Evaluation of the AnnAGNPS model for prediction of runoff and sediment yields in St Lucia watersheds

The Annualised Agricultural Non-Point Source Pollution Model (AnnAGNPS) was used to predict runoff and sediment losses from a forested and an agricultural watershed of St. Lucia Island in the Caribbean. Digital elevation models (DEM) of the agricultural and forested watersheds were generated from digitised topographic data. Based on the critical source area (CSA) and minimum source channel length (MSCL) specifications, the agricultural watershed was discretised into eight cells and three channel reaches, and the forest watershed into 12 cells and five channel reaches. The weighted curve numbers (CNs) were used for the cells with multiple land uses. The CN was observed to be the most sensitive parameter in the prediction of runoff and was adjusted during calibrations. The calibration runs of the AnnAGNPS resulted in errors between observed and simulated values of 0-33% for the agricultural watershed and 3.3-46.2% for the forested watershed for runoff prediction. The sediment yield prediction error percentage during the calibration run varied from 18.2% to 40.5% for the agricultural watershed and 9.1% to 50% for the forest watershed. However, validation of the calibrated model for different rainfall events resulted in errors of 6.7-36% and a value for the coefficient of prediction (C-p'A) of 0.028 (agricultural watershed) and 3.4% to 36% with a value for the C-p'A of 0.23 (forested watersheds) for runoff prediction. Validation of sediment loss for both the watersheds resulted in higher errors (23% to S5%) and C-p'A value of 0.341 for both the watersheds. Also, the model prediction of average annual runoff and sediment loss revealed that the agricultural watershed sediment loss (73.3 t ha(-1) year') was significantly higher than for the forest watershed (7.2 t ha(-1) year(-1)). Further, the validated model was used to simulate the runoff and sediment losses under a recommended land management regime for the agricultural watershed, which resulted in an 18.5% reduction in runoff and a 63% reduction in sediment loss as compared to the current management practice. This study revealed that the AnnAGNPS can be successfully applied for assessment of runoff and sediment losses and subsequent land use planning to conserve the natural resources in the watersheds of St. Lucia. (c) 2007 IAgrE. All rights reserved. Published by Elsevier Ltd