Submitted to: Agricultural Water Management
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/12/2009
Publication Date: 6/11/2009
Citation: Hunsaker, D.J., El-Shikha, D.M., Clarke, T.R., French, A.N., Thorp, K.R. 2009. Using ESAP Software for Predicing the Spatial Distributions of NDVI and Transpiration of Cotton. Agricultural Water Management. 96(9):1293-1304 Interpretive Summary: Remote sensing provides a way to view crop vegetation differences within an entire field. Such information, called a remote sensing vegetation index, is an important tool used for managing irrigation water to fields more efficiently. Remote sensing from satellites and aircraft can provide the vegetation index information. However, due to the number of times the vegetation index data is needed during the season, frequent aerial acquisitions would not be cost-effective. In this study, a method was evaluated which used vegetation index data collected from a few sensors placed within a cotton field to predict the vegetation index at all locations within the field. A statistical software program was used to determine the best locations for the field sensors. The results indicate that using a few sensors in the field with a few well-timed aircraft collections of vegetation index could provide an effective method for guiding cotton irrigation water scheduling and management. This information will be of interest to farmers, irrigation consulting firms, and water resource agencies.
Technical Abstract: The normalized difference vegetation index (NDVI) has many applications in agricultural management, including monitoring real-time crop coefficients for estimating crop evapotranspiration (ET). However, frequent monitoring of NDVI as needed in such applications is generally not feasible from aerial or satellite platforms due to both cost of image acquisition and the need for clear sky conditions. An alternative monitoring approach would be to combine continuous NDVI data obtained from fixed sensors, deployed at strategic field locations, and occasional high-resolution NDVI spatial imagery. Determining the optimal locations for the fixed sensors is, therefore, a key aspect for the approach. In this study, we used ESAP (ECe Sampling, Assessment, and Prediction) software to select a small set of NDVI sampling locations within an irrigated cotton field. A full set of field NDVI values (>17,000 pixels) acquired from aerial imagery on six days during two cotton growing seasons were applied in ESAP to select sample designs having 6, 12, and 20 locations. The primary objective was to evaluate ESAP predictions of the spatial distribution of NDVI based on NDVI calibration data obtained at the sampling locations selected by ESAP. A second objective was to evaluate NDVI predictions as influenced by the number of sampling locations. A third objective was to assess the effectiveness for using ESAP-predicted rather than observed NDVI from imagery for estimating ET. The ESAP-predicted NDVI means for each of the six dates were statistically equal to observed. However, the smaller than observed standard deviations corresponded to less effective predictions at the lower and upper extremes of the observed NDVI range. Regressions of predicted versus observed NDVI resulted in r2 values from 0.48 to 0.75 over the six dates. Calibration models and prediction results were similar for all three sample designs indicating little benefit for using more than 6 sampling locations. Higher r2 and lower root mean square errors occurred during periods having nearly full cotton cover than for periods of sparse crop cover. This suggests that when the field contains a large portion of exposed bare soil the spatial predictions of NDVI may be somewhat less accurate than when higher vegetative conditions exist. The errors associated with using crop coefficients based on ESAP-predicted NDVI rather than observed NDVI for calculating cumulative ET varied from 3 to 10%, where the higher ET errors occurred during early season periods. Generally, ET errors on the order of less than 10% would be acceptable for most irrigation management applications. Using ESAP procedures in conjunction with a few well-timed imagery acquisitions of NDVI could provide an effective method to estimate crop water needs for guiding cotton irrigation water scheduling and management.