CoChat

Hybrid physics-AI outperforms numerical weather prediction for extreme precipitation nowcasting

Open access
Das, Puja, Posch, August, Barber, Nathan, Hicks, Michael, Duffy, Kate, Vandal, Thomas, Singh, Debjani, Werkhoven, Katie van, Ganguly, Auroop R. · 2024 · npj Climate and Atmospheric Science

The DOI matches its Crossref record; 55 of the 55 references it cites were found in a registry or an index.

Abstract

Precipitation nowcasting, which is critical for flood emergency and river management, has remained challenging for decades, although recent developments in deep generative modeling (DGM) suggest the possibility of improvements. River management centers, such as the Tennessee Valley Authority, have been using Numerical Weather Prediction (NWP) models for nowcasting, but they have been struggling with missed detections even from best-in-class NWP models. While decades of prior research achieved limited improvements beyond advection and localized evolution, recent attempts have shown progress from so-called physics-free machine learning (ML) methods, and even greater improvements from physics-embedded ML approaches. Developers of DGM for nowcasting have compared their approaches with optical flow (a variant of advection) and meteorologists' judgment, but not with NWP models. Further, they have not conducted independent co-evaluations with water resources and river managers. Here we show that the state-of-the-art physics-embedded deep generative model, specifically NowcastNet, outperforms the High Resolution Rapid Refresh (HRRR) model, which is the latest generation of NWP, along with advection and persistence, especially for heavy precipitation events. Thus, for grid-cell extremes over 16 mm/h, NowcastNet demonstrated a median critical success index (CSI) of 0.30, compared with median CSI of 0.04 for HRRR. However, despite hydrologically-relevant improvements in point-by-point forecasts from NowcastNet, caveats include overestimation of spatially aggregate precipitation over longer lead times. Our co-evaluation with ML developers, hydrologists and river managers suggest the possibility of improved flood emergency response and hydropower management.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41612-024-00834-8

Cites

55 of 55 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Map depicting the Tennessee Valley Authority (TVA) service area and the locations of key electricity generating assets within the region. The figure provides an overview of the geographical coverage of TVA's operations and highlights the distribution of major power generation facilities as reported by the Government Accountability Office (GAO) in their 2023 Report to Congressional Requesters62.
Fig. 1 Map depicting the Tennessee Valley Authority (TVA) service area and the locations of key electricity generating assets within the region. The figure provides an overview of the geographical coverage of TVA's operations and highlights the distribution of major power generation facilities as reported by the Government Accountability Office (GAO) in their 2023 Report to Congressional Requesters62.CC BY-NC-ND · PubMed Central
Fig. 2 Comparison of accumulated precipitation forecasts from High-Resolution Rapid Refresh (HRRR) model at the McEwen precipitation gauge during the Waverly event provides insights into the forecast bias. Precipitation forecasts (in inches) on August 21, 2021, displayed in Coordinated Universal Time (UTC) from the High-Resolution Rapid Refresh (HRRR) model, are shown with red, yellow, and green lines, while observations are shown with a blue line. The McEwen precipitation accumulation data utilized in this comparison is derived from TVA-collected data. The plot illustrates the discrepancy between the accumulated precipitation forecasts and the actual observations at the McEwen gauge.
Fig. 2 Comparison of accumulated precipitation forecasts from High-Resolution Rapid Refresh (HRRR) model at the McEwen precipitation gauge during the Waverly event provides insights into the forecast bias. Precipitation forecasts (in inches) on August 21, 2021, displayed in Coordinated Universal Time (UTC) from the High-Resolution Rapid Refresh (HRRR) model, are shown with red, yellow, and green lines, while observations are shown with a blue line. The McEwen precipitation accumulation data utilized in this comparison is derived from TVA-collected data. The plot illustrates the discrepancy between the accumulated precipitation forecasts and the actual observations at the McEwen gauge.CC BY-NC-ND · PubMed Central
Fig. 3 Comparison of precipitation forecasts from NowcastNet and HRRR for the Waverly flood event (August 21, 2021): spatial accuracy and power spectral density analysis. Precipitation forecasts (in mm/h) from NowcastNet (1 km spatial resolution) and HRRR (3 km spatial resolution) at different lead times (T + 1 h, T + 2 h, and T + 3 h) with MRMS QPE32 for the Waverly flood event on August 21, 2021 (T = 8:00 UTC) within the TVA area. The precipitation images cover a spatial extent of 384 km × 384 km. The base map shows US state boundaries. NowcastNet predicts the MRMS precipitation patterns more closely than HRRR does, in terms of the spatial distribution and intensity of the precipitation. The last row depicts the PSD at different wavelengths, at different lead times (T + 1 h, T + 2 h, and T + 3 h).
Fig. 3 Comparison of precipitation forecasts from NowcastNet and HRRR for the Waverly flood event (August 21, 2021): spatial accuracy and power spectral density analysis. Precipitation forecasts (in mm/h) from NowcastNet (1 km spatial resolution) and HRRR (3 km spatial resolution) at different lead times (T + 1 h, T + 2 h, and T + 3 h) with MRMS QPE32 for the Waverly flood event on August 21, 2021 (T = 8:00 UTC) within the TVA area. The precipitation images cover a spatial extent of 384 km × 384 km. The base map shows US state boundaries. NowcastNet predicts the MRMS precipitation patterns more closely than HRRR does, in terms of the spatial distribution and intensity of the precipitation. The last row depicts the PSD at different wavelengths, at different lead times (T + 1 h, T + 2 h, and T + 3 h).CC BY-NC-ND · PubMed Central
Fig. 4 Metrics comparison across 30 heavy precipitation events. Comparison of precipitation forecast accuracy between NowcastNet, HRRR, Persistence, and Advection against MRMS QPE (at 3 km spatial resolution) for 30 heavy precipitation events across the geography of interest. Metrics include Critical Success Index (CSI), Probability of Detection (POD), False Alarm Ratio (FAR), and Fractions Skill Score (FSS); all at thresholds (t) of t > 0.1 mm/h, t > 16 mm/h and t > 32 mm/h for different lead times (T + 1 h, T + 2 h, and T + 3 h). Upward arrow indicates higher score is better and downward arrow indicates lower score is better.
Fig. 4 Metrics comparison across 30 heavy precipitation events. Comparison of precipitation forecast accuracy between NowcastNet, HRRR, Persistence, and Advection against MRMS QPE (at 3 km spatial resolution) for 30 heavy precipitation events across the geography of interest. Metrics include Critical Success Index (CSI), Probability of Detection (POD), False Alarm Ratio (FAR), and Fractions Skill Score (FSS); all at thresholds (t) of t > 0.1 mm/h, t > 16 mm/h and t > 32 mm/h for different lead times (T + 1 h, T + 2 h, and T + 3 h). Upward arrow indicates higher score is better and downward arrow indicates lower score is better.CC BY-NC-ND · PubMed Central
Fig. 5 Precipitation prediction discrepancies from NowcastNet at varying lead times for four extreme rainfall events. Comparison of precipitation prediction discrepancies (in mm/h) from NowcastNet model at different lead times (T + 10 min, T + 1 h, T + 2 h, and T + 3 h) with MRMS for four extreme rainfall events on August 21, 2021 (T = 8:00 UTC), February 17, 2022 (T = 18:00 UTC), February 16, 2023 (T = 12:00 UTC) and March 15, 2024 (T = 3:00 UTC), within the TVA area. The basemap shows US state boundaries. Blue shades represent underestimation, while red shades represent overestimation of precipitation. With increasing lead time, discrepancies between MRMS and NowcastNet predictions become more pronounced.
Fig. 5 Precipitation prediction discrepancies from NowcastNet at varying lead times for four extreme rainfall events. Comparison of precipitation prediction discrepancies (in mm/h) from NowcastNet model at different lead times (T + 10 min, T + 1 h, T + 2 h, and T + 3 h) with MRMS for four extreme rainfall events on August 21, 2021 (T = 8:00 UTC), February 17, 2022 (T = 18:00 UTC), February 16, 2023 (T = 12:00 UTC) and March 15, 2024 (T = 3:00 UTC), within the TVA area. The basemap shows US state boundaries. Blue shades represent underestimation, while red shades represent overestimation of precipitation. With increasing lead time, discrepancies between MRMS and NowcastNet predictions become more pronounced.CC BY-NC-ND · PubMed Central
Fig. 6 Comparison of observed and forecast precipitation patterns and error decomposition using the contiguous rain area (CRA) method. A Illustration of the CRA formation by aligning the isohytes between observed (MRMS) and forecast (NowcastNet) fields, highlighting the displacement required for optimal alignment34,35. B Spatial distributions of observed and forecasted precipitation at various lead times (T + 1 h, T + 2 h, and T + 3 h) with identified pattern and displacement errors. C Error decomposition into volume, displacement, and pattern errors for different lead times, quantified as RMSE in mm/h. D Summary of CRA verification metrics (with threshold of 16 mm/h), including Pearson correlation coefficients (CC), RMSE values, and error decomposition percentages, across different lead times, demonstrating the dominance of pattern errors in forecast accuracy.
Fig. 6 Comparison of observed and forecast precipitation patterns and error decomposition using the contiguous rain area (CRA) method. A Illustration of the CRA formation by aligning the isohytes between observed (MRMS) and forecast (NowcastNet) fields, highlighting the displacement required for optimal alignment34,35. B Spatial distributions of observed and forecasted precipitation at various lead times (T + 1 h, T + 2 h, and T + 3 h) with identified pattern and displacement errors. C Error decomposition into volume, displacement, and pattern errors for different lead times, quantified as RMSE in mm/h. D Summary of CRA verification metrics (with threshold of 16 mm/h), including Pearson correlation coefficients (CC), RMSE values, and error decomposition percentages, across different lead times, demonstrating the dominance of pattern errors in forecast accuracy.CC BY-NC-ND · PubMed Central
Fig. 7 Multisource integration and predictive analytics in precipitation forecasting. A demonstrates the reduction of prediction skill or information content of precipitation forecasts as lead time (shown in logarithmic scale) increases, comparing (a) persistence, (b) nowcasting, (c) mesoscale and (d) synoptic scale numerical weather prediction (NWP), (e) merged approach within the boundary of (f) limit of predictability44,63–65. Merged forecasts can be a combination of nowcasting, NWP models, satellite information, etc. B demonstrates generation of precipitation forecasts using a deep generative model (DGM). The proposed DGM combines observed remotely sensed data from radar and geostationary satellites, ground sensors, ancillary information from terrain properties, physics of precipitation22,66 and NWP state variables to enhance forecast accuracy.
Fig. 7 Multisource integration and predictive analytics in precipitation forecasting. A demonstrates the reduction of prediction skill or information content of precipitation forecasts as lead time (shown in logarithmic scale) increases, comparing (a) persistence, (b) nowcasting, (c) mesoscale and (d) synoptic scale numerical weather prediction (NWP), (e) merged approach within the boundary of (f) limit of predictability44,63–65. Merged forecasts can be a combination of nowcasting, NWP models, satellite information, etc. B demonstrates generation of precipitation forecasts using a deep generative model (DGM). The proposed DGM combines observed remotely sensed data from radar and geostationary satellites, ground sensors, ancillary information from terrain properties, physics of precipitation22,66 and NWP state variables to enhance forecast accuracy.CC BY-NC-ND · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 55 references

Research works printed but not linked: 12

  1. Genevois, R. & Tecca, P. R. The vajont landslide: state-of-the-art. Ital. J. Eng. Geol. Environ. 6, 15–39 (2013).
  2. Marchuk, G. Numerical Methods in Weather Prediction (Elsevier, 2012).
  3. Ganguly, A. R. & Bras, R. L. Distributed quantitative precipitation forecasting using information from radar and numerical weather prediction models. J. Hydrometeorol. 4, 1168–1180 (2003). · doi:10.1175/1525-7541(2003)004 · Paper page
  4. Germann, U. & Zawadzki, I. Scale-dependence of the predictability of precipitation from continental radar images. Mon. Weather Rev. 130, 2859–2873 (2002). · doi:10.1175/1520-0493(2002)130
  5. Alexander, C., Carley, J. & Pyle, M. The rapid refresh forecast system: looking beyond the first operational version. In 28th Conference on Numerical Weather Prediction (2023).
  6. Goodfellow, I et al. Generative Adversarial Nets (Advances in Neural Information Processing Systems) 2672–2680 (Curran, 2014).
  7. Murphy, A. H. What is a good forecast? An essay on the nature of goodness in weather forecasting. Weather Forecast. 8, 281–293 (1993). · doi:10.1175/1520-0434(1993)008
  8. Doswell, C. H. A. R. L., Davies-Jones, R. & Keller, D. L. On summary measures of skill in rare event forecasting based on contingency tables. Weather Forecast. 5, 576–585 (1990). · doi:10.1175/1520-0434(1990)005
  9. Heidke, P. Calculation of the success and quality of wind force forecasts in the storm warning service. Geogr. Ann. 8, 301–349 (1926).
  10. Harris, D., Foufoula-Georgiou, E., Droegemeier, K. K. & Levit, J. J. Multiscale statistical properties of a high-resolution precipitation forecast. J. Hydrometeorol. 2, 406–418 (2001). · doi:10.1175/1525-7541(2001)002
  11. Zipser, E. Rainfall predictability: when will extrapolation-based algorithms fail. In 8th Conference on Hydrometeorology, American Meteorological Society (1990).
  12. Houze Jr, R. A. Stratiform precipitation in regions of convection: a meteorological paradox? Bull. Am. Meteorol. Soc. 78, 2179–2196 (1997). · doi:10.1175/1520-0477(1997)078

Cited by

See all 58

Metadata

Authors
Das, P., Posch, A., Barber, N., Hicks, M., Duffy, K., Vandal, T., Singh, D., Werkhoven, K. V., & Ganguly, A. R.
Affiliations
  • Northeastern University (US)
  • Tennessee Valley Authority (US)
  • Oak Ridge National Laboratory (US)
  • RTI International (US)
Corresponding author
Ganguly, Auroop R.
Venue
npj Climate and Atmospheric Science
Published
2024-11-18
Volume
7
Issue
1
Pages
282-282
ISSN
2397-3722 (ISSN-L)
OpenAlex
W4404482519
Cited by
58
References
55
Citations by year
2026: 35 · 2025: 23
FWCI
9.37
Type
journal article
Language
en
Open access
Licence
Locations
  • npj Climate and Atmospheric Science (publishedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • arXiv (Cornell University) (cc-by)
Funders
  • National Aeronautics and Space Administration (80NSSC22K1138)
  • Nuclear Safety and Security Commission
Topics
  • Physical Sciences › Earth and Planetary Sciences › Atmospheric Science › Meteorological Phenomena and Simulations
  • Physical Sciences › Environmental Science › Environmental Engineering › Hydrological Forecasting Using AI
  • Physical Sciences › Earth and Planetary Sciences › Atmospheric Science › Precipitation Measurement and Analysis
Keywords
precipitation nowcasting, heavy rainfall, deep generative models, numerical weather prediction, HRRR model, flood disaster response, river management, critical success index, hydropower management, advection

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Hybrid physics-AI outperforms numerical weather prediction for extreme precipitation nowcasting. https://app.cochat.ai/papers/10.1038/s41612-024-00834-8
The paper
Das, P., Posch, A., Barber, N., Hicks, M., Duffy, K., Vandal, T., Singh, D., Werkhoven, K. V., & Ganguly, A. R. (2024). Hybrid physics-AI outperforms numerical weather prediction for extreme precipitation nowcasting. npj Climate and Atmospheric Science. https://doi.org/10.1038/s41612-024-00834-8

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.