## Selected Habitat Suitability-Index Model Evaluations > [!Cite]- > Terrell, James W, and Jeanette Carpenter. “Selected Habitat Suitability-Index Model Evaluations,” 1997. > [online](http://zotero.org/users/local/kycSZ2wR/items/758YWG6C) [local](zotero://select/library/items/758YWG6C) [pdf](file://C:\Users\erikt\Zotero\storage\EGUXAZLY\Terrell%20and%20Carpenter%20-%20Selected%20Habitat%20Suitability-Index%20Model%20Evaluatio.pdf) ## Notes %% begin notes %% %% end notes %% %% begin annotations %% ## Annotations | evernote 2016.04.25 The single most important implication of the model evaluation program is that if an HSI model does not provide an explicit description of expected wildlife response to changes in HSI, it is difficult to determine what the model output represents. This ambiguity forces model users (and model testers) to provide their own, possibly different, definitions of what the model output means and may be an underlying cause for the lack of HSI model test results in the refereed literature reported by Brooks (1997). Although specific problem areas and solutions varied for individual model evaluations, two basic approaches to improving the efficiency of future cycles of habitat model development and testing emerge from the model evaluation program: (1) defining specifically the time, spatial scale, and range of variation of the response(s) represented by the model; and (2) using statistical metrics that incorporate the concept of limiting factors to develop models and to compare model predictions to wildlife responses. However, the meaning of individual SI graphs would be less ambiguous if a second Y-axis were added to the graph identifying exactly what the SI represents. Examples of this approach can be found in Layher and Brunson (1992) where mean standing crop of fish (kg/ ha) associated with an SI is identified on a second Y-axis for individual SI graphs and documentation is provided on how to collect data to develop graphs. Well-defined SI’s will not solve the problem of choosing the best response (as discussed by van Horne 1983 and Hobbs and Hanley 1990) for rating “habitat quality” or the problem of what to do when the temporal and spatial scales of a response (e.g., microhabitat selection by individual animals during a short time period) do not match those of the perturbation (e.g., a large scale habitat alteration that will last for years). Well-defined SI’s will bring these problems into the open, where they can be recognized and the most appropriate response(s) selected (e.g., Minns et al. 1990). Clear identification of the responses represented by individual SI’s should make them more useful in the type of hierarchical analyses suggested by Rabeni and Sowa (1996) who noted that effective habitat conservation requires recognizing the relative influence of each habitat variable and the spatial scale over which each operates. A habitat variable (or combination of habitat variables) can be a limiting factor without being strongly correlated with a species’ (or group of species or species life stage) response. Thomson et al. (1996) describe in detail this fundamental problem in the interpretation of ecological data and note that correlation analysis may be shortsighted, or even blind to informative aspects of ecological data sets where data points are widely scattered beneath a ceiling imposed by a limiting factor. Heteroscedastic variance patterns should be evaluated as evidence supporting the occurrence of limiting factors. Statistical techniques that define the upper limits and internal structure of data should be used to develop and test explicit models of habitat characteristics that act as limiting factors. Models that describe limiting factors or other patterns of ecological associations are only a first step to understanding ecological processes. The problem faced by managers - predicting the impact of active manipulation of habitat - is better solved by an understanding of ecological processes than reliance on the repeatability of observed patterns of association. %% end annotations %% %% Import Date: 2024-08-15T10:29:53.278-06:00 %%