Applying different inversion techniques to retrieve stand variables of summer barley with PROSPECT + SAIL |
| |
Authors: | M Vohland S Mader W Dorigo |
| |
Institution: | 1. Remote Sensing and Geoinformatics, Faculty of Geography and Geosciences, University of Trier, Campus II, 54286 Trier, Germany;2. Remote Sensing Department, Faculty of Geography and Geosciences, University of Trier, Campus II, 54286 Trier, Germany;3. Institute of Photogrammetry and Remote Sensing, Vienna University of Technology, Gusshausstraße 27-29, 1040 Vienna, Austria |
| |
Abstract: | This study describes the retrieval of state variables (LAI, canopy chlorophyll, water and dry matter contents) for summer barley from airborne HyMap data by means of a canopy reflectance model (PROSPECT + SAIL). Three different inversion techniques were applied to explore the impact of the employed method on estimation accuracies: numerical optimization (downhill simplex method), a look-up table (LUT) and an artificial neural network (ANN) approach. By numerical optimization (Num Opt), reliable estimates were obtained for LAI and canopy chlorophyll contents (LAI × Cab) with r2 of 0.85 and 0.94 and RDP values of 1.81 and 2.65, respectively. Accuracies dropped for canopy water (LAI × Cw) and dry matter contents (LAI × Cm). Nevertheless, the range of leaf water contents (Cw) was very narrow in the studied plant material. Prediction accuracies generally decreased in the order Num Opt > LUT > ANN. This decrease in accuracy mainly resulted from an increase in offset in the obtained values, as the retrievals from the different approaches were highly correlated. The same decreasing order in accuracy was found for the difference between the measured spectra and those reconstructed from the retrieved variable values. The parallel application of the different inversion techniques to one collective data set was helpful to identify modelling uncertainties, as shortcomings of the retrieval algorithms themselves could be separated from uncertainties in model structure and parameterisation schemes. |
| |
Keywords: | Canopy reflectance modelling PROSPECT SAIL Numerical optimization Artificial neural network Look-up table inversion |
本文献已被 ScienceDirect 等数据库收录! |
|