Retrieval of spruce leaf chlorophyll content from airborne image data using continuum removal and radiative transfer

We investigate combined continuum removal and radiative transfer (RT) modeling to retrieve leaf chlorophyll a & b content (Cab) from the AISA Eagle airborne imaging spectrometer data of sub-meter (0.4 m) spatial resolution. Based on coupled PROSPECT-DART RT simulations of a Norway spruce (Picea abies (L.) Karst.) stand, we propose a new Cab sensitive index located between 650 and 720 nm and termed ANCB650–720. The performance of ANCB650–720 was validated against ground-measured Cab of ten spruce crowns and compared with Cab estimated by a conventional artificial neural network (ANN) trained with continuum removed RT simulations and also by three previously published chlorophyll optical indices: normalized difference between reflectance at 925 and 710 nm (ND925&710), simple reflectance ratio between 750 and 710 nm (SR750/710) and the ratio of TCARI/OSAVI indices. Although all retrieval methods produced visually comparable Cab spatial patterns, the ground validation revealed that the ANCB650–720 and ANN retrievals are more accurate than the other three chlorophyll indices (R2 = 0.72 for both methods). ANCB650–720 estimated Cab with an RMSE = 2.27 μg cm− 2 (relative RRMSE = 4.35%) and ANN with an RMSE = 2.18 μg cm− 2 (RRMSE = 4.18%), while SR750/710 with an RMSE = 4.16 μg cm− 2 (RRMSE = 7.97%), ND925&710 with an RMSE = 9.07 μg cm− 2 (RRMSE = 17.38%) and TCARI/OSAVI with an RMSE = 12.30 μg cm− 2 (RRMSE = 23.56%). Also the systematic RMSES was lower than the unsystematic one only for the ANCB650–720 and ANN retrievals. Our results indicate that the newly proposed index can provide the same accuracy as ANN except for Cab values below 30 μg cm− 2, which are slightly overestimated (RMSE = 2.42 μg cm− 2). The computationally efficient ANCB650–720 retrieval provides accurate high spatial resolution airborne Cab maps, considerable as a suitable reference data for validating satellite-based Cab products.