Using GIS to predict habitat in lakes: An example using nearshore substrate categories

Habitat is an integral component of lake ecosystems and threatened by anthropogenic alterations. Quantifying habitat is typically done with labor intensive and spatially limited surveys (i.e., transects) or with surveys requiring specialized field equipment combined with computer analyses (i.e., sonar). These approaches are limited to inventorying habitat condition and do not directly describe processes that influence habitat distributions. We developed a framework that utilizes geographic information systems (GIS) to answer habitat‐based questions and allows for an understanding of the factors that influence its distribution. This framework uses GIS‐derived data to describe factors input as predictive variables. We tested this framework by predicting nearshore substrate composition in Minnesota lakes. Substrate composition was measured during the summers of 2014–2016 along transects in 28 lakes across Minnesota. Composition was then grouped into three size categories (muck, sandy gravel, and coarse). For each transect, we obtained GIS‐derived data describing exposure (fetch), riparian height, maximum depth, wind power, and bathymetric aspect to use as predictor variables in a classification tree model. We randomly selected 15% of transects to validate the model. Using repeated sampling with replacement, we determined this model predicted substrate composition with 62.0–71.0% accuracy. We then used data from all study lakes and examples from Belle Lake, Minnesota, to determine sources of error. Our results demonstrate that a GIS framework can describe both the distribution of a habitat component and the factors structuring it. This framework can enhance communication effectiveness and decision‐making processes regarding habitat protection, management, and restoration.