TRU-NET: A Deep Learning Approach to High Resolution Prediction of Rainfall

Abstract

Climate models (CM) are used to evaluate the impact of climate change on the risk of floods and strong precipitation events. However, these numerical simulators have difficulties representing precipitation events accurately, mainly due to limited spatial resolution when simulating multi-scale dynamics in the atmosphere. To improve the prediction of high resolution precipitation we apply a Deep Learning (DL) approach using an input of CM simulations of the model fields (weather variables) that are more predictable than local precipitation. To this end, we present TRU-NET (Temporal Recurrent U-Net), an encoder-decoder model featuring a novel 2D cross attention mechanism between contiguous convolutionalrecurrent layers to effectively model multi-scale spatio-temporal weather processes. We use a conditional-continuous loss function to capture the zero-skewed patterns of rainfall. Experiments show that our model consistently attains lower RMSE and MAE scores than a DL model prevalent in short term precipitation prediction and improves upon the rainfall predictions of a state-of-the-art dynamical weather model. Moreover, by evaluating the performance of our model under various, training and testing, data formulation strategies, we show that there is enough data for our deep learning approach to output robust, high-quality results across seasons and varying regions.

High-resolution Probabilistic Precipitation Prediction for use in Climate Simulations

“…Further improvement in prediction can be achieved via imposing non-linear dependence of the parameters of compound Poisson distribution on the input variables, compared to the linear dependence considered here. A larger stencil of grid points for model input could also provide improvements to our models, observed in a second study of the same dataset that was based on the use of deep neural networks and was able to achieve even better results for deterministic predictions when taking also input vectors for neighbourhood locations into account (Adewoyin et al, 2020). However, the deep learning study could not provide reasonable representations of forecast uncertainty.…”