The Value of Weather Station Metadata

The Oklahoma Mesonet is a network comprising over 110 automated weather stations. During the past 11 years, Mesonet personnel have learned the crucial importance of obtaining accurate weather station metadata and have adopted several principles for maintaining and recording those data. The guidelines described here are essential to the end-to-end quality assurance (QA) system at the Oklahoma Mesonet.

There are strict guidelines for station names, geographic data, and land owner information. If a station is moved, even as few as 100 meters, a new station name is assigned to ensure an accurate climate record. Latitude, longitude, elevation, and legal description of the station are recorded. All contact information for landowners is obtained and updated as required. Another essential component of network metadata is a history of site photographs. Panoramic pictures are taken every five years and additional pictures are taken three times per year during routine maintenance visits. The naming convention of the digital photographs is self-descriptive and includes the site name, date, and panoramic direction toward which the picture was taken.

Sensor and equipment metadata include, but are not limited to, an associated serial number, vendor, manufacturer, model, and sampling interval. Historical calibration and coefficient data for sensors are also recorded. Visits to sites also provide a wealth of metadata. Whether for routine maintenance or for emergency repairs, each site visit is documented by a report that details the date and time of the visit, as well as the type of work performed. During routine site visits, sensor inter-comparisons are performed. Statistics describing the difference between the station sensor and reference sensor observations are noted in a digital report. For emergency site repairs, a unique “trouble ticket” is generated that describes the sensor problem and subsequent repair in detail.

Each measured variable (e.g., air temperature and pressure) has a number of associated automated QA tests that are performed in real-time. These tests and the accompanying sensor-specific thresholds are well documented. A unique QA flag is archived with each observation to indicate whether the data has been deemed “good”, “suspect”, “warning”, or “failure”.