The West Texas Mesonet (WTM) currently maintains 44 automated surface observation sites that provide 232,320 daily observations. Observations include information about 15 meteorological parameters and 10 agricultural parameters at temporal resolutions of 5 minutes and 15 minutes, respectfully.  Such a volume of data requires automated and effective quality assurance tests to extract suspicious and/or bad observations and improve the overall quality of the real time and/or archived dataset. The WTM employs various tests including the: Range Test, Step Test, Like Instrument Test, Persistence Test and Spatial tests.  These tests and subtests provide different flag levels depending on the legitimacy of the observations. While an initial set of tests were developed, largely based on the historical efforts of the Oklahoma Mesonet, modifications are still under consideration in hopes of improving the performance of each test for the WTM domain.

 

When the QA/QC tests are performed on the 2002 dataset, 4,307 observations of various parameters are flagged as suspicious or failing.  This number represents ~0.01% of the 59,605,920 total observations.  The flagged observations are examined manually to confirm the level of authenticity of the observation. Additionally, various analyses are applied for unflagged observations of different parameters.  This process is conducted to verify the QA/QC process and seek improvements to the automated system. For instance, all of the unflagged 2002 wind speed observations at 2 m are plotted versus their 10 m counterpart in Figure 1. The plot indicates an expected band where wind speed at 10 m is greater than the wind speed at 2 m. On the other hand, some perturbations from the band take place at low wind speeds.  These data point represent time periods where the 5 minute average wind speed at 2 m is greater than the 5 minute average wind speed measured at 10 m. Closer examination of these observations indicate that majority occurred in December 2002 while temperature at 1.5 m and 9 m were below the freezing.  Review of the data and subsequent field observations revealed that the 10 m wind sensor froze up first relative to the 1.5 m sensor on many occasions.  Armed with this new information, the Like Instrument Test for wind speed at 2 m and 10 m were enhanced by examining the temperatures at 1.5 m and 9 m curing the same time periods.

 

Figure 1.  Scatter plot of unflagged 2 m versus 10 m wind speed measurements as observed by the WTM during 2002.