Amdar Optimization Studies at The Earth System Research Laboratory / Global Systems Division

Abstract

Commercial aircraft now provide more than 130,000 observations per day of temperature and winds aloft over the contiguous US. The general term for these data is AMDAR (Aircraft Meteorological Data Reports). See, e.g., Moninger et al., 2003 for more information about AMDAR. These data have been shown to improve both short- and long-term weather forecasts. Recently, additional atmospheric variables—water vapor, turbulence, and icing—have become available as well. While not yet ingested into numerical weather prediction (NWP) models, these new data are proving useful to weather forecasters. The costs of gathering these observations (primarily communications costs) are large and will increase as the number of observations increases. Before 2004 the US airlines providing the data bore nearly all of these costs. Since then the U.S. government has begun paying 50% of the costs and in the future will be asked to cover a larger share. Moreover, new companies with new business models are offering to sell to the government data from sensors and communication systems that the companies control. The new TAMDAR (Tropospheric AMDAR) sensor system (Daniels et al., 2006) is an example of this. As a result of these developments, there is considerable interest in finding the optimum number and spatial/temporal distribution of observations that need to be taken and transmitted (that is, purchased by the government) under various weather regimes. The Global Systems Division (GSD) of NOAA's Earth System Research Laboratory has, under FAA sponsorship, completed a study to begin to determine the optimal amount of data needed. We approached this problem in several ways. First, we studied how human forecasters currently use AMDAR data directly in their bench forecasting. Second, we studied how GSD's Rapid Update Cycle (RUC) forecast skill changed as we varied the amount of AMDAR data ingested. Third, we have begun to develop a method to measure the sensitivity of individual RUC analyses to various data inputs. Finally, we are developing a proposed grid (RUC) based, three dimensional, aircraft data need field, which might be used to target AMDAR observations to locations where those observations are expected to have the most impact on forecasts.