One weakness of the current AMDAR data set is the absence of data below 25,000 ft between major airline hubs and the almost complete absence of water vapor data. To address this weakness, a sensor called TAMDAR, developed by AirDat, LLC, under NASA sponsorship, has been deployed on approximately 50 regional turboprop aircraft operated by Mesaba airlines flying over the U. S. Midwest. Like the rest of the AMDAR fleet, TAMDAR measures winds and temperature. But unlike most of the rest of the fleet, TAMDAR measures humidity, turbulence, and icing. By 2008, AirDat expects to have more than 400 aircraft operating in the contiguous U. S. and Alaska.
Over the past 2.5 years, NOAA/ESRL/GSD has evaluated TAMDAR's data quality (as compared with traditional AMDAR measurements) and its impact on RUC forecasts. To measure TAMDAR impact, we run two identical RUC cycles in real-time: one with TAMDAR and one withoutotherwise both use the same input data. These cycles use up-to-date assimilation/model techniques (generally corresponding to the 13km RUC, but run at 20km resolution), and complete assimilation of all observation types (as used in the RUC13), including cloud analysis (GOES and METAR), full METAR assimilation with effects of boundary-layer depth, GOES precipitable water, all other aircraft, profiler (NOAA and 915-MHz boundary layer), and rawinsondes. With its hourly assimilation and full use of other observations, the RUC provides a stringent assessment for forecast value added by TAMDAR. The parallel models are strictly controlled to isolate the effects of TAMDAR data, including a resetting of common initial conditions every 48h to ensure a full control.
Forecast skill in these parallel RUC cycles is evaluated by verification against rawinsondes and surface observations. Results generally show that TAMDAR has a positive impact on 3h forecasts of temperature, wind, and relative humidity in the TAMDAR domain, during times when TAMDAR-equipped aircraft fly. More recently, we have studied the impact of TAMDAR on ceiling and precipitation forecasts, and will report these results at the conference.
In addition to the real-time paired cycles, we have run the RUC retrospectively for an active Midwest weather case (26 Nov. 2006 to 6 Dec. 2006) using a variety of TAMDAR data assimilation strategies. Because TAMDAR's error characteristics differ in some ways from traditional AMDAR data, we are studying how to optimally ingest TAMDAR data.
These TAMDAR RUC experiments, along with the case study investigations discussed by Szoke et al. at this meeting, have helped to improve the TAMDAR data quality and improve experimental forecasts for aviation.
We believe these studies are particularly important as the U.S. government considers paying a larger portion of the costs associated with aircraft-measured meteorological data. In this new era, the government will have to more carefully monitor the quality of data from a variety of aircraft fleets, as well as their impact on model forecast skill. We will show how our approaches can help facilitate this function.