GROUND-TRUTHING A PREDICTIVE MODEL OF SEASONAL PONDS ON THE CHIPPEWA NATIONAL FOREST, MINNESOTA

Seasonal ponds form a key hydrologic and ecologic component of National Forests within post-glacial landscapes in the upper Midwest. While forest guidelines outline specific management practices for seasonal ponds, a comprehensive inventory of these ephemeral hydrologic features is not always feasible. However, seasonal ponds form in subtle topographic depressions that can be remotely detected within recent high-resolution LIDAR datasets. Using these data, the Chippewa National Forest employed a GIS model to predict the occurrence of seasonal ponds within targeted regions in north-central Minnesota. We present results from field-based ground-truthing of this model, completed during a GeoCorps America term of service. Surveys indicate the model achieved an accuracy rate of 80-85% (error of commission = 15-20%) with an approximate error of omission of 5%. False positives are correlated to ecosystem type, suggesting further refinement of the model may help improve accuracy. Results indicate that LIDAR and GIS-based detection of seasonal ponds on the Chippewa National Forest is a viable, efficient method to remotely characterize the location, size, and density of these features. The model will be particularly valuable in prescribing management actions during timber sales and pinpointing regions of biologic interest. More broadly, these findings demonstrate the utility of remote sensing techniques to inform the management of hydrologic resources on public lands.