Introduction
Data Management
Data Quality Control
Data Analysis
Summary
Acknowlegements
Author Information
In order to perform these functions, AFCCC uses a database of 1.3 terrabytes. This database contains nine gridded data sets at varying resolutions (i.e. 2 1/2 degrees by 2 1/2 degrees) with eight complimentary point data sets having over 6000 specific observations sites and information from the surface to 60,000 feet above the earth. The data is stored in a relational database on an open architecture platform. Typical working extracts from the database are 300,000 to 5 million records in length with 5 to 15 separate variables or a file size of 50 to 150 Mbytes.
To process these data for management, analysis, or quality control several software applications are used. ARCVIEW GIS and Spatial Analyst are used as tools to view and manipulate both gridded and point data. Both tools enable the user to dynamically link multiple tables of raw and post-processed data in order to audit and analyze particular observation sites, geographic/political regions, or ad-hoc bounded areas.
To perform a complete data audit, an analyst will first search a metadata file for all the stations within his area of interest or stations which meet a certain atmospheric threshold. For example, a search may be done on all available reporting stations in Southwest Asia or sites throughout the world which report daily precipitation totals, since some sites may not normally report precipitation. Or a combination of the reporting stations in Southwest Asia which have precipitation data may be searched. In this case, the search is done by displaying and selecting the sites within Southwest Asia's political boundaries then querying the attributes of these sites to highlight locations with precipitation data. Once sites in Southwest Asia with precipitation are highlighted, an analyst may need to continue to audit the data to ensure he has enough records for a particular site or region. For instance, if he needed to compute the wettest months for Southwest Asia, he would need a large sample size of monthly precipitation records so the statistics would accurately represent the occurrences of precipitation over the years. For this type of audit, the analyst will either query or page through the attribute table's record counts and the period of record.
Once the audited sites are selected, they can then be exported as text delimited files to be reused by ARCVIEW for combing with other data files or used by additional statistical software.
A project requiring the identification of areas prone to lightning strikes around an Army base in central Georgia, Fort McClellan, used ARCVIEW and Spatial Analyst to grid over 500,000 lightning point data observations then summarize their frequency of occurrence. This enabled the analyst to see "pockets" of areas prone to strikes during the month of August.
Similarly, a project collected, displayed, and contoured point precipitation data over Eastern Russia. The analyst was quickly able to identify the dry continental climate regions over the interior of central Russia. These can be easily identified by the light blue shading. It was particularly interesting to note the area within this region, denoted by hash marks, receiving slightly more precipitation for the month of March. By performing cursory analysis like this, the analyst was able to identify then further interrogate the data to explain this embedded, slightly wetter area.
Another type of data visualization uses built-in symbology to display multiple attributes of observation sites. Wind data is normally summarized by combining direction and speed, wind barbs or wind vectors with speed values. This is done to easily identify general wind flows and their associated speeds. Wind data was summarized over the Persian Gulf to quickly get an idea of regional wind speeds and directions. This picture showed a general northerly surface flow over the central and western sections of the peninsula and an offshore flow along the southern coast.
Lastly, a project involving a comparison of gridded data sets used Spatial Analyst to quickly subtract the wind speeds and directional differences at each grid point to estimate model differences. This study involved comparing temperature and wind values at 1000 mb of an old numerical weather prediction model, HIRAS, with a newer numerical analysis technique, REANAL. The goal was to identify geographic areas, where the two models varied the greatest. To do this, first each parameter for both models were plotted independently. They were then subtracted from each other to produce a third set of graphs. Two of the parameters, wind speed and wind direction, differences are shown below as an example. From this example, it is easy to see HIRAS has a low wind speed bias with directional differences varying nearly 180 degrees in South America-an area where the models' differences seem most pronounced.
I wish to acknowledge Mr Michael Squires and Capt Matthew Doggett, AFCCC/CCX, for their help in preparing the graphs and products used in this paper.
Title and Office: DoD Modeling and Simulation Executive Agent for Air and Space, Air Force Combat Climatology Center, Air Force Weather Agency
Address: 151 Patton Avenue, Rm 120, Asheville, NC 28715
Phone, FAX, email: (704)-271-4403, Phone; (704)-271-4334, FAX; rabaydaa@afccc.af.mil