An official website of the United States government
Accurate prediction of precipitation intensity is crucial for both human and natural systems, especially in a warming climate more prone to extreme precipitation. Yet, climate models fail to accurately predict precipitation intensity, particularly extremes. One missing piece of information in traditional climate model parameterizations is subgrid-scale cloud structure and organization, which affects precipitation intensity and stochasticity at coarse resolution. Here, using global storm-resolving simulations and machine learning, we show that, by implicitly learning subgrid organization, we can accurately predict precipitation variability and stochasticity with a low-dimensional set of latent variables. Using a neural network to parameterize coarse-grained precipitation, we find that the overall behavior of precipitation is reasonably predictable using large-scale quantities only; however, the neural network cannot predict the variability of precipitation ( R 2 ∼ 0.45) and underestimates precipitation extremes. The performance is significantly improved when the network is informed by our organization metric, correctly predicting precipitation extremes and spatial variability ( R 2 ∼ 0.9). The organization metric is implicitly learned by training the algorithm on a high-resolution precipitable water field, encoding the degree of subgrid organization. The organization metric shows large hysteresis, emphasizing the role of memory created by subgrid-scale structures. We demonstrate that this organization metric can be predicted as a simple memory process from information available at the previous time steps. These findings stress the role of organization and memory in accurate prediction of precipitation intensity and extremes and the necessity of parameterizing subgrid-scale convective organization in climate models to better project future changes of water cycle and extremes.
more »
« less
Data-driven multiscale modeling of subgrid parameterizations in climate models
Subgrid parameterizations, which represent physical processes occurring below the resolu- tion of current climate models, are an important component in producing accurate, long-term predictions for the climate. A variety of approaches have been tested to design these com- ponents, including deep learning methods. In this work, we evaluate a proof of concept illustrating a multiscale approach to this prediction problem. We train neural networks to predict subgrid forcing values on a testbed model and examine improvements in prediction accuracy that can be obtained by using additional information in both fine-to-coarse and coarse-to-fine directions.
more »
« less
- Award ID(s):
- 1901091
- PAR ID:
- 10455950
- Date Published:
- Journal Name:
- ICLR Workshop on ML for Climate
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract We propose a multiscale approach for predicting quantities in dynamical systems which is explicitly structured to extract information in both fine-to-coarse and coarse-to-fine directions. Our approach improves model accuracy and stability with minimally increased computation compared to non-multiscale approaches with analogous network architecture. We evaluate our approach on an idealized fluid subgrid parameterization (known as closure) task in which our multiscale networks correct chaotic underlying models to reflect the contributions of unresolved, fine-scale dynamics.more » « less
-
Compound flooding, the concurrence of multiple flooding mechanisms such as storm surge, heavy rainfall, and riverine flooding, poses a significant threat to coastal communities. To mitigate the impacts of compound flooding, forecasts must represent the variability of flooding drivers over a wide range of spatial scales while remaining timely. One approach to develop these forecasts is through subgrid corrections, which utilize information at smaller scales to “correct” water levels and current velocities averaged over the model scale. Recent studies have shown that subgrid models can improve both accuracy and efficiency; however, existing models are not able to account for the dynamic interactions of hydrologic and hydrodynamic drivers and their contributions to flooding along the smallest flow pathways when using a coarse resolution. Here, we have developed a solver called CoaSToRM (Coastal Subgrid Topography Research Model) with subgrid corrections to compute compound flooding in coastal systems resulting from fluvial, pluvial, tidal, and wind-driven processes. A key contribution is the model’s ability to enforce all flood drivers and use the subgrid corrections to improve the accuracy of the coarse-resolution simulation. The model is validated for Hurricane Eta 2020 in Tampa Bay, showing improved prediction accuracy with subgrid corrections at 42 locations. Subgrid models with coarse resolutions (R2 = 0.70, 0.73, 0.77 for 3-, 1.5-, 0.75-km grids) outperform standard counterparts (R2 = 0.03, 0.14, 0.26). A 3-km subgrid simulation runs roughly 50 times faster than a 0.75-km subgrid simulation, with similar accuracy.more » « less
-
Abstract A promising approach to improve climate‐model simulations is to replace traditional subgrid parameterizations based on simplified physical models by machine learning algorithms that are data‐driven. However, neural networks (NNs) often lead to instabilities and climate drift when coupled to an atmospheric model. Here, we learn an NN parameterization from a high‐resolution atmospheric simulation in an idealized domain by accurately calculating subgrid terms through coarse graining. The NN parameterization has a structure that ensures physical constraints are respected, such as by predicting subgrid fluxes instead of tendencies. The NN parameterization leads to stable simulations that replicate the climate of the high‐resolution simulation with similar accuracy to a successful random‐forest parameterization while needing far less memory. We find that the simulations are stable for different horizontal resolutions and a variety of NN architectures, and that an NN with substantially reduced numerical precision could decrease computational costs without affecting the quality of simulations.more » « less
-
Abstract Global climate models parameterize a range of atmospheric‐oceanic processes, including gravity waves (GWs), clouds, moist convection, and turbulence, that cannot be sufficiently resolved. These subgrid‐scale closures for unresolved processes are a substantial source of model uncertainty. Here, we present a new approach to developing machine learning (ML) parameterizations of small‐scale climate processes by fine‐tuning a pre‐trained AI foundation model (FM). FMs are largely unexplored in climate research. A pre‐trained encoder‐decoder from a 2.3 billion parameter FM (NASA and IBM Research's Prithvi WxC)—which contains a latent probabilistic representation of atmospheric evolution—is fine‐tuned (or reused) to create a deep learning parameterization for atmospheric gravity waves (GWs); a process unseen during pre‐training. The parameterization captures GW effects for a coarse‐resolution climate model by learning the fluxes from an atmospheric reanalysis with 10 times finer resolution. A comparison of monthly averages and instantaneous evolution with a machine learning model baseline (an Attention U‐Net) reveals superior predictive performance of the FM parameterization throughout the atmosphere, even in regions excluded during pre‐training. This performance boost is quantified using the Hellinger distance, which is 0.11 for the baseline and 0.06 for the fine‐tuned model. Our findings emphasize the versatility and reusability of FMs, which could be used to accomplish a range of atmosphere‐ and climate‐related applications, leading the way for the creation of observations‐driven and physically accurate parameterizations for more earth system processes.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
