Predicting stream temperatures: geostatistical model comparison using alternative distance metrics

Abstract

The Beaverkill Watershed in the Catskill Mountains of New York, U.S.A., shows evidence of high summer stream temperatures throughout critical trout habitat. Because habitat quality, as characterized by stream temperature, dramatically influences trout communities, it is important for biologists to identify and map these characteristics and to monitor how they change over time. Stream temperatures were recorded over time throughout the Beaverkill Watershed and were used to identify thermal refugia and areas of thermal stress. Seventy-two temperature loggers were placed throughout the watershed during the summer of 2000. Three geostatistical metrics for predicting temperature across the watershed were constructed and evaluated. The first metric utilized the shortest path between temperature loggers without using stream network information. A second metric used distances calculated along the stream network in both upstream and downstream directions. Our final metric was a modified network system in which the distances were weighted by stream order. Each metric was found to provide predictive capability, with added complexity improving the accuracy of represented stream temperatures.