A newly developed multiple linear regression model outperformed two operational hybrid models used by the National Oceanic and Atmospheric Administration in forecasting Atlantic seasonal hurricanes, according to a recent study. Researchers Zeng-Zhen Hu, Wanqiu Wang, Hui Wang, Daniel Harnos, Matthew Rosencrans, and Wassila Thiaw designed the statistical framework to help enhance predictive skill for seasonal tropical cyclone activity, addressing major meteorological disasters that cause substantial damage and loss of life and property across the United States.
Each year, NOAA prepares for the upcoming Atlantic hurricane season by issuing official outlooks for expected seasonal tropical cyclone activity. A suite of dynamical, statistical, and hybrid models currently supports these outlooks. Given the operational need to enhance predictive skill, efforts to build models of varying complexity remain necessary. Within this context, the researchers introduced the multiple linear regression model and evaluated its performance against two operational hybrid models developed and used at NOAA.
The multiple linear regression model is constructed using two key observed predictors. The first predictor consists of observed antecedent vertical zonal wind shear anomalies between 200 hPa and 850 hPa over the tropical North Atlantic. The second predictor is based on observed sea surface temperature anomalies in the hurricane Main Development Region of the North Atlantic.
When evaluated across forecasts and hindcasts from 1982 to 2024 using June observations, the multiple linear regression model overall outperformed the two operational hybrid models within NOAA’s seasonal hurricane forecast tool suite. In operational application, predictions from the regression model incorporating March and June 2025 predictors contributed directly to NOAA’s official seasonal outlooks issued in May 2025 and August 2025, respectively. These seasonal model predictions proved consistent with the observed storm counts.
Although the model demonstrated superior comparative skill, the authors noted an important methodological caveat regarding the findings. The validation period is relatively short, which may impact the robustness of skill comparisons to some extent. Nevertheless, the encouraging results of the multiple linear regression model demonstrate the importance of continued efforts to explore optimal predictors to improve the skill of the Atlantic hurricane season outlook.
Subject of Research: Statistical modeling for Atlantic seasonal hurricane activity forecasting
Article Title: A statistical model for Atlantic seasonal hurricane forecasting and its operational applications
Article References: Hu, Z.-Z., Wang, W., Wang, H., Harnos, D., Rosencrans, M., & Thiaw, W. (2026). A statistical model for Atlantic seasonal hurricane forecasting and its operational applications. PLOS Climate, 5(11), e0001096. https://doi.org/10.1371/journal.pclm.0001096
Image Credits: AI Generated
DOI: 10.1371/journal.pclm.0001096
Keywords: Atlantic hurricanes, seasonal hurricane forecasting, multiple linear regression, tropical cyclone activity, NOAA hurricane outlooks, operational forecast models, vertical zonal wind shear, sea surface temperature anomalies, Main Development Region, tropical North Atlantic, operational hybrid models, hurricane hindcasts, storm count predictions, meteorological forecasting, seasonal outlook updates, forecast skill evaluation
Cite Scienmag News
Reid Dalton. (November 17, 2026). New Statistical Model Outperforms Operational NOAA Tools in Atlantic Hurricane Forecasts. Scienmag. https://scienmag.com/new-statistical-model-outperforms-operational-noaa-tools-in-atlantic-hurricane-forecasts/
Reid Dalton. "New Statistical Model Outperforms Operational NOAA Tools in Atlantic Hurricane Forecasts." Scienmag, 17 November 2026, https://scienmag.com/new-statistical-model-outperforms-operational-noaa-tools-in-atlantic-hurricane-forecasts/. Accessed 22 November 2026.
Reid Dalton. "New Statistical Model Outperforms Operational NOAA Tools in Atlantic Hurricane Forecasts." Scienmag. November 17, 2026. https://scienmag.com/new-statistical-model-outperforms-operational-noaa-tools-in-atlantic-hurricane-forecasts/








