A CART Based Sub-pixel Method to Map Spatial and Temporal Patterns of Prairie Pothole Lakes with Climatic Variability

The Prairie Pothole Region in the United States contains millions of seasonal, semi-permanent, or permanent lakes and wetlands that typically range in size from 0.1 to 10 hectares. These lakes and wetlands are vulnerable to climate change, especially in our study area in South Dakota, in which a period of deluge following a sharp drought considerably expanded the areal extent of prairie pothole lakes during the last decade of the twentieth century. Preliminary estimates of lake areas, determined using Landsat 5 and 7 images, had appreciable errors especially for the smallest of these lakes. This paper describes a new sub-pixel approach integrated with a CART (Classification and Regression Tree) model using GIS (Geographical Information System) to quantify mixed water pixels along lake boundaries to improve area estimations for pothole lakes. Errors in estimated area were typically 10 percent or less for lakes greater than 1 hectare in size. An analysis of lakes in our study area demonstrates how lake area changed with the transition from drought to deluge. Small lakes exhibited a distinct seasonal variation in contrast to large lakes that tended to follow precipitation trends more broadly. The total water area of lakes is consistent with broad variation in rainfall.