Suitability of YieldTracker for Predicting Corn Yields in Western Kansas.

YieldTracker is a mathematical model that simulates growth and yield of graminoid crops using weather and plant canopy observations (leaf area index, LAI). LAI can be predicted from surrogates derived from remotely sensed images or hyperspectra. We tested YieldTracker for accuracy using corn data from western Kansas. Projecting in-season land productivity has potential for optimizing irrigation in the Ogallala Aquifer region. Three years of yield data for four replications of three treatments--rainfed and subsurface drip irrigation (SDI) at 3.8 mm d^-1 and 7.6 mm d^-1--(36 model runs) were compared to simulated yields. Results indicated YieldTracker has promise as a decision aide for managing irrigated corn, but has insufficient mechanistic complexity to simulate yields of water stressed corn. YieldTracker projected canopy development well, but LAI does not necessarily correlate with canopy efficiency in capturing solar radiation and converting it to biomass and then partitioning biomass to grain under conditions of limiting soil water. Remotely sensed normalized difference vegetation index (NDVI) is frequently used as a predictor for LAI, but for intensively managed corn, where LAI values approaching 6 are common, traditional formulations of NDVI are problematic as they tend to saturate at LAI>3. Hyperspectral reflectance data were used to find wavelength pairs that more nearly linearized the response of NDVI to LAI.