Impacts of atmospheric river reconnaissance dropsondes on ECMWF Integrated Forecasting System precipitation forecasts

September 8, 2025

A paper titled “Impacts of atmospheric river reconnaissance dropsondes on ECMWF Integrated Forecasting System precipitation forecasts” by Jia Wang (CW3E), David Lavers (ECMWF), Luca Delle Monache (CW3E), Bruce Ingleby (ECMWF), Minghua Zheng (CW3E), Xingren Wu (NCEP), F. Martin Ralph (CW3E), and Florian Pappenberger (ECMWF), was recently published in the Quarterly Journal of the Royal Meteorological Society. This study represents a collaborative effort among CW3E, ECMWF, and NCEP to continue monitoring the impacts of Atmospheric River Reconnaissance (AR Recon) observations in operational models, and supports multiple priorities identified in the CW3E 2025-2029 Strategic Plan, including Atmospheric Rivers and Extreme Precipitation Research, Prediction, and Applications; Novel Observations; and Advanced Precipitation and Streamflow Prediction.

This study focuses on the influence of AR Recon dropsonde observations on ECMWF Integrated Forecasting System (IFS) precipitation forecasts during the 2023 and 2024 field seasons. Dropsonde impacts are evaluated through Observing System Experiments (OSEs). Two sets of experiments are conducted: control runs, which assimilate all operationally available observations, and data denial runs, which assimilate all observations except dropsondes. Validation is performed using stage IV precipitation data across three domains: California (CA), the Pacific Northwest (PNW), and the western United States (WEST; defined as west of 107°W).

Results (Fig. 1) show that dropsondes provide forecast improvements over CA during two periods: 24–36 h and 72–96 h. Improvements during the latter period are larger, with more values passing the significance test. We hypothesize that these two periods of improvements are associated with aircraft operations from two bases: California and Hawaii. Over the PNW, dropsondes increase forecast skill mainly at 48–72 h, with secondary improvements at 96 h and 120 h. The difference in timing of improvements between PNW and CA in 2023 is associated with the southward propagation of error reductions in the large-scale flow (Fig. 2). Across the WEST domain, dropsonde assimilation is overall beneficial through 96 h, with the exception around 48 h.

Dropsonde impacts are also assessed in the NCEP Global Forecast System (GFS) for 2023 (Fig. 3). Notable consistency emerges between the two operational systems: (1) both exhibit two periods of improvement over CA, with greater benefits in the second period; (2) both show that the skill enhancement over the PNW peaks around 72 h. The consistency in the timing of improvements between the IFS and GFS, the alignment of impacts on both large-scale flow and precipitation forecasts, and the temporal coherence among three evaluation metrics collectively strengthen confidence that the dropsonde benefits are physically meaningful.

Figure 1. The percentage differences (%) between the control experiments (CTRL) and the denial experiments (noDROP) for (a–c) Gilbert Skill Score (GSS), (d–f) Spearman’s rank correlation coefficient (SCC), and (g–i) root-mean-squared error (RMSE) for the two seasons combined. Half-filled cells represent absolute percentage differences smaller than 0.5%. For the GSS and SCC, the difference is CTRL − noDROP, while for the RMSE, the difference is noDROP − CTRL. Thus, green represents improvement from dropsondes, and brown indicates degradation. The left, middle, and right columns correspond to the California (CA), Pacific Northwest (PNW), and WEST verification domains, respectively. The x-axis represents thresholds for 24-hour accumulated precipitation. Values with frames are those passing the significance test, using the 90% confidence interval. Figure credit: Figure 5 of Wang et al. (2025).

Figure 2. The differences [m] in mean absolute errors for 500-hPa geopotential height between the control experiments (CTRL) and the denial experiments (noDROP) (CTRL − noDROP), for (a–g) 2022/23, across lead times from 48 to 120 hours. Dotted areas indicate statistically significant differences at the 90% confidence level, in which uncertainty is estimated as ±1.645×(σ/√n), with σ representing the standard deviation and n the sample size. Contours are averaged 500-hPa geopotential height from the CTRL analyses. Contour interval is 80 m. Figure credit: Figure 8 of Wang et al. (2025).

Figure 3. As in Figure 1a–c, but are dropsonde impacts in Global Forecast System (GFS) 2022/23 experiments. Half-filled cells represent absolute percentage differences smaller than 0.5%. Figure credit: Figure 9 of Wang et al. (2025).

Wang, J., Lavers, D. A., Delle Monache, L., Ingleby, B., Zheng, M., Wu, X., Ralph, F. M. & Pappenberger, F. (2025). Impacts of atmospheric river reconnaissance dropsondes on ECMWF Integrated Forecasting System precipitation forecasts. Quarterly Journal of the Royal Meteorological Society (published online ahead of print 2025), e70019. https://doi.org/10.1002/qj.70019

Ralph, F. M., Cannon, F., Tallapragada, V., Davis, C. A., Doyle, J. D., Pappenberger, F., Subramanian, A., Wilson, A. M., Lavers, D. A., Reynolds, C. A., Haase, J. S., Centurioni, L., Ingleby, B., Rutz, J. J., Cordeira, J. M., Zheng, M., Hecht, C., Kawzenuk, B., & Delle Monache, L. (2020). West Coast forecast challenges and development of atmospheric river reconnaissance. Bulletin of the American Meteorological Society, 101(8), E1357-E1377. https://doi.org/10.1175/BAMS-D-19-0183.1

Zheng, M., Torn, R., Delle Monache, L., Doyle, J., Ralph, F. M., Tallapragada, V., Davis, C., Steinhoff, D., Wu, X., Wilson, A., Papadopoulos, C., & Mulrooney, P. (2024). An Assessment of Dropsonde Sampling Strategies for Atmospheric River Reconnaissance. Monthly Weather Review, 152(3), 811–835. https://doi.org/10.1175/MWR-D-23-0111.1

Zheng, M., Ralph, F. M., Tallapragada, V., Wilson, A. M., Babbitt, S. H., Bartlett, S. M., Cao, B., Centurioni, L., Cordeira, J. M., Davis, C., Monache, L. D., Doyle, J. D., Elless, T. J., Feuer, S., Haase, J. S., Hathaway, N., Hutchinson, T., Iniguez, P., Kawzenuk, B., Knappe, E., Lavers, D. A., Lundry, A., Michaelis, A., Pappenberger, F., Reynolds, C. A., Rickert, R., Roj, S., Rutz, J. J., Subramanian, A. C., Torn, R. D., Wang, J., Wu, K., & Wu, X. (2025). Atmospheric River Reconnaissance: Mission Planning, Execution, and Incorporation of Operational and Science Objectives. Bulletin of the American Meteorological Society (published online ahead of print 2025), BAMS-D-24-0160.1. https://doi.org/10.1175/BAMS-D-24-0160.1