Hans R. Zuuring

Relationships among grizzly bears, roads and habitat in the Swan Mountains, Montana

Relationships between grizzly bears, habitat, and roads were investigated between 1990 and 1994 in the Swan Mountains, Montana. Relationships were examined at three levels of resource selection. Differences existed between habitat and road features within, and those outside, the multi-year composite female grizzly bear home range. Using logistic regression, large resource selection probability functions were obtained for the subalpine zone within multiple-use lands having no roads. Selection probability was zero for private lands and declined as total road density increased. Within seasonal ranges, most grizzly bears favoured low temperate and temperate elevation zones over the subalpine zone during all seasons. Relative to forested habitats, avalanche chutes were positively selected for during all seasons, but especially in spring. Shrub lands and cutting units were important to most bears during summer and autumn. Grizzly bears were more closely associated with higher total road densities during spring than during other seasons. When in low temperate habitats, most bears used habitats with lower total road density than occurred randomly. Seasonal use by grizzly bears of areas within a 0.5 km buffer surrounding roads was evaluated. Most grizzly bears exhibited either neutral or positive selection for buffers surrounding closed roads and roads receiving 10 vehicles per day. Between 1988 and 1994, eight grizzly bears were killed by humans. These deaths were directly influenced by road access and unnatural food sources. These deaths, in addition to natural mortality, were too great to promote local population growth.

Prediction of neotropical tree and liana species richness from soil and climatic data

We present an analysis of local species richness in neotropical forests, based on a number of 0.1 ha samples of woody plants collected by the late Alwyn Gentry. For each of 69 forests, soils were analysed and climatic data were collated. Using transformed independent variables and interaction terms, multiple regression equations were developed that explained the greatest possible amount of variation in species richness, and the best equations were selected on the basis of regression diagnostics. The best models are presented for (a) all neotropical forests, (b) forests west of the Andes (transandean) and (c) east of the Andes (cisandean), and for various subsets based on elevation and annual rainfall. For the whole dataset, and for most subsets, annual rainfall and rainfall seasonality were the most important variables for explaining species richness. Soil variables were correlated with precipitation — drier forests have more nutrient-rich soils. After the inclusion of rainfall variables, available soil nutrient concentrations contributed little to explaining or accounting for additional variation in species numbers, indicating that tropical forest species richness is surprisingly independent of soil quality. The results are consistent with the hypothesis that plants in mature tropical forests may obtain nutrients through the process of direct cycling, in which mineral nutrients are extracted from litterfall before they enter the soil. The strong relationship between community species richness and rainfall patterns has implications for biodiversity conservation. Wet forests with an ample year-round moisture supply harbour the greatest number of woody plant species and should be a focus of conservation efforts.

A method for quantifying vertical forest structure

Vertical forest structure is an attribute of forests that is of interest to many disciplines and is consistently discussed in the context of ecosystem management. The vertical stratification of tree crowns is a forest attribute that influences both tree growth and understory community structure. Therefore, it should be considered when making management decisions that affect the structure of stands. However, current methods of quantifying vertical structure are either arbitrarily-defined and do not represent natural stratification patterns of stands or forests, or are too time consuming for landscape analyses. The program, TSTRAT, was developed to place trees into vertical strata in a structural classification of forest vegetation developed for the Inland Northwest (USA). The primary classification criteria were cover types and classes of stand development described by structural criteria. The TSTRAT algorithm defines strata on the basis of an assumption related to a competition cut-off point among tree crowns in a given area. The predicted strata assignments of trees closely approximated vertical strata that were visually identified, in addition to those identified through cluster analysis. TSTRAT assigns each tree to a stratum, produces various descriptive statistics by vertical stratum, and quantifies overstory tree species diversity and inequality of tree heights. Because TSTRAT simulates the natural vertical arrangement of tree crowns, it is potentially useful in identifying strata that are biologically-related to processes that determine natural vertical stratification patterns.