Submitted to: Agronomy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: January 5, 2010
Publication Date: March 26, 2010
Citation: Colaizzi, P.D., O'Shaughnessy, S.A., Gowda, P., Kustas, W.P., Anderson, M.C., Evett, S.R., Howell, T.A. 2010. Radiometer footprint model to estimate sunlit and shaded components for row crops. Agronomy Journal. 102(3):942-955. Interpretive Summary: Remote sensing is useful for crop management because the reflectance and surface temperature of crops can be measured by non-contact means over large areas. Reflectance and temperature are measured by devices called radiometers, and reflectance and temperature can be related to crop health, water and nutrient status, and other properties important for farm management and profitability. Interpretation of reflectance and temperature of vegetated surfaces; however, can be challenging because these are usually a mixture of sunlit vegetation, shaded vegetation, sunlit soil, and shaded soil that appear in the radiometer footprint. A mathematical model was developed to more accurately predict the relative proportion of each sunlit and shaded component that would appear to a radiometer. The model will improve the accuracy of existing remote sensing algorithms used for crop management. This is important because it will allow farm managers to more precisely match crop inputs to actual needs, and it will increase the impact of remote sensing on farm profitability.
Technical Abstract: This paper describes a geometric model for computing the relative proportion of sunlit vegetation, shaded vegetation, sunlit soil, and shaded soil appearing in a circular or elliptical radiometer footprint for row crops, where the crop rows were modeled as continuous ellipses. The model was validated using digital photographs of row crops, where each component was determined by supervised classification. Root mean squared errors (RMSE) between modeled and observed components were 35, 49, 29, and 44% of observed means for sunlit vegetation, shaded vegetation, sunlit soil, and shaded soil, respectively. Mean bias errors (MBE) were, respectively, -5.6, 16.6, -4.0, and -0.5% of observed means. The continuous ellipse model was compared to the commonly used clumping index model, where the latter does not account for radiometer footprint shape dimensions, and estimates total vegetation and total soil, but does not resolve these into their sunlit or shaded components. The continuous ellipse model resulted in RMSE for vegetation and soil of 22 and 19%, respectively, whereas the clumping index model resulted in respective RMSE of 37 and 31%. The continuous ellipse model had MBE of 3.3 and -2.6% for vegetation and soil, respectively, which was slightly greater than the respective MBE of -1.5% and 1.4% for clumping index model. Given the model sensitivity and uncertainty of LAI, the RMSE and MBE resulting from the continuous ellipse model would not be expected to be less than 20% of the observed means, and model performance was therefore deemed reasonable in this study.