Provenance & quality
Given the large geographic scale of the study, range of variables to be covered, long time series coverage and high resolution, the daily ERA-Interim reanalysis dataset, produced at the European Centre for Medium-Range Weather Forecasts (ECMWF), was selected as a source of climate data (Dee et al., 2011). Daily aggregates (means, maxima, minima, or totals) were generated from raw 3 or 6 hourly data for temperature, dew point temperature, soil moisture, snow depth, maximum wind gust and precipitation. Due to uncertainties in precipitation data from this source, interpolated gauge-based products from the Global Precipitation Climatology Centre (GPCC), which are independent of ERA data, were also used, although these were available only from 1988 onwards (Schamm et al., 2014).
Absolute NDVI values at calving/MME sites were obtained using a dataset of 7 day filtered and smoothed MODIS NDVI products from the University of Natural Resources and Life Sciences, Vienna (Vuolo et al., 2012). These data are included in the attached dataset but were not used in final models, they cover the period 2001-2016 and cover a subset of 94 sites.
NDVI anomaly data based on SPOT and PROBA-V data were supplied by the Université Catholique de Louvain-LifeWatch Wallonia-Brussels (WB) project (Radoux et al., 2015). Snow anomaly data (unusual values at the date of observation compared to mean probability of snow presence) were used to look for unusual variation in winter (e.g. winter length, severity, sudden cold shocks) and were derived by LifeWatch WB from filtered MODIS data (Rousseau et al., 2015). These datasets cover periods from 1998-2016 and 2000-2016 respectively and thus include only 2015 MME sites and a set of controls - covering a subset of 96 & 95 sites.