Automating air temperature Siting Classification of meteorological stations according to the World Meteorological Organization guidelines

Abstract

This research thesis investigated the automatisation of the air temperature Siting Classification (SC) according to its World Meteorological Organization (WMO) guidelines. Instead of going into the field, the Automated SC (ASC) developed in this research thesis uses the Actueel Hoogtebestand Nederland (AHN) and Basisregistratie Grootschalige Topografie (BGT) geospatial open data datasets to determine the classifications for air temperature. The algorithm was first tested at the Automatic Weather Station (AWS) in De Bilt, after which it was applied to all the other 33 AWS sites owned and maintained by the Royal Netherlands Meteorological Institute (KNMI). The ASC model was developed using the programming language of R and made use of a decision tree to determine an air temperature SC Class 1 through 5 for each AWS. The usability of the model was tested through visual validations based on photos or satellite imagery to check on its accuracy, and comparison validations based on the current Manual Siting Classification (MSC) procedure done by the KNMI. Results shows that an ASC model corresponds for 55.9% with the MSC values and that the model can be applied for a reproducible and consistent SC if the data is determined to be correct and if a manual validation of the outcomes is executed. The ability of performing an ASC can contribute for further research in improving the air temperature SC of existing, or determining completely new locations. Furthermore, the model can also be used to determine the air temperature SC of the vast of other sensors of third-party networks. Moreover, the model further substantiates the need to specify the different air temperature SC guideline criteria even more so that the surroundings are taken better into account.