Assessment of five different methods for the estimation of surface photosynthetically active radiation from satellite imagery at three sites – application to the monitoring of indoor soft fruit crops in southern UK

摘要

This paper assesses several methods for the retrieval of Photosynthetically Active Radiation?(PAR) from satellite imagery. The results of five different methods are compared to coincident in-situ measurements collected at three sites in southern?UK. PAR retrieval methods are separated into two distinct groups. The first group comprises three methods that compute PAR by multiplying the satellite-retrieved solar broadband irradiance at the surface?(SSI) by a constant coefficient. The two methods in the second group are based on more sophisticated modelling of the radiative transfer in the atmosphere involving advanced global aerosol property analyses and physically consistent total column water vapour and ozone produced by the Copernicus Atmosphere Monitoring Service?(CAMS). Both methods compute a cloud modification factor from satellite-retrieved SSI. The five methods have been applied to two satellite-retrieved SSI datasets: HelioClim-3 version?5?(HC3v5) and CAMS Radiation Service?(CAMS-Rad). Except at the seashore site, Group?2 methods combined with the cloud extinction from the HC3v5 dataset deliver the best results with small biases of ?5?to 0 μmol m?2 s?1 (?1 % to 0 % relative to the mean of the measurements), root mean square errors of 130 μmol m?2 s?1 (28 %) and correlation coefficients exceeding?0.945. For all methods, best results are attained with the HC3v5 data set. These results demonstrate that all methods capture the temporal and spatial variability of the PAR irradiation field well, although several methods require a posteriori bias adjustments for reliable results. Combined with such an adjustment, the Udo et Aro method is a good compromise for this geographical area in terms of reliability, tractability and its ability to run in real-time. Overall, the method performing a spectral discretization in cloud-free conditions, combined with the HC3v5 dataset, outperforms other methods and has great potential for supporting an operational system.