The Arctic Samoylov site is extensively described in Boike et al. (2013) and more shortly in Gouttevin et al. (2017) where the present dataset is exploited.Air temperature and relative humidity were obtained from the Samoylov met-station using an HMP45A air temperature and humidity sensor . Unfortunately the sensor became saturated at temperatures below -40 °C From 2013-02-01 to 2013-03-15 air temperatures dropped below -40°C, and TTT in the dataset is replaced by the 2 m air temperature from the ERA-interim (ERA-i) reanalysis Project (Dee et al., 2011). The ERA-i data were linearly interpolated from their native 3 hourly temporal resolution (analysis and forecast fields) to a 30 min time-series. This substitution seems appropriate since for the rest of the 2012-2013 winter period, ERA-i temperatures show a high correlation with Samoylov observations (r2=0.97) and a low bias (-0.9°C). ERA-i fields were also proven to be a high quality source of driving variables to simulate the evolution of the Northern Eurasian snowpack including Siberia (Brun et al., 2013), with minor differences between station data and grid-field over large, rather flat areas like the Lena Delta. Finally, a comparison of ERA-i with locally acquired meteorological data from earlier years at Samoylov confirmed this validity for the skin surface temperature, which responds very sensitively to differences in the driving variables (Langer et al., 2013). Incoming shortwave and longwave radiation were measured at the Samoylov meteorological station with an NR01 Hukseflux 4-Component Net Radiation Sensor. Wind speed was measured at 3 m above ground with a 05103 Young Wind Monitor. Wind speed was (together with air temperature) the only meteorological field for which likely instrumental failure was detected, characterized by periods of a few hours to a few days with null wind-speed. These likely failures were not corrected for in the present dataset.Snow height was continuously recorded by an SR50 sensor (Campbell Scientific, ±1 cm accuracy, ±1 cm precision) located in the Samoylov reference polygon, in the topographically lower center (see Gouttevin et al., 2017 for details). For the "snow h" field provided here, the SR50 raw data was treated by a applying a moving average filter with a window size of 72 hours, thus removing spikes partly due to blowing snow events. Note this "snow h" field does not include the correcting factor used by Gouttevin et al., 2017. Raw SR50 data can be provided on request.