Comparison of a novel machine learning approach with dynamical downscaling for Australian precipitation

Abstract

Dynamical downscaling, and machine learning (ML) based techniques have been widely applied to downscale global climate models and reanalyses to a finer spatiotemporal scale, but the relative performance of these two methods remains unclear. We implement an ML regression approach using a multi-layer perceptron (MLP) with a novel loss function to downscale coarse-resolution precipitation from the Bureau of Meteorology Atmospheric high-resolution Regional Reanalysis for Australia from grids of 12-48 km to 5 km, using the Australia Gridded Climate Data observations as the target. A separate MLP is developed for each coarse grid to predict the fine grid values within it, by combining coarse-scale time-varying meteorological variables with fine-scale static surface properties as predictors. The resulting predictions (on out-of-sample test periods) are more accurate than dynamical downscaling in capturing the rainfall climatology, as well as the frequency distribution and spatiotemporal variability of daily precipitation, reducing biases in daily extremes by 15-85% with 12-km prediction fields. When prediction fields are coarsened, the skill of the MLP decreases—at 24 km relative bias increases by ~ 10%, and at 48 km it increases by another ~ 4%—but skill remains comparable to or, for some metrics, much better than dynamical downscaling. These results show that ML-based downscaling benefits from higher-resolution driving data but can still improve on dynamical downscaling (and at far less computational cost) when downscaling from a global climate model grid of ~50 km.

A Robust Generative Adversarial Network Approach for Climate Downscaling and Weather Generation

“…Recently, computationally efficient statistical/empirical algorithms have been explored for RCM emulation, including simple multiple linear regression (Holden et al, 2015), multilayer perceptron (Chadwick et al, 2011;Hobeichi et al, 2023;Nishant et al, 2023), statistical analogues (Boé et al, 2023), and normalizing flows (Groenke et al, 2020). In both RCM emulation and other downscaling applications, there has been a shift towards regression-based deep learning computer vision algorithms such as CNNs (Babaousmail et al, 2021;Bano-Medina et al, 2023;Doury et al, 2022;van der Meer et al, 2023).…”

Residual Diffusion Modeling for Km-scale Atmospheric Downscaling

“…Since these predictions are augmented by data assimilation from ground-based precipitation radar and other sensors, good estimates of regional km-scale atmospheric states exists [13]. Such dynamical downscaling is computationally expensive, which limits the number of ensemble members used to quantify uncertainties [38]. A common inexpensive alternative is to learn a statistical downscaling from these dynamical downscaling simulations and observations [60].…”

“…In this context, ML downscaling enters as an advanced (non linear) form of statistical downscaling with potential to emulate the fidelity of dynamical downscaling. Several ML methods have been previously used for downscaling [9,50,18,57,38,59,2]. Convolutional Neural Networks, which reduce the input dimensions, have shown promise in globally downscaling climate (100km) data to weather scales (25km) [35,47,3,45].…”