Automated quality control of meteorological observations

Overview

The main automated quality control program that is used today in the area of meteorologocial observations is the MADIS program. MADIS stands for Meteorological Assimilation Data Ingest System.

History

To understand how MADIS works there needs to be some understanding of how quality control systems work and how they have progressed. One of the main and simplest forms of quality control is the check of probability (Gandin 1988). This check simply throws out observations that could not possibly exist. Examples of this would be if the dew point is higher than the temperature or if any of the numbers in the data fields are over acceptable ranges, such as if the temperature was recorded as over 200 degrees Fahrenheit. Another basic quality control check is to have the data compared to preset geographic extremes (DeGaetano 1996). Also having diurnal variation built into this as well. However this process only flags the data as uncertain because the station could be reporting correctly but there is no way to know. A better way of doing quality control checks is to have the system check with the previous observation as well as the other simple checks (Miller 1991). This method uses one hour persistence to check the quality of the current observation. This method makes continuity of observations better since the system is able to make better judgments on whether the current observations are bad or not.

Current

The current type of quality control such as systems such as MADIS use a three pronged approach (Graybeal 2003). This approach is much better mainly because it has more information to compare the current observation to. The first part of the process is the limit check. As already described the program checks whether the observation is within pre determined limits that are set according to whether they can physically exist or not. The second part is the temporal check which compares the station it its closest surrounding stations. The third part of the check is internal check. This check compares the observation to previous ones and sees whether it makes sense or not. It also takes into account present weather conditions so that the data is not considered bad just because the system is set for fair weather.

MADIS uses this current three pronged approach for its quality control tests. They are organized into three different levels of checks (Miller 2007). Level one is the validity tests, level two is the internal checks and also statistical spatial tests and level three is the spatial test. The level two statistical spatial test tests whether or not the station has failed any quality control check more than 75% of the time during the previous seven days. Once this has happened the station will continue to fail until it improves to failing only 25% of the time. The spatial check for the MADIS program also uses a reanalysis procedure. What this does is if there is a large difference between the station being checked and the station that it is being checked against then one of them is wrong. Instead of assuming that the station being check is wrong the program then moves onto the other stations that are near the one being checked. If the station that is being checked still is way off compared to most of the stations surrounding it then it is flagged as bad. However it the station is close to all of the other ones except for one then that one is bad.

References

DeGaetano, A., 1997: A Quality-Control Routine for Hourly Wind Observations. Journal of Atmospheric and Oceanic Technology, 14, 308-317.

Gandin,L., 1988: Complex Quality Control of Meteorological Observations. Monthly Weather Review, 116, 1137-1156.

Graybeal, D., A. DeGaetano, and K. Eggleston, 2004: Improved Quality Assurance for Historical Hourly Temperature and Humidity: Development and Application to Environmental Analysis. Journal of Applied Meteorology,43, 1722-1735.

Miller, P. and S. Benjamin, 1992: A System for the Hourly Assimilation of Surface Observations in Mountainous and Flat Terrain. Monthly Weather Review,120,2342-2359.