Web-based verification of numerical model data using GIS, in-situ and remotely sensed observations

Verification of umerical prediction data is a time-consuming and somewhat tedious task requiring significant effort to align observation sites with model output. Automation, involving acquisition of in-situ observations, and appropriate georeferencing with regard to model data outputs is both feasible and straighforward. Use of this approach allows incorporation of the observations data into a geospatial database. Use of geospatially registered model output graphics, and geospatially registered in situ data, when incorporated into a geographic information system (GIS), allows display of both the predictive data and the temporally associated observational component. Finally, incorporation of remotely sensed data, including multispectral sensors, doppler weather radar, vertical doppler wind profilers and radiosondes allows another set of verifcation data which can be incorporated rapidly using GIS techniques. Finally, creating a web-enabled service capable of providing these data on demand, and allowing data retrieval using well-documented web-services for inter-institutional comparison.

Texas A&M University is engaged in modeling efforts using the community MM5 version 3.7 for predictions used in the Texas Air Quality Field Study (TexAQS-II). This study has high-impact areas in southeastern Texas (Houston-Galveston area) and north-central Texas (Dallas-Fort Worth). The goals of of this study include documentation of modeling skill, and precise field studies of particulate, organic and photochemical elements of airborne pollution. MM5 is run once daily, starting at 00Z, with hourly predictive output for 54 hours. Graphical output is created “on-the-fly” as each hourly output file is written. Three domains are currently employed: a 36 kilometer (km) coarse grid, an intermediate grid, 12 km, and a 4-km grid covering essentially Texas. Streaming in-situ observations are received over the intermediate and fine gridded domains from a variety of data providers, using the Unidata Local Data Manager (LDM). Texas A&M receives all available Level II radar data from the National Weather Service's (NWS) WSR-88D (Weather Surveillance Radar 88 Doppler) network. These data are cached for a period of 31 days.

The Texas Mesonet has been providing data using a web-mapping system based on on the University of Minnesota's Mapserver software, PostgreSQL, and PostGIS. We have evolved from creating shapefile-format GIS data files, to using on-the-fly data requests to the PostGIS database for maps and in-situ data. We have been utilizing a georegistered Level II mosaic image to present near-real-time radar data. Thus, we can now leverage this experience to provide map-layer data overlaid to promote near-real-time verification in a highly public, reviewable and scrutinizable manner.

This presentation will detail the methods and procedures used to present these data, and initial results of verification using these tools. A real-time, web-based demonstration will be available.