An ensemble of meteorological stations for estimating daily air temperature time series at Jungfraujoch since 1864

Air temperature is a pivotal factor influencing numerous chemical, physical, and biological processes. However, there is a notable scarcity of long-term data, especially at high elevations, exceeding 2000 m a.s.l. This study focuses on reconstructing the daily maximum, mean, and minimum temperatures at Jungfraujoch (3571 m a.s.l.) since 1864. The approach involves daily data from 10 meteorological stations within the ECA&D (6) and Meteo Swiss (4) databases. All selected stations are situated above 2000 m a.s.l. (in the range 2140-3109 m a.s.l.), providing uninterrupted observations spanning at least from 1961 to 2022. The methodology includes these steps: 1) for each meteorological station, in the calibration period 1980-1999, it was modeled the daily temperature at Jungfraujoch as the sum of the temperature at the selected station plus a deterministic and a stochastic component; the deterministic component is the product of the temperature lapse rate (TLR) and the elevation difference between the reference and selected station, and the stochastic component is a “noise” which comes from the statistical distribution of the residuals. The seasonality requires parameters with monthly variability which are different considering minimum, mean and maximum temperature. The calibration phase consists in the estimation of TLR and the statistical distribution of residuals (among Normal, GEV, Stable and Tlocscale distributions). The evaluation of model performances was based on the calculation of Pearson correlation coefficients (ρ_P) and Root mean squared errors (RMSE) within the two validation periods (1961-1979 and 2000-2022). High correlation coefficients (greater than 0.9 in both calibration and validation periods) and low values of RMSE (from 1.56°C to 3.32 °C in the calibration period and from 1.56°C to 3.42°C in the validation) confirm the model’s accuracy. The same high performances were found before (1961-1979) and after (2000-2022) the calibration period, for every meteorological stations. 2) Then the 10 simulated time series at Jungfraujoch were sorted according to the lowest values of the RMSE, and the first three was mediated to define an “ensemble” daily temperature time series, which was able to obtain these performances: (ρ_P=0.96,0.98,0.97; RMSE=1.97,1.46,1.68 °C respectively for max, mean and min temperature). The study was then extended from the year 1864. Comparing the results with the existing literature we highlighted: i) high performances without the need of modeling the observed trend due to the climate change (subjected to high uncertainty in the future), ii) very parsimonious model without the need of any other variables (relative humidity, cloud cover, wind velocity, weather patterns); iii) the importance of selecting high stations elevations (above 2000 m a.s.l.) rather than considering closer stations but subjected to the thermal inversion phenomena; iv) maximum temperature is affected by higher errors, especially from 2000-2022 which is probably due to the higher increasing of the summer and winter temperatures at high elevation accordingly to an elevation warming dependence; v) This method could be easily extended in many regions of the world and these results could be used to make a back ward analysis of many environmental processes (glacio-hydrological and permafrost), within the Jungfrau-Aletsch UNESCO World Heritage Site.