*3.1. Precipitation and Temperature Variability*

The inter-annual variability in precipitation across the study area is high and frequently exceeded ±1 standard deviation (in 30% of the cases), less often ±2 standard deviations (5% of the cases), (Figure 2; for zoomed-in versions see Supplementary Information, Figure S5). Mean absolute deviation; for zoomed-in versions see Supplementary Information, Figure S5). Mean absolute deviation was 154 mm for annual precipitation, and negative precipitation anomalies were more frequent but had lower magnitudes as compared to the positive precipitation anomalies. The annual mean of all products was 656 mm, with a standard deviation (SD) of 197 mm and a coefficient of variation (CV) of 32%. The data products often, but not always, agreed on the direction of the anomaly (70% of the cases), but generally disagreed on the magnitude of the anomaly across all years (Figure 2). The precipitation products in greatest disagreement with the others were JRA-55 and MERRA-2 reanalysis. These products showed positive anomalies when most of the other products agreed on negative anomalies in 1993, 1995–1996 (both), 1999 (MERRA-2), 2000–2001, 2003, 2009–2010 (JRA-55), and 2013–2014, 2016 (MERRA-2), or negative anomalies in case of otherwise widespread agreement on positive anomalies in 1988 (both), 1989, 1994 (JRA-55), 2002 (MERRA-2), and 2015 (JRA-55). These two products also turned out the least correlated with other products and the measurement stations (Figure S3; correlations between 0.5 and 0.8). The greatest inter-product agreement was found in the years 1987, 1991 (negative anomalies), 1997 (positive anomaly), 2004–2005 (negative anomalies), 2006 (positive anomaly) and 2007–2008 (negative anomalies). The greatest disagreement was found in the years 1989, 1992–1993, 1995, 1998, and 2001, and the more recent years of 2010, 2012–2013, and 2015–2016.

**Figure 2.** Annual precipitation anomalies as conveyed by the different data products.

There was less variation in Tmin/Tmax compared to precipitation (mean = 29.97 ◦C, SD = 0.86 ◦C, CV = 1.74%; mean = 17.59 ◦C, SD = 0.30 ◦C, CV = 1.67% for Tmax and Tmin, respectively). The Tmin/Tmax products were more similar in inter-annual pattern than magnitude (Figure 3). KMD and MERRA-2 largely agreed both in terms of Tmax pattern and magnitude, whereas CRU and JRA-55 showed a similar pattern but lower values (see Table S1 for means and CV). For Tmin, the products were largely in agreement in terms of pattern, but not in magnitude, with mean Tmin decreasing in the order KMD, JRA-55, MERRA-2, and CRU (see Figure 3 and Table S1). The agreement in pattern could also be seen in the correlation analysis (Figure S4; all coefficients greater than 0.8).

**Figure 3.** Annual maximum (Tmax) and minimum (Tmin) temperature time series as conveyed by the different data products.
