1. Description of the Geographic Information System (GIS) Analysis Used in the Development of HFQLG Alternatives.

BASE LAYERS. The following GIS coverages were obtained from Vestra Resources of Redding, California: Offbase, Deferred, and "Available for Group Selection." These coverages became the starting point for the development of Alternatives 2-5. The coverages were then updated with each forest's latest ownership information (February 1999).

Spotted Owl PAC/SOHA coverage and stream coverage were created using each forest's current GIS data sets (also in Febuary 1999). Stream coverages were buffered according to SAT guidelines. SNEP datasets for LSOG's and ALSE's were obtained from Alsea Geospatial of Corvallis OR, which is the data source for all SNEP data.

ALTERNATIVE 1. This GIS coverage was provided by the Sierra Nevada Conservation Framework GIS department and represents the current Forest Plan of each Forest.

ALTERNATIVE 2. As defined in the HFQLG bill, the pilot project area is that area designated as "Available for Group Selection", according to coverage obtained from Vestra Resources. All proposed resource management activities in alternatives 2-5 occur within the boundaries of the pilot project area. This area was edited by removing PAC/SOHA's, LSOG 4/5, and perennial stream buffers. The resulting coverage is the area available for resource management activities for Alternative 2.

Each Ranger District Fuels Specialist designed a strategic DFPZ network for their district and the resulting networks were then appended together into one DFPZ coverage for the HFQLG project area. This network was edited by removing completed DFPZ's, offbase/deferred areas, PAC/SOHA's, LSOG 4/5, and perennial stream buffers. This became the Alternative 2 - "Defensible Fuel Profile Zones."

ALTERNATIVE 3. Alternative 3 area available for resource management activities is the same as Alternative 2. The highest priority (most strategic) DFPZ's were selected from Alternative 2 based on proximity to urban areas, protection of communities, high-use roadways and elevations less than 6,500 feet. Next an "area treatment" fuel strategy was added that is based on southerly aspects from 90 - 270 degrees and mid to upper slopes within the pilot project area.

ALTERNATIVE 4. Starting with the Alternative 3 area available for resource management activities, SNEP-defined ALSE areas were removed and this became Alternative 4's area available for resource management activities.

The Alternative 3 DFPZ network was edited and those DFPZ's falling in ALSE's were removed. The Alternative 3 area treatments that fell within ALSE's were also removed, and in addition smaller, isolated area treatments were removed to define Alternative 4's areas available for resource management activities.

ALTERNATIVE 5. The offbase/deferred land allocations used in Alternatives 2-4 were also applied to Alternative 5. The area fuel treatment strategy used in Alternative 3 was used in this alternative. No DFPZ's are proposed in this alternative. In addition, the SNEP (Kondolf) riparian strategy is used in this alternative, not SAT guidelines.

A new set of GIS data, obtained from J. Fites of the Sierra Nevada Framework, was then applied to determine areas available for resource management activities. Those datasets included: Old Forest Emphasis Areas, Great Gray Owl buffers, Kondolf riparian inner zone buffers, Kondolf variable width outer zone buffers, matrix lands (used for Alternative 5 areas "available for group selection"), Pacific Rivers aquatic diversity areas, ecologically significant roadless areas greater than 1,000 acres in size, unsuitable/unavailable lands as defined by Foresst Service Remote Sensing Lab, and SNEP Ecologically Significant areas.

The only areas where any resource management activities (ie. area fuel treatments, group selection) are prohibited within the pilot project area under this alternative include Kondolf riparian inner zone buffers and PAC/SOHA's.

2. National Forest Modeling for the HFQLG Act

Forest ecosystems are complex. Analyzing the impacts of management activities on forest ecosystems can be made easier through the use of computer-based decision support systems such as GIS and linear programming models. The linear programming model SPECTRUM, developed by K. Norman Johnson was used to analyze the different alternatives considered in this Environmental Impact Statement.

Linear programming is very useful in formulating and quantifying the outputs from each alternative considered, including the "No Action" of "projection of the current management direction" alternitive. These results provide the basis for both the analysis of estimated effects and evaluation of tradeoffs between different alternatives (Kent 1989).

Inputs to the SPECTRUM model include vegetation maps derived from satellite data, California Wildlife Habitat Relationship (CWHR) categories, forest growth and yield data, economic data (such as costs for sale preparation, road construction, and benefits for sawtimber and biomass), and proposed activity locations derived from GIS data. The GIS system has digital maps of forest resources with attendant databases describing attributes of the maps.

Forest Vegetation Simulator (FVS) is a model that starts with inventories of existing primary vegetation and then simulates estimates of the future states of the primary vegetation, given certain actions. FVS was used to calculate CWHR categories, based on forest inventory data, and to project volume and CWHR through time, based on different prescriptions and timing.

SPECTRUM allocates resources among competing activities (timber harvest, wildlife habitat, riparian reserves) in an optimal manner. The modeler identifies the objective function or "criterian of optimality", such as "minimize habitat fragmentation" or "maximize present net value (PNV)" by assigning a numerical value to each solution. Different prescriptions and timing of activities are then applied, based on the objective of each alternative. Additional constraints can be entered into the model, such as maximum tree diameter limits available for harvest or minimum basal area to retain.

Some of the outputs from the SPECTRUM modeling for this project include resulting areas of CWHR categories, production of board-feet of timber greater than 15 inches, and the costs and benefits of each alternative.

The major strength of the SPECTRUM model is its ability to model the effects of constraints on outputs and effects over time. The major limitation of this model is that its schedule of activities and projected effects are not spatially explicit and they are deterministic, that is, they do not consider variability and uncertainty in input data.

The HFQLG team considered a range of management alternatives. A SPECTRUM analysis was made for each alternative and option by each forest within the HFQLG analysis area. All the information needed for SPECTRUM analysis is entered into a number of data files. The SPECTRUM matrix generator then creates a matrix of rows and columns that is then solved by linear programming software. A report is then generated from the solution.

2.1 The Data Input Process:

A number of different data input sources were used to develop the core biological model. Forest vegetation types maps for the national forests were developed using remote sensing technology. The vegetative units were stratified into classes based on their vegetative types, the size of the vegetation, and its density. Plantation data, updates and changes in detection processes, and silvicultural stand records were used to update and modify these maps. The resulting product was a spatially explicit map of vegetation across the pilot project area.

The Region 5 Forest Service FIA [Forest Inventory and Analysis] databases provided sampling data to describe the various vegetative strata. This data is also the input used in the FVS. The growth simulators allow us to grow the tree portion of FIA plots and apply various management treatments and disturbances (fire, insects, disease, etc.) agents to these stands and see the effects on growth over time. Data output from these simulators include yield tables, which show how various attributes of the stand change over time based on growth, treatment, and disturbance. Also these data allow us to classify the vegetation into CWHR categories, old growth rankings, and different types of specialized habitat. Table #1 illustrates the various outputs and classification being tracked by this analysis and used in SPECTRUM.

Information on regeneration success in plantations (summarized in the Silvicultural Accomplishment Report), estimates of insect and disease activities based on change detection, and analysis of the last 25 years of fire history were used to develop the mortality model used in FVS and SPECTRUM. Economic data, such as costs for various treatments such as thinning and biomass removal were based on a large part on data contained in the Forest Service Timber Sale Program Information Reporting System (TSPIRS) databases. Projections of future values and cost was also done.

2.2 Identifying the Land Base:

The SPECTRUM model used a strata-based approach to modeling. The acres in each national forest are divided into different analysis areas based on: (1) vegetative types described above; (2) management areas or zones that define where activities are permitted, modified, or restricted; (3) constraints that must to be met in order to do activities, and/or meet desired conditions; and (4) other physiographic or management breaks where the same activity under the same prescription can be expected to produce significantly different outputs. Acres in each unique analysis area are assumed to respond similarly to given management activities and produce the same outputs and effects regardless of their location on the forest. Each analysis area is defined by up to 7 levels of classification illustrated in Tables #1 and #2.

2.3 Modeling Multiple Outcomes:

There are two basic approaches in using SPECTRUM to model vegetation over time, Model I and Model II. In the Model I formulation, each decision variable represents a possible management sequence for an acre over the planning horizon. With this approach, an acre of land is assumed to stay within the same analysis area with the same prescription schedule through the entire planning horizon (150-years in this analysis). In the Model II formulation, each decision variable (unique treatment schedule of an analysis area) is allowed to have multiple outcomes at each treatment point. They might switch or transfer to other analysis areas or to new analysis areas based on a set of user-defined rules. This structure allows the analyst to model the situation more realistically. An example of this flexibility is in plantation created by a harvest. A proportion of the acres can be assigned to a class where regeneration is successful, a portion to where planting failed, and another portion where the site is captured by brush or hardwoods. Allowing acres to transfer at any age allows the modeling of explicit mortality due to fire, disease, insects or other causes.

2.4 Management Prescription:

Management prescriptions describe the activities that can occur in a stand on a per acre basis or within a subunit of the naational forest, such as a landscape or watershed. In the per acre case, many of the management constraints and desired future conditions are hard-wired into the prescription. These constraints and goals are modeled directly into yield and habitat tables developed by the vegetative simulators. In the latter case, the model is allowed to select from a suite of per acre prescriptions but must meet landscape constraints and move toward desired conditions at the larger scale unit and not the acre. Also, prescriptions that do not optimize a desired condition on an acre may be selected if they, in combination with the other prescription-acre combinations, meet the desired condition at landscape better than single choice. Both approaches have been used to model HFQLG alternatives. Table #3 is list of Per-Acre prescriptions modeled in this analysis along with key assumptions.

The ID team has defined which set of prescriptions are "permitted'' on each acre of land, by alternative. This assignment was mapped and overlaid with vegetative strata and other layers to form unique analysis areas for each alternative.

2.5 Objective Functions:

The objective function in a linear programming model specifies the goals to be minimized or maximized subject to a set of constraints. Up to five objective functions can be specified but each must be solved individually. All the alternatives run multiple goals but they are ordered differently. A "rollover'' technique is used where the first goal is maximized or minimized. Once this objective has been optimized, it is then set as a constraint and the next goal is optimized and then set as a constraint, until all the objectives have been optimized based on utilizing any slack left in the model. Also, maximization of Present Net Value (proxy for net public benefit) is always the last objective function analyzed unless it is already used as a prior goal. This financial objective ensures that if there are choices available on how the objectives can best be met, that the most economically-efficient choice would be selected. In Alternative 2, the first goal optimized is maximizing Present Net Value, then minimizing stand replacement acres, followed by maximizing timber volume.

2.6 Constraints [Standards and Guides]

Constraints, or standards and guidelines, are factors added to the linear programming model that limit the means of optimizing attainment of the desired condition or other objectives. In many cases, these constraints have already been imbedded in the development of yield and habitat tables. For example, the California Spotted Owl Technical Assessment (CASPO) basal area, diameter, and crown density retention rules are built into the CASPO prescriptions themselves. However, maintaining owl habitat over a landscape area as in Alternative 3, that meets a predefined size classes or conditions is modeled as constraints. There are three main types of constraints used in SPECTRUM. Absolute constraints are used to control a known amount of activity or output of effect for a given time period. For example, under Alternative 2, there is required to have a minimum of 200,000 acres of DFPZ acres treated. Flow constraints are used to model Forest Service policy related to non-declining yields. This requirement specifies that the flow of regulated timber volume must not decrease over the planning horizon. General relational constraints allow the user to specify proportional relationship between two quantities. These constraints can be based on acres (i.e. acres of group selection regeneration must be 5.7% of the acres available to this practice as in Alternative 2) or on an output or effects (i.e. owl nesting habitat must be at least 20% of the area within a watershed.)

2.7 Outputs from the SPECTRUM Analysis:

A number of different output reports can be generated from the SPECTRUM system. This includes files that are compatible with most databases and spreadsheet software. The solve file from the C-WHIZ LP solver (the mathematical optimizer used by SPECTRUM) is also generated and saved for each alternative. This output lists the optimal objective functions and information associated with each row and column in the solution. This file is formatted so that the SPECTRUM report writer and the user can interpret the Lp solution.

The SPECTRUM report includes a number of different summaries about each alternative run. The major portion of the report contains information on scheduling of different activities, the amount of outputs and effects produced by these activities, and their financial effects. Some of the activities and outputs reported include items such as acres of mortality by severity classes, volume of timber removed by various products, acres of the forest in various CWHR categories or old growth states, costs of different management activities, and the inventory, growth / mortality, and removal of various stand attributes such as snags, dead and down material, and large trees. Information can be broken down into classes by any of the 8 levels of identifiers and treatment classes used to define the analysis areas and management prescription.

Files can also be produced that can be utilized by other software programs to provide additional reports, charts, and graphs and databases. For the HFQLG analysis, results from SPECTRUM runs were entered into EXCEL spreadsheets and ACCESS database programs for further manipulation. Graphic software in EXCEL was used to develop charts depicting the change in CWHR states over time for different forest types, number of acres in specialized habitats such as owl nesting, and change in mortality acres. Tabular printouts are also provided. This step was taken to make the results from SPECTRUM more easily visualized and understood. Another reason for moving some of the data into EXCEL format is to allow us to use post processors such as @RISK, which analyzes the risk related to the uncertainty associated with much of the data used in this project. @RISK does Monte Carlo simulations of the results based on the estimated variability associated with the input data.

Table #1: ANALYSIS AREA IDENTIFIERS USED TO STRATIFY LANDS FOR SPECTRUM ANALYSIS

TABLE No: 02: MANAGEMENT PRESCRIPTIONS AND TREATMENTS CLASSES FOR SPECTRUM ANALYSIS

Table #3: ACTIVITIES, OUTPUTS, AND ENVIRONMENTAL EFFECTS TRACKED IN SPECTRUM ANALYSIS

3. Modeling at the PER ACRE Scale using GAMMA and FVS.

3.1 Forest Growth and Treatment Simulation Modeling for HFQLG Analysis.

In order to provide projections of future forest conditions resulting from different management activities to support HFQLG analysis, a forest growth and management modeling application called Gamma was developed. Gamma's design objectives include:

• generate spectrum yield streams for forest strata
• execute tree growth and mortality functions on individual plots, rather than running entire strata as single composite stands
• incorporate an episodic stand mortality model based on the Applegate Forest Watershed Study, also run at plot level
• store individual tree data in a database to facilitate post-priori analysis and data mining
• be able to address information at multiple spatial scales (strata, plot, point) in treatment simulations and output reporting
• retrieve data using database functions for easier programming
• use full power programming language and database functions for treatment simulations to enable precise simulation of any set of treatment rules.

3.2 AWFS Mortality

Many FVS users have found that long-term FVS projections of strata within the Sierran ecosystem generate stands at densities that are neither within common experience nor expectation. This is especially true of pine and to a lesser degree mixed conifer types. The reason is that the FVS model does not account for drought induced insect attack and episodic windthrow, two factors that strongly influence stand development. While stands within a small area may avoid insect attack and thus maintain high density, this is unlikely at the level of vegetation strata. The vegetation modeling for the California Spotted Owl Revised Draft EIS incorporated a relatively simplistic keyword based submodel to simulate bark beetle attack. A more sophisticated beetle attack/windthrow model was developed for the Applegate Forest Watershed Study. This model is incorporated, with some modifications into Gamma.

In the Applegate model, occurrence and effects of windthrow are predicted according to Table 4.

Stands subject to possible wind disturbance were at least:

• 4000 feet elevation,
• 120 square feet of basal area per acre
• 50 feet tall
• azimuth 45 through 315

Insect attacks in the AWFS occur in response to drought events for stands above certain basal area thresholds. Two types of drought events, mild and severe, are modeled. AWFS assumed mild drought occurs in 20 percent of 5-year growth periods and severe drought occurs in 5 percent of 5-year growth periods. In addition to a drought trigger, stands must be above a specified basal area threshold for insect attack to occur. The effects (i.e. the number of trees killed) in susceptible stands is a function of tree species and drought severity.

Most of the treatment scenarios for vegetation simulations for the HFQLG analysis were done in sets of 5, with the first treatment entry scheduled in periods 1 - 5. In order to avoid confounding the effects of randomly-generated drought or windthrow events with these first entry timing choices, the drought and windthrow effects are modeled as chronic conditions rather than episodic. In other words, the mean expected mortality is applied each period, subject to basal area or other thresholds. The annual expected mortality from windthrow in susceptible stands is 0.19 percent from combined light, moderate, and severe wind events. Mortality due to catastrophic wind events, an additional 0.5 percent annual loss, is not included. In Gamma, AFWS mortality is applied after treatment, stand growth, and FVS-based competition induced mortality and is applied plot-by-plot. The following tables show insect attack thresholds and percent tree kill used for HFQLG. Plots with site index 0, 1, and 2 are in the "high" Site group, site indexes 3, 4, and 5 are in the "low" site group.

The effects of incorporating the AWFS mortality model into forest growth projections are significant. Compared with trial runs, in which drought and windthrow events were allowed to trigger randomly, stand density peaks and "crashes" are reduced, especially at the plot level. Strata that are currently at high density see little basal area growth over the 200-year projections, as growth and mortality remain in balance. Projected numbers of large trees increases, and the understory gradually disappears, but stand basal area remains in the 220 - 260 square feet range, depending on site class and other factors. In the absence of AWFS mortality, stand density steadily increases to maximum stand density index limits.

3.3 Stand Treatments for HFQLG Analysis:

The Spectrum planning model, when using a Model 2 formulation, cannot accommodate partial treatment of strata, in which only a subset of plots meet the conditions for treatment. It's all or nothing. Therefore stand treatments modeled using Gamma are applied to entire strata, and not plot-by-plot. That is, the "stand" for purposes of treatment simulation is the strata, while the "stand" for purposes of growth and mortality is the plot.

All treatments, including fires, trigger an ingrowth event and hardwoods, if present, sprout immediately following the treatment. Ingrowth events are also triggered on a plot-by-plot basis whenever past-accumulated mortality exceeds 3000 board feet per acre.

Fire. In simulating fire, the tree-killing algorithms in FOFEM are applied to each tree in the stand. The factors that affect tree mortality in FOFEM include scorch height and bark thickness. Gamma calculates a scorch height based on user-supplied flame length using relations in FOFEM. Bark thickness is calculated using Region 5 species-specific equations found in Wessin and Icasca source code. For HFQLG, a 3.5 foot flame length (19.7 feet scorch height) is used for all Fire treatments, except for the lethal event. Fire recurs every 20 years. Dead and down material < 11" is reduced by 95%, dead and down material >= 11" is reduced by 50%. Snags are reduced by 50%.

CASPO. Follows the recommendations of the CASPO Technical report:

Selected Timber Strata. For M4G, M4N, and P4G strata (as well as F3G and W3G on the Lassen and Tahoe Basin), the treatment specifications are:

• thin from below, retain 40 percent of stand basal area,
• remove no trees 30 inches dbh or larger,
• if the post-treatment stand would have less than 40 percent canopy cover, leave enough additional trees larger than 12 inches dbh to make up the deficit.

• thin from below, keep 30 percent of stand basal area or 50 square feet, whichever is larger.
• remove no trees 30 inches dbh or larger.

DFPZ.

• Thin from below, retaining 40 percent canopy cover. Tree selection is in order of increasing height to bottom of crown base. Repeat every 20 years.
• Dead and downed material of all sizes and snags are reduced by 90 percent.

• Thin from below to 40 percent of maximum stand density index. Maximum SDI for the stand is calculated as the basal area by species weighted average of maximum SDI values used in the growth model. Tree selection is in order of increasing height to bottom of crown base. Repeat every 20 years.
• Dead and down material < 11" is reduced by 75 percent. Dead and down material >= 11" and snags are reduced by 25 percent.

• determine dbh limit, trees per acre requirement, and canopy closure requirement for LSOG rank 3, which varies by vegetation type. In cases where rank 3 is defined by more than one set of conditions, the set of conditions with the highest canopy closure requirement is used. The snags per acre criterion are not currently being used. If stand exceeds thresholds, proceed.

• Set up q-factor stand table for portion of stand below dbh limit based on 2 inch dbh classes and q = 1.3. Trees per acre (tpa) target in largest class is 1, for successively smaller dbh classes, target tpa is previous target multiplied by q. Calculate dbh class basal area targets using trees per target and class mid-point dbh.
• Remove excess large trees (those at and above LSOG dbh limit) proportionally. Post-treatment large tree numbers will be equal to LSOG 3 threshold. Diameter and species distribution will be reflection of pre-treatment distribution.
• For the remainder of the stand, beginning with largest dbh class, if the basal area exceeds the class basal area target, remove excess trees; otherwise add the deficiency to the next lowest class. Work from largest to smallest dbh class, removing excess trees.
• If harvest by above procedure is less than 2800 board feet per acre, cancel treatment.

• Commercial thinning.
• Calculate basal area by species weighted maximum SDI using growth model settings.
• If the stand is at or above 55 percent of the maximum SDI, reduce the basal area by 35 percent.
• One half of basal area removal is from below, selecting in order of increasing crown volume.
• The other half of basal area removal is proportional.

Lethal Event. Simulates stand replacement fire and recovery. To simulate stand replacement fire, tree mortality is predicted using FOFEM equations, assuming a 180 foot scorch height. Current inventories of dead and down material and snags are reduced to 0. Recovery is based on a lagged natural regeneration model:

• Hardwoods, if present sprout immediately, with number of sprouts a function of tree dbh as follows:

• The lagged natural regeneration model establishes conifer regeneration after a vegetation type specific lag time has occured. Number and size of conifer regeneration is also a function of vegetation type:

• Species composition of the conifer component is derived from pre-fire stand. Species distribution is dependent on hardwood occupancy, with higher hardwood occupancy favoring tolerant species.

Mixed-lethal events are simulated as 2 severe droughts occurring during the same period. Tree mortality is based on HFQLG insect attack thresholds and effects in above tables.

SUMMARY OF WILDLIFE GIS DATA

The following GIS coverages were used for Wildlife related analysis:

Ecological Zones (Eastside, Crest/Transition, Westside) - derived from Ecological Units of California.

PAC/SOHA's - derived by each affected forest.

Spotted Owl Home Ranges - obtained from Sierra Nevada Conservation Framework (J. Fites). Represents percent of suitable habitat (as defined below) within each Spotted Owl's home range.

Carnivore Network - the most current network as derived by each affected forest.

Goshawk Territories - the most current territories as mapped by each affected forest.

Red-Legged Frog Core Areas - California state watersheds (CALWATER 2.0), designated by the US Fish and Wildlife Service as being core areas. This was obtained from US Fish and Wildlife Service.

California Wildlife Habitat Relationship (CWHR) - derived from GIS vegetation layers based on LANDSAT imagery. The Interdisciplinary Team decided to use CWHR types for reporting vegetation summary analysis. In order to derive a CWHR coverage for the HFQLG area, each forest's vegetation inventory was crosswalked from 'strata type' to CWHR. FVS assigned CWHR types based on tree lists from each forest's latest inventory plots (the actual cross-walks available upon request in planning file). Once CWHR classifications (see table below) had been assigned, the forest coverages were merged.
 

Goshawk Habitat was defined as follows: Nesting = 4M, 4D, 5M, 5D. Foraging = 3M, 3D, 5S, 5P, 6.

Spotted Owl Habitat was defined as follows: Nesting = 5M, 5D, 6. Foraging = 4M, 4D.

Carnivore Habitat was defined as follows:

Westside and Transition Denning = 4D, 5D, 6
Westside and Transition Foraging = 4M, 5M
Eastside Denning = 4M, 4D, 5M, 5D, 6
Eastside Foraging = 4P, 5P

Return to Table of Contents