Modelling of downwelling longwave radiation over multiple agro-climate settings of India under foggy sky conditions – A neural network approach

Estimation of LWin under overcast conditions like fog, has been one of the prime challenges of modelling the radiation budget. This is due to the lack of instrumentation and regular measurements at different space-time scales, especially in tropical and sub-tropical regions. In this study, two Multi-Layer Perceptron (MLP) based Artificial Neural Network (ANN) models were developed for LWin flux estimation under foggy sky during daytime and nightime. The models were developed using half-hourly flux measurements over different agro-climatic settings and several atmospheric parameters, from satellite observations and numerical model outputs. A comparative evaluation was made between Radiative Transfer (RT) computations and ANN-based models. The ANN models were found to have consistent performance across different sites except slightly less accuracy in sub-humid or humid climate. The ANN models showed overall RMSE of 4.4% and 4.6% with correlation coefficient of 0.9 and 0.86 for daytime and nightime, respectively with reference to in situ measurements.