Deep Learning for Daily Precipitation and Temperature Downscaling

Abstract

Understanding the impact of climate variability and change is of great importance for developing adaptation and mitigation strategies. Coarse resolution data sets (such as climate reanalysis, numerical weather predictions, satellite products, and simulations of general circulation models) are important for reconstructing historical climate and predicting the future. However, scale discrepancy limits the coarse resolution data sets from being directly used for impact assessments and decision making. One solution for bridging this gap is to downscale coarse resolution data to the local scale (

Generating Fine-Scale Aerosol Data through Downscaling with an Artificial Neural Network Enhanced with Transfer Learning

“…The downscaled results of ASDM, ASDM with transfer enhancement (ASDMTE), SRDRN, dissever framework with GAM and LM as regressors were compared on the same test sets with the above metrics. The structure of SRDRN can be found in Wang et al ( 2021) [26].…”

“…Leveraging temporal replicates, they fit separate linear regression models independently to each fine scale grid, ignoring spatial and temporal associations in either the fine-or coarse-scale data. Recently, deep learning approaches have been used that address spatial features, such as Wang et al (2021) [26], who developed a CNN-based method, Super Resolution Deep Residual Network (SRDRN), to downscale precipitation and temperature from coarse resolutions (25, 50 and 100 km) to fine resolution (4 km) by learning the between-scale image-to-image mapping function. However, they ignore the temporal associations between images.…”

“…The convolutional neural network (CNN) is a popular method for downscaling due to its ability to learn spatial features from large gridded data [25]. Recently, a CNN-based model called the Super Resolution Deep Residual Network (SRDRN), which utilized convolutional layers and residual networks, was developed to downscale daily precipitation and temperature [26]. Autoencoder-like models with residual connections and parameter sharing have also been used to downscale by incorporating an iterative training strategy to force spatial value consistency [27].…”

“…Autoencoder-like models with residual connections and parameter sharing have also been used to downscale by incorporating an iterative training strategy to force spatial value consistency [27]. Networks with transfer learning have been used in a spatial context to generalize the empirical associations within one region to apply downscaling in a different region, showing notable improvement compared to classical statistical downscaling methods [26,28].…”

An Improved Tandem Neural Network Architecture for Inverse Modeling of Multicomponent Reactive Transport in Porous Media

“…Related applications of DNN are available in various fields of hydrology and water resources, such as water storage measurements (Eltner et al., 2021; Irrgang et al., 2020; A. Y. Sun, Scanlon, et al., 2019), meteorological forecasting (Pudashine et al., 2020; F. Wang et al., 2021), flood predictions (Berkhahn et al., 2019; Kabir et al., 2020), and geological parameterizations (Laloy et al., 2018; Tang et al., 2021).…”