Statistical reconstruction of daily precipitation and temperature fields in Switzerland back to 1864

Abstract

Abstract. Spatial information on past weather contributes to better understanding the processes behind day-to-day weather variability and to assessing the risks arising from weather extremes. For Switzerland, daily resolved spatial information on meteorological parameters is restricted to the period starting from 1961, whereas prior to that local station observations are the only source of daily long-term weather data. While attempts have been made to reconstruct spatial weather patterns for certain extreme events, the task of creating a continuous spatial weather reconstruction dataset for Switzerland has so far not been addressed. Here, we aim to reconstruct daily high-resolution precipitation and temperature fields for Switzerland back to 1864 with an analogue resampling method (ARM) using station data and a weather type classification. Analogue reconstructions are post-processed with an ensemble Kalman fitting (EnKF) approach and quantile mapping. Results suggest that the presented methods are suitable for daily precipitation and temperature reconstruction. Evaluation experiments reveal excellent skill for temperature and good skill for precipitation. As illustrated with the example of the avalanche winter of 1887/88, these weather reconstructions have great potential for various analyses of past weather and for climate impact modelling.

Synthetic Weather Diaries: Concept and Application to Swiss Weather in 1816

“…For all analyses, we should note that rainfall is much more difficult to reconstruct by the analog method than temperature due to its very high spatial variability (Pfister et al, 2020). Moreover there is a large representativity error and arguably also a large observation error.…”

“…Note that a much improved reconstruction (relying on much more data and using a post-processing step that additionally corrects the best analog towards the observations) is available from 1864 onward (Pfister et al, 2020).…”

“…There are several causes for discrepancies: Errors in the reconstruction, errors in the observations, lack of representativity of observations for a grid cell, and inadequate translation. The error in the reconstructions can be partly assessed by comparison with similar analyses in a period after 1864, when better reconstructions are available (Pfister et al, 2020). In these reconstructions, the error of precipitation was assessed by subsampling in the 1961-2010 period.…”

“…In addition to data assimilation, providing global weather data at coarse resolution back to the early 19th century (Slivinski et al 2019), also other techniques such as analog resampling of regional weather fields (Caillouet et al, 2019;Devers et al, 2020;Pfister et al, 2020) have been used to reconstruct local daily weather 150-200 years back in time, with the potential to go even further back. These reconstructions provide a resource not just for climate science, but also for historians.…”

Synthetic weather diaries: concept and application to Swiss weather in 1816

“…In addition to data assimilation, providing global weather data at a coarse resolution back to the early 19th century (Slivinski et al, 2019), other techniques such as analogue resampling of regional weather fields (Caillouet et al, 2019;Devers et al, 2020;Pfister et al, 2020) have also been used to reconstruct local daily weather 150-200 years back in time, with the potential to go even further back. These reconstructions provide a resource not just for climate science but also for historians.…”