Impact of the North Sea–Caspian pattern on meteorological drought and vegetation response over diverging environmental systems in western Eurasia

Emerging drought stress on vegetation over western Eurasia is linked to varying teleconnection patterns. The North Sea–Caspian Pattern (NCP) is a relatively less studied Eurasian teleconnection pattern, which has a role on drought conditions and the consequence of changing conditions on vegetation. Between 1981 and 2015, we found that the Standardized Precipitation Index (SPI) and the Normalized Difference Vegetation Index (NDVI) have different trend patterns over various parts of western Eurasia. Specifically, the vegetation greenness is linked with wetter conditions over Scandinavia, and vegetation cover decreases over a drying central Asia. However, western Russia and Franceare paradoxically becoming greener under drier conditions. Using the Budyko framework, such paradoxical patterns are found in energy‐limited environmental systems, where vegetation growth is primarily promoted by warmer temperatures. While most studies focused on the impacts of the North Atlantic Oscillation (NAO), we test whether the NCP explains better the variability of meteorological drought and vegetation response over western Eurasia. We hypothesised that the positive phases of the NCP are correlated to high pressure anomalies over the North Sea, which can be associated with weakening onshore moisture advection, leading to warmer and dryness conditions. These conditions are driving vegetation greening, as western Eurasia is mainly energy limited. However, we show that as the climate is warming along with the teleconnection impacts, the future ecosystem over western Eurasia will be transferred from energy‐limited to water‐limited systems. This suggests that the observed vegetation greening over past three decades is unlikely to sustain in the future.

western Eurasia is mainly energy limited. However, we show that as the climate is warming along with the teleconnection impacts, the future ecosystem over western Eurasia will be transferred from energy-limited to water-limited systems. This suggests that the observed vegetation greening over past three decades is unlikely to sustain in the future.

Significance statement
Meteorological droughts affect vegetation cover differently across western Eurasia depending on water and energy supply. Droughts hinder vegetation development in water-limited areas, but promote vegetation growth in energy-limited areas due to rising temperatures. The teleconnections that influence precipitation and temperature patterns are linked to these patterns of vegetation responses. The North Sea-Caspian Pattern (NCP) explains more variance between meteorological drought and vegetation response over western Eurasia than the North Atlantic Oscillation (NAO).
The observed vegetation greening over the last three decades is unlikely to persist, given the effects of rising temperatures and NCP changes.  Lee et al., 2020), East Atlantic (EA; Ionita, 2014;Toreti et al., 2010;Toši c & Putnikovi c, 2021;Ulbrich et al., 2012), El Nino Southern Oscillation (ENSO; R. W. Lee et al., 2019;Mezzina et al., 2020) and South Asian monsoon (SAM; Alpert et al., 2006). Despite that teleconnections between these different modes of climate variability and hydroclimate conditions are well-documented in previous studies (Lled o et al., 2020;Rousi et al., 2020;Rust et al., 2018), the impact of the North Sea-Caspian Pattern (NCP) remains largely unexplored. The NCP is an upper-level atmospheric teleconnection with centres of active variability over the North Sea and the Caspian Sea regions. These seas are regional water bodies, which affect the common locations of local pressure systems due to seasonal circulation variations driven by temperature gradients caused by differences of thermal capacity between land and water (Dippner et al., 2012;Sibley et al., 2015). To date, NCP effects on temperature and precipitation have been studied mainly for the East Mediterranean and its surrounding regions Kutiel & Türkeş, 2005;Sezen & Partal, 2019). However, given that the NCP centres of action are farther north, it might strongly impact precipitation patterns, hence drought occurrence and vegetation conditions over the entire Eurasia (Brunetti & Kutiel, 2011;Ça glar et al., 2021).
Droughts are traditionally quantified by water availability indices, such as Standardized Precipitation Index (SPI) (Cancelliere et al., 2007;Livada & Assimakopoulos, 2007;McKee et al., 1993). Given the direct effects of droughts on vegetation, vegetation indices (van Hateren et al., 2021) have also been used as drought indicators (Buitink & Swank, 2020;Hu et al., 2019;Sepulcre-Canto et al., 2012;Di Wu et al., 2015). However, there are various factors, such as warming temperatures (Peng et al., 2011;X. Wang et al., 2017;G. Xu et al., 2014), increasing solar irradiation (Teuling, 2013), and increasing CO 2 fertilization (Lian et al., 2021), which may lead the vegetation to be less affected by the declining water availability. According to previous studies, the vegetation responses to drought conditions are divergent for different climate environments, and the similarity and differences in water and vegetation variability need to be further explored (Denissen et al., 2020;Peled et al., 2010;van van Hateren et al., 2021). To explore the long-term interactions among climate, regional hydrology and vegetation cover, which is related to the water-limited and energy-limited environment, the Budyko framework is applied (Abera et al., 2019;Li et al., 2013;D. Zhang et al., 2018).
In Section 2, we introduce the study area, data sources and methods. In Section 3, we examine the similarity and disparity in meteorological and vegetation variations over western Eurasia. Then, we test whether NCP can be a better contributor than NAO to explain changes in meteorological and vegetation conditions and describe the mechanisms associated with NCP teleconnections over various environmental systems. In Sections 4 and 5, we discuss their wider implications and summarize the main results.
2 | DATA AND METHODS

| Study area
The study is centred on western Eurasia (i.e., 10 W-70 E and 35 N-72 N; Figure 1). The western part of the study region has an oceanic/ maritime climate with temperate summers and mild winters, whereas the central and eastern parts have a continental climate with hot summers and cold winters (Stampoulis & Anagnostou, 2012). Due to different climate zones of western Eurasia, water and vegetation are unevenly distributed. The total amount of precipitation ranges from 600 to 900 mmÁyear À1 over the western part to 500 mmÁyear À1 in the eastern part of western Eurasia (Mikolaskova, 2009). The main vegetation cover types vary from woody savannas and mixed forest in the high latitudes to croplands and grassland in the mid-latitudes ( Figure 1).

| Normalized Difference Vegetation Index data
As a proxy of vegetation greenness, the Normalized Difference Vegetation Index (NDVI) is extracted from the National Oceanographic and Atmospheric Administration (NOAA) Global Inventory Monitoring and Modeling System (GIMMS), version number 3g.v1 (Cai et al., 2014;Pinzon & Tucker, 2014;Donghai Wu et al., 2014). This dataset provides bimonthly data ($every 14 days), with a 1/12 resolution, between July 1981 and December 2015. In order to validate the GIMMS NDVI datasets, we use the monthly NDVI dataset from the

Moderate
Resolution Imaging Spectroradiometer (MODIS MOD13C2), which is available at 0.05 resolution (Fensholt & Proud, 2012;Solano et al., 2010). For comparison, the GIMMS NDVI data were made to be monthly by averaging the two values in each month. We found that both datasets are largely similar in terms of seasonal and interannual variations ( Figure S1).

| Climate data
For hydroclimate variables (precipitation, potential evapotranspiration, temperature), we used the ERA5-Land data, which is derived from the fifth generation European Centre for Medium-Range Weather The land cover types over the western continental Eurasia between 10 W-70 E and 35 N-72 N Forecasts (ECMWF) atmospheric reanalysis of the global climate, with a spatial resolution of 0.1 Â 0.1 and monthly temporal resolution (Smith et al., 2021;Tarek et al., 2020). By using a higher spatial resolution (0.1 Â 0.1 ) than that of its driven climate reanalysis data (ERA5; 0.25 Â 0.25 ), ERA5-Land reanalysis datasets provide an improved representations of land surface processes (Muñoz, 2019).
To calculate the NAO and NCP indices, sea-level pressure (SLP) and 500 hPa geopotential height are derived from ERA5 datasets, with a spatial resolution of 0.25 Â 0.25 . Here, the NAO index is calculated using the 1st principal component (PC) of SLP anomalies over the North Atlantic (90 W-20 E, 20 -80 N) (Hurrell, 1995;Hurrell & Deser, 2009). The NCP index is the difference in the 500-hPa level geopotential height between the North Sea (0 , 55 N; 10 E, 55 N) and northern Caspian Sea (50 E, 45 N; 60 E, 45 N) region Sezen, 2017). Here, ERA5-Land and ERA5 data have been extracted between 1981 and 2015 for consistency with the NDVI datasets.
We provide a comparison of precipitation patterns between the ERA5 and observation data from the Climate Research Unit (CRU: https://crudata.uea.ac.uk/cru/data/hrg/) in Figure S2. Generally, the annual summer and winter precipitation trends derived from ERA5 are consistent with that from CRU, with drying trends over eastern part of western Eurasia and wetting trends over western part ( Figure S2). However, the CRU precipitation variability could be lower over the eastern part of western Eurasia, as the CRU dataset has fewer station data than ERA5 in some regions (Harris et al., 2020).
Since we focus on the mechanism explaining co-variations in the NCP and dry/wet conditions in western Eurasia, such potential shortcomings should be neglectable.

| Drought and vegetation indices
To quantify droughts and vegetation conditions, the SPI (McKee et al., 1993) and the NDVI are used. The SPI is calculated by assuming that precipitation is Gamma distributed (Chun, 2010). As the growing season generally ranges from April to September (total 6 months) over western Eurasia, we use a 6-month aggregated SPI (i.e., SPI-6) (van and NDVI, the NDVI is also standardized (hereafter called NDVI anomalies). The NDVI value less than À1 is also characterized as a significant vegetation decline.

| Trend and the GLS regression
The Mann-Kendall (MK) test is used to quantify the significance of linear temporal trends nonparametrically (Kendall, 1975;Mann, 1945).
Previous work argued that the results of the MK trend test could be misleading if serial correlations and outliers are ignored (Guo et al., 2018;Hamed, 2008;Hamed & Ramachandra Rao, 1998;Klaus et al., 2015;Sang et al., 2014). Thus, we use the modified MK test (Hamed & Ramachandra Rao, 1998) to examine trends of NDVI and hydroclimate variability. Trend intensity is estimated based on Thiel-Sen's slope, which is robust to outliers (Sen, 1968).
To explore the relationships between teleconnections and water and vegetation distributions, generalized least square (GLS) regressions are used, as in previous hydroclimate analysis (He et al., 2021;Klaus et al., 2015). The analytic GLS regressions can be written as where Y is the climate variables or vegetation index, and X is the NCP index value. β 0 and β 1 are regression coefficients.
To consider spatial autocorrelation and the problem of multiplicity of spatial trend and regression results, we used a global significance test based on false discovery rate (FDR) as recommended in previous studies (Wilks, 2006(Wilks, , 2016. After estimating the p values of trend and GLS regressions, the FDR test at p = 0.05 is used to control the expected proportion of locally significant tests that are actually true, i.e., not occurring by chance due to spatial autocorrelation and the problem of multiplicity. A local significance test is rejected if the local p value is no greater than: where p local ( j) denotes the jth smallest (out of k) local p values, and the α FDR is the chosen control level for the FDR (0.05 for this study).

| Budyko framework
The Budyko framework has been used to examine long-term interactions among climate, regional hydrology and vegetation cover (Abera et al., 2019; Li et al., 2013;D. Zhang et al., 2018). There are different types of Budyko framework (Budyko, 1974;Choudhury, 1999;Fu, 1981;H. Yang et al., 2008), which are summarized as follows: Fu Choudhury 1999 ð Þ;H:Yang et al: 2008 where ET, P and PET are land surface evapotranspiration, precipitation (water supply) and potential evapotranspiration (water demand), respectively, and ω is an empirical coefficient related to land patterns and characteristics.

| Impact of NAO and NCP on meteorological drought and vegetation
We examine the relationships between two teleconnection patterns (i.e., NAO and NCP) and regional drought conditions over western Eurasia (Figure 3). There are no significant relationships between the DI, EI and the NAO, whereas the NCP is positively related to the DI and EI over western Eurasia (Figure 3). This indicates that positive NCP phases may better contribute to drier conditions over the region than the NAO. Moreover, based on the Budyko space ( Figure S3b), the positive phases of NCP contribute to increasing dryness, thus causing the region to become more water-limited.
The NCP has significant negative impacts on precipitation over most western Eurasia, but positive impacts over central Asia ( Figure S4). Regarding the temperature, NCP shows a negative relationship over the whole Eurasia ( Figure S4). Such negative relationship with NCP is found over most parts of western Eurasia for PET, ET and NDVI variations. Generally, the NCP has negative impacts on meteorological and vegetation conditions over the region.
To further investigate how NCP affects regional drying/wetting conditions related to moisture circulations, we examine the regressed vertically integrated moisture flux associated with NCP variations ( Figure 4). Positive NCP phases appear to promote anti-cyclonic circulation over the North Sea (closely adjacent to the continental western Eurasia), but cyclonic circulation anomalies over the Caspian Sea ( Figure 4). Such circulation patterns are associated with drier condition over most part of the study region, but wetter condition over Central Asia.

| Trends in meteorological and vegetation conditions
The mean SPI-6 values are positive and near to zero, suggesting that there is generally slightly humid air over most western Eurasia

| Concomitance of meteorological drought and vegetation decline
To further explore the relationships between water deficits and vegetation growth, we examine the concurrence of meteorological drought and vegetation decline in Figure 6. On the basis of the SPI-6, we can identify five drought events that are particularly widespread (the area percentage is more than 50%; Figure S5) : 1988, 1996, 2003, 2006 and 2010. Among these widespread drought events, the 1996 drought event appears as the most severe one, and therefore, it is selected to analyse whether the meteorological drought and vegeta- Based on the above results (Figure 6), western Russia and France tend to become drier, but greener (Figure 5b,d). Here, we found that these regions are generally energy limited (Figure 7), indicating that vegetation growth is mainly restricted by energy supply, and not by water supply (Gokmen et al., 2013;Parsons & Abrahams, 1994).
Therefore, even though the region is drier, warmer temperatures promote vegetation growth. Scandinavia and Baltic regions are mainly water limited (Figure 7), and thus the vegetation is very responsive to water deficits, and the NDVI declines as soon as water deficits occur  (Wilks, 2006(Wilks, , 2016 4 | DISCUSSION 4.1 | The role of teleconnection patterns on vegetation during droughts NAO has been broadly accepted as the major mode causing hydrological and climate variability over western Eurasia (Lled o et al., 2020;Tsonis et al., 2008;W. Zhang et al., 2019). However, here, we demonstrated that NCP is a much better indicator for regional environmental changes over this region. The positive phases of NCP are related to an anticyclonic anomaly centred over the North Sea, and a cyclonic anomaly is centred over the Caspian Sea region. NCP is indeed associated with moisture flux divergence, causing dryness over most parts of western Eurasia. Moreover, the positive phases of NCP are found to promote the drier continental airflow from northern Asia that affect the study region (Kutiel & Benaroch, 2002), contributing to the increased dryness. Such a negative relationship between NCP and winter precipitation has been previously reported over Turkey (Sezen, 2017). Sezen (2017), however, suggested that NCP is positively correlated with summer precipitation in some parts of Turkey.
The difference of NCP impacts on precipitation could be related to seasonal variances. The negative relationships between NCP and precipitation in winter may be too strong to mask the positive relationship in summer. Therefore, we found that the NCP has negative impacts on total precipitation of the whole year in this study. The seasonal impacts of NCP on regional climate variability require further investigation in future research work. In addition, the NCP has significant impacts on temperature (Brunetti & Kutiel, 2011), which may also contribute to vegetation growth. In addition, future studies should also examine the potential contribution of other large-scale climate modes of variability.

| Similarity and disparity between drought conditions and vegetation patterns
Previous studies examined the relationship between droughts and vegetation growth in water-and energy-limited environments over Europe (Denissen et al., 2020;Peled et al., 2010;van Hateren et al., 2021). show results that are statistically significant at p ≤ 0.05 according to the MK trend test, and to the FDR global significance test (Wilks, 2006(Wilks, , 2016. For trend maps, the colorbar refers to the slope value of trend analysis vegetation to decline in response to such water deficits and thus NDVI values to decrease. However, this is not what we observed: NDVI values increased. France and western Russia are energy limited according to the Budyko framework ( Figure 5), which explains why vegetation is greening instead of suffering from the lack of water. In an energy-limited region, warmer temperatures promote vegetation growth, which is hypothesised to be due to enhancing photosynthesis activity (Dusenge et al., 2019;L. Xu et al., 2013;D. Yang et al., 2021).
However, if the precipitation deficits persist, we expect that vegetation growth will eventually be negatively affected. For instance, in 1996, the NDVI decreased in August and September over France after four months (April-July) of continuous precipitation deficits. The reason for the delayed vegetation responses might be due to large aquifer systems for water storage in this region (de Lavenne et al., 2021). Therefore, over France, the vegetation takes around four months before starting to decline after a precipitation deficit.
Moreover, there are mostly croplands over France, and the irrigation for agriculture may also be the reason for the delayed vegetation responses (Foudi & Erdlenbruch, 2012;Sidibé et al., 2012). will have a transition to water-limited ecosystems and be more sensitive to drought risk (Bhuiyan et al., 2017;Tello-Garca et al., 2020).

| Future implication for vegetation
Therefore, the greening trend of vegetation will be unlikely to be continued sustainably in the future. Similarly, Rajczak and Schär (2017) projected the decrease in precipitation in the future over central Europe and France, and the increased droughts negatively affect the forestry (Maracchi et al., 2005). However, based on the IPCC report

| CONCLUSIONS
In this study, we aim at exploring the similarity and disparity between F I G U R E 7 The Dryness Index pattern over WE between 1981 and 2015. Warm colours denote water-limited regions (i.e., DI > 1), and cool colours denote energy-limited regions (i.e., DI < 1)