EMC Experimental Rawinsonde Network Project Details

Experiments are currently underway at EMC to evaluate the impact on numerical forecasts of reducing the number of rawinsonde sites in the United States. One such experiment involves using the full rawinsonde network at 12z and no reporting sites at 00z. Another experiment involves using half the rawinsonde sites at 00z and the other half at 12z. The graphics on this page are designed for a third experiment, which uses NCEP ensemble forecast data to eliminate and select sites from a 50% US rawinsonde network at 00z and 12z. The motivation behind this method is to eliminate sites for the next raob reporting cycle in areas where the NCEP ensemble forecast indicates a relatively high degree of confidence, and add sites where the NCEP ensemble forecast indicates a low degree of confidence. The method used in this system is to take the spread of the 24-hour ensemble forecast for the 300 - 700 hPa thickness and, for each site, multiply the spread value at that site by a "rawinsonde density factor". This density factor gives an objective, quantitative description of how many other reporting rawinsonde sites are already nearby a given site; the greater the number of nearby sites that are reporting and the weaker the value of the ensemble spread at that site, then the less likely it is that that site would be needed to report for that cycle. Likewise, a site that has sparse rawinsonde coverage nearby and has a large ensemble spread value would be more likely to be picked to report for that cycle. The system iteratively eliminates the "least needed" sites for that cycle, and then once a specified number of sites have been eliminated, the system iteratively adds the "most needed" sites. The only sites included in this selection process are US NWS sites, and no changes are made to any US military, Canadian, Mexican or Caribbean sites for any of these experiments.

Results of forecasts using any of these methods are not yet available.

UPDATE (11 Sep 1997):

The experiments have now changed so that we start with a full raob network instead of a half-raob network, and only elimination of sites is done (no sites are subsequently added, as was done with the half-raob system). It remains an iterative procedure, and currently the system is set up to eliminate 20% of the network on a daily basis.