Field-scale Leaf Area Index estimation using IRS-1D LISS-III data

A study was carried out to estimate field-scale Leaf Area Index (LAI) using fine resolution polar orbiting IRS-1D LISS-III sensor data at 23 m spatial resolution. Three cloud-free scenes on 8 January, 2002 (D1), 30 January, 2002 (D2) and 15 January, 2003 (D3) over two sites in Gujarat, India, were acquired. The subscenes encompassing the study area were extracted and geo-registered. Surface reflectances in the red (0.62 - 0.68 mu m) and near-infrared (NIR) (0.77 - 0.86 mu m) bands were generated using 6S atmospheric correction code and coincident ground measurements on aerosol and water vapour. Normalized difference vegetation index (NDVI), simple ratio (SR) and soil-adjusted vegetation index (SAVI) were computed from the reflectances in the red and NIR. A total of 70 mean measured LAI datasets on wheat and tobacco were used for regression analysis and empirical models were developed between LAI and three vegetation indices ( VI). Both exponential and power models gave R-2 between 0.53 and 0.61 except for D2 ( R 2 between 0.04 and 0.11) when wheat was mostly at the postanthesis stage and the VI - LAI relation seems to be influenced by canopy geometry and angular distribution of leaves. The analysis indicated that with the soil type of the study sites being different, the SAVI-based model had a smaller rms. error (R-2 = 0.496 and rms. error = 0.685) in estimation of LAI when compared with the SR- (R-2 = 0.478, rms. error = 0.698) and NDVI- (R-2 = 0.491, rms. error = 0.689) based models. The LAIs for the study region were estimated by inversion of empirical models and validated against ground-measured data.