Needle-leaf chlorophyll content (Cab) of a Norway spruce stand was estimated from CHRIS-PROBA images using the canopy reflectance simulated by the PROSPECT model coupled with two canopy reflectance models: 1) discrete anisotropic radiative transfer model (DART); and 2) PARAS. The DART model uses a detailed description of the forest scene, whereas PARAS is based on the photon recollision probability theory and uses a simplified forest structural description. Subsequently, statistically significant empirical functions between the optical indices ANCB670-720 and ANMB670-720 and the needle-leaf Cab content were established and then applied to CHRIS-PROBA data. The Cab estimating regressions using ANMB670_720 were more robust than using ANCB670-720 since the latter was more sensitive to LAI, especially in case of PARAS. Comparison between Cab estimates showed strong linear correlations between PARAS and DART retrievals, with a nearly perfect one-to-one fit when using ANMB670-720 (slope = 1.1, offset = 11 µg · cm-2). Further comparison with Cab estimated from an AISA Eagle image of the same stand showed better results for PARAS (RMSE = 2.7 µg · cm-2 for ANCB670-720; RMSE = 9.5 µg · cm-2 for ANMB670_720) than for DART (RMSE = 7.5 µg · cm-2 for ANCB670-720; RMSE = 23 µg · cm-2 for ANMB670-720). Although these results show the potential for simpler models like PARAS in estimating needle-leaf Cab from satellite imaging spectroscopy data, further analyses regarding parameterization of radiative transfer models are recommended.
|Journal||IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing|
|Publication status||Published - 2015|
- photon recollision probability
- area index
- spectral invariants