Paper on forecast improvements due to the use of dropwindsonde data released online in Weather and Forecasting – Hurricane Research Division

A regional hurricane model is used to conduct the most comprehensive  assessment of the impact of dropsondes on tropical cyclone (TC) forecasts to date. Dropsonde data can improve many aspects of TC forecasts if they are assimilated with sufficiently advanced techniques. Particularly notable is the impact of dropsonde data on TC-significant-wind-radii forecasts, since improving those forecasts leads to more effective TC-hazard forecasts.

Fig 1. (left) An image of one U.S. Air Force reconnaissance aircraft, (middle) an image of two NOAA reconnaissance aircraft, and (right) a cartoon of a dropsonde (courtesy NCAR Earth Observing Laboratory).

Reconnaissance aircraft (Fig. 1, left and middle) are flown into and around tropical cyclones (TCs) to collect data so that hurricane specialists can understand what is happening and for use in computer forecast models so that they have the best start to make the best forecast. One instrument, the Global Positioning Satellite dropsonde (Fig.1, right), is a single-use instrument launched out of aircraft to record data like pressure, temperature, humidity, and wind velocity as they fall to the Earth’s surface. They are valuable since they enable data collection in regions where it is too dangerous to fly (e.g., turbulent regions near the surface) and in regions that are not able to be “seen” by satellites (e.g., within clouds and at the surface). Previous research has generally found that, overall, dropsonde data improve TC track and intensity forecasts in a wide variety of computer models. Yet, a major shortcoming of those studies is that there are rarely enough forecasts to get a fully accurate picture of the impact of the dropsonde data, especially with year-to-year, TC-to-TC, and even forecast-to-forecast differences. 

To address these shortcomings, this study conducts the most comprehensive assessment of the overall impact of dropsonde data on TC forecasts of track, intensity, and significant wind radii to date. In doing so, it also represents the most comprehensive assessment of the impact of any airborne-observing system on TC forecasts to date. We compare two sets of computer forecasts: one that ingested, and another that did not ingest, dropsonde observations. These experiments used an experimental version of the  Hurricane Weather Research and Forecasting model (HWRF) and covered active North Atlantic basin periods during the 2017–2020 hurricane seasons. As the only difference between these experiments was whether dropsonde data were ingested, differences in forecasts highlight their impact. The study demonstrates the importance of using a large, diverse sample of forecasts and closely examining the forecasts when evaluating the impact of observing systems, new modeling systems, or model upgrades on TC-forecast performance.

Fig. 2. The impact of dropsonde data on TC forecasts by forecast lead time for five variables: (row 1) track – TRK; two measures of TC intensity: (row 2) maximum sustained wind speed at 10-m altitude – VMAX  and (row 3) minimum sea level pressure – PMIN; and three significant surface wind-speed radii reported by NHC: (row 4) 34-kt wind-speed radii – R34, (row 5) 50-kt wind-speed radii – R50, and (row 6) 64-kt wind-speed radii – R64. Shaded boxes indicate forecast lead times where results showed improvement (green) or degradation (rust).

Important Conclusions:

• Improving how observational data are ingested into forecast models and reducing limitations in observing systems would likely enhance dropsonde-data impacts.
• Directly sampling TCs with dropsondes can improve many aspects of TC forecasts if the data are ingested in the model using sufficiently advanced techniques (a process called data assimilation; Fig. 2).
• Forecasts of the extent of the tropical-storm-force (34-kt) wind and storm-force (50-kt) wind had the most consistent improvement from the dropsonde data. This is particularly notable since improving those forecasts leads to more effective TC-hazard forecasts
• Forecasts of the hurricane-force (64-kt) wind-speed radii were degraded by the dropsonde data if the area within 150 km of the TC center was not adequately observed.  This suggests ways to improve how TCs are sampled by aircraft in the future.

For more information, contact aoml.communications@noaa.gov. The full manuscript can be found at https://doi.org/10.1175/WAF-D-22-0102.1. This research was carried out in part under the auspices of the Cooperative Institute for Marine and Atmospheric Studies (CIMAS), a Cooperative Institute of the University of Miami and the National Oceanic and Atmospheric Administration, cooperative agreement NA20OAR4320472 while the lead author (Sarah Ditchek) was supported by the FY18 Hurricane Supplemental (NOAA Award ID NA19OAR0220188).