INVERSION OF PADDY LEAF AREA INDEX USING BEER-LAMBERT LAW AND HJ-1/2 CCD IMAGE

Monitoring crop leaf area index (LAI) timely and accurately by remote sensing is crucial to assess crop growth, manage field water-fertilizer and predict yield. The Huaihe River Basin was chose as study area to carry out field survey. By using decision tree classification and HJ-1/2 CCD image, the spatial distribution of paddy was identified. The extinction coefficient of paddy surface was confirmed with in-situ samples. The Beer-Lambert law was introduced to develop the inversion model of paddy LAI. The accuracy of inversion model was evaluated with in-situ samples, including coefficient of determination (R-2), RMSE and overall accuracy, while contrasting with the model of single-variable and multi-variables. Results showed that the inversion model based on Beer-Lambert law reached highest accuracy with the average R-2 of 0.684 and the average RMSE of 0.592. The average R-2 of multi-variables was 0.636, while the average RMSE was 0.661. The model of single-variable has lowest accuracy with average R-2 of 0.595 and average RMSE of 0.732. It indicated that the retrieval accuracy of LAI was improved with more variables inputted. The model based on Beer-Lambert law simulated the physical process of radiative transfer of paddy that differed from the two other models. The overall accuracy of Beer-Lambert law model exceeded 95 percent, while those of the two other models were 91.0 percent and 88.2 percent respectively. So the inversion model of paddy LAI based on Beer-Lambert law could eliminate the influence of water background and improve the accuracy of paddy LAI by remote sensing.