Homogenisation and interpolation of daily snow depth time series in Slovenia from 1950 to 2020

Slovenian Environment Agency (ARSO) has recently renewed the homogenisation and interpolation of daily snow depth time series in Slovenia. After conducting quality control, the data was aggregated into monthly sums, resulting in 242 time series of new snow and 278 of total snow depth. Homogenisation was performed using the HOMER software tool, separately for three climate regions: littoral, interior, and higher elevations. Inhomogeneities were identified and removed through several iterations using various detection methods and station metadata.

Most of the time series were found to be homogeneous, with only a few stations exhibiting more than one break. Approximately one-third of the breaks are of unknown origin, while nearly half are due to site relocation. The relative data correction for inhomogeneities ranged from 5–30 % for new snow depth and 7–36 % for total snow depth.

Following monthly homogenisation, the original daily data was adjusted based on the ratio between the monthly homogenised data and the original data. Missing daily values were replaced with spatially interpolated values. The results of ordinary spatial interpolation were partially adjusted to better match the probability density function of the last homogeneous period.

Since the procedure was conducted separately for new and total snow depth, discrepancies between the two variables emerged. Consequently, the daily time series of both variables were harmonized both between the variables and within the time series. For new snow depth, the precipitation sum was also considered as a control variable. We aimed to align the snow cover depth with the precipitation amount and the new snow depth, while also accounting for the settling of the snow cover. During the alignment process, we prioritized corrections of interpolated values over measured ones. Harmonization was performed iteratively, converging towards the final result.

Homogenised time series of new and total snow depth show a very large and statistically significant negative trend of approximately –8 % and –9 % per decade, respectively, over the period from 1950 to 2020 throughout Slovenia. In spring and autumn, the trend is more pronounced at higher elevations, whereas in winter and on an annual level, it is stronger in the mountains.

Homogenised and interpolated time series have been used to calculate climate normals for the latest WMO standard reference period (1991–2020) and are planned to be used to improve climate projections for Slovenia.

Keywords: snow depth, climate change, homogenisation, spatial interpolation