Influences of Adjacent Forest Management Activities on Migratory Elk of Mount Rainier National Park
Kurt J. Jenkins, Edward E. Starkey

CH. 3: ELK HABITAT MODELING

Habitat modeling is an important tool used in assessing influences of land-use developments on wildlife populations. Simulation procedures that link models of secondary forest succession and carrying capacity (the number of animals that can be supported per unit area of habitat) are particularly useful for predicting the long-term effects of forest management activities on big-game populations and habitats (Hett et al. 1978, Raedeke and Lemkuhl 1985, Jenkins and Wright 1987). Such models are generally very tractable, cost-efficient, and they enable managers to predict responses of wildlife to a variety of forest management activities without long-term and costly field experimentation (Raedeke and Lemkuhl 1985).

Two models of forest succession/carrying capacity have been developed for assessing responses of elk populations to forest harvesting activities near Mount Rainier National Park (MORA), Washington. One model, HABSIM, was developed for the National Park Service to determine possible influences of forestry practices on elk populations that summer within the MORA (Raedeke and Lemkuhl 1984, 1985). Another model, FORPLAN, was developed by the U.S. Forest Service to assist with long-term planning of timber and wildlife resources on the Gifford Pinchot National Forest adjacent to MORA (B. Ruediger, Gifford Pinchot National Forest, pers. comm.). Although the overall objectives and structure of the two models were similar, they produced markedly different predictions of elk population responses to logging. While such discrepancies cast obvious doubts on the predictive accuracy of one or both models, further examination of these differences would fulfill an important role of ecological modeling -- identifying deficiencies in ecological understanding and important areas of future research.

The purpose of this study was to review existing models of elk responses to forest harvesting and secondary succession in the MORA ecosystem, and attempt to resolve discrepancies between the two models. In so doing, we developed a new model, a hybrid of its predecessors that we believe draws the best features from both models using site-specific data obtained from this study. Unlike its predecessors, however, our model forecasts forage values of elk winter ranges rather than elk densities. That approach reflects our perspective that forage, although having an important influence on habitat quality, is a poor predictor of actual elk numbers. Lastly, we demonstrate several applications of our model by assessing the influences of harvesting rates, rotation lengths, thinning rates, hardwood conversions and winter snowfall on forage values in the White River drainage adjacent to MORA.

Review of Previous Models

Both HABSIM and FORPLAN were developed as tools for estimating future Potential Carrying Capacities (PCC) of elk winter ranges adjoining MORA (additionally FORPLAN has very broad capabilities of estimating a variety of future environmental qualities in the managed forest). PCC is defined as the maximum density of elk that a habitat can support at a specified time based on the available forage supplies and food-requirements of elk. Actual densities of elk may be less than PCC whenever factors other than forage supplies limit the elk population (e.g., hunting, predation, disease, density independent mortality).

The concept of PCC used in HABSIM and FORPLAN differs from the interactive forage-based concept of ecological carrying capacity (ECC) described by Caughley (1976:217). ECC is the equilibrium density of herbivores that is established after a lengthy period of mutual interaction between herbivore and food supply. PCC, in contrast, is the density of herbivores capable of being supported by the food-supply at any specified time.

The structures of HABSIM and FORPLAN are very similar. Both models use a computer algorithm as a simple book-keeping tool to track acreages of several forest age-classes during the simulation period. The modeler must provide data on the initial acreages of each forest age-class, and an estimate of PCC of each. After each simulated time-step, each acre is advanced one unit of age, and a specified proportion of acreage exceeding rotation age is set to age 0, thus simulating harvest. PCC of the region is determined for each time-step by multiplying acreage of each forest age-class by its estimated PCC (i.e., elk/acre). Both models assume that PCC of forest age-classes is constant during the simulation period. Hence, each model implies that PCC is uninfluenced by environmental factors such as winter severity, human disturbance, spatial arrangements of habitat patches, or any interaction between elk and their food supplies (e.g., retrogressive succession, accelerated succession, response of plant production to herbivory).

Despite the conceptual similarities, HABSIM and FORPLAN consistently produce different patterns of elk population responses to timber harvest on winter ranges adjacent to MORA. HABSIM predicts that PCC of winter range will decline steadily following the initiation of logging (Raedeke and Lumkuhl 1984), whereas FORPLAN predicts that PCC will increase for approximately 20 years following harvest (B. Ruediger, GPNF, pers. comm.). FORPLAN, like HABSIM, predicts that PCC decreases after 21-60 years following logging; i.e., after regenerating overstories reach complete crown closure and shade out understory forages. The major discrepancy between models, therefore, occurs in the early stages of forest regeneration when HABSIM predicts declining habitat quality and FORPLAN predicts an increase in habitat quality.

Reasons for this discrepancy can best be appreciated by examining the sequences of post-logging forest succession incorporated in each model (Fig. 3.1). HABSIM incorporates six age-classes in the successional sequence (Fig. 3.1). HABSIM assumes that PCCs of forested age-classes are maximum in mature forests (60-150 years post-logging) and old-growth timber (stands > 150 years old), and that PCCs of immature, regenerating stands are less than in older forests. Given those relative rankings of carrying capacity, the model will always predict a decline in PCC whenever mature or old-growth stands are harvested.

FORPLAN, in contrast, allows for only three stages of forest succession after logging (Fig. 3.1). Immediately following logging, PCC of early-seral clearcuts is presumed to be 2.6-13.0 times greater than PCC of mature forests (Fig. 3.1). Variable PCC of clearcuts reflects variable production of elk forage associated with three different management practices. FORPLAN assumes that PCCs of clearcuts declines to zero at 21 years post-logging, and then increases in mature stands of second growth (60 years +). High PCC weightings given to clearcuts causes FORPLAN to predict an increase in habitat quality for a limited time following logging.

Figure 3.1. Comparison of post-logging successional pathways and relative carrying capacities of forest age-classes used in the habitat models, HABSIM and FORPLAN. Relative carrying capacities in parentheses are expressed as fractions or multiples of PCC of mature forests ( > 60 years old).

Differences in PCC weightings used in the two models reflect different estimation techniques, as well as broadly different assumptions regarding factors limiting elk populations in the Pacific Northwest. Habitat weightings used in HABSIM are based largely on professional "guesses" at carrying capacity. Those weightings assume implicitly that deep snow eliminates many forage values of early seral clearcuts during winter, and that elk are limited primarily by forage available in mature coniferous stands during severe winters. Raedeke and Lemkuhl (1984) recognized that these assumptions present a conservative view of population limitation of elk in the western Cascades, and that elk may benefit in actuality from forage produced in clearcuts during mild winters. They concluded, however, that populations are limited in the long term by abilities of habitats to carry elk through severe winters.

Habitat rankings used in FORPLAN were based on forage production in the Siuslaw National Forest in the Coast Range of Oregon (B. Ruediger, GPNF, pers. comm.). PCC of each forest age-class was determined from estimated biomass of consumable forage in each age-class and estimated forage consumption by elk aver a 120-day winter. Those computations of PCC assume implicitly that forage production in the Cascade Mountains of Washington is the same as in the Coast Range of Oregon. Furthermore, the procedure assumes that forages in clearcuts are available to elk during winter. The model fails to account for seasonal reductions of forage in clearcuts due to deep accumulations of snow and increased cover requirements of elk during severe winters.

Each of the models described above presents a simplistic picture of complex elk-range interrelationships in the Cascade Mountains. Comparisons of model predictions to observed elk populations in the White River do not support the simple modeling assumptions. Elk populations in MORA's northern elk range more than tripled from 1974-1984 (unpublished data, S. Schlegel, Mount Rainier National Park) during a period when predictions of PCC declined (Raedeke and Lemkuhl 1984:62). Raedeke and Lemkuhl suggested that opposing patterns were caused by an uncoupling of elk population and habitat trends, i.e., that elk populations colonized an understocked range at the time PCC of the winter range was declining. An alternative explanation is that elk populations increased in response to low levels of sport-harvest and increased forage available in clearcuts during a series of mild winters in the 1970s and early 1980s. Additionally, elk have been observed wintering at elevations as high as 3000' and above during recent mild winters (typically transitional spring range), which has enabled elk to exploit forage produced in clearcuts at elevations that are not usually accessible during normal winter (pers. comm., R. Spencer, Washington Dept. of Wildlife). Although these explanations would tend to support the assumptions of FORPLAN over those of HABSIM, we stress that elk population trends following a series of severe winters may more closely resemble predictions of HABSIM than FORPLAN. Densities of elk wintering in the Cascade Mountains reflect a complex interplay between poorly understood forage successional patterns, cover requirements of elk, winter severity, and sport-harvest.

The above examination of model assumptions and performances fulfills an important function of ecological simulation. It points out deficiencies in our understanding of interacting components of ecological systems. And it points to specific improvements of existing models that would enhance model realism and predictive capabilities. Clearly additional research is needed on relationships among overstory canopy, snow depth, availability of preferred forages, and rates of nutrient acquisition in a variety of forest habitat types fauna in the western Cascade Mountains. Additional modeling efforts in the MORA ecosystem should incorporate site-specific measures of forage availability, more realistic patterns of post-logging forage succession, and influences of stochastic snowfall on elk habitat qualities.

Model Development

The model developed below, like HABSIM and FORPLAN, links models of successional patterns with habitat quality to permit long-term assessment of forest management activities on elk habitats in the MORA ecosystem. Unlike either of its predecessors, the model incorporates site-specific data on elk distribution, forest successional patterns and forage availability. We make no attempt to simulate actual carrying capacities (i.e., potential numbers of elk) of the MORA ecosystem; rather, this model simulates changes in forage values of habitats resulting from past and present logging activities. It is tempting to equate variations of forage values caused by forest management activities with corresponding changes in carrying capacity of elk ranges. Carrying capacity, however, is defined by complex interactions among availability of high quality forages, availability of forested cover used for energy conservation, human disturbance factors, and interspersion of required habitat components. Changes in forage values, therefore, would influence carrying capacity only if forage were limiting. Other important limiting factors, including cover, winter weather, and hunting will be discussed qualitatively. The model is deterministic in the sense that winter severity is held constant to allow the user to compare influences of several forest management scenarios without the confounding influence of variable winter severity. We have, however, added an optional stochastic element in the model to also permit an assessment of potential effects of winter weather.

The model was developed for assessing forage values on specific winter and spring home ranges used by migratory elk in the White River drainage (Fig. 1.2-1.3). The first step in model construction, therefore, was to describe site-specific patterns of forest succession found within the home ranges of elk. We identified two broadly different successional pathways along the White River, one describing a mesic sere in bottomlands on the valley floor, and another describing a xeric sere on elevated river terraces and uplands (Fig. 3.2). For both seres, two stages of clearcuts were recognized; 0-10 year-old clearcuts, which corresponded to grass-, forb- and shrub-dominated stages of early succession, and 11-20 year-old clearcuts, which corresponded to shrub- and conifer-dominated stages that occur before complete crown closure is reached. Successional pathways were variable after age 20 in both seres. Mesic stands developed closed canopies of red alder (ALRU, Fig. 3.2) on mesic to hydric soils of the Tsuga heterophylla/Polystichum munitum and Thuja plicata habitat types (Henderson and Peter 1984). Grass/sedge stands formed an hydric soils on Thuja plicata habitat types. Mid-seral stages of succession in xeric seres included thinned and unthinned Douglas-fir stands (PSME, Fig. 3.2), sparse PSME stands on xeric, shallow soils, and PSME-Black Cottonwood (POTR, Fig. 3.2) stands on shallow alluvium.

Figure 3.2. Post-logging successional pathways for mesic and xeric seres along the White River, Washington. Sold lines represent forest successional pathways and dashed lines represent timber harvesting options in the model. Transition probabilities shown for 20-year-old forests were determined from aerial photographs and thinning records.

System feed-backs were incorporated in the model by allowing rotation-stands to be converted to clearcuts as a result of timber harvesting activities (Fig. 3.2). Rotation lengths of 60 and 100 years were assumed for xeric and mesic seres, respectively. Longer rotation periods in the mesic sere simulate the effects of approximately 40-60 years of competition between regenerating conifers and either red alder overstories or grass-sedge understories.

Forest successional patterns observed along the White River were simulated using a compartment-based model similar to that described by Shugart et al. (1973) and applied to elk habitats by Hett et al. (1978). The modeling algorithm treats each vegetation class as an environmental compartment and land-area as a variable that flows between compartments. The rate of flow of land between compartments during each simulated time-step is controlled by a series of transition probabilities. In this model we used a time increment of 5 years, and during each iteration we allowed 100% of area within each 5-year age-compartment to flow to the next older compartment. Following each iteration, land areas contained within each of the vegetation and age-classes shown in Fig. 3.2 were determined by summing land areas within the appropriate 5-year intervals. Flow of land-areas among various mid-successional pathways were governed by transition probabilities determined from aerial photographs. Rotation-aged stands were allowed to remain in the same cover class during an iteration or they were converted to clearcuts at a user-defined transition rate. From an ecological viewpoint, transition probabilities allowed seral communities to age five years during each iteration, and they allowed rotation-aged stands to either remain the same, or to be harvested and converted to a clearcut.

The successional model was used to simulate future patterns of forest succession. To provide a complete record of past and future habitat composition, historical patterns were reconstructed from logging records of each land-ownership in the study area. Historical patterns linked with future projections provided a continuous record of past and future vegetational composition of elk winter and spring ranges in the White River drainage.

The successional model was developed for making projections of elk forage values on winter and spring ranges in the White River drainage. This was accomplished by estimating a forage value index (FVI) for each vegetation class and weighting these indices by acreages of each vegetation class after each model iteration. As discussed in chapter 2, FVI was defined as the sum of digestible dry matter of forage classes in each vegetation class weighted by relative forage preferences of elk (see Fig. 2.5). An examination of Fig. 2.5 reveals that FVI of each vegetation class varies seasonally in response to changing snow depths and forage availabilities. Therefore, it was necessary to derive mean estimates of FVI that described relative forage values of vegetation classes averaged over winter. Mean FVIs were computed as the mean of weekly estimates of FVI obtained between 1 November 1986-15 April 1987 (Fig. 2.5). Those indices reflected average forage values for a winter that received approximately 12" of snow for a two-week period. In actuality, however, FVIs vary between years in relation to depth and duration of snowpacks. Therefore, we also estimated mean FVIs of each vegetation class for a variety of snowpacks possible in winter (Table 3.1). Mean FVIs were estimated for variable snowpacks based upon known height distributions of twigs in each vegetation class (K. Jenkins, unpublished data). Unless otherwise specified, all simulations of forage values were based on snowfree winter conditions. An optional stochastic feature permitted evaluation of influences of random snowpacks on forage values. To simulate the effects of stochastic snowfalls, a sub-routine was used to randomly select (equal probabilities) one column vector of FVIs corresponding to a specific combination of depth and duration of snowpacks.

Simulation of FVI as a function of successional change required the use of several simplifying assumptions. First, it was assumed that rates of forest succession were constant during the simulation. Secondly, it was assumed that forage values were constant within each vegetation class and throughout the simulation period, and there was no interaction between elk and their food supplies.

Model Applications

The forage succession model was used to simulate changes in elk forage values resulting from past and future forest management activities in the White River watershed. Forage values were simulated separately for winter and spring ranges of the White River elk herd. Results of the model, therefore, apply only to these specific ranges and may not reflect regional trends.

Reconstruction of past logging histories indicated that forest management activities have influenced forage values profoundly for both the winter and spring ranges of elk in the White River (Fig. 3.3). Forage values increased sharply on winter ranges during the 1960's, reflecting the rapid liquidation of low elevation old-growth forests during the late 1950's and 1960's (Table 1.3). Forage values of winter ranges peaked in the late 1960's and declined from the 1970's to the present. Declining forage values have resulted primarily from development of complete crown closure in dense regenerating stands of Douglas-fir.

Figure 3.3. Simulated response of Forage Value Indices (FVI) to past forest harvesting activities and forest succession on winter range, spring range and combined range in the White River watershed. Combined FVI is the sum of FVIs for winter and spring ranges.

Forage values on spring ranges also increased in response to forest harvesting during the 1950's and 1960's (Fig. 3.3). Forest harvesting, however, occurred more gradually on higher elevation spring ranges than on low elevation winter ranges. Consequently, the increase in forage values seen on spring ranges occurred more gradually than on winter ranges and has been more persistent. Forage values on spring ranges appear to have increased steadily from the time logging was initiated until the present.

Four different forest management activities were simulated using the forest transition matrices, the 1985 plant community composition, and forage value indices. The first set of simulations examined the influence of different rates of forest harvesting on winter and spring ranges. Harvest rates referred to the percentage of rotation-aged stands that were cut every five years. Computationally, these are the transfer rates governing the flow of land area from second-growth and old-growth model compartments to the clearcut compartment during each model iteration (Fig. 3.2).

Three harvest levels, corresponding to 0%, 20%, and 40%, were simulated on elk winter ranges along the White River (Fig. 3.4). Predicted forage values declined steadily from 1985 levels until approximately 2015 in each simulation. Declining forage values reflected overstory development and shading in areas clearcut during the 1950's and 1960's. Different harvest rates had very little effect on forage values before 2015 because only a few stands came of rotation age before then. Beyond that date, forage values increased at a rate depending upon cutting intensity. Simulation of a no-harvest option produced a pattern of declining forage values well into the next century.

Figure 3.4. Simulated response of Forage Value Indices (FVI) to three levels of forest harvesting on elk winter range in the White River watershed. FVIs are scaled relative to baseline 1985 levels.

Two harvest levels were simulated on elk spring ranges, corresponding to 20% and 40% five-year harvest rates (Fig. 3.5). As on winter ranges, estimated forage values of spring ranges declined until the existing second-growth stands reached rotation age. Simulated forage values decreased rapidly during the 1990's as the extensive areas harvested in the early 1970's developed a closed canopy. Stands that were harvested in the 1940's reached rotation-age in the early 2000's. Simulated harvest of those stands helped stabilize forage values of spring range during the early 21st century.

Figure 3.5. Simulated response of Forage Value Indices (FVI) to two levels of forest harvesting on elk spring range in the White River watershed. FVIs are scaled relative to baseline 1985 levels.

A second set of simulations examined the influence of precommercial thinning on forage values in 20-40 year-old Douglas-fir stands. Two thinning scenarios were compared on elk winter ranges: 0% thinning rate in which no stands were thinned, and 100% thinning in which all stands were thinned every five years. Variable thinning rates were simulated by varying the transition probabilities which governed the flow of land area into thinned versus unthinned model compartments (Fig. 3.2).

Simulated thinning activities had little influence on forage values of elk winter ranges in the White River (Fig. 3.6). The small influence of thinning reflected narrow differences of forage values between thinned versus unthinned stands (Table 3.1).

Figure 3.6. Simulated response of Forage Value Indices (FVI) to two levels of precommercial thinning of 20 year-old Douglas-fir forests on elk winter range in the White River waterhsed. FVIs are scaled relative to baseline 1985 levels.

A third set of simulations compared forage values resulting from various harvesting rates of second-growth alder communities. We simulated 0%, 20% and 60% five-year harvest rates in alder communities by adjusting transition probabilities governing flow of land between alder and clearcut model compartments.

Simulated harvest of alder communities produced a minor influence on forage values of elk winter range (Fig. 3.7). Even under very intensive harvesting, forage values generated from logging were insufficient to offset declining forage values overall. Although forage values increased following logging in alder communities, land area was too small compared to that of Douglas-fir communities for logging in alder stands to have a strong influence on forage values overall.

Figure 3.7. Simulated response of Forage Value Indices (FVI) to three levels of forest harvesting of 30-year-old red alder stands on elk winter range in the White River watershed. FVIs are scaled relative to baseline 1985 levels.

A fourth simulation examined the influence of rotation-length on forage values. A shortened rotation length of 45 years was simulated by adjusting transition probabilities of 45-55 year-old stands to allow 20% harvest every five years.

Shortened rotation length produced an appreciable increase in forage values of winter range after the year 2010 (Fig. 3.8). Increased forage production due to shortened rotation was sufficient to increase forage values to approximately the 1985 level by 2020.

Figure 3.8. Simulated response of Forage Value Indices (FVI) to two harvest rotation lenghts on elk winter range in the White River. FVIs are scaled relative to baseline 1985 levels.

Each of the above simulations assumed a constant, negligible influence of snow, which seemed desirable for the sake of making comparisons among forest management activities. In an additional simulation, however, forest management activities were held constant and severity of winter snowpack was adjusted randomly during each model iteration. Incorporating stochastic snowfall in the model permitted an assessment of the influences of winter severity on forage values.

Influences of stochastic snowfall are demonstrated in Fig. 3.9. Forage values for random snow depths were always less than the baseline values because the baseline represented snowfree conditions. The annual percentage change in forage value due to snowfall averaged 29%, compared to 9% due to forest succession in the base run, indicating that stochastic variation in snowfall would have an appreciable influence on forage values. The influence of random winter severity did not obscure underlying successional patterns, but it added a high level of annual variability.

Figure 3.9. Simulated response of Forage Value Indices (FVI) to snowfree conditions versus random snow depths on elk winter range in the White River watershed. FVIs are scaled relative to baseline 1985 levels.

Discussion

Our model required the use of several simplifying assumptions. The primary assumption was that forage production and successional trends remained constant during the simulation period. Forage production and successional trends appear to have been constant in the recent past. Successional pathways and forage values, however, may change appreciably in the future due to modern forest management practices. Forage seeding and new thinning practices in clearcuts are just two examples of management practices that could enhance forage values of regenerating forests. Because we were unable to anticipate and model these and other possible management activities, results of our simulations are most useful for evaluating immediate or short-term influences of selected management activities.

Simulation results suggest that forage values of elk winter and spring ranges have been altered appreciably by recent forest management activities. There is little question that logging practices have improved forage conditions on both the spring and winter ranges of elk that migrate from MORA, but that forage conditions are now declining on both ranges as a result of forest succession and overstory closure. It is important to recognize that the anticipated decline on managed lands is due largely to the rapid rate with which old-age stands in this drainage were converted. The population increase that this brought about in recent years, aided by mild winters, simply cannot be sustained. Long-lived seral communities in bottomlands, such as red alder and grass/sedge communities, have prolonged forage benefits derived from logging somewhat, but they are not extensive enough to offset declining forage values on the winter range overall.

Our simulations also suggest that opportunities are limited at the present for improving forage conditions in the near future through normal silvicultural practices. The majority of stands are approximately 30 years old on the primary winter range. Thus they will require an additional 15-30 years until they reach harvestable age. There are opportunities for harvesting red alder at present, however, forage values of existing red alder communities are already high, and harvestable alder stands are scarce compared to Douglas-fir, so harvesting alder communities will have little effect on forage values. Lastly, accumulation of slash and low forage production in thinned stands appear to limit possibilities of improving forage values appreciably through extensive thinning.

Changes in forage values of an elk range would produce changes in carrying capacity only when forage is limiting. Primevally, elk populations west of the Cascades appeared to be limited largely by the availability of forage in seral and old-growth riparian corridors along the major river systems (Raedeke and Taber 1979, Starkey et al. 1982). It is likely, therefore, that improved foraging conditions due to logging would have increased elk carrying capacities during the early phases of logging in the White River. At present, following extensive logging of bottomland forests, cover used for energy conservation during severe winters and security from harassment may be more limiting than forage. Elk populations limited by cover may he expected to exhibit rapid growth during successive mild winters and to exhibit large density-dependent and -independent winter mortality during periodic severe winters. Under present circumstances, therefore, we consider forage to be an unreliable indicator of elk population trends.

Simulation models presented above, based an site-specific data, improve upon the previously existing, more speculative models. Results of our simulation models support the notion that increased densities of elk within Mount Rainier National Park during the 1970's may be related to increased availability of forage resulting from intensive logging outside the park. Based on the current stand age-structures and successional patterns, however, we anticipate that forage values of winter and spring ranges outside MORA will decline until the next century. Future reductions of available forage, together with reductions of mature forested cover, will reduce potential carrying capacities of winter and spring ranges adjoining the park. Reductions of mature forested cover, as suggested above, may also increase density-dependent and independent winter mortality of elk and could stabilize population numbers.