Skip to main content
ARS Home » Pacific West Area » Tucson, Arizona » SWRC » Research » Publications at this Location » Publication #357540

Research Project: Understanding Water-Driven Ecohydrologic and Erosion Processes in the Semiarid Southwest to Improve Watershed Management

Location: Southwest Watershed Research Center

Title: Multi-parameter regression modeling for improving the quality of measured rainfall and runoff data in densely instrumented watersheds

item Meles, Menberu
item Goodrich, David - Dave
item Demaria, Eleonora
item Heilman, Philip - Phil
item Nichols, Mary
item LEVICK, L. - University Of Arizona
item Unkrich, Carl
item Kautz, Mark

Submitted to: Journal Hydrologic Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/7/2019
Publication Date: 7/26/2019
Citation: Meles, M.B., Goodrich, D.C., Demaria, E.M., Heilman, P., Nichols, M.H., Levick, L., Unkrich, C.L., Kautz, M.A. 2019. Multi-parameter regression modeling for improving the quality of measured rainfall and runoff data in densely instrumented watersheds. Journal Hydrologic Engineering. 24(10).

Interpretive Summary: Long-term precipitation records often span changes in measurement, data processing, or data recording methodologies that potentially affect data quality. Legacy precipitation and runoff data recorded in the long-term database for the Walnut Gulch Experimental Watershed showed some errors related to event occurrence time and magnitude. The errors were due to instrument malfunction and human errors during data processing and storage. Compared to the analog period, there were more erroneous numbers of precipitation and runoff events than during the digital period. To develop data quality control procedures and flag suspect runoff and rainfall observations, a multiple regression model was created using the causal relationship between runoff and rainfall properties, watershed properties, and the antecedent conditions in the watershed. The multi-regression model based quality assessment and control procedure revealed a total of 25 erroneous runoff and rainfall events for two sub-watersheds evaluated. This approach showed good potential for identifying and flagging suspect rainfall and runoff observations in densely instrumented watersheds.

Technical Abstract: The Walnut Gulch Experimental Watershed is a semi-arid experimental watershed and Long Term Agro-ecosystem Research (LTAR) site managed by the USDA-ARS Southwest Watershed Research Center for which high-resolution, long-term hydro-climatic data are available across its 149 km2 drainage area. We present the analysis of 50 years of hourly rainfall data to develop criteria for assessing data quality. A multiple regression model was developed to relate rainfall and watershed properties to runoff characteristics in 12 sub-watersheds ranging in area from 0.002 – 94 km2. The rainfall properties were event depth, maximum intensity, duration, location of the storm center with respect to the watershed outlet, and storm size. The watershed properties include contributing area, slope, shape, channel length, stream density, channel flow area, and percent area behind stock ponds for each of the nested watersheds. The interaction between rainfall and runoff was evaluated through antecedent moisture condition (AMC), antecedent runoff condition (ARC), as well as runoff depth and duration for each rainfall event. A regression model was developed based on 18 predictor variables which we evaluated using basic and categorical statistics to show its predictive skills. The regression model produced correlation coefficients ranging from 0.4-0.94, and Nash efficiency coefficients up to 0.76. The model predicted 92% of runoff events and 86% of no-runoff events across all the sub-watersheds considered in the study. The regression model was used to predict runoff given measured rainfall characteristics. The model is a complement to existing QAQC procedures and provides a specific method for ensuring that rainfall and runoff data in the Walnut Gulch database are consistent and contain minimal error. The model also has the potential for making runoff predictions in similar hydro-climatic environments where high-resolution ground-based radar-rainfall estimates are available.