Considerations on the use of third party data in spatial interpolation for climatological applications [article]

Cristian Lussana, Thomas Nils Nipen, Ivar Ambjørn Seierstad, Ole Einar Tveito, Ketil Tunheim, Line Båserud
2019 Figshare  
12th EUMETNET Data Management Workshop, 6-8 Novemebr 2019, De Bilt (The Netherlands)MET Norway is using amateur weather observations and remote sensing data to post-process numerical model output so to deliver better automatic forecasts for temperature and precipitation on The temperature post-processing is operational since March 2018. The precipitation post-processed datasets are currently shared with hydropower companies, which are assessing their potential for hydrological
more » ... ological applications. MET Norway has demonstrated that third party (or non-conventional) data can be used to improve the weather forecasts.Our contribution to the workshop begins describing the lessons learned so far in the use of a massive amount of non-conventional observations for monitoring the weather. Depending on the geographical area considered, the number of amateur observations can be up to two orders of magnitude larger than the number of conventional observations. New challenges must be addressed to effectively take advantage of non-conventional data. In particular, we will discuss the issues related to: data archiving; quality control and spatial interpolation.Public institutions have no control at all on third party data, as a consequence this data often lacks proper metadata. Furthermore, non-conventional stations can be easily relocated from one place to another. This poses a challenge on the data model used to store observations in climatological archives. We will present a possible solution, which is currently under development at MET Norway.Quality control is a key-component of any system that aims at making use of observations. For non-conventional data, the expected frequency of gross measurement errors is much larger than for observational networks managed by national MET services. Non-ideal sensor placements may cause significant representativity issues in the observation errors. On the other hand, the redundancy of non-conventional data allows the quality control to be extremely reliable in data-dense [...]
doi:10.6084/m9.figshare.10279637 fatcat:qcsfogvjt5dircna54xoiufwsy