An official website of the United States government
Abstract The traditional method for estimating weather forecast sensitivity to initial conditions uses adjoint models, which are limited to short lead times due to linearization around a control forecast. The advent of deep‐learning frameworks enables a new approach using backpropagation and gradient descent to iteratively optimize initial conditions, minimizing forecast errors. We apply this approach to the June 2021 Pacific Northwest heatwave using the GraphCast model, yielding over 90% reduction in 10‐day forecast errors over the Pacific Northwest. Similar improvements are found for Pangu‐Weather model forecasts initialized with the GraphCast‐derived optimal, suggesting that model error is an unimportant part of the perturbations. Eliminating small scales from the perturbations also yields similar forecast improvements. Extending the length of the optimization window, we find forecast improvement to about 23 days, suggesting atmospheric predictability at the upper end of recent estimates.
more »
« less
This content will become publicly available on July 1, 2027
Atmospheric Predictability beyond 30 Days with Machine Learning
Abstract Atmospheric predictability research has long held that rapid error growth at small spatial scales imposes an intrinsic limit of roughly 2 weeks on deterministic weather forecast skill. We challenge this limit using GraphCast, a machine learning weather model, by optimizing initial conditions for twice-daily forecasts spanning 2020. This approach yields an average error reduction of 86% at 10 days relative to control forecasts from reanalysis initial conditions, with skill lasting beyond 30 days. Mean optimal initial-condition perturbations reveal large-scale, spatially coherent corrections primarily reflecting an intensification of the Hadley circulation. Forecasts using GraphCast-optimal initial conditions in the Pangu-Weather model achieve a 21% error reduction, peaking at 4 days, indicating that analysis corrections reflect adjustments that target both model and analysis error. These results demonstrate the existence of initial conditions producing skillful deterministic forecasts far beyond 2 weeks. Whether such initial conditions can be identified in real time for improving operational weather forecasts remains a topic of future research. Significance StatementOur results contest the long-standing view that rapid error growth at small spatial scales fundamentally limits reliable weather forecasts to about 2 weeks. By applying machine learning–based optimization to atmospheric initial conditions, we achieve skillful deterministic hindcasts to at least 33 days—more than twice the conventional limit. Validating these optimized initial conditions in an independent weather model leads to significant but smaller improvements, highlighting the importance of model error in the procedure. On average, the optimal corrections reveal physically consistent large-scale patterns that align with an intensified tropical overturning circulation. These results demonstrate the existence of initial conditions that considerably extend the current established limit of predictability. Real-time determination of such initial conditions for operational forecasting remains an open question.
more »
« less
- Award ID(s):
- 2501400
- PAR ID:
- 10701209
- Publisher / Repository:
- American Meteorological Society
- Date Published:
- Journal Name:
- Artificial Intelligence for the Earth Systems
- Volume:
- 5
- Issue:
- 3
- ISSN:
- 2769-7525
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
null (Ed.)Abstract Although useful at short and medium ranges, current dynamical models provide little additional skill for precipitation forecasts beyond week 2 (14 days). However, recent studies have demonstrated that downstream forcing by the Madden–Julian oscillation (MJO) and quasi-biennial oscillation (QBO) influences subseasonal variability, and predictability, of sensible weather across North America. Building on prior studies evaluating the influence of the MJO and QBO on the subseasonal prediction of North American weather, we apply an empirical model that uses the MJO and QBO as predictors to forecast anomalous (i.e., categorical above- or below-normal) pentadal precipitation at weeks 3–6 (15–42 days). A novel aspect of our study is the application and evaluation of the model for subseasonal prediction of precipitation across the entire contiguous United States and Alaska during all seasons. In almost all regions and seasons, the model provides “skillful forecasts of opportunity” for 20%–50% of all forecasts valid weeks 3–6. We also find that this model skill is correlated with historical responses of precipitation, and related synoptic quantities, to the MJO and QBO. Finally, we show that the inclusion of the QBO as a predictor increases the frequency of skillful forecasts of opportunity over most of the contiguous United States and Alaska during all seasons. These findings will provide guidance to forecasters regarding the utility of the MJO and QBO for subseasonal precipitation outlooks.more » « less
-
Abstract Pure artificial intelligence (AI)-based weather prediction (AIWP) models have made waves within the scientific community and the media, claiming superior performance to numerical weather prediction (NWP) models. However, these models often lack impactful output variables such as precipitation. One exception is Google DeepMind’s GraphCast model, which became the first mainstream AIWP model to predict precipitation, but performed only limited verification. We present an analysis of the ECMWF’s Integrated Forecasting System (IFS)-initialized (GRAPIFS) and the NCEP’s Global Forecast System (GFS)-initialized (GRAPGFS) GraphCast precipitation forecasts over the contiguous United States and compare to results from the GFS and IFS models using 1) grid-based, 2) neighborhood, and 3) object-oriented metrics verified against the fifth major global reanalysis produced by ECMWF (ERA5) and the NCEP/Environmental Modeling Center (EMC) stage IV precipitation analysis datasets. We affirmed that GRAPGFSand GRAPIFSperform better than the GFS and IFS in terms of root-mean-square error and stable equitable errors in probability space, but the GFS and IFS precipitation distributions more closely align with the ERA5 and stage IV distributions. Equitable threat score also generally favored GraphCast, particularly for lower accumulation thresholds. Fractions skill score for increasing neighborhood sizes shows greater gains for the GFS and IFS than GraphCast, suggesting the NWP models may have a better handle on intensity but struggle with the location. Object-based verification for GraphCast found positive area biases at low accumulation thresholds and large negative biases at high accumulation thresholds. GRAPGFSsaw similar performance gains to GRAPIFSwhen compared to their NWP counterparts, but initializing with the less familiar GFS conditions appeared to lead to an increase in light precipitation. Significance StatementPure artificial intelligence (AI)-based weather prediction (AIWP) has exploded in popularity with promises of better performance and faster run times than numerical weather prediction (NWP) models. However, less attention has been paid to their capability to predict impactful, sensible weather like precipitation, precipitation type, or specific meteorological features. We seek to address this gap by comparing precipitation forecast performance by an AI model called GraphCast to the Global Forecast System (GFS) and the Integrated Forecasting System (IFS) NWP models. While GraphCast does perform better on many verification metrics, it has some limitations for intense precipitation forecasts. In particular, it less frequently predicts intense precipitation events than the GFS or IFS. Overall, this article emphasizes the promise of AIWP while at the same time stresses the need for robust verification by domain experts.more » « less
-
Abstract This study investigates skill enhancement in operational seasonal forecasts of Beijing Climate Center’s Climate System Model through regional Climate-Weather Research and Forecasting (CWRF) downscaling and improved land initialization in China. The downscaling mitigates regional climate biases, enhancing precipitation pattern correlations by 0.29 in spring and 0.21 in summer. It also strengthens predictive capabilities for interannual anomalies, expanding skillful temperature forecast areas by 6% in spring and 12% in summer. Remarkably, during seven of ten years with relative high predictability, the downscaling increases average seasonal precipitation anomaly correlations by 0.22 and 0.25. Additionally, substitution of initial land conditions via a Common Land Model integration reduces snow cover and cold biases across the Tibetan Plateau and Mongolia-Northeast China, consistently contributing to CWRF’s overall enhanced forecasting capabilities. Improved downscaling predictive skill is attributed to CWRF’s enhanced physics representation, accurately capturing intricate regional interactions and associated teleconnections across China, especially linked to the Tibetan Plateau’s blocking and thermal effects. In summer, CWRF predicts an intensified South Asian High alongside a strengthened East Asian Jet compared to CSM, amplifying cold air advection and warm moisture transport over central to northeast regions. Consequently, rainfall distributions and interannual anomalies over these areas experience substantial improvements. Similar enhanced circulation processes elucidate skill improvement from land initialization, where accurate specification of initial snow cover and soil temperature within sensitive regions persists in influencing local and remote circulations extending beyond two seasons. Our findings emphasize the potential of improving physics representation and surface initialization to markedly enhance regional climate predictions.more » « less
-
On average, modern numerical weather prediction forecasts for daily tornado frequency exhibit no skill beyond day 10. However, in this extended-range lead window, there are particular model cycles that have exceptionally high forecast skill for tornadoes because of their ability to correctly simulate the future synoptic pattern. Here, model initial conditions that produced a more skillful forecast for tornadoes over the United States were exploited while also highlighting potential causes for low-skill cycles within the Global Ensemble Forecasting System, version 12 (GEFSv12). There were 88 high-skill and 91 low-skill forecasts in which the verifying day-10 synoptic pattern for tornado conditions revealed a western U.S. thermal trough and an eastern U.S. thermal ridge, a favorable configuration for tornadic storm occurrence. Initial conditions for high skill forecasts tended to exhibit warmer sea surface temperatures throughout the tropical Pacific Ocean and Gulf of Mexico, an active Madden–Julian oscillation, and significant modulation of Earth-relative atmospheric angular momentum. Low-skill forecasts were often initialized during La Niña and negative Pacific decadal oscillation conditions. Significant atmospheric blocking over eastern Russia—in which the GEFSv12 overforecast the duration and characteristics of the downstream flow—was a common physical process associated with low-skill forecasts. This work helps to increase our understanding of the common causes of high- or low-skill extended-range tornado forecasts and could serve as a helpful tool for operational forecasters.more » « less
