Submitted to: Transactions of the ASAE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/1/2005
Publication Date: 5/1/2005
Citation: Zhang, X.J. 2005. Generating correlative storm variables for CLINGEN using a distribution-free approach. Transactions of the ASAE. 48(2):567-575.
Interpretive Summary: Storm data such as storm amount, storm duration, and peak rainfall intensity are needed by many agricultural system models for simulating rainwater runoff and soil erosion. CLIGEN is the only weather generator that generates these data; however, CLIGEN-generated storm variables are not well correlated for a particular storm. The lack of correlation may limit the predictability of many agricultural system models. The objectives of this work were to (1) test a mathematical method for inducing desired correlation between generated storm amount and storm duration, and (2) evaluate the improvement of runoff and soil loss prediction resulting from the new method. Measured storm data as well as mathematically manipulated CLIGEN-generated storm data were used. The Water Erosion Prediction Project (WEPP) model was used to predict runoff and soil loss for the measured and generated storm conditions. Results showed that the method was simple to use and capable of inducing desired correlation between storm amount and storm duration. Runoff and soil loss prediction were generally improved when correlated CLIGEN storm data were used (following manipulation) rather than the original uncorrelated CLIGEN data. This work provides a useful means to modelers for improving the CLIGEN model and to professionals for improving runoff and soil loss prediction using agricultural system models such as WEPP.
Technical Abstract: CLIGEN is the only weather generator that generates internal storm patterns, which are required by many agricultural system models such as the Water Erosion Prediction Project (WEPP) model. The lack of correlation between CLIGEN-generated storm variables may limit those models' ability to predict surface runoff and soil erosion. The objectives of this study were to (1) test a distribution-free method for inducing desired rank correlation between generated storm variables, and (2) compare WEPP-predicted runoff and soil loss using measured vs. variously generated storm patterns on eight U.S. sites. Four climate files containing four storm patterns (measured, original uncorrelated CLIGEN output, correlated CLIGEN output, and correlated output with exponentially generated storm duration), along with measured soil, slope, and crop management on each site, were used to run WEPP. The distribution-free approach was simple to use and capable of inducing desired rank correlation between storm amount and duration and consequently between storm amount and relative peak intensity. Original CLIGEN output after inducing desired correlation considerably improved WEPP average annual runoff and soil loss predictions on most sites where strong correlation between storm amount and duration existed. The use of exponentially distributed storm duration instead of the original CLIGEN output only slightly improved WEPP runoff prediction but somewhat worsened soil loss prediction. Overall results indicate that for better runoff and soil loss prediction, correlated CLIGEN output should be used on sites where strong correlation between storm amount and duration exists.