All,

I am also not aware of systematic study that measured the value of radiosonde data on the human 
forecast process.  As Russ alluded to, observational data have been fundamental to the 
analysis and forecasting processing "since the beginning" of the Signal Corps days in 
the 1870s, and radiosondes became part of this process starting in the 1940s in the US.  
Given the historical short-term focus of SPC, observational data continue to play a key 
role in our forecasting process although NWP have clearly taken on increasing importance 
in recent decades.  The latter have allowed us to extend our forecast products farther out 
in time, but observational data continue to play key roles in short-term severe weather prediction.

One item that may be useful is the results from September 2013 NOSIA-II (NOAA Observing System 
Integrated Analysis) data call that SPC participated in.  This was a systematic survey across 
the NWS that evaluated the impact of various data sources, including NWP model and observational 
data, on key products and services.  For SPC, key products included Convective Outlooks, Mesoscale 
Discussions, Convective Watches, Watch Status Reports, and Fire Weather Outlooks. The Day 1 
Convective Outlooks were evaluated separately from those for Day 2 and beyond owing to the 
increased use of observational data during the Day 1 period.
 
A key part of the NOSIA analysis was to quantitatively estimate in a subjective manner the overall 
quality of each product and then assess the quantitative impact, if any, that the loss of a specific 
data source alone would have on the product quality.  This was done for 56 data sources that 
were considered relevant to the SPC forecast mission.  For the Day 1 Outlooks and Convective 
Watches, representative of important SPC products for the near-term when observational data 
would be expected to have the most impact, radiosondes were among highest rated sources of 
information.  Here are the specific results of the analysis for these two products:
 
Data Sources Having Highest Impact on Product Quality – 
Equal Impact Sources (Ties) Listed in Alphabetical Order
 
A. Day 1 Convective Outlook
Data Source and Reduction of Product Quality (%) Without Data Source
1. GOES Imager (50%)
2. Surface Observations (44%)
3 (tie). Atmospheric Models (38%; note – especially important for initial Outlooks)
3 (tie). UA Soundings – Radiosondes (38%)
3 (tie). Weather Radar - Reflectivity (38%)
 
B.  Convective Watches
Data Source and Reduction of Product Quality (%) Without Data Source
1 (tie). GOES Imager (40%)
1 (tie). SPC Hourly Mesoscale Analysis (40%)
1 (tie). Surface Observations (40%)
1 (tie). Weather Radar - Reflectivity (40%)
2 (tie). NEXRAD VAD Winds (27%)
2 (tie). UA Soundings – Radiosondes (27%)

Finally, here are two papers (one a bit old) that examined the role of various data sources 
on model performance and the value of radiosondes is documented.    

Zapotocny, Tom H., and Coauthors, 2000: A Case Study of the Sensitivity of the Eta Data 
Assimilation System. Wea. Forecasting, 15, 603–621.

doi: http://dx.doi.org/10.1175/1520-0434(2000)015<0603:ACSOTS>2.0.CO;2

Benjamin, Stanley G., Brian D. Jamison, William R. Moninger, Susan R. Sahm, Barry E. 
Schwartz, Thomas W. Schlatter, 2010: Relative Short-Range Forecast Impact from Aircraft, 
Profiler, Radiosonde, VAD, GPS-PW, METAR, and Mesonet Observations via the RUC Hourly 
Assimilation Cycle. Mon. Wea. Rev., 138, 1319–1343.

doi: http://dx.doi.org/10.1175/2009MWR3097.1

Steve W