildlife habitats into the future. This is done inside of the growth and yield simulator, which is an individual-tree, distance-independent model. This type of model starts with a list of trees and then projects diameter growth, height growth, changes in live crown ratio, and mortality over time under simulations of different management activities. This limits the data available that can be used to predict habitats.
The primary information used to determine habitat type comes from a list of tree records for each veg/site quality combination within the planning problem for each planning period (usually a decade). Each record contains tree species, diameter, total height, live crown ratio, and tree weight (number of trees per acre or hectare). Crown area and percent crown closure are calculated for each tree record. Arrays are then created which summarize basal area, height, crown closure and trees per acre by tree species and diameter class. This information is then used to classify habitats based upon species, size (diameter class), and density (crown closure). There are many different ways in which these classifications can be made.
Dominant size can be determined using either quadratic mean diameter, basal area, or crown closure. Species classification can be made based upon maximum crown closure by species or species group for all size classes or the dominant cover in the dominant size class. Density can also be calculated in several different ways. Multi-storied habitats are detected by analyzing the average height and crown closure in different size classes.
Figure 4. WHR
Habitat -- Quadratic Mean Diameter at Breast Height (QMDBH) Size
Class Determination.
Figure 5. WHR Habitat -- Multi-storied Size Class Determination.
Figure 4 shows the logic for determining the size class for a single-storied stand. Figure 5 shows the more complex logic used to determine if a stand meets the WHR multi-storied criteria.
Habitat changes over time are presented as a series of watershed maps and tables. VESTRA is also experimenting with draping habitat maps over 3-D terrain models to make computer animations showing habitat development over a 100-year planning horizon.
Figure 6: An example of a stand visualization from the PALCO planning process utilizing World Construction Set software to create the image.
Airola, D.A. 1988. A Guide for the California Wildlife Habitat
Relationships System. California Dept. of Fish and Game,
Wildlife Investigations Lab. Rancho Cordova, CA.
Mayer, K.E. and W.F. Laudenslayer, Jr. (eds.). 1988. A Guide to the Wildlife Habitats of California. California Dept of Forestry and Fire Protection. Sacramento, CA.
Zeiner, D.C., W.F. Laudenslayer, Jr., and K.E. Mayer. 1990. California's Wildlife. Vol. I-3. California Dept. of Fish and Game. Sacramento, CA.
VESTRA Resources, Inc.
962 Maraglia St.
Redding, California 96002
Craig M. Olson, Biometrician
Scott P. Holmen, Senior Developer
Dean P. Angelides, Vice President/Senior GIS Analyst
Tel. (916) 223-2585
Fax: (916) 223-1145
E-mail: mapmaker@vestra.com
Web Site: www.vestra.com