Accuracy and sensitivity analyses of SAIL model-predicted reflectance of maize

Abstract

Bidirectional reflectance of six maize genotypes was measured on 16 dates in 1984 at the Nebraska Sandhills Laboratory using a modular multiband radiometer. On 10 of the dates, measurements were repeated during the day to obtain a range of solar zenith angles. Plants were sampled approximately weekly to determine leaf area index (LAI) and smoothed to provide LAI values for each of the 560 reflectance measurements. Reflectance in the red and near-infrared (NIR) was predicted by three versions of the SAIL model, a one-dimensional version, a two-dimensional version that took rows into account, and a modified two-dimensional version that accounted for shading caused by rows and crop height, and then evaluated using an accuracy analysis that considered bias, regression, and random errors. The results indicated that the 1-D SAIL model underestimated red and overestimated NIR reflectance at partial canopies because it did not account for exposed bare soil. The 2-D version, which accounted for ground cover, generally overcame this problem. The 2-D′ version, which accounted for shading of soil and canopy due to increasing canopy height, did not improve the results of the 2-D version enough to warrant the increased complexity. Sensitivity analyses, in which the main inputs required to fit the data were varied, revealed that the same leaf angle distribution and the same single leaf reflectance and transmittance values could be used for the whole season. Diurnal, multitemporal reflectance combined with the SAIL model adequately described the radiative transfer properties of the maize crop canopies studied.