Regional climate model emulator based on deep learning: concept and first evaluation of a novel hybrid downscaling approach

There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.

Abstract

Providing reliable information on climate change at local scale remains a challenge of first importance for impact studies and policymakers. Here, we propose a novel hybrid downscaling method combining the strengths of both empirical statistical downscaling methods and Regional Climate Models (RCMs). In the longer term, the final aim of this tool is to enlarge the high-resolution RCM simulation ensembles at low cost to explore better the various sources of projection uncertainty at local scale. Using a neural network, we build a statistical RCM-emulator by estimating the downscaling function included in the RCM. This framework allows us to learn the relationship between large-scale predictors and a local surface variable of interest over the RCM domain in present and future climate. The RCM-emulator developed in this study is trained to produce daily maps of the near-surface temperature at the RCM resolution (12 km). The emulator demonstrates an excellent ability to reproduce the complex spatial structure and daily variability simulated by the RCM, particularly how the RCM refines the low-resolution climate patterns. Training in future climate appears to be a key feature of our emulator. Moreover, there is a substantial computational benefit of running the emulator rather than the RCM, since training the emulator takes about 2 h on GPU, and the prediction takes less than a minute. However, further work is needed to improve the reproduction of some temperature extremes, the climate change intensity and extend the proposed methodology to different regions, GCMs, RCMs, and variables of interest.

Related collections

Most cited references 60

Record: found
Abstract: found
Article: not found

Deep learning.

Yann LeCun, Yoshua Bengio, Geoffrey E Hinton (2015)

Deep learning allows computational models that are composed of multiple processing layers to learn representations of data with multiple levels of abstraction. These methods have dramatically improved the state-of-the-art in speech recognition, visual object recognition, object detection and many other domains such as drug discovery and genomics. Deep learning discovers intricate structure in large data sets by using the backpropagation algorithm to indicate how a machine should change its internal parameters that are used to compute the representation in each layer from the representation in the previous layer. Deep convolutional nets have brought about breakthroughs in processing images, video, speech and audio, whereas recurrent nets have shone light on sequential data such as text and speech.

0 comments Cited 10034 times – based on 0 reviews      Review now

Bookmark

Record: found
Abstract: found
Article: not found

An Overview of CMIP5 and the Experiment Design

Karl E. Taylor, Ronald J Stouffer, Gerald Meehl (2012)

The fifth phase of the Coupled Model Intercomparison Project (CMIP5) will produce a state-of-the- art multimodel dataset designed to advance our knowledge of climate variability and climate change. Researchers worldwide are analyzing the model output and will produce results likely to underlie the forthcoming Fifth Assessment Report by the Intergovernmental Panel on Climate Change. Unprecedented in scale and attracting interest from all major climate modeling groups, CMIP5 includes “long term” simulations of twentieth-century climate and projections for the twenty-first century and beyond. Conventional atmosphere–ocean global climate models and Earth system models of intermediate complexity are for the first time being joined by more recently developed Earth system models under an experiment design that allows both types of models to be compared to observations on an equal footing. Besides the longterm experiments, CMIP5 calls for an entirely new suite of “near term” simulations focusing on recent decades and the future to year 2035. These “decadal predictions” are initialized based on observations and will be used to explore the predictability of climate and to assess the forecast system's predictive skill. The CMIP5 experiment design also allows for participation of stand-alone atmospheric models and includes a variety of idealized experiments that will improve understanding of the range of model responses found in the more complex and realistic simulations. An exceptionally comprehensive set of model output is being collected and made freely available to researchers through an integrated but distributed data archive. For researchers unfamiliar with climate models, the limitations of the models and experiment design are described.