Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Environmental Microbial & Food Safety Laboratory » Research » Publications at this Location » Publication #266189

Title: Information and complexity measures for hydrologic model evaluation

Author
item PAN, FENG - University Of Maryland
item WAGENER, THORSTEN - University Of Pennsylvania
item Pachepsky, Yakov
item HILL, ROBERT - University Of Maryland

Submitted to: BARC Poster Day
Publication Type: Abstract Only
Publication Acceptance Date: 3/15/2011
Publication Date: 4/27/2011
Citation: Pan, F., Wagener, T., Pachepsky, Y.A., Hill, R. 2011. Information and complexity measures for hydrologic model evaluation. BARC Poster Day. 39.

Interpretive Summary:

Technical Abstract: Hydrological models are commonly evaluated through the residual-based performance measures such as the root-mean square error or efficiency criteria. Such measures, however, do not evaluate the degree of similarity of patterns in simulated and measured time series. The objective of this study was to apply information content and complexity measures to evaluate and compare performance of hydrological models in simulations of streamflow time series. Five watersheds were selected in Texas, North Carolina, and West Virginia to represent a range of hydro-climatic conditions. Simulated daily streamflow was obtained from eight hydrological models having different structures. Information content measures were the metric entropy and mean information gain, and complexity measures were the fluctuation complexity and effective measure complexity. These measures were computed based on the binary encoding of 10-year daily time series of measured and simulated streamflow. Measured streamflow time series had smaller information content in arid watersheds than in humid watersheds, and were more complex. In arid watersheds, the information content and complexity measures of simulated time series from multi-buckets models were larger than from single-bucket models. The information content decreased with the increase of model complexity in humid watersheds. The best models for each watershed were selected based on the comparisons of information content and complexity measures of measured and simulated streamflow time series. The best models were different for different hydro-climatic conditions. Overall, more complex models were needed to simulate the daily streamflow in arid watersheds than in humid watersheds.