| Summary: | Ground surface temperatures (GST) are widely measured in mountain permafrost areas, but their time series data can be interrupted by gaps. Gaps complicate the calculation of aggregates and indices required for analysing temporal and spatial variability between loggers and sites. We present an algorithm to estimate daily mean GST and the resulting uncertainty. The algorithm is designed to automatically fill data gaps in a database of several tens to hundreds of time series, for example, the Swiss Permafrost Monitoring Network (PERMOS). Using numerous randomly generated artificial gaps, we validated the performance of the gap‐filling routine in terms of (1) the bias resulting on annual means, (2) thawing and freezing degree‐days, and (3) the accuracy of the uncertainty estimation. Although quantile mapping provided the most reliable gap‐filling approach overall, linear interpolation between neighbouring values performed equally well for gap durations of up to 3–5 days. Finding the most similar regressors is crucial and also the main source of errors, particularly because of the large spatial and temporal variability of ground and snow properties in high‐mountain terrains. Applying the gap‐filling technique to the PERMOS GST data increased the total number of complete hydrological years available for analysis by 70 per cent (>450‐filled gaps), likely without exceeding a maximal uncertainty of ± 0.25 °C in calculated annual mean values. Copyright © 2016 John Wiley & Sons, Ltd.
|
|---|