Improving the inversion of PRI estimating light use efficiency using the BRDF model

Abstract: The Photochemical Reflectance Index (PRI) has great potential for estimating Light Use Efficiency (LUE). Mixed canopy structure under different lighting and observation geometries can affect the remote sensing observations of the PRI, leading to some errors in estimating vegetation photosynthesis. Multi-angle observation is an effective method for resolving differences in observation geometry. Fitting multi-angle observations to a Bidirectional Reflectance Distribution Function (BRDF) can normalize PRI to a common solar-observation target position, resulting in a significant increase in the correlation between canopy PRI and LUE, and has been well used in forest ecosystems. However, studies on the standardization of multi-angle PRI by BRDF models on crop canopies are still not well developed. To investigate the ability of the improved PRI by BRDF model to estimate the LUE of rice canopies, this paper used multi-angle rice canopy reflection data and contemporaneous flux data from the National Meteorological Station in Shouxian, Anhui, a semi-empirical, kernel-driven BRDF model was used to standardize the multi-angle PRI for each half-hourly from 09:00 to 15:00 during the observation period. This paper explored the BRDF model simulated PRI values and the model parameter characteristics under different physiological and non-physiological conditions. The results showed that the BRDF model fitting was poorly at low Clearness Index (CI) and became better as the CI increased, with the best performance at CI = 0.9 (R2 = 0.41). The parameters of the BRDF model are influenced by light conditions and vegetation status, where the isotropic weights (ki) correlated well with LUE for different CI ranges (linear regression coefficients R2 > 0.3 for all CI ranges), with the best correlation at 0.6 ≤ CI < 0.7 (R2 = 0.63); the daily variation of volume scattering weights (kv) showed a "U-shape" during the observation period, which was consistent with the daily variation of LUE. After correction by the BRDF model, the R2 of the exponential regression between PRI and LUE increased from 0.46 (P<0.05) to 0.8 (P<0.01). Compared with the PRI from simple multi-angle remote sensing observations, the angular correction by the BRDF model improved the ability of PRI estimating LUE; this study also demonstrated the feasibility that multi-angle canopy spectral observations can use the BRDF model to improve vegetation indices tracking plant physiological activity. This study has reference value for processing multi-angle vegetation remote sensing observation data and for remote sensing estimation of physiological activities related to photosynthesis in regional vegetation. The BRDF model can be practiced in multi-angle satellite observations to achieve the monitoring of photosynthetic capacity in a large area of farming area. Subsequent experiments in different farming areas will be carried out to calibrate the wider band spectral observations available from satellite data to achieve a comprehensive understanding of the correlation between LUE variation and spectral reflectance in lutein and fluorescence-related absorption characteristics, and to provide some reference for the calibration of parameters in agricultural areas in the global productivity model.