Table S1: List of models used in the study, including the GCMs and RCMs Figure S1: (Top) Evolution of the recorded crop yields of corn, maize, and grape in Moldova and (Bottom) the annual average anomalies for the current and previous 5 years. Figure S2: Evolution of the anomalies in AED calculated from the FAO56-Penman-Monteith and the Hargreaves equations. Figure S3: Spectral characteristics of precipitation time series recorded during the period 1853–2020 at Chisinau station. (a) Time series of monthly precipitation records; (b) Global wavelet periodogram expressed as percentage of total global power summed over all periods. Time-scales beyond the dominant annual time-scale are identified and magnified in the bottom figures; (c) Seasonal-annual response; (d) Short-term annual response; and (e) Long-term-annual response. Figure S4: Local wavelet power spectrum of the monthly precipitation time series. The vertical axis is the wavelet time-scale and the horizontal axis is the time position during the period 1853–2020 at Chisinau station. The thick white dashed curve depicts the cone of influence below which the edge effects on the amplitude of the local power spectrum are negligible.61 Table S2: Summary of the temporal variance fraction accounted by each dominant time-scale determined based on the local and global power spectra displayed in Figures S1 and S2. The variance fractions were obtained by applying the standard statistical variance formula to each wavelet time series. Percentages in parentheses give incremental percentage of such temporal variance as the sum of the time-scales. Figure S5: Temporal scale-by-scale decomposition of monthly precipitation time-series anomalies (mm/month). Positive and negative anomalies were calculated with respect to the mean values. The three nonoverlapping time-scales bands contain the major power peaks and were determined based on the global power spectra displayed in Figure S1. Seasonal-to-annual time-scales represent a time band of 0.083 to 1.5 years; short-term-annual time-scale depicts a time band of 1.5−10.5 years; and long-term-annual time-scale is related with a time band of 10.5−50 years. Figure S6: Evolution of the SPI in the five meteorological stations used in this study and from the average precipitation series. Figure S7: Monthly correlations between maximum temperature accumulated over different temporal scales and the crop yields of maize, sunflower, and grape in Moldova. Figure S8: Monthly correlations between Penman-Monteith reference evapotranspiration accumulated over different temporal scales and the crop yields of maize, sunflower, and grape in Moldova. Figure S9: Pearson's r correlations between the maize yield and the monthly meteorological variables accumulated over different temporal scales for the period 2000–2020. Figure S10: Pearson's r correlations between the sunflower yield and the monthly meteorological variables accumulated over different temporal scales for the period 2000–2020. Figure S11: Pearson's r correlations between the grape yield and the monthly meteorological variables accumulated over different temporal scales for the period 2000–2020. Figure S12: Evolution of annual precipitation, maximum and minimum temperature from 43 climate models considering the historical period (from 1970 to 2005) and the RCP8.5 scenario (2006−2100). Gray lines represent the individual models and color lines the average of the different models. Black line represents the regression line and dark colors the evolution of observations for Moldova. Figure S13: Distribution of monthly maximum temperature from the different climate models (gray lines) for the periods 1970–2020 and 2070–2100. Dashed black lines represent the average of the models and red line the observed temperature for the period 1970–2020.

Peer reviewed