Comprehensive models for agrichemical transport necessarily include runoff predictions to partition rainfall between infiltration and runoff as this ability is fundamental to predictions of chemical runoff and leaching. We compared GLEAMS, Opus, andPRZM-2 model runoff predictions with runoff measured in a precisely controlled field site used for chemical runoff studies. In 1992 and 1993, two 14.5 m×42.9m corn (Zea mays, L.) field plots with 3 slope on Tifton loamy sand (fine-loamy, siliceous,thermic Plinthic Kandiudult) received six severe, artificial rainfall events over the growing season with each event consisting of a 25 mm h{sup}-1 rainfall for 2 h. Runoff was monitored continuously using a collector and flume. Model performance criteria included sensitivity analysis, graphical comparison and statistical analysis including mean, ratio of means, root mean square error (RMSE), and a paired difference t-test. Observed runoff averaged 20 of added rainfall. Lowest values occurred with freshly plowed soil or full canopy cove,; while 24 to 34 runoff occurred when nearly bare soils had crusted over. Using an initial moisture condition-II curve number (CN) of 85, GLEAMS and Opus predicted runoff within 10, overall, and produced a pattern of high and low runoff that closely followed observed. PRZM-2 overpredicted runoff by 90, overall, and predicted its highest runoff when observed runoff was lowest. Paired difference t-tests indicated a significant difference between measured and predictedrunoff for PRZM-2 (p<0.001 at a = 0.05), but none for GLEAMS (p = 0.761) or Opus (p = 0.194). Mean, ratio of means, and RMSE showed that GLEAMS and Opus performed better than PRZM-2. All three models were very sensitive to CN values which were empiricaland subjective, but less sensitive to measurable soil physical properties. With careful parameterization, GLEAMS and Opus could be used to simulate runoff from similar row-crop and soil conditions.
展开▼
