Impacts of climate change and meteo-solar parameters on photosynthetically active radiation prediction using hybrid machine learning with Physics-based models

Abstract

The photosynthetically active radiation (PAR) ratio of 0.45 has been applied in numerous ecological studies to convert global solar radiation (H) to PAR by ordinary multiplication. The ratio is essentially useful and convenient, particularly when in situ measurement of climate data sets are not readily available due to inadequate instrumentation network and experience required. Up till now, there is no general agreement on whether the ratio is 0.45 and how the solar fluxes are distributed in Nigeria or around the world. Numerous empirical ratios have been reported in literature. This study proposed five (5) machine learning models (bagging, boosting, artificial neural network (MLP), automatic regression integrated moving average (ARIMA), controlled ARIMA), and 5 Physics-based models (Physics-based Gumbel model (PGM), popularly applied ratio (PCR), observed PAR ratio (POR), empirical, and PGM-CARIMA hybrid models. The 10 models were used to predict PAR using generalized datasets in Nigeria and the PGM-CARIMA model emerged as the best performing fit based on the R2, RMSE, MAPE, and BIC error metric indicators. PGM-CARIMA reported a better estimate for predicting PAR compared to POR and PCR with a relative error ranging from 0.00002 to 0.0503% compared to POR and PCR which recorded 4.657% and 6.178%. This implies that the use of PGM-CARIMA is more suitable for predicting the PAR ratio as it introduces a lower marginal error of approximately 0.00002–0.0503% when applied for the calculation of net primary productivity (NNP) and of climate change with a significant reduction from 6.17% and 4.65% was achieved in using the ratio of PCR and POR, respectively. Using the proposed PGM-CARIMA evolutionary hybrid model, we demonstrated that the annual mean of the ratio yielded a range of 0.4377 to 0.4712 in Nigeria. The authors also proposed the Physics-based classical climate change model (PCM) to assess the impact of climate change on PAR productivity in Nigeria. The findings revealed that changes in PAR productivity as a result of the impact of climate change may accelerate food production and crop yields in Southern Nigeria and depreciate overall food production and crop yields in Northern Nigeria in 2030, 2040, and 2021–2040. The results of the climate analysis also revealed that north-central Nigeria often referred to as the nation's food basket is most negatively impacted by the impact of climate change externalities over the periods studied. The authors recommended mitigation and adaptation plans such as delayed planting dates, intercropping, routine planting of tropical forages, especially livestock that feed on plants, could be key to climate change mitigation, also, expansion of the planting area as well as the incorporation of improved seedlings that are highly resistant to extreme temperatures will accelerate food security across the country within the periods studied.