A grid dataset of daily sunshine duration for Austria since 1961

Grid datasets of sunshine duration at high spatial resolution and extending over many decades are required for many quantitative applications in regional climatology and environmental change, including the modelling of droughts and snow/ice covers, the evaluation of clouds in numerical models, the mapping of solar energy potentials, and many more. Here, we present a new gridded dataset of relative sunshine duration, and derived absolute sunshine duration, developed for the territory of Austria at a grid spacing of 1 km, and extending back until 1961 at daily time resolution. The dataset complements the collection of SPARTACUS climate monitoring datasets, operationally produced by the national meteorological service of Austria. The presentation will outline the efforts made into a consistent station dataset, the methodological solution adopted in this complex topography region, and it will highlight key results from an evaluation with a user-orientated viewpoint.

The big challenges in the construction of the dataset were (a) issues of consistency in the available station data, notably the inhomogeneities related to changes in measurement devices and practices, and (b) limitations posed by the scarcity of long series and the high variation of cloudiness in the study region. The challenges were addressed by special efforts to correct evident breaks in the station series and by adopting an analysis method that combines the station data with information from satellite data. The methodology merges the data sources non-contemporaneously, i.e. using statistical patterns distilled over a short period, which allowed involving satellite data even for the early periods of our dataset.

The resulting fields contain plausible mesoscale structures, which could not be resolved by the station network alone. On average, the daily analysis explains 81 % of the spatial variance in daily sunshine duration at the stations (cross-validation). Comparison to other datasets showed that the new dataset meets a similar standard in temporal consistency, in spite of the limited data quality over such a long period. Our evaluation revealed a slight systematic underestimation (−1.5 %) and a mean absolute error of 9 %. The average error is larger during winter, at high altitudes and around the 1990s. We also find that the dataset exhibits a conditional bias, related to the involved uncertainties, which can lead to considerable systematic errors (up to 15 %) when calculating sunshine-related climate indices.