A drying Europe
The majority of Central Europe (CEU) and Mediterranean Europe (MED) showed a significant negative trend in the yearly mean soil moisture (SM) in the topsoil (0–7 cm; April to September), while Northern Europe (NEU) has showed substantial soil wetting as indicated by the positive trend in SM (Fig. 1a) from 1950–2021. Except for southwestern Finland, yearly mean VPD (April to September) showed a significantly increasing trend between 1950 and 2021, with parts of southern Spain showed the highest positive trend of more than 0.1 kPa/decade (Fig. 1c). Additionally, we also explored the trends of the yearly extreme thresholds of SM and VPD, i.e., yearly SM10P (10th percentile SM of each year) and VPD90P (90th percentile VPD of each year). The patterns of yearly SM10P and VPD90P trends (Fig. 1b & 1d) were spatially similar but more pronounced than those of the yearly mean SM and VPD trends (Fig. 1a & 1c). The trends of SM10P and VPD90P were about 35% and 80% higher than those of yearly mean SM and VPD between 1950 and 2021, respectively (indicated by slope of the linear regression in Figure S2). This indicated that the rate of intensification of extreme soil and air drying was higher than that mean drying. Therefore, we observed development of compound dry conditions characterized by both decreasing trend of SM and increasing trend of VPD across most of the CEU and MED over last 72 years (1950–2021).
Changes in extreme soil dryness and air dryness
Compared to the reference period, the SM10P threshold (indication of intensity of extreme soil dryness) of the present period in CEU and MED was typically 15–25% lower (i.e., intensity of extreme soil dryness increased by 15 to 25%). In contrast, the SM10P in NEU was about 10% higher than that during the reference period (Fig. 2a), implying a 10% decrease in intensity of extreme soil dryness. The spatial pattern of change in frequency was similar to that of change in intensity of extreme soil dryness (Fig. 2a, b). The frequency of extreme soil dryness increased 1.2-fold[0.8,1.6] (median[10th, 90th percentile] (Fig. 2b) across Europe (compared to the reference), with most of CEU and MED showed more than a 1.5-fold increase in frequency of extreme soil dryness. Both increased in intensity and frequency of extreme soil dryness was prominent for urban areas, croplands, as well as broadleaved and mixed forests (Fig. 2c, d).
Except for Finland, the VPD90P threshold (indication of intensity of extreme air dryness) of the present period was higher than that of the reference period across Europe (Fig. 3a). Overall, the increase in intensity of extreme air dryness across Europe was about 15%, with more than 50% increase in intensity for majority of MED. The spatial pattern of change in frequency was similar to that of change in intensity of extreme air dryness (Fig. 3a, b). The frequency of extreme air dryness largely increased across Europe (compared to the reference), with about 1.6-fold [1,2.3] increase (median[10th, 90th percentile] over Europe (Fig. 3b) and about one-quarter of Europe showing more than a two-fold increase in frequency of extreme air dryness during the present period in comparison to the reference period. Both increased in intensity and frequency of extreme air dryness was prominent (intensity > 20% and frequency > two-fold) for urban areas, croplands, as well as broadleaved forests (Fig. 3c, d).
Daily SM-VPD coupling
Compound extreme dryness, i.e., the co-occurrence of extreme soil and air dryness not only relate to changes in either SM and VPD contributing to the compound extreme, but also to the relationship between SM and VPD. Daily topsoil SM and VPD values were significantly negatively correlated, indicating strong (negative) SM-VPD coupling across most of Europe during the reference and the present period (Figure S3). Weak SM-VPD coupling [absolute r(SM,VPD) < 0.2] was observed at higher latitudes (> 65°N), particularly at higher elevations (NEU), and in the Alpine region (CEU; Figure S3a,b). Compared to the reference period (median r(SM,VPD) of -0.55), the present period showed a stronger SM-VPD coupling dependence (median r(SM,VPD) of -0.61), with more than 80% of Europe showing stronger SM-VPD coupling, largely consistent across land cover types (Fig. 4). This increase in strength of SM-VPD coupling during the present period was highest in NEU and the Alpine region of CEU (Fig. 4).
To quantify the impact on daily SM-VPD coupling on the frequency of occurrence of compound extreme dryness, we calculated the probability multiplication factor (PMF). The PMF indicated the increased probability (or frequency of occurrence) of compound extreme dryness compared to that expected when SM and VPD are independent (i.e., P = 0.1 × 0.1 = 0.01; see Methods). The PMF across Europe during the reference period was 3.6 [2.5, 4.2] (median [10th percentile, 90th percentile]), indicating that the frequency of co-occurrence of soil and air dryness (i.e., compound extreme dryness) during the reference period was 3.6 time more than if SM and VPD would have been independent (Figure S4a). As expected, due to increased SM-VPD coupling over large parts of Europe (Fig. 4), the PMF during the present period increased to 4 [2.9, 4.6] across Europe (Figure S4b). This increase in present day PMF compared to the reference PMF was largest over NEU, Alpine region, and southern Spain (more than 1.5-fold; Fig. 5a). However, across France and southern Italy, the PMF decreased during the present period (Fig. 5a). Among different land cover types, the highest observed increase in PMF was over shrublands and grasslands (mean of 1.2-fold; Fig. 5b). The relationship between daily SM and VPD coupling, as indicated by r(SM, VPD), and PMF for compound extreme dryness was largely linear, with an increase of PMF with increase in negative coupling in CEU and NEU (Fig. 6). However, in MED, we observed a decrease in PMF for r(SM, VPD) < -0.6 as shown in Fig. 6. Furthermore, the relationship between PMF and r(SM, VPD) was significantly different between reference and present period over MED (for r(SM, VPD) < -0.6) and NEU (for r(SM, VPD) < -0.2), with higher PMF values during the present period compared to reference period (Fig. 6). However, across CEU, the PMF vs r(SM, VPD) relationship remained unchanged during present and reference period (Fig. 6).
Change in frequency of compound extreme dryness
The probability of the occurrence, which indicates frequency, of compound extreme dryness (PCD) across Europe during the reference period is also equal to the PMF of the reference period, i.e., 3.5 ± 0.7% (mean ± sd) as shown in Fig. 7a. Using the SM10P and VPD90P thresholds from the reference period, the PCD increased to 6.0 ± 2.4% during the present period (Fig. 7b), thereby showing a 1.7-fold [0.9,2.5] (median[10th, 90th percentile]) increase overall across Europe and more than 2-fold increase for more than one-quarter of the European land area (Fig. 7c). The increase in PCD was highest in the MED (more than 4-fold increase), whereas a decrease in occurrence was observed in some areas of NEU (i.e., Finland, Ireland, and the western part of the UK), comprising about 12% of the study area (Fig. 7c). To understand if this increase in PCD was due to increase in SM-VPD coupling or due to decreasing SM and/or increasing VPD trend from reference to present period, we calculated ∆PCD due to SM-VPD coupling (ratio of PMF in present and reference period) and due to SM and VPD trend (ratio of PCD and PMF of reference period; see Methods). Our results indicate that the increase in PCD across CEU (excluding the Alpine area) and MED was dominantly due to the decreasing SM and/or increasing VPD trend from reference to the present period (Fig. 8a). Whereas, for much of the NEU and Alpine region in CEU, the ∆PCD was due to the increased SM-VPD from reference to the present period. Overall, the change in SM-VPD coupling (as shown in Figs. 4 & 5) resulted in a ∆PCD of 1.1-fold [0.9,1.4], whereas decreasing SM and/or increasing VPD trend resulted in a ∆PCD of 1.5-fold [0.9, 2.3] (Figures S5 & 8b). Among different land cover types, we observed a mean increase in the frequency by more than 2-fold over evergreen broadleaved forests, croplands, and urban areas during the present period in comparison to the reference period (Fig. 7d).
Future projections of compound extreme dryness
Climate projections indicated a further compound drying (both soil and air drying) trend across Europe. Compared to the reference period, the decrease in average SM10P across Europe was only marginal, i.e., 1%, 3% and 3.5% decrease during the present period, mid 21st century (2030–2065), and late 21st century (2066–2100), respectively (Figure S6a) with largest decrease in MED (Figure S7a). The VPD90P however showed an average increase across Europe by 12%, 35% and 68% compared to the reference period, during the present period, mid 21st century, and late 21st century, respectively, as simulated by the five RCMs (Figure S6b), with largest increase in CEU (Figure S7b). Furthermore, the ensemble means value (mean from all five RCMs) of the Pearson coefficient correlation – r(SM, VPD) indicated an significantly increasing SM-VPD from 1950–2100 (larger negative correlations from reference to future periods; Figure S8). The RCM models, however, underestimate the SM-VPD coupling as the mean correlation during reference and present period obtained from RCM models was − 0.33 and − 0.35 (Figure S8) which is significantly lower than correlation obtained from E-OBS and ERA5-Land data (-0.5 and − 0.54 during reference and present period, respectively as shown in Figure S3). Furthermore, the increased SM-VPD coupling simulated by the RCMs did not significantly increase the PMF of compound extreme dryness across MED, CEU and NEU, with similar PMF and r(SM, VPD) relationship from reference to future periods (Figure S9).
Owing to the underestimation of SM-VPD coupling in the RCM models, the PCD was also lower (2 ± 0.7% and 3.6 ± 2.1% during reference and present period, respectively; Figure S10a) than what was calculated based on the in-situ and reanalysis data (E-OBS and ERA5-Land) which was 3.5 ± 0.7% and 6.0 ± 2.4% during reference and present period, respectively (Figure S4). However, the change in PCD during present period in comparison to reference period, obtained by in-situ and reanalysis and the RCM ensembles was similar, i.e., 72% increase by the former dataset and 75% increase by latter dataset (Figure S10b). This similar increase in probability during the present period was also spatially consistent across different land cover types (Figures S10c).
For the future, the RCM ensembles showed a 3.4-fold [2.0,6.5] (median[10th percentile, 90th percentile]) increase in the frequency of compound extreme dryness across Europe during the mid 21st century (2030–2065 period) compared to reference period (Fig. 10a) with the largest increase in MED, some parts of NEU (high latitudes of Norway, Sweden, and Finland) and surrounding Alpine region in CEU (Fig. 9a). All land cover types are projected to experience on average of more than three times the frequency of compound extreme dryness by the mid 21st century as compared to the reference period, with the highest increase in probability for open shrublands (Fig. 9c). During late 21st century, the projections indicated a further increase in the frequency of occurrence of compound extreme dryness of 4.2-fold [2.0,10.8] in comparison to the reference period (Figure S10b), with a majority of spatial pattern similar to that of the mid 21st century (Fig. 9b). Only northern part of CEU (northern Germany and Poland) indicated decrease in frequency of compound extreme dryness during late 21st century in comparison to mid 21st century, mostly likely due to increase in SM10P (Figure S7a). Among different land cover types, open shrublands, grasslands and broadleaved forests are projected to experience more than five times more frequent compound dryness extremes during the late 21st century as compared to late 20th century (Fig. 9d). Finally, the increase in frequency of compound extreme dryness during mid 21st century and late 21st century compared to the reference period is entirely and dominantly driven by decreasing SM and/or increasing VPD trend from reference to future periods throughout Europe.