Abstract
Abstract. This study develops a neural-network-based approach for emulating high-resolution modeled precipitation data with comparable statistical properties but at greatly reduced computational cost. The key idea is to use combination of low- and high-resolution simulations (that differ not only in spatial resolution but also in geospatial patterns) to train a neural network to map from the former to the latter. Specifically, we define two types of CNNs, one that stacks variables directly and one that encodes each variable before stacking, and we train each CNN type both with a conventional loss function, such as mean square error (MSE), and with a conditional generative adversarial network (CGAN), for a total of four CNN variants. We compare the four new CNN-derived high-resolution precipitation results with precipitation generated from original high-resolution simulations, a bilinear interpolater and the state-of-the-art CNN-based super-resolution (SR) technique. Results show that the SR technique produces results similar to those of the bilinear interpolator with smoother spatial and temporal distributions and smaller data variabilities and extremes than the original high-resolution simulations. While the new CNNs trained by MSE generate better results over some regions than the interpolator and SR technique do, their predictions are still biased from the original high-resolution simulations. The CNNs trained by CGAN generate more realistic and physically reasonable results, better capturing not only data variability in time and space but also extremes such as intense and long-lasting storms. The new proposed CNN-based downscaling approach can downscale precipitation from 50 to 12 km in 14 min for 30 years once the network is trained (training takes 4 h using 1 GPU), while the conventional dynamical downscaling would take 1 month using 600 CPU cores to generate simulations at the resolution of 12 km over the contiguous United States.
Highlights
Earth system models (ESMs) integrate the interactions of atmospheric, land, ocean, ice, and biosphere and generate principal data products used across many disciplines to characterize the likely impacts and uncertainties of climate change (Heavens et al, 2013; Stouffer et al, 2017)
Informed by the physics of precipitation, we include, in addition to the low-resolution precipitation and high-resolution topography data used by Vandal et al (2017) for their SR model, the vertically integrated water vapor (IWV) or precipitable water, sea level pressure (SLP), and 2 m air temperature (T2) as inputs (Table 1) since we find these variables show high pattern correlations with precipitation along the time dimension
We evaluate the efficacy of the five convolutional neural network (CNN) methods by comparing their predictions with the original WRF output at a grid spacing of 12 km (Ground Truth), the output of a 12 km bilinear interpolation from the 50 km data (Interpolator), and the state-of-the-art SR-conditional generative adversarial network (CGAN) model output
Summary
Earth system models (ESMs) integrate the interactions of atmospheric, land, ocean, ice, and biosphere and generate principal data products used across many disciplines to characterize the likely impacts and uncertainties of climate change (Heavens et al, 2013; Stouffer et al, 2017). DNN-based SR (Dong et al, 2014; Yang et al, 2014) describes various algorithms that take one or more low-resolution images and generate an estimate of a high-resolution image of the same target (Tian and Ma, 2011), a concept closely related to downscaling in climate modeling They employ a form of DNN called a convolutional neural network (CNN; LeCun et al, 1998) in which node connections are configured to focus on correlations within neighboring patches. Since this is an interdisciplinary study, different terms are used in different contexts but refer to the same meanings
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
