Predictive models for photosynthetic active radiation irradiance in temperate climates

This research evaluated 10 different empirical models designed for predicting Photosynthetically Active Radiation (PAR) at higher latitudes, addressing atmospheric conditions specific to these regions. The research introduces the Musleh-Rahman (MR) model, which substitutes Diffuse Horziontal Irradiance (DHI) with Clear Direct Normal Irradiance (DNI), Ozone and Aerosol Optical Depth at 550 nm (AOD550) sourced for satellite reanalysis data, achieving a Mean Bias Deviation (MBD) of 0.19 % and Root Mean Square Error (RMSE) of 12.42 W/m2. Furthermore, when applied to six untested locations, results demonstrate that the MR model outperformed the best performing empirical model with an MBD improvement of 3.68 % and an RMSE of 4.28 W/m2, whereas, when compared to machine learning models, the Light Gradient Boost Model (LGBM), had an MBD of −3.85 %. The MR model also maintained consistency across seasonal and density evaluations, attaining an R2 value as high as 0.9709, thereby highlighting the significant benefits of integrating satellite-sourced atmospheric data into PAR prediction models. Moreover, the research illustrated that substituting DHI with Clear DNI, Ozone, and AOD550 not only reduces MBD and boosts R2 values but also amplifies the model's applicability and accuracy in capturing early PAR peaks and reducing overestimations through precise adjustments in Ozone and AOD550 levels. This highlights the benefits of incorporating satellite-derived atmospheric data into PAR predictions models. •Analysis of different parameters to Photosynthetically Active Radiation (PAR).•New model that uses satellite data to predict PAR with updated input parameters.•Evaluating seasonal effects on empirical and ML models in temperate climates.