Point data for four case studies related to testing of multi-order hydrologic position

Publication Date

2019-12-16

Time Period

2019

Belitz, K., Arnold, T.L., and Sharpe, J.B., 2019, Point data for four case studies related to testing of multi-order hydrologic position: U.S. Geological Survey data release, https://doi.org/10.5066/P9LVCANT.

Type: Related Primary Publication

An important motivation for developing the concept of MOHP is the recognition that machine learning methods are useful for understanding and assessing hydrologic systems, and that there is a need for wall-to-wall data to support these efforts. Towards that end, four case studies were developed. Two of the case studies are based on categorical response variables (random forest classification): geomorphic provinces in California’s Central Valley and physiographic provinces in the U.S. Although the area of the Central Valley is less than 1% of the conterminous U.S., the performance of the random forest classification models is high for both (Belitz and others, 2019). The random forest classification models show that MOHP provides a set of metrics that are applicable at scales ranging from regional to national. Two of the case studies are based on depth to the water table, which is a continuous variable (random forest regression): simulated values in the Fox-Wolf Peshtigo area, which is located in the eastern part of Wisconsin, and observed values for the entire state. The two case studies are of similar scale when compared to the size of the U.S., but the data are dissimilar: simulated values that are uniformly distributed as compared to observed values that are clustered and presumably affected by a wider range of factors than are incorporated into the simulation model. The random forest regression models provide comparable results for both case studies, even though the data are of different type (Belitz and others, 2019).

• National Water-Quality Assessment Project
• USGS Data Release Products