An official website of the United States government
Abstract An ensemble postprocessing method is developed to improve the probabilistic forecasts of extreme precipitation events across the conterminous United States (CONUS). The method combines a 3D vision transformer (ViT) for bias correction with a latent diffusion model (LDM), a generative artificial intelligence (AI) method, to postprocess 6-hourly precipitation ensemble forecasts and produce an enlarged generative ensemble that contains spatiotemporally consistent precipitation trajectories. These trajectories are expected to improve the characterization of extreme precipitation events and offer skillful multiday accumulated and 6-hourly precipitation guidance. The method is tested using the Global Ensemble Forecast System (GEFS) precipitation forecasts out to day 6 and is verified against the Climatology-Calibrated Precipitation Analysis (CCPA) data. Verification results indicate that the method generated skillful ensemble members with improved continuous ranked probabilistic skill scores (CRPSSs) and Brier skill scores (BSSs) over the raw operational GEFS and a multivariate statistical postprocessing baseline. It showed skillful and reliable probabilities for events at extreme precipitation thresholds. Explainability studies were further conducted, which revealed the decision-making process of the method and confirmed its effectiveness on ensemble member generation. This work introduces a novel, generative AI–based approach to address the limitation of small numerical ensembles and the need for larger ensembles to identify extreme precipitation events. Significance StatementWe use a new artificial intelligence (AI) technique to improve extreme precipitation forecasts from a numerical weather prediction ensemble, generating more scenarios that better characterize extreme precipitation events. This AI-generated ensemble improved the accuracy of precipitation forecasts and probabilistic warnings for extreme precipitation events. The study explores AI methods to generate precipitation forecasts and explains the decision-making mechanisms of such AI techniques to prove their effectiveness.
more »
« less
Customized deep learning for precipitation bias correction and downscaling
Abstract. Systematic biases and coarse resolutions are major limitations ofcurrent precipitation datasets. Many deep learning (DL)-based studies havebeen conducted for precipitation bias correction and downscaling. However,it is still challenging for the current approaches to handle complexfeatures of hourly precipitation, resulting in the incapability ofreproducing small-scale features, such as extreme events. This studydeveloped a customized DL model by incorporating customized loss functions,multitask learning and physically relevant covariates to bias correct anddownscale hourly precipitation data. We designed six scenarios tosystematically evaluate the added values of weighted loss functions,multitask learning, and atmospheric covariates compared to the regular DLand statistical approaches. The models were trained and tested using theModern-era Retrospective Analysis for Research and Applications version 2(MERRA2) reanalysis and the Stage IV radar observations over the northerncoastal region of the Gulf of Mexico on an hourly time scale. We found thatall the scenarios with weighted loss functions performed notably better thanthe other scenarios with conventional loss functions and a quantilemapping-based approach at hourly, daily, and monthly time scales as well asextremes. Multitask learning showed improved performance on capturing finefeatures of extreme events and accounting for atmospheric covariates highlyimproved model performance at hourly and aggregated time scales, while theimprovement is not as large as from weighted loss functions. We show thatthe customized DL model can better downscale and bias correct hourlyprecipitation datasets and provide improved precipitation estimates at finespatial and temporal resolutions where regular DL and statistical methodsexperience challenges.
more »
« less
- PAR ID:
- 10404877
- Date Published:
- Journal Name:
- Geoscientific Model Development
- Volume:
- 16
- Issue:
- 2
- ISSN:
- 1991-9603
- Page Range / eLocation ID:
- 535 to 556
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Climate and weather data such as precipitation derived from Global Climate Models (GCMs) and satellite observations are essential for the global and local hydrological assessment. However, most climatic popular precipitation products (with spatial resolutions coarser than 10km) are too coarse for local impact studies and require “downscaling” to obtain higher resolutions. Traditional precipitation downscaling methods such as statistical and dynamic downscaling require an input of additional meteorological variables, and very few are applicable for downscaling hourly precipitation for higher spatial resolution. Based on dynamic dictionary learning, we propose a new downscaling method, PreciPatch, to address this challenge by producing spatially distributed higher resolution precipitation fields with only precipitation input from GCMs at hourly temporal resolution and a large geographical extent. Using aggregated Integrated Multi-satellitE Retrievals for GPM (IMERG) data, an experiment was conducted to evaluate the performance of PreciPatch, in comparison with bicubic interpolation using RainFARM—a stochastic downscaling method, and DeepSD—a Super-Resolution Convolutional Neural Network (SRCNN) based downscaling method. PreciPatch demonstrates better performance than other methods for downscaling short-duration precipitation events (used historical data from 2014 to 2017 as the training set to estimate high-resolution hourly events in 2018).more » « less
-
Abstract Extreme precipitation across multiple time scales is a natural hazard that creates a significant risk to life, with a commensurately large cost through property loss. We devise a method to create 14-day extreme-event windows that characterize precipitation events in the contiguous United States (CONUS) for the years 1915–2018. Our algorithm imposes thresholds for both total precipitation and the duration of the precipitation to identify events with sufficient length to accentuate the synoptic and longer time scale contribution to the precipitation event. Kernel density estimation is employed to create extreme-event polygons that are formed into a database spanning from 1915 through 2018. Using the developed database, we clustered events into regions using a k -means algorithm. We define the “hybrid index,” a weighted composite of silhouette score and number of clustered events, to show that the optimal number of clusters is 15. We also show that 14-day extreme precipitation events are increasing in the CONUS, specifically in the Dakotas and much of New England. The algorithm presented in this work is designed to be sufficiently flexible to be extended to any desired number of days on the subseasonal-to-seasonal (S2S) time scale (e.g., 30 days). Additional databases generated using this framework are available for download from our GitHub. Consequently, these S2S databases can be analyzed in future works to determine the climatology of S2S extreme precipitation events and be used for predictability studies for identified events.more » « less
-
As precipitation analysis reveals critical statistical characteristics, temporal patterns, and spatial distributions of rainfall and snowfall events, it plays an important role in planning urban drainage systems, flood forecasting, hydrological modeling, and climate studies. It helps engineers design climate-resilient infrastructure capable of withstanding extreme weather events, which is becoming increasingly important as precipitation patterns change over time. With precipitation analysis, multiple valuable information can be determined, such as storm intensity, duration, and frequency. To enhance understanding of precipitation data and analysis results, researchers often use graphical representation methods to show the data in visual formats. Although existing precipitation analysis and basic visual representations are helpful, it is critical to have a comprehensive analysis and visualization system to detect significant patterns and anomalies in high-resolution temporal precipitation data more effectively. This study presents a visual analytics system enabling interactive analysis of hourly precipitation data across all U.S. states. Multiple coordinated visualizations are designed to support both single and multiple-station analysis. These visualizations allow users to examine temporal patterns, spatial distributions, and statistical characteristics of precipitation events directly within visualizations. Case studies demonstrate the usefulness of the designed system by evaluating various historical storm events.more » « less
-
Abstract Climate change impacts on precipitation characteristics will alter the hydrologic characteristics, such as peak flows, time to peak, and erosion potential of watersheds. However, many of the currently available climate change datasets are provided at temporal and spatial resolutions that are inadequate to quantify projected changes in hydrologic characteristics of a watershed. Therefore, it is critical to temporally disaggregate coarse-resolution precipitation data to finer resolutions for studies sensitive to precipitation characteristics. In this study, we generated novel 15-minute precipitation datasets from hourly precipitation datasets obtained from five NA-CORDEX downscaled climate models under RCP 8.5 scenario for the historical (1970–1999) and projected (2030–2059) years over the Southeast United States using a modified version of the stochastic method. The results showed conservation of mass of the precipitation inputs. Furthermore, the probability of zero precipitation, variance of precipitation, and maximum precipitation in the disaggregated data matched well with the observed precipitation characteristics. The generated 15-minute precipitation data can be used in all scientific studies that require precipitation data at that resolution.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.5194/gmd-16-535-2023
