Crop fraction estimation from casi hyperspectral data using linear spectral unmixing and vegetation indices

Abstract

It is important to estimate vegetation fraction for forecasting regional weather and in precision agriculture for assessing crop performance during emergence and early growth phases. In this study, two approaches, linear spectral unmixing and vegetation indices, were reviewed and evaluated for the estimation of crop fraction from hyperspectral data. Compact Airborne Spectrographic Imager (casi) hyperspectral data were acquired three times in the 2001 growing season over four agricultural fields to monitor crop growth conditions and develop procedures for delineating major subunits for crop management. Crops planted in these fields included corn, soybean, and wheat. End-member spectra were extracted from casi data and used for linear spectral unmixing. Various vegetation indices, including the normalized difference vegetation index (NDVI), soil-adjusted vegetation index (SAVI), optimized soil-adjusted vegetation index (OSAVI), modified soil-adjusted vegetation index (MSAVI), transformed soil-adjusted vegetation index (TSAVI), and recently developed modified triangular vegetation index (MTVI2) and VI700 and VIgreen indices, were evaluated with casi data and with simulated spectra using coupled PROSPECT and SAILH models. All these indices were highly correlated with measured crop fractions. A comparison study based on simulated spectra showed that MTVI2 maintained adequate sensitivity up to a higher crop coverage. A high coefficient of determination (R2 = 0.90) and a low root mean square error (RMSE = 0.10) were obtained between measured and estimated crop fraction using MTVI2. The crop fraction derived from linear spectral unmixing was also highly correlated with the measured crop fraction (R2 = 0.94; RMSE = 0.08). However, determining end-member spectra in the linear spectral unmixing method remains a challenge. Using vegetation indices is a convenient method for crop fraction estimation with satisfactory accuracy.