Producing a nearest neighbour check for station sunshine data as part of a station-satellite blended sunshine duration dataset for the UK

Sunshine duration (SD) is an important climate variable in the UK, with UK Met Office records extending back to the 1800s. However, the station network is spatially and temporally inhomogeneous, and the number of stations declined considerably between 1990-2010. A project to create a blended UK sunshine duration dataset is currently underway and aims to address these issues through combining satellite and in situ measurements, with the goal to improve spatial representation and reduce uncertainties. Initial comparisons between the satellite and station sunshine duration values identified some days where the agreement is poor. Further investigation highlighted that the satellite data sometimes misidentify snow-covered ground as clouds, leading to clear-sky days having low sunshine duration, and that both satellite and station data struggled with Saharan dust events. Additionally, the comparisons indicated that there were some erroneous station sunshine values, and this highlighted the need for a preliminary quality control check for the station data. This study presents a nearest neighbour check that identifies incorrect sunshine duration values in the station data. The check compares stations with their nearest eight neighbours within 80km, limiting neighbours to those with similar elevations and proximity to the sea. This version of the nearest neighbour check uses conservative parameters to minimise false positives, with future work aimed at increasing the flexibility of these parameters. In the process of developing the nearest neighbour check, six stations were identified that had around a decade of sunshine duration values offset by a day, resulting in them being out of step with their nearest neighbours. The correction of these stations in the Met Office database has resulted in improved quality of the UK historical climate record, highlighting the benefit of utilising both station and satellite measurements to identify errors in both datasets.