Analyses for High‐Resolution Projections Through the End of the 21st Century for Precipitation Extremes Over the United States

Zobel, Zachary; Wang, Jiali; Wuebbles, Donald J.; Kotamarthi, V. R.

doi:10.1029/2018ef000956

Cited by 36 publications

(25 citation statements)

References 67 publications

(109 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This study focuses on precipitation, which is highly variable in time and space and is often the most difficult to compute in ESMs (Legates, 2014). Downscaling of ESMs with RCMs has generally reduced the bias in precipitation projections, often due to an increase in the model spatial resolution that allows for resolving critical terrain features such as changes in topography and coastlines (Wang et al, 2015;Zobel et al, 2018;Chang et al, 2020). In addition, precipitation data at high spatial resolution is needed for a variety of climate impact assessments, ranging from flooding risk to agriculture (Maraun et al, Figure 1.…”

Section: Methodsmentioning

confidence: 99%

Fast and accurate learned multiresolution dynamical downscaling for precipitation

Wang¹,

Liu²,

Foster³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

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 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 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 not as close as ground truth. 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 km to 12 km in 14 min for 30 years once the network is trained (training takes 4 hours using 1 GPU), while the conventional dynamical downscaling would take 1 months using 600 CPU cores to generate simulations at the resolution of 12 km over contiguous United States.

show abstract

Section: Methodsmentioning

confidence: 99%

Fast and accurate learned multiresolution dynamical downscaling for precipitation

Wang¹,

Liu²,

Foster³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Section: Methodsmentioning

confidence: 99%

Comment on gmd-2020-412

2021

View full text Add to dashboard Cite

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 highresolution simulations 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 not as close as 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 km to 12 km in 14 min for 30 years once the network is trained (training takes 4 hours 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 contiguous United States. IntroductionEarth 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). The computationally demanding nature of ESMs, however, limits their spatial resolution mostly to between 100 and 300 km. Such resolutions are not sufficient to resolve critical physical processes such as clouds, which play a key role in determining the Earth's climate by transporting heat and moisture, reflecting and absorbing radiation, and producing rain. Moreover, ESMs cannot assess stakeholder-relevant local impacts of significant changes in the attributes of these processes (at scales of 1-10 km; Gutowski Jr et al., 2020). Higher-resolution simulations covering the entire globe 1

show abstract

“…But Zobel's model requires a lot of computing power, so the team looked only at the continental United States, not the entire globe. Even using a supercomputer, it took the better part of a year for the model to crunch through its calculations 6 . But in the end, the researchers got a detailed look at how different parts of the country would be affected, in both temperature and precipitation extremes, if greenhouse-gas emissions continue to remain high out until 2094.…”

Section: More Moisturementioning

confidence: 99%

“…"That will complicate how farmers deal with planting their crops. " The work appeared last month in Earth's Future 6 .…”

Section: Boom or Bustmentioning

confidence: 99%

Why extreme rains are gaining strength as the climate warms

Witze¹

2018

Nature

View full text Add to dashboard Cite

, when the centre of Hurricane Florence was still hundreds of kilometres from North Carolina's coast. As the giant storm lurched towards land, officials ordered more than 1.5 million people to evacuate, warning of "lifethreatening" damage. On 15 September, Florence finally crashed into the United States, where it slowed to a crawl and unleashed even heavier rains. In some places, the deluge continued non-stop for four days. By the time it was all over, Florence had dumped record amounts of rain-including nearly one metre in the town of Elizabethtown, North Carolina-and caused catastrophic flooding. Dozens of people died, and the storm racked up tens of billions of dollars in damages. Even now, months later, the area is struggling to recover. The story of how Florence brought a thriving region to its knees is about to get a lot more familiar. Climate scientists expect that as global temperatures rise, much more rain will fall in extreme storms. The warmer the atmosphere, the more moisture it can hold, which means storms can get wetter. Even before Florence made landfall, a team based at Stony Brook University in New York predicted that the hurricane's heaviest rains would dump at least 50% more precipitation than would have happened people not warmed the planet. Extreme rains-along with the flooding, landslides and other devastation they cause-are some of the deadliest weather events worldwide. This year, heavy rains in the Indian state of Kerala killed more than 470 people, and flooding in southwestern Japan left more than 200 dead. In the United States, flooding, severe storms and tropical cyclones account for 9 of the 11 natural disasters that have topped US$1 billion in damages so far this year. But forecasting how the most punishing rains might change in the future has been notoriously difficult, because scientists can't easily simulate these storms in computer models. Now, many research teams are making advances in understanding the future of extreme precipitation across the world, thanks to models with very high resolution that can provide insight into how storms evolve. Some of the most sophisticated forecasts suggest that as the globe warms, more rains will fall in severe, intermittent storms rather than in the kind of gentle soaking showers that can sustain crops. Other research indicates that the ways in which thunderstorms organize The latest climate simulations are showing that storms will get wetter and more erratic as the world warms.

show abstract

Analyses for High‐Resolution Projections Through the End of the 21st Century for Precipitation Extremes Over the United States

Cited by 36 publications

References 67 publications

Fast and accurate learned multiresolution dynamical downscaling for precipitation

Fast and accurate learned multiresolution dynamical downscaling for precipitation

Comment on gmd-2020-412

Why extreme rains are gaining strength as the climate warms

Contact Info

Product

Resources

About