by Ralph Wells, Centre for Applied Conservation Biology, UBC

Eric Valdal, Invermere Forest District

Christoph Steeger, Pandion Ecological Research

and Pierre Vernier, Centre for Applied Conservation, UBC

INTRODUCTION

A key objective of the Invermere Enhanced Forest Management Pilot Project (EFMPP) is to develop socio-economic, wood supply and environmental projections over a harvest rotation (BCMoF 1998). To meet the goal of projecting environmental effects of Pilot harvest scenarios, we initiated a habitat analysis project in 1998, using the SIMFOR habitat model to evaluate to harvest scenarios projected by the BCMoF forest estate model FSSIM.

Historically, habitat modeling has often been based on qualitative information, based on the opinion of various habitat specialists. As a result these models have been subject to criticism because results are considered too subjective, difficult to test and hard to compare among (e.g. Van Horne and Wiens 1991). We believe that verifiable, quantitative approaches are essential to over come these limitations, and so we have developed two quantitative approaches to habitat modeling for this project.

We present two examples of these approaches in this report. In the first approach, we examine the effect of forest harvesting on the availability of nesting and foraging habitat of the Three-toed Woodpecker (Habitat Supply Modeling). In the second approach we use broad forest level variables to predict probability of occurrence of the Orange-crowned Warbler, (Habitat Association Modeling).

The Three-toed Woodpecker (Picoides tridactylus), is a year-round resident of British Columbia forests. In the B.C. interior, the species is common in lodgepole pine dominated stands, and preys extensively on the larvae Dendoroctonus ponderosae) and associated wood borers (Steeger and Dulisse 1997). We chose the Three-toed Woodpecker because of its association with late seral elements of the forest (nesting and foraging snags), which are affected by forest management. Further, the Three-toed Woodpecker is an aggressive predator of MPB and may play a role in the natural regulation of beetle populations (Steeger et al. 1998). It may also be vulnerable to a province wide policy of harvesting infested and susceptible stands of lodgepole pine in areas such as the pilot study area, where lodgepole pine stands are common. However, few studies have examined the potential effects of this policy on Three-toed Woodpecker habitat.

   Figure 1: Three-toed Woodpecker

The Orange-crowned Warbler (Vermivoa celata), is a neotropical migrant which breeds in many areas of B.C. This species is known to prefer habitats with shrubs and low vegetation that provide cover for its nest on or near the ground (Sogge et al. 1994). As such it is a representative of species requiring elements normally associated with early seral stages, and provides a good contrast to the late seral focus of the Three-toed Woodpecker model.

Figure 2: Orange-crowned Warbler

The EFMPP study area is a 258 000 hectare area southeast of Invermere, encompassing the White and Lussier River drainages (Anderson 1997).

1. To use the SIMFOR habitat model to evaluate harvest scenarios projected by the BCMoF forest estate model FSSIM.

2. To evaluate two quantitative approaches to habitat modeling (Habitat Supply Modeling and Habitat Association Modeling).

3. To investigate the effect of two harvest scenarios on the Three-toed Woodpecker and Orange-crowned Warbler.

METHODS

Nest and forage tree characteristics for Three-toed Woodpecker were developed from data collected in Deer Creek watershed (Wells and Steeger 1998), in the ICHdw subzone (Braumandl, and Curran 1992). Dominant stand types in Deer Creek are similar to the pilot study area, including lodgepole pine, Douglas-fir and western larch. In Deer Creek, Three-toed Woodpeckers were observed nesting primarily in lodgepole pine and western larch (Table 1). While the foraging behaviour of Three-toed Woodpeckers is not fully understood, they are know to be aggressive foragers of MPB and associated wood borers in lodgepole pine. In Deer Creek, nearly 80% of foraging observations were on lodgepole pine; most of these trees had indications of MPB infestation (Steeger and Dulisse 1997). Therefor, for our modeling exercise, we assumed that lodgepole pine stands harbouring MPB represent potential forage opportunity for Three-toed Woodpecker.

We determined the availability of potential nest trees for stand types and age classes found in the Pilot area using the Invermere Forest District cruise inventory database (e.g. Figure 3) which was developed from over 5000 plots and had snag classifications analogous to decay classes 2-4. Analysis units (AU; forest cover derived stand groups used for FSSIM modeling) in which abundant potential lodgepole pine nest trees were found include: AU 1-3 (Douglas-fir good, medium and poor sites); AU 4-6 (western larch good, medium and poor sites); AU 11-13 (lodgepole pine good, medium and poor sites). Analysis units with abundant western larch trees included AU 4-6 (western larch good, medium and poor sites). For habitat analysis, we used relative abundance of snags; high abundance was considered to be >75% of maximum abundances observed for the two tree species (>150 stems/ha for lodgepole pine; > 38 snags/ha for western larch).

We assumed snag densities to be < 10 stems/ha (well under thresholds used for habitat analysis) for all stands younger than 30 years because few plots were available to estimate densities in these stands; where data were available, low densities were observed. There are large areas in the study area that are of fire origin that still retain a legacy of snags recruited from fire mortality. We did not have data to estimate snag densities for these stands, and so they were not modeled in this study. Thus our results are likely conservative, underestimating available nesting habitat. We also did not develop trajectories for partial cut stands (AU 19-21) because few data exist to estimate snag availability under partial cutting regimes. Further, WCB regulations require removal of snags before harvesting, therefor we assume that without special management to retain snags, snag densities will remain low in these harvesting regimes. Finally, a major assumption of this approach is that stands developing after harvesting will generate similar snag densities as occurred in natural systems where cruise data were collected. This is not an unreasonable assumption, as snag densities do not reach high levels until older ages (e.g. >120 years), where the legacies of the stand initiating disturbance are expected to have less influence on snag densities than current tree mortality.

We predicted availability of foraging habitat using the Mountain Pine Beetle (MPB) susceptibility model of Shore and Safranyik 1992 for lodgepole pine (e.g. Figure 4). We used stand susceptibility as a surrogate for endemic MPB populations. Susceptibility is not a determination of risk, which requires knowledge of local beetle populations. As such, we are not attempting to predict outbreaks of MPB. Rather, the susceptibility results are intended to represent endemic levels of foraging opportunity. For this modeling exercise, we assume that endemic MPB populations represent minimum availability of foraging opportunity required to sustain populations of Three-toed Woodpecker between MPB outbreaks. In reality, Three-toed Woodpeckers will have some other foraging opportunities (in beetle infested spruce stands, for example), so our estimates are conservative.

Stands with less than 50% susceptibility were considered to have no endemic MPB, stands of 50-75% susceptibility were considered to have moderate levels of endemic MPB and stands >75% were considered to have high levels of endemic MPB (Emile Begin, p. comm.). Susceptibility was projected into the future using harvest schedules for the two harvest scenarios linked to stand density tables created from Invermere Forest District cruise data. We did not model succession; lodgepole pine stands were assumed to remain lodgepole pine after harvest. We also did not model susceptibility for stand types designated for partial cutting, but we assume that MPB susceptibility (and endemic beetle populations) will remain relatively low in these stands as lodgepole pine is not expected to be a major component.

Vernier et al. 1997 developed multiple regression models for a number of bird species based on data collected in the Beaverfoot and Kootenay River watersheds in 1993 and 1994. We used a logistic regression model for predicting probability of occurrence of the Orange-crowned Warbler based on the local ‘neighbourhood’ of broad forest level characteristics, such as seral stage, edge and road density and patch evenness. For our analyses we used raster maps of 1ha cell resolution; we define a neighbourhood as the average of 16 hectares surrounding a given cell (e.g. Figure 5). For Orange-crowned Warbler, the significant neighbourhood characteristics in the logistic regression model were based on the following two variables:

Newcut: % of neighbourhood in recent clearcuts (0-5 years)

Eveness: patch type evenness of neighbourhood based on Simpson’s index (0-1; 1 represents a neighbourhood with equal representation of all patch types). Patches were based on four seral stages: 0-5 years; 6-30 years; 31-90 years; >90 years). Non-forested patches were not included.

This model was successful in predicting non-occurrence 91% of the time and correctly predicted occurrence 39% of the time for the data used to generate the model. Thus model results are conservative, underestimating good habitat.

Habitat analysis for Orange-crowned warbler was limited to the MS zone, because the model was developed from data collected from plots in MS zone forests. Partial cut stands were also excluded, since no partial cutting occurred in stands where model data were collected.

The harvest scenarios evaluated in this study are "Base Case" and Strategy 98" flowed runs. The management objectives and model assumptions for these scenarios are described in detail in the EFMPP Management Strategy Report (BCMoF 1998). We outline key differences here: in the Strategy 98 scenario, there is immediate partial cutting in some Forest Ecosystem Networks (FENs), compared to a 40 year deferral in Base Case; for Strategy 98, adjacency constraints and green-up requirements in pine stands are removed and changes are made to Ungulate Winter Range definitions.

The two harvest scenarios evaluated in this study were developed by the Invermere Forest District using FSSIM, the B.C. Forest Service forest estate model. Typically FSSIM is a non-spatial model (i.e. specific harvest locations are not tracked), however for EFMPP runs, FSSIM was linked to GIS so that individual polygons harvested are tracked (e.g. Figure 6). This spatial tracking allows spatial habitat modeling to be undertaken with the SIMFOR habitat model. However, there are limits to this approach: while spatial locations of harvest are tracked, polygons may not represent realistic harvest units in an operational sense (polygons based on harvested cutblocks are an exception). This is reasonable for strategic analyses over extended time periods, but should not be considered a realistic representation of harvest for a specific place at a specific time.

A common problem with spatially explicit forest harvesting models is lack of natural disturbance modeling. While this may be reasonable for areas where harvesting is the dominant disturbance type, this is very unrealistic in inoperable areas. In the absence of disturbance these stands continue to age in the model, leading to seral stage estimates increasingly biased towards older seral stages as stands age during model runs.

We developed a disturbance model for inoperable stands based on an analysis of fire over the last 100 years in inoperable stands in the study area. FSSIM "harvested" an equivalent area to that disturbed over the previous 100 years in inoperable stands for a modeling time horizon of 100 years. We also applied a bimodal distribution of disturbed area, based on the observed historical distribution.

Relative abundance of potential pine and larch nesting were similar for Base Case and Strategy 98 harvest scenarios over 50 years (Figure 7a). By 50 years, potential nesting habitat has declined by 4 % for the Base Case scenario and by 7% for Strategy 98. Differences in potential nesting sites for Three-toed Woodpecker for a portion of the pilot area are shown in Figure 8.

Without accounting for natural disturbance in inoperable stands, the model would have overestimated nesting habitat by 8% for the Strategy 98 scenario (Figure 7a). Some of this disturbed area would likely provide nesting habitat due to snag recruitment, but this was not included in the model because of data limitations (see Methods). In this sense, our results are conservative, possibly underestimating available habitat in inoperable stands.

We also evaluated potential nesting habitat for Three-toed Woodpecker for a reconstruction of age classes in the absence of harvesting developed by BCMoF Research Branch for the study area (D. Morgan p. comm.). We note that this landscape represents only one estimation of many possible landscapes in an unmanaged scenario. While these results should be considered preliminary, they do show that 79 % of habitat area associated with old seral stages remains in a landscape with an intensive harvesting history compared to one estimation of a landscape in an unmanaged state (Figure 7a).

Foraging habitat results suggest that stands likely to maintain endemic populations of MPB are much less abundant than nesting habitat for the Three-toed Woodpecker. Currently, less than 1000 ha of the study area is estimated to have a high foraging potential (MPB susceptibility >75%; Figure 7b) and moderate foraging potential in less than 10000 ha (susceptibility 50-75%; Figure 7c). High potential forage habitat is projected to increase by 34% and 24% respectively for Base Case and Strategy 98 (Figure 7b), and 18% and 20% respectively for the moderate foraging class (Figure 7c). Differences between harvest scenarios in potential nesting sites for Three-toed Woodpecker are shown in Figure 9 for a portion of the pilot area.

We modeled probability of occurrence of the Orange-crowned Warbler in non-partial cut stands in the MS zone of the study area. We predicted probability of occurrence >0.5 for 1780 ha (or 7%) the 26730 ha included in this portion of the study area for year 1 of the planning period. By year 50, this had declined to 307 ha (1%) for the Base Case and to 670 ha (3%) for Strategy 98. Differences in probability of occurrence of the Orange-crowned Warbler for a portion of the pilot area are shown in Figure 11 for the two harvest scenarios.

We met our objective of evaluating FSSIM harvest scenarios with the SIMFOR habitat model for two different quantitative approaches to habitat analysis and for two terrestrial vertebrates. For Three-toed Woodpecker we found that while potential nesting habitat seems abundant for both the Base Case and Strategy 98 harvest scenarios, foraging habitat may be limiting. This result deserves further investigation because studies elsewhere suggest that limited forage opportunity may be reducing reproductive success of Three-toed Woodpecker (Steeger and Dulisse 1997). Our results also show that the neotropical migrant Orange-crowned Warbler, an early seral specialist, may have limited habitat in MS zone forests of the study area with substantial declines in habitat area under both harvesting scenarios.

For both models results should be considered preliminary until they have been verified in the field. The paucity of Orange-crowned Warbler habitat may be related to the evenness component of the regression model. High ranking for this species occurs when a range of seral stages are closely associated with each other (Evenness) and an early seral patch (Newcut). Orange-crowned Warblers may be associated with young patches surrounded by later seral stages because of the feeding habit of gleaning from tree foliage and the males preference for vocalizing from canopy trees (Sogge et. al. 1994). These conditions may be rare in the area analysed in this study. A final caution for predictions of occurrence based on neighbourhood models: since these models are highly spatially explicit, accuracy of predictions will be limited by the spatial accuracy of FSSIM harvests, to the extent that FSSIM harvest polygons represent realistic harvest units.

More generally we can draw some conclusions about the two approaches to habitat analysis. First, we believe that habitat supply models are most useful for species that depend on quantifiable habitat components that are likely to be affected by forest management Cavity-nesting species with their dependence on snags are ideal species for this approach. Habitat supply modeling is also amenable to modeling species groups, where different species use of similar habitat components can be quantified. Again, cavity-nesting species are ideal for modeling as species groups, because of similarities in snag use among species and because secondary users make use of the nests constructed by primary nesters. We have also shown that forest inventory databases can be used to develop estimates of habitat supply.

The results of Vernier et al. 1997 suggest that for some species, good regression models can be developed with coarse, forest level variables, easily obtained from forest cover mapping. These models may be of use for determining general trends in habitat for species not amenable to habitat supply modeling. This includes species which do not have strong associations to specific structural habitat components, or for which these associations remain uncertain. However we stress that to obtain reliable results, regression models must be applied in areas of similar composition, treatments and ranges of variability to the areas from which the models were developed. Until proven reliable, results should always be verified by field monitoring. More generally, simple presence of a species (Habitat Association) or an important habitat component (Habitat Supply) does not necessarily mean that a species is viable. Either approach to modeling habitat does not negate the need for ongoing research and monitoring of long term persistence of species.

Finally, because both approaches are quantitative, they are verifiable by field sampling (i.e. we can measure the stems/ha of a snag type in different stand types and age classes). Where snag densities are found to be present as predicted, use by cavity nesters can be evaluated. As such they are testable and subject to improvement. As a result they can form an integral part of an adaptive management strategy.

1. We plan to refine and test nesting and foraging habitat assumptions of the Three-toed Woodpecker and other cavity nesting birds based on field data recently collected from the Invermere Pilot area (Steeger and Quesnel 1998). We plan to apply regression models for other neotropical migrants such as the Townsend’s Warbler (Dendroica townsendi), often associated with mature stands. Opportunity to test and refine regression models for songbirds also exists, because of research currently underway in the Pilot study area (Kari Stuart-Smith 1998). Finally, we hope to develop habitat supply models for Northern Goshawk and Bobcat in the study area.

2. We intend to expand our modeling to include a broader range of cavity-nesting birds by utilizing Terrain Ecosystem Mapping (TEM) data and other sources to estimate deciduous availability (an important nesting and foraging source for many cavity nesters). We also plan to develop models predicting patches of Armillaria infected Douglas-fir, important for forage habitat for many cavity-nesters, and nesting habitat for weak-primary cavity nesters such as the Red-breasted Nuthatch (e.g. Steeger and Hitchcock 1998).

3. We believe that model predictions for current landscape patterns are best treated as hypotheses. In this study, we present hypotheses about nesting and foraging habitat for the Three-toed Woodpecker, and likelihood of occurrence of the Orange-crowned Warbler. As such they should be tested by field verification, and models should incorporate new findings.

4. We further that believe habitat objectives should be incorporated into harvest strategies for species whose habitat models have been field validated. For example, our models could be used to develop harvest plans that retain of patches of Three-toed Woodpecker nesting and foraging habitat, once models have been verified and refined. These management treatments could then be monitored as part of an adaptive management strategy for Three-toed Woodpecker.

We gratefully acknowledge Forest Renewal B.C. funding provided by the Invermere Forest District EFMPP. We further thank Greg Anderson for his support of this project. Thanks also to Russ Hendry for providing the harvest schedules, Emile Begin for discussions on mountain pine beetle modeling and Peter Holmes for comments on habitat results. Finally, special thanks to Susan Shirkoff for assistance in model development and runs and Fred Bunnell for helpful comments and support. This represents research report R-28 of the Centre for Applied Conservation Biology, University of British Columbia.

Anderson, G. 1996. Enhanced Forest Management Pilot Project. Project Summary. B.C. Ministry of Forests, Invermere Forest District. Report No. 1.

B.C. Ministry of Forests. 1998. Enhanced Forest Management Pilot Project Management Strategies Report. Invermere Forest District. Invermere, B.C. 76pp.

Web address: www.for.gov.bc.ca/nelson/district/invermer/pilot/strategy/strategy.htm

Begin, Emile. Personal communication. Forest health Officer. Invermere Forest District, Invermere, B.C.

Braumandl, T.F. and M.C. Curran (eds.). 1992. A guide to site identification and interpretation for the Nelson Forest Region. British Columbia Ministry of Forests. Nelson, B.C.

Morgan, Don. Personal communication. B.C. Ministry of Forests Research Branch. Victoria, B.C.

Shore, T.L. and L. Safranyik. 1992. Susceptibility and risk rating systems for the mountain pine beetle in lodgepole pine stands. Canadian Forest Service Pacific and Yukon Region. Information Report BC-X-336.

Sogge, M.K., W.M. Gilbert and C.V. Ripper III. 1994. Orange-crowned Warbler (Vemivora celata). North America, No. 101. (A. Poole and F. Gill, Eds). Philadelphia: the Academy of Natural Sciences; Washington, D.C.: The American Ornithologists Union.

Steeger, C. and J. Dulisse. 1997. Ecological interrelationships of Three-toed Woodpeckers with bark beetles and pine trees. B.C. Ministry of Forests, Nelson Forest Region. Research Summary RS-035.

Steeger, C., M.M. Machmer and B. Gowens. 1998. Impact of insectivorous birds on bark beetles: a literature review. Contract report submitted to B.C. Ministry of Forests. Victoria, B.C.

Steeger, C. and C.L. Hitchcock. 1998. Influence of forest structure and diseases on nest-site selection by red-breasted nuthatches. Journal of Wildlife Management. 62: 1349.

Steeger, C. and H. Quesnel. 1998. Impacts of partial cutting on old-growth forests in the Rocky Mountain Trench: Interim Report. Enhanced Forest Management Pilot Project Report No. 9. Invermere Forest District. Invermere, B.C.

Stuart-Smith, K. 1998. Ecological comparisons between wildfire and logging, the East Kootenay songbird project: Interim Report. Enhanced Forest Management Pilot Project No.8. Invermere Forest District. Invermere, B.C.

Van Horne, B. and J.A. Wiens. 1991. Forest Bird Habitat Suitability Models and the Development of General Habitat Models. In Fish and Wildlife Research Report No. 8. United States Department of the Interior Fish and Wildlife Service.
31 p.

Vernier, P, Ann Chan-McLoed and F.L. Bunnell. 1997. Testing and Refining Biodiversity Approaches. Forest Renewal B.C. Final Report.

Wells, R.W. and C. Steeger. 1998. Species-habitat relationships and habitat supply for cavity-nesting birds in a dry Interior Cedar Hemlock watershed. Unpublished Report. Centre for Applied Conservation Biology, University of British Columbia.