D.W. Meek, J.L. Hatfield, K.J. Cole, R.J. Jaquis, and J.H. Prueger USDA ARS MWA National Soil Tilth Laboratory 2150 Pammel Dr. Ames, IA 50011-4420
TEL: (515) 294-2246; FAX: (515) 294-8125; EMAIL: meek@nstl.gov
Abstract
Data quality control for observations from automated weather stations (AWS) is a critical part of routine operation. This paper reviews the experience of screening almost a decade long period of record for several stations via several automated and graphical procedures. The basis for the routines can be from nearby long term climatic record, physical theory, a second nearby station, and/or instrument specifications. The automated procedures are mostly dynamic adaptations of standard quality control methods for hourly and daily data. Minimally range limits and rate-of-change limits, both of which can be cast graphically, are possible for most meteorological data. Recording the maintenance history and values of regular independent sensor checks are highly recommend practices. Questionable vapor pressure values at or near saturation are the most common problem. The hourly solar radiation rule applied to a record at one site showed that sunrise values were regularly exceeding the extraterrestrial limit. A trained observer found that the site was close by to a large embankment, so following standard siting recommendations is essential. Climatic based rules are best developed from high quality databases like SAMSON. Bias problems were found with solar radiation data from two experiment stations. Climatic based rules are preferable to sensor based rules. Combining automated data processing rules with exploratory graphics works better than just using either procedure alone. Paneling graphs of related data, especially from the same sensor like wind speed and direction, is generally very insightful. The detection of a sensor drift requires and independent and regular record from either another close by station or the field maintenance log from hand held spot comparisons at the site in question. Data processing rules are only one part of trying to assure and assess the quality of AWS data but their routine use will result in a more reliable record as well as help with data exploration, further analysis, and modeling.
![](images/flag.png)
![](images/blue_ribbon.png)