Automated Quality Control of Daily Temperature Observations in Northern Climates
Abstract
High-quality observational data are the foundation of spatial analysis of climate variability and climate model evaluation. As the number of stations, instrumental technology, and monitoring frequency change, the challenge for data quality control increases. Many automated quality assurance/control (QA/QC) procedures have been proposed to increase the consistency of the observations and reduce human error when processing large numbers of observations. Those procedures lessen the amount of manual intervention or inspection needed, but they often have limitations when applied to real world observations with, for example, short observatory periods or missing periods of data, or are located in mountainous or northern continental areas. In this study, we further develop QC procedures that we apply to historical daily temperature observations from 1173 weather stations in British Columbia, Canada, located in both coastal and interior mountainous areas, for the 1991-2020 climate normal period. We demonstrate a modified approach to detecting climatological outliers that reduces the rate of identification of false positives and improves the consistency of true outlier detection by considering varying characteristics of observational data. These often arise because when QA procedures implicitly assume that day-to-day temperature variability is Gaussian in nature. In areas or seasons where the distribution of temperature variability is skewed, this has the effect of creating unequal numbers of false outliers or undetected outliers depending on the direction of the deviations. Therefore, one of the aims of our improvements is to be able to detect outliers in either direction with a similar probability of false detection. We also propose other modifications to ensure that the real phenomena, such as extreme events, are not removed and to enable the retention of shorter records, thereby increasing the validity of QC'd daily temperature observations and their spatial coverage.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.A45K1983W