Leaf carbon-based constituents of temperate forest species retrieved using PROSPECT-PRO

Abstract

The retrieval of leaf carbon-based constituents of vegetation species and their separation from the overall leaf mass per area using radiative transfer models was historically challenging, until the recent re-calibration of the PROSPECT-PRO model. Nevertheless, it remains unexplored for temperate tree species. This study evaluated the retrieval of carbon-based constituents of fresh leaf samples from four European temperate tree species using the PROSPECT-PRO model. We collected a comprehensive dataset of 249 fresh leaf samples obtained from the top canopy of temperate forest species in Germany and the Netherlands. Measurements of the carbon-based constituents were conducted in the lab using a novel customisation of the traditional method, and the spectral measurements of leaf samples were obtained with the ASD FieldSpec-3 and integrating sphere. Employing a look-up table approach, the PROSPECT-PRO model was inverted across the 800 to 2400 nm wavelength range, and the model was also applied with optimal bands selected for leaf carbon-based constituents. The retrieval of carbon-based constituents yielded reasonable accuracy for the four species: European beech (R² = 0.53, NRMSE = 0.36), English oak (R² = 0.45, NRMSE = 0.34), Scots pine (R² = 0.63, NRMSE = 0.36) and Norway spruce (R² = 0.62, NRMSE = 0.26). Slight improvements were observed in the retrieval accuracies with the identified optimal spectral bands. As such, the NRMSE values decreased by 0.03 and 0.05 for European beech and English oak; however, they slightly worsened for Norway spruce and Scots pine by 0.05 and 0.01, respectively. This study highlights the effect of considering individual constituents in laboratory measurements and during the calibration of absorption coefficients within the model. This could have a more substantial influence on the retrieval accuracy of carbon-based constituents than the influence of water interference or solely applying optimal spectral bands.