Diverse Responses of Vegetation Growth to Meteorological Drought

Agricultural and Forest Meteorology 262 (2018) 1–13

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology

journal homepage: www.elsevier.com/locate/agrformet

Diverse responses of vegetation growth to meteorological drought across T climate zones and land biomes in northern China from 1981 to 2014 ⁎ Hao-jie Xua, , Xin-ping Wangb, Chuan-yan Zhaoa, Xue-mei Yangc a State Key Laboratory of Grassland Agro-ecosystems, Key Laboratory of Grassland Livestock Industry Innovation, Ministry of Agriculture and Rural Aﬀairs, College of Pastoral Agriculture Science and Technology, Lanzhou University, Lanzhou, 730020, China b Shapotou Desert Research and Experiment Station, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou, 730000, China c State Key Laboratory Breeding Base of Desertiﬁcation and Aeolian Sand Disaster Combating, Gansu Desert Control Research Institute, Lanzhou, 730070, China

ARTICLE INFO ABSTRACT

Keywords: Improving our understanding of present and future impacts of drought on the vegetation in northern China is Climatic extremes heightened by expectations that drought would increase its vulnerability and subsequently accelerate land de- Dryland ecosystems gradation. The response of vegetation activity to drought and the underlying mechanisms are not well known. By Landscape vulnerability using the third-generation Normalized Difference Vegetation Index (NDVI) and the Standardized Precipitation Functional diversity Evapotranspiration Index (SPEI), we investigated the relationship between NDVI and SPEI, across different Water balance climate regimes and land cover types, and determined the dominant time-scales at which different biome types respond to drought during the period of 1981–2014. Our results showed that biome response is coupled with drought trends in most regions of northern China. The highest correlation between monthly NDVI and SPEI at different time scales (1–48 months) assessed the impact of drought on vegetation, and the time scales resulting in the highest correlation were an effective indicator of drought resistance, which was related to the interactive roles of mean water balance and divergent drought survival traits and strategies. Diverse responses of vegetation to drought were critically dependent on characteristic drought time-scales and different growing environments. This study highlighted the most susceptible ecosystem types to drought occurrence under current climate, including temperate steppes, temperate desert steppes, warm shrubs and dry forests. Given that drought will be more frequent and severe under future climate scenarios, it may threaten the survival of mesic ecosystems, such as temperate meadows, alpine grasslands, dwarf shrubs, and moist forests not normally considered at drought risk. We propose that future research should be focused on arid and semi-arid ecosystems, where the strongest impact of drought on vegetation is occurring and the need for an early warning drought system is increasingly urgent.

1. Introduction Ahlström et al., 2015; Huang et al., 2016). Therefore, understanding the response of dryland ecosystems to drought is important for assessing More frequent and severe drought has been forecast in the 21st vegetation vulnerability to climate extreme events (Smith et al., 2014) century, particularly in the mid-latitudes (Sheffield and Wood, 2008). and has strong implications for enhancing drought mitigation and Increases in drought are driven primarily by decreased precipitation preparedness (Wilhite et al., 2007). with increased evapotranspiration from higher temperatures (Trenberth Northern China is located in the mid-latitude East Asia with arid, et al., 2014). Drought is recognized as the world’s most costly and semi-arid and dry sub-humid regions accounting for approximately pressing natural hazard that influences water resource systems, agri- 34%, 27% and 16% of its total land area, respectively. The Taklimakan cultural production and natural ecosystems (Mishra and Singh, 2010). and Gobi Deserts constitute two major dust sources over East Asia Water availability acts as the main driver of vegetation distribution and (Tanaka and Chiba, 2006). The vegetation in northern China plays a productivity in arid and semi-arid regions (Neilson, 1995; Churkina and pivotal role in the prevention and control of land degradation (Wang Running, 1998). Evidence is accumulating that semi-arid ecosystems et al., 2015), which has a significant effect on the national ecological dominate the trend and inter-annual variability in the land CO2 sink, security (Miao et al., 2015). Soil water availability is a primary con- and are highly sensitive to drought trends (Zhao and Running, 2010; straint on the survival of sand-binding vegetation (Li et al., 2013).

⁎ Corresponding author. E-mail address: [email protected] (H.-j. Xu). https://doi.org/10.1016/j.agrformet.2018.06.027 Received 8 December 2017; Received in revised form 17 June 2018; Accepted 25 June 2018 0168-1923/ © 2018 Elsevier B.V. All rights reserved. H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13

Under global warming, severe and extreme droughts have been more (Begueria et al., 2010). The satellite-derived Normalized Difference frequent in the eastern and central portions of northern China since the Vegetation Index (NDVI) is considered a robust indicator for vegetation late 1990s (Yu et al., 2014; Zhang et al., 2016), and a wetting trend in greenness and vigor at regional and continental scales (Carlson and the western region has weakened since the 1980s (Ma and Fu, 2006). In Ripley, 1997). addition, an increased incidence of intensified drought is expected by The present work analyzed the relationship between NDVI and SPEI the middle of the 21st century (Wang and Chen, 2014; Leng et al., in northern China, across different climate regimes and land cover 2015). In this context, the need to evaluate the potential impact of types, and determined the drought time-scales at which different eco- drought on vegetation growth is becoming increasingly urgent system types respond to persistent water deficit. The objective of this (Reichstein et al., 2013). However, examining the impact of drought on study was to examine the drought responses and the drought resistance northern China’s vegetation is difficult given the great diversity of of northern China’s vegetation by using remote sensing data. As future landscape and climate. Drought characteristics in terms of magnitude, drought will be more intense and greater in spatial extent during the duration and spatial extent could be quantified by a drought index, 21 st century (Wang and Chen, 2014; Leng et al., 2015), this research allowing observation of water deficit through time and space (Vicente- identifies the most susceptible ecosystem types to meteorological Serrano et al., 2010). The application of remote sensing data shows drought and thus supports drought mitigation to reduce land de- great potential to establish the complicated relationship between ve- gradation (Wilhite et al., 2007). getation vigor and water availability at large temporal and spatial scales (McDowell et al., 2015). 2. Materials and methods Drought-related damage on vegetation can be characterized by slow growth, low productivity and an increase in plant mortality (Lotsch 2.1. Study area et al., 2005; McAuliffe and Hamerlynck, 2010; Zhao and Running, 2010). Xu et al. (2012) found that decreased vegetation growth of Northern China spans 31°23′ N–53°34′ N in latitude and 73°29′ northern China in the 2000s was associated with an increased fre- E–135°04′ E in longitude, with an area of approximately quency of extreme drought. Drought-induced water stress caused a 5.62 × 106 km2 (almost 58% of China's total land area). It comprises reduction in terrestrial gross primary production over northern China Northeast China, North China and Northwest China, covering 15 pro- from 1999 to 2011 (Yuan et al., 2014). Liu et al. (2013) indicated that vincial administrative units (Fig. 1a). From the east to the west, ele- an increase in tree mortality could be expected under future global vation increases from the North China and Northeast Plains to the Inner warming in the semi-arid region of northern China. Nevertheless, the Mongolian, Loess and Qinghai Plateaus. Temperature and precipitation magnitude of the response of vegetation to drought remains uncertain have a gradient from east to west part and from basins to mountains due to drought complexity and intrinsic drought sensitivity among mainly due to regional differences by longitude and topography (Li vegetation types (Smith, 2011). The intensity, duration and timing of et al., 2016). By classification criteria of temperature zone and arid/ drought partly determine the effect of drought on vegetation pro- humid region (Zheng et al., 2013), northern China can be divided into ductivity (Ruppert et al., 2015; Zeiter et al., 2016). In the eastern region 18 climatic sub-regions under the scheme of climate regionalization for of northern China, moderate drought with higher temperatures might the period of 1981–2010 (Fig. 1b). Rich climate diversity contributes to increase net primary production (NPP), while severe drought caused a a variety of vegetation communities and soil types. Grasslands and delayed response of NPP to precipitation (Sun et al., 2016). Lei et al. deserts are major land cover types (Xu and Wang, 2016). In this region, (2015) provided evidence that NPP reduction increased along a drought the desert is composed of sparse vegetation and closed to open shrubs. gradient in Inner Mongolia grasslands of China. The capacity for re- Deserts and grasslands dominate most of arid and semi-arid lands. sistance and resilience of land biomes to drought stress changes in Croplands are located in main parts of Northeast China and North different ecosystem types (West et al., 2012; Craine et al., 2013) and China, and in the eastern portion of Northwest China (Deng et al., environmental conditions such as climatic and edaphic factors (Vicente- 2006). Natural forests are distributed mainly in the Changbai, Greater/ Serrano et al., 2013; Gouveia et al., 2017). The vegetation in arid and Lesser Khingan and Qinling Mountains (Liu et al., 2013). Wide-spread semi-arid environments chronically exposed to water limitations shows desert and agro-pastoral ecotone are the most serious areas in northern high resistance to drought as well as great resilience when drought China that suffer from sandstorms (Zou and Zhai, 2004). To reduce and conditions end due to its morphological and phenological strategies combat sandstorms, the Chinese government has launched many eco- (Knapp et al., 2015). Grasslands were most responsive to precipitation logical restoration projects since the 1980s (Lu et al., 2015). The ve- deficit, followed by forests and deserts in the arid region of northern getation in the desertified areas of northern China still has a weak self- China (Xu and Wang, 2016). Hua et al. (2017) suggested that the ve- regulation capacity and poor stability at the present time (Wang et al., getation activity in the eastern region of northern China was more re- 2015). Accordingly, increased drought risk may have significant im- sponsive to drought than that in the western region, even in areas with pacts on vegetation establishment and recovery. the same land cover type. Despite an increasing recognition of the negative effects of drought 2.2. Data sources and preprocessing on natural ecosystems and food security (Barriopedro et al., 2012; Yuan et al., 2014; Zhang et al., 2016), few studies have focused on the way by To observe inter-annual variations of vegetation activity, we used which drought determines vegetation activity across climate zones and the third-generation NDVI (NDVI3g) data derived from the Advanced land biomes in northern China. Previous studies indicated that vege- Very High Resolution Radiometer (AVHRR) sensor by the Global tation types had different resistance to drought (Wang et al., 2017), and Inventory Modeling and Mapping Studies (GIMMS) group. This dataset there were diverse vegetation responses to drought in different regions was provided by the Ecological Forecasting Lab at NASA Ames Research over northern China (Hua et al., 2017). Moreover, the study into the lag Center (http://ecocast.arc.nasa.gov/). Currently, it has proven to be the effect of drought on vegetation productivity and the vegetation re- best NDVI product for the monitoring of long-term vegetation dynamics sponses to drought at various time-scales is scarce in northern China (Li in arid and semi-arid lands (Beck et al., 2011; Tian et al., 2015). The et al., 2015; Zhang et al., 2017). The concept of drought time-scale GIMMS NDVI3g dataset is a global bimonthly composite product that refers to the time lag between drought occurrence and its consequences spans from July 1981 to December 2014 with a spatial resolution of (Vicente-Serrano et al., 2013). The Standardized Precipitation Evapo- 8 km. It has been corrected for orbital drift, sensor calibration, viewing transpiration Index (SPEI) is a widely used multi-scalar drought index geometry, volcanic aerosols, atmospheric water vapor and cloud cover, based on meteorological data that quantifies different drought types, and other errors unrelated to vegetation change (Zeng et al., 2013; and captures the impact of increased temperature on evaporation Pinzon and Tucker, 2014). To further reduce the effects of cloud and

2 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13

Fig. 1. The spatial distribution of meteorological stations (a) and climate zones (b) over northern China for the period of 1981–2010. The climate zoning is extracted from Zheng et al. (2013).

3 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13 haze contamination, we aggregated the original NDVI series into scales was normalized at length into a log-logistic probability dis- monthly time steps by using the Maximum Value Composite (MVC) tribution to gain SPEI series. Details on the calculation of SPEI can be technique (Holben, 1986). For each pixel, the mean growing-season found at Vicente-Serrano et al. (2010). A negative value of SPEI sug- NDVI < 0.1 for the entire study period indicated barren and/or spar- gests water deficit, whereas a positive value indicates water surplus. sely vegetated areas, which were excluded from the analysis to mini- Compared with other common meteorological drought indices (e.g., the mize soil background influences and enhance green vegetation signals Palmer Drought Severity Index and the Standardized Precipitation (Liu et al., 2010). Previous studies demonstrated that there was an Index), the SPEI shows improved capability to measure drought impacts overall acceptable agreement between GIMMS NDVI3g and multi-sa- on agricultural and ecological response variables (Vicente-Serrano tellite NDVI products on regional to global scales (Fensholt and Proud, et al., 2012), but might overestimate the drought severity in arid and 2012; Marshall et al., 2016). semi-arid regions (Yang et al., 2017). Meteorological data, including monthly mean air temperature and We performed the Spearman correlation analysis between monthly precipitation, were collected from 350 stations across northern China values of NDVI and SPEI for time-scales from 1 to 48 months. The re- from 1977 to 2014 (Fig. 1a), and were obtained from the National lationship between NDVI and SPEI might be nonlinear, and the Climate Center of China Meteorological Administration (http://data. Spearman correlation coefficient identifies monotonic relationships (de cma.cn/). Given that 48-month scale SPEI was considered in this study, Beurs et al., 2015). For each grid cell, the maximum correlation coef- meteorological data of 4 years before the time span of GIMMS NDVI3g ficient (Max CorrelCoeff) was retained and also the SPEI time-scale at dataset were also analyzed. All meteorological data were processed into which the Max CorrelCoeff was obtained. A P-value < 0.05 was con- monthly raster layers at 8 km spatial resolution by using thin plate sidered significant for Spearman correlation coefficients. Occasionally smoothing splines, measuring the effect of topography on spatial cli- wet and dry conditions in summer were common over drylands because mate interpolation, such as elevation and aspect (Hutchinson, 1995). To of high evaporative demand and inter-annual variability in rainfall (von evaluate the performance of interpolated climate raster layers at the Wehrden et al., 2010). Temperature and solar radiation are key climatic monthly to seasonal scale, the Moderate Resolution Imaging Spectro- drivers of spring and autumn vegetation growth in northern China (Xu radiometer (MODIS) Land Surface Temperature (LST) product et al., 2017), and summer drought has a more dramatic impact on ve- (MOD11C3) at 0.05° spatial resolution from 2000 to 2014 (https:// getation productivity than spring and autumn droughts (Zhang et al., lpdaac.usgs.gov/) and the Tropical Rainfall Measurement Mission 2016; Hua et al., 2017). Furthermore, the lack of greenness may pre- (TRMM) product (TRMM 3B43) at 0.25° spatial resolution for the clude the application of NDVI in arid regions (Tueller, 1987). Over period of 1998–2014 (http://trmm.gsfc.nasa.gov/) were used. Both the northern China, most of the vegetated land surface (61%) had the MODIS LST and TRMM precipitation data were interpolated to match highest value of NDVI in August (Fig. S1). Vegetation activity is more the 8 km spatial resolution by using a bilinear interpolation technique responsive to soil water availability in reproductive growth stages than (Metzger et al., 2005). during leaf green-up and senescence periods (Ji and Peters, 2003). Digital elevation model (DEM) dataset with a spatial resolution of Hence, we correlated August NDVI with SPEI series for all the time- 1 km was downloaded from the NASA Shuttle Rader Topographic scales ranging from 1 to 48 months in each grid cell during the period of Mission (SRTM) website (http://www.glcf.umd.edu/). We resampled 1981–2014 to take short-, medium-, and long-term droughts into ac- the DEM data to 8 km spatial resolution, using a cubic spline method count. The subsequent analysis was performed with the main focus on (Xu et al., 2017). It supported spatial climate interpolation through the Max CorrelCoeff and SPEI time-scale for different climate zones and provision of topographical information. Land use data were the 500 m biome categories. MODIS land cover type product (MCD12Q1) produced for the year 2013 based on the University of Maryland (UMD) global vegetation 3. Results classification scheme derived from the Land Processes Distributed Ac- tive Archive Center (https://lpdaac.usgs.gov/). We resampled it to 3.1. Validation of the spatially interpolated climate layers match the 8 km spatial resolution by using a majority function in the Resample Tool of ArcGIS. The land surface in northern China was re- The spatial interpolation results of temperature and precipitation presented by the following 4 ecosystem types, including forests, grass- were compared with MODIS LST and TRMM precipitation products, lands, croplands and deserts. Separating the effect of land cover change respectively. General compatibility between the spatially interpolated on NDVI and its response to climate change, the 8 km AVHRR land climate layers and the satellite-derived products was observed among cover classification product was also employed, which was acquired seasons (Fig. 2). The seasonal correlation between interpolated tem- between 1981 and 1994 and was provided by the Global Land Cover perature and MODIS LST exceeds 0.89, and the correlation coefficient Facility (http://www.landcover.org/). At the global scale, the overall of interpolated precipitation and TRMM precipitation was 0.93, 0.95, accuracy of MODIS and AVHRR land cover type data reached 75% and 0.96 and 0.87 in spring, summer, autumn and winter, respectively. 65–82%, respectively (Hansen et al., 2000; Friedl et al., 2010). 3.2. Spatial patterns of the NDVI–SPEI relationship 2.3. Data analysis Biome responses were coupled with drought trends in most areas of The SPEI was used to indicate water balance at different time-scales. northern China (Fig. 3a). Large areas of northern China (43%) were The calculation of SPEI was based on a climatic balance between pre- dominated by a significant positive correlation between NDVI and SPEI cipitation and atmospheric evaporative demand, using monthly mean (e.g., most parts of North China and Northwest China). For 24.8% of air temperature and precipitation as the input data. Specifically, the northern China the significance level of the positive NDVI–SPEI corre- SPEI uses the monthly difference between precipitation and potential lations exceeded 0.01, such as the Inner Mongolian Plateau, the Hexi evapotranspiration (PET) to obtain simple climatic measurements of the Corridor, the Qaidam Basin, and northern Xinjiang. However, the area water surplus or deficit (D). The monthly PET was estimated at first with significant negative correlations between NDVI and SPEI was following the method of Thornthwaite (1948) that only required small (0.08%), mostly in the northern Greater Khingan Mountains. monthly mean temperature data, since spatial patterns of the relative The NDVI–SPEI correlations of each climate zone differed con- humidity, solar radiation and wind speed were unreliable. The second siderably (Fig. 4a). A positive correlation between NDVI and SPEI was step was to measure the accumulated D using time-scales between 1 and detected in most climate zones of northern China, except for the humid 48 months, for the purpose of assessing the response of vegetation to area of the cold temperate zone, where NDVI was negatively correlated water deficit of different duration. The value of D at different time- with SPEI. In addition, the vegetation in arid and semi-arid regions had

4 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13

Fig. 2. The validation of satellite-derived LST and precipitation variables with spatial interpolation results (a–d. temperature, e–h. precipitation) in different seasons. higher positive correlations than in sub-humid and humid ones re- 3.3. Spatial patterns of the drought time-scale gardless of different temperature environments. Positive correlations were particularly strong in the arid and semi-arid regions of the sub- The vegetation activity responded predominantly to short and temperate zone, having 68.5% of the vegetated land surface showing medium drought time-scales in northern China, although spatial significant relationships. This percentage was 42.8%, 45.7% and variability was high (Fig. 3b). August NDVI was responsive to the SPEI 54.2%, respectively, for the warm temperate, plateau sub-cold and time-scale of 1–12 months in 37.2% of the vegetated land surface, plateau temperate zones. It was also found that the highest correlation particularly in the Inner Mongolian Plateau, the central Loess Plateau, between NDVI and SPEI occurred in the semi-arid area of the above- and the arid western China. NDVI was positively correlated with SPEI at mentioned climatic sub-regions. medium time-scales (12–24 months) in most regions of the North China Noticeable differences between NDVI and SPEI correlations existed Plain, the Greater Khingan and Tianshan Mountains (28.0%). The re- for various biome types (Fig. 5a). Only areas with a constant land cover sponse of vegetation to long-term drought (> 24 months) was mainly type during the period of 1981–2014 were used to analyze the eco- found in the Northeast Plain, the Changbai and Lesser Khingan Moun- system types more prone to drought and the dominant time-scale at tains, the Qinling Mountains, the eastern Loess Plateau, and the Qinghai which drought affected vegetation (Fig. S2). The Max CorrelCoeff was Plateau (34.8%). higher for grasslands (median, 0.39) and deserts (median, 0.35), but Fig. 4b showed that, in a sub-temperate climate, the SPEI time-scale lower for croplands (median, 0.24) and forests (median, 0.14). The area decreased from humid to arid regions. In a warm temperate climate, the percentage of biome types with significant positive correlations be- SPEI time-scale was shorter in arid and sub-humid regions than that in tween NDVI and SPEI was 60.4% for grasslands, 51.4% for deserts, humid and semi-arid regions. The difference in the median values of the 29.4% for croplands, and 16.6% for forests. Furthermore, the lowest SPEI time-scale was not significant in plateau sub-cold regions. It was inter-quartile range was 0.22 for deserts, occupying the smallest pro- likely to identify general patterns that the humid biomes inclined to portion of the vegetated areas, while the highest inter-quartile range respond at longer SPEI time-scales than the arid, semi-arid and sub- (0.28) was observed for forests. humid biomes in a plateau temperate climate. Diverse responses of vegetation activity to water availability were The dominant time-scales at which drought affected different biome found in different climate regimes of northern China (Fig. 6). Although types were presented in Fig. 5b. The SPEI time-scale (unit: months) was NDVI was positively correlated with SPEI in most of the vegetated land longer for forests (median, 24), but shorter for croplands (median, 16), surface, the negative NDVI–SPEI correlation occurred in cold temperate grasslands (median, 14) and deserts (median, 12). The inter-quartile forests and grasslands. The sub-temperate and warm temperate forests range of the SPEI time-scale was 12–35 months for forests. The SPEI located in arid and semi-arid lands had a higher correlation between time-scales for the majority of the croplands and grasslands ranged NDVI and SPEI than that in humid and sub-humid regions. The same between 11 and 27 months, and between 10 and 25 months. Most de- pattern was seen in croplands. The influence of the SPEI was lower in serts had the SPEI time-scale of 4–25 months. humid and sub-humid regions (including sub-temperate and alpine The sub-temperate and warm temperate forests in arid areas re- grasslands) than in arid and semi-arid areas. Response of NDVI to SPEI sponded to drought at shorter time-scales when compared with those in for warm temperate grasslands displayed no significant differences humid areas (Fig. 7). With increasing aridity, the drought time-scale from arid to sub-humid regions. For the deserts in northern China, shortened for sub-temperate and warm temperate grasslands but vegetation activity in a sub-temperate region was more determined by showed no significant changes for alpine grasslands despite different drought than in warm temperate and cold regions. Additionally, semi- temperature and precipitation environments. This pattern was fairly arid deserts were more responsive to drought than arid deserts. evident in agricultural regions. There was a quicker reaction of desert vegetation to drought in arid regions than that in semi-arid regions

5 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13

Fig. 3. The spatial pattern of the maximum Spearman correlation coefficient between NDVI and SPEI for the period of 1981–2014 (a). The values of 0.34 and 0.44 indicate 5% and 1% significant levels of the correlation coefficient. The time-scales at which the maximum Spearman correlation coefficient exists between NDVI and SPEI are obtained (b). Climate region boundaries are shown in dark solid lines.

6 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13

Fig. 4. The NDVI–SPEI correlations (a) and the SPEI time-scales (b) varying in different climate regimes. Boxplot elements: box = values of 25th and 75th per- centiles; horizontal line = median; whiskers = ± 1 Standard Deviation (SD). regardless of a warm or cold climate. indicated by high positive correlations with SPEI. For grasslands, the highest positive correlation between NDVI and SPEI emerged in regions 3.4. Influence of water balance with water balance approaching zero, but in areas with the most positive or negative water balance, the positive NDVI–SPEI correlation The influence of water balance on the response of vegetation to was weaker in general. In semi-arid and sub-humid regions, the positive drought was analyzed (Fig. 8). The spatial pattern of annual average correlations between cropland NDVI and SPEI were relatively strong in water balance defined as the difference between precipitation and po- contrast with arid and humid regions. There was a clear gradient of tential evapotranspiration was established for the period of 1981–2014 drought impacts on the dynamics of desert vegetation as a function of (Fig. S3). The results suggested that the moist forests in a warm tem- the annual mean water balance. It was found that the drought legacy perate region with positive water balance had a low positive correlation effect on vegetation growth was more apparent as the average water between NDVI and SPEI. In cold temperate areas, little influence of the balance increased, when the positive NDVI–SPEI relationship was de- SPEI on moist forest NDVI was found, resulting in negative correlations. veloped independently. By contrast, the dependence of the SPEI time- Dry forests were distributed in semi-arid and sub-humid lands, where scale with water balance was greater for desert vegetation (R2 = 0.28) the control of vegetation growth by water availability was effective, as and forests (R2 = 0.23) than grasslands (R2 = 0.15) and croplands

7 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13

Fig. 5. The NDVI–SPEI correlations (a) and the SPEI time-scales (b) among biome types.

(R2 < 0.01). NDVI and SPEI might indicate the resistance to drought of a given ecosystem. In northern China, our finding showed that forests had the 4. Discussion strongest resistance to drought, followed by croplands, grasslands and desert vegetation (Fig. 5b). Over the temperate Northern Hemisphere, 4.1. Ecosystem resistance to drought Wu et al. (2018) found that forests represented a drought legacy response through to 4 years, while shrubs and grasses had the maximum Ecosystem resistance to drought could be characterized by the span drought legacy of 2 years and 1 year, respectively. Across the Medi- of time over which water deficit must persist before ecosystem variables terranean region, the shorter drought time-scale was observed for dry show negative anomalies (Begueria et al., 2010). In this study, the time- vegetation communities than the vegetation types over the temperate scale at which the maximum correlation coefficient existed between oceanic and continental regions (Gouveia et al., 2017). Forests are deep

Fig. 6. The NDVI–SPEI correlations across climate regimes and biome types. Dot is mean value, and error bar is ± 1/2 SD.

8 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13

Fig. 7. The SPEI time-scales across climate regimes and biome types. rooted species that can tap the water stored in deep soil layers under xylem systems of herbaceous plants are less resistant to drought for severe drought conditions (Davidson et al., 2000), and the time lag their low storage capacity for water and carbon (Craine et al., 2013). eﬀect in deep soil moisture with changes in precipitation may lead to Desert vegetation is located in hyper-arid to arid environments, where prolonged drought periods and hence delayed tree growth responses plants respond immediately to water shortage and prevent drought- (Anderegg et al., 2015). By comparison with forests, the shallow rooted related damage through physiological, anatomical and functional stra- grass normally absorbs soil moisture from top-middle layers that re- tegies (Vicente-Serrano et al., 2013). To summarize the resistance of sponds quickly to rainfall variability (Knapp et al., 2015), and the natural ecosystems to drought might be related to diverse rooting habits

Fig. 8. The relationships between the NDVI–SPEI correlations and the average water balance (a), and between the SPEI time-scales and the average water balance (b) for the cluster regions with diﬀerent ecosystem types.

9 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13 and associated water use strategies, and different hydraulic responses. Khingan Mountains, the Qinling Mountains, and the eastern Loess In the temperate region of northern China, we found a low dependence Plateau, most parts of these regions were climatologically humid and of the drought time-scale with mean water balance in agriculture re- were occupied by temperate meadows and moist forests, respectively, gions (Figs. 7 and 8b), indicating that the drought resistance of crop- exhibiting a prolonged response to drought. Regarding most alpine lands could be affected by anthropogenic activity. For instance, crop grasslands in the Qinghai Plateau, rainfall and the melting of accumu- improvement and water-saving irrigation technologies would increase lated snow and frozen soil contributed to high water availability and water use efficiency and drought tolerance capacity (Deng et al., 2006). slow drought responses (Fig. S3). Our results suggested a negative effect Climate and vegetation characteristics might lead to differences in of drought on the growth of alpine grasslands (Fig. 6). This was in line drought resistance among ecosystem types (Fig. 7). We found that dry with previous researches, which provided evidence that experimental forests were less resistant to drought than moist forests over the tem- warming-induced environmental drying significantly reduced above- perate region of northern China. Our results also revealed a quicker ground biomass and NPP in alpine grasslands on the Tibetan Plateau reaction of desert vegetation growth to drought in arid environments (Klein et al., 2007; Fu et al., 2018). In contrast with temperate steppes, than in semi-arid environments. Recent studies indicated that the alpine grasslands responded to drought at longer time-scales, possibly ability of woody plants to survive from periods of sustained drought because of different hydroclimatic conditions (Fig. 7). was largely determined by their embolism resistance (McDowell et al., 2008). It is common that woody plants growing in areas of deficient 4.2. Drought impacts on vegetation activity rainfall have low embolism resistance and therefore experience high risk of hydraulic dysfunction (Choat et al., 2012). Isohydric woody Our finding that northern China’s vegetation was sensitive to me- plants maintain constant water potential by reducing stomatal con- teorological drought (Fig. 3a) was consistent with previous studies ductance under drought conditions, whereas anisohydric woody plants (Yuan et al., 2014; Zhang et al., 2016; Hua et al., 2017), and the keep their stomata open and photosynthetic rates high for longer per- drought impacts (assessed using the maximum correlation coefficient iods, even at very low water potential. Indeed, woody plants growing in between NDVI and SPEI) were most prominent in a semi-arid climate more arid environments usually show isohydric stomatal behavior that (Fig. 4a). Higher correlations between vegetation vigor and drought maintains stem water potential above the critical level by limited gas also emerged over Northern Hemisphere semi-arid regions with lower exchange during drought (Pivovaroff et al., 2016). In contrast with total available water capacity due to decreased water storage (Ji and isohydric species, woody plants displaying anisohydric stomatal beha- Peters, 2003; Wu et al., 2018). Moist forests in a temperate region could vior in mesic ecosystems maintain gas exchange at the expense of de- be tolerant to drought and cold climates (Sun et al., 2016; Zhang et al., creasing water potential (Roman et al., 2015). The rapid response to 2017), as indicated by negative correlation coefficients of NDVI with water deficit showed by isohydric woody plants allows them to avoid SPEI in the northern Greater Khingan Mountains. Precipitation could drought stress and minimize risk of xylem cavitation, but may lower contribute to soil erosion, which decreased soil organic matter content, their ability to survive moderate stress of prolonged drought. Thus, and hence moist forest growth. In addition, photosynthesis activity woody plants in xeric environments (e.g. dry forests) were more vul- might decrease during the wet season in response to a decrease in in- nerable to fatal xylem embolisms and have lower survival rates than coming photosynthetically active radiation and vice versa. Our results anisohydric woody plants (e.g. moist forests) under severe drought. further revealed that drought impacts were largely determined by dif- Drought legacy effects on grassland growth in the sub-temperate and ferences in ecosystem types (Fig. 5a). The response of vegetation ac- warm temperate zones were more prominent with increasing water tivity to drought was most evident in grasslands as herbaceous plants balance (Fig. 7). However, this pattern was not obvious in alpine were the vegetation types more prone to drought due to shallow root grasslands, causing a comparatively weak dependence of the drought systems and low xylem hydraulic resistance (Craine et al., 2013; Knapp time-scale with water balance (Fig. 8b). Over northern China, tempe- et al., 2015). Warm shrubs were on average less responsive to drought rate meadows mainly appear in humid and sub-humid regions, where stress than grasslands since they showed a differentiating water use perennial aquatic or marsh plants are the dominant species. It was pattern, with increased use of surface soil moisture under non-drought thought that temperature was the principle factor for growth of tem- conditions and a tendency of absorbing water from deeper soil during perate meadows (Piao et al., 2006). Abundant precipitation and rela- drought (West et al., 2012). Deserts and grasslands dominated most of tively low evaporation might account for the tardy response of alpine arid and semi-arid lands, which might account for the most prevalent grasslands to drought (Zhang et al., 2017). Temperate steppes are drought impacts in these regions. It appeared that forests and croplands dominated by herbaceous perennials occurring in a semi-arid climate, were less influenced by drought and thus the drought impacts in humid while temperate desert steppes are mainly composed of annual and and sub-humid regions were weak. perennial herbs in an arid area. Annual herbs in most cases respond With an increasing drought time-scale, we found that the more rapidly to severe drought than perennial grass species (Copeland NDVI–SPEI correlations were higher in forests and croplands but lower et al., 2016). In summary, xylem structures and stomatal regulation in grasslands and deserts (Fig. 9). These results suggested that forests strategies might illustrate the differences in drought resistance within and croplands must suffer from extreme and prolonged drought con- forests and shrubs in northern China. The resistance of different ditions. Wider safety margins in temperate forests did not imply that grassland types to drought might be dependent on different growth they were immune to the threat of hydraulic failure induced by drought environments and the effect of lifespan. (Anderegg et al., 2015). Pervasive forest growth declines since the mid- The spatial heterogeneity of ecosystem resistance to drought was 1990s have been detected in the semi-arid region of northern China (Liu attributable to differences in hydrological background and vegetation et al., 2013). Forest ecosystems across a wide range of rainfall en- composition. In the Inner Mongolian Plateau, the central Loess Plateau, vironments were similar in their vulnerability to drought (Choat et al., and the arid western China, the vegetation activity had a rapid response 2012). Severe and sustained drought between May 2009 and July 2010 to drought (Fig. 3b). Most of the above-mentioned regions appeared in caused a widespread decline in vegetation activity in the large cropland arid and semi-arid climates (Fig. 4b), where temperate steppes, tem- sectors of northern China (Barriopedro et al., 2012). Yuan et al. (2014) perate desert steppes and warm desert shrubs were the main ecosystem found a dramatic decrease in maize yield in northern China attributed types. The vegetation activity showed an intermediate response to by a multiyear precipitation reduction from 1999 to 2011, particularly drought occurrence in the North China Plain, the Greater Khingan and in Northeast China. Regarding grasslands and deserts, short-term Tianshan Mountains, where large areas of the vegetated land surface drought exerted stronger influence on vegetation growth than long- were dominated by crops and dry forests with moderately positive term drought as they comprised drought-avoiding species (Fig. 5b). mean water balance. In the Northeast Plain, the Changbai and Lesser All vegetation types followed a general pattern that drought impacts

10 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13

Fig. 9. The relationship between the NDVI–SPEI correlations and the SPEI time-scales in various ecosystem types. were more profound in a semi-arid region but become relatively weak agriculture (Fig. 9). These findings will offer general guidance on eco- with mean water balance approaching more positive or negative logical assessments and coping strategies for drought on vegetation. (Fig. 8a). We also found that, even in regions dominated by the same However, the vegetation climatology analysis in humid regions should ecosystem type, drought impacts changed in different climate regimes be interpreted with caution (Fig. 3). Further progress in exploring the (Fig. 6). Dry forests were more responsive to drought over the tempe- mechanisms underlying differences in drought vulnerability across rate northern China. Drought impacts were more significant in tempe- ecosystem types requires long-term field experiments and process-based rate steppes and desert steppes than temperate meadows and alpine modeling studies (Smith et al., 2014; Knapp et al., 2015; McDowell grasslands due to water availability. Rain-fed agriculture might be more et al., 2015). affected by drought than irrigation agriculture as there were differences in crop types, seasonal precipitation dependence and hydrological conditions (Deng et al., 2006; Gong et al., 2017). The vegetation ac- 5. Conclusions tivity in arid deserts was less responsive to drought than in semi-arid deserts, probably by the use of groundwater, root biomass allocation, The maximum NDVI–SPEI correlation and the corresponding SPEI stem water storage and reduced leaf area (McDowell et al., 2008; time-scale were effective for modeling the response and resistance of Vicente-Serrano et al., 2013; Pivovaroff et al., 2016). Based on the northern China’s vegetation to drought, indicating that remotely sensed analysis of the response and resistance of northern China’s vegetation to metrics could be used in a prospective mode in examining the effects of meteorological drought, the most susceptible ecosystem types to water drought on vegetation activity at large spatial scales. The drought-re- stress under current climate comprised temperate steppes, temperate sistant capacity was closely associated with water balance and vege- desert steppes, warm shrubs and dry forests. Because of the suscept- tation characteristics (e.g., the diversity of drought survival traits and ibility of these ecosystem types to drought occurrence and their im- strategies). The impact of drought on vegetation activity was largely portance in reducing sandstorms and preventing land degradation (Zou determined by differences in drought resistance among ecosystem types and Zhai, 2004; Miao et al., 2015), the need for an early warning and drought stress levels. Our results highlighted the dramatic impacts drought system covering arid and semi-arid ecosystems is emphasized. of drought on the arid and semi-arid ecosystems of northern China. We Given that drought has the potential to become more frequent and se- suggest that management strategies in these regions are urgently vere under future climate scenarios (Wang and Chen, 2014; Leng et al., needed to maintain ecosystem services, such as windbreak and sand 2015), we hypothesize that it would threaten the survival of temperate fixation. As future drought will be more frequent and severe in large meadows, alpine grasslands and dwarf shrubs, and moist forests across areas of northern China under climate change scenarios, its impacts northern China. In the absence of human management, an increase in might further expand to humid and sub-humid regions, resulting in drought risk could be expected for croplands, especially for rain-fed widespread declines in vegetation growth.

11 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13

Acknowledgements Huang, L., He, B., Chen, A., Wang, H., Liu, J., Lu, A., Chen, Z., 2016. Drought dominates the interannual variability in global terrestrial net primary production by controlling semi-arid ecosystems. Sci. Rep. 6, 24639. This work was supported by the Fundamental Research Funds for Hutchinson, M.F., 1995. Interpolating mean rainfall using thin plate smoothing splines. the Central Universities [grant numbers lzujbky-2018-4], the National Int. J. Geogr. Inf. Sci. 9, 385–403. Natural Science Foundation of China [grant numbers 41530750], and Ji, L., Peters, A.J., 2003. Assessing vegetation response to drought in the northern Great Plains using vegetation and drought indices. Remote Sens. Environ. 87, 85–98. the Natural Science Foundation of Gansu Province [grant numbers Klein, J., Harte, J., Zhao, X., 2007. Experimental warming, not grazing, decreases ran- 17JR5RA061]. We thank anonymous referees for their comments that geland quality on the Tibetan Plateau. Ecol. Appl. 17, 541–557. significantly improved this paper. Knapp, A.K., Carroll, C.J.W., Denton, E.M., La Pierre, K.J., Collins, S.L., Smith, M.D., 2015. Differential sensitivity to regional-scale drought in six central US grasslands. Oecologia 177, 949–957. Appendix A. Supplementary data Lei, T., Wu, J., Li, X., Geng, G., Shao, C., Zhou, H., Wang, Q., Liu, L., 2015. A new framework for evaluating the impacts of drought on net primary productivity of – Supplementary material related to this article can be found, in the grassland. Sci. Total Environ. 536, 161 172. Leng, G., Tang, Q., Rayburg, S., 2015. Climate change impacts on meteorological, agri- online version, at doi:https://doi.org/10.1016/j.agrformet.2018.06. cultural and hydrological droughts in China. Glob. Planet. Change 126, 23–34. 027. Li, X., Zhang, Z., Huang, L., Wang, X., 2013. Review of the ecohydrological processes and feedback mechanisms controlling sand-binding vegetation systems in sandy desert regions of China. Chin. Sci. Bull. 58, 1483–1496. References Li, Z., Zhou, T., Zhao, X., Huang, K., Gao, S., Wu, H., Luo, H., 2015. Assessments of drought impacts on vegetation in china with the optimal time scales of the climatic – Ahlström, A., Raupach, M.R., Schurgers, G., Smith, B., Arneth, A., Jung, M., Reichstein, drought index. Int. J. Env. Res. Pub. He. 12, 7615 7634. M., Canadell, J.G., Friedlingstein, P., Jain, A.K., Kato, E., Poulter, B., Sitch, S., Li, R., Wang, J., Zhao, T., Shi, J., 2016. Index-based evaluation of vegetation response to – Stocker, B.D., Viovy, N., Wang, Y.P., Wiltshire, A., Zaehle, S., Zeng, N., 2015. The meteorological drought in Northern China. Nat. Hazards 84, 2179 2193. dominant role of semi-arid ecosystems in the trend and variability of the land CO Liu, S., Wang, T., Guo, J., Qu, J., An, P., 2010. Vegetation change based on SPOT-VGT 2 – sink. Science 348, 895–899. data from 1998-2007, northern China. Environ. Earth Sci. 60, 1459 1466. Anderegg, W.R.L., Schwalm, C., Biondi, F., Camarero, J.J., Koch, G., Litvak, M., Ogle, K., Liu, H., Park Williams, A., Allen, C.D., Guo, D., Wu, X., Anenkhonov, O.A., Liang, E., Shaw, J.D., Shevliakova, E., Williams, A.P., Wolf, A., Ziaco, E., Pacala, S., 2015. Sandanov, D.V., Yin, Y., Qi, Z., Badmaeva, N.K., 2013. Rapid warming accelerates Pervasive drought legacies in forest ecosystems and their implications for carbon tree growth decline in semi-arid forests of inner Asia. Glob. Change Biol. 19, – cycle models. Science 349, 528–532. 2500 2510. Barriopedro, D., Gouveia, C.M., Trigo, R.M., Wang, L., 2012. The 2009/10 drought in Lotsch, A., Friedl, M.A., Anderson, B.T., Tucker, C.J., 2005. Response of terrestrial eco- China: possible causes and impacts on vegetation. J. Hydrometeorol. 13, 1251–1267. systems to recent Northern Hemispheric drought. Geophys. Res. Lett. 32, L06705. Beck, H.E., McVicar, T.R., van Dijk, A.I.J.M., Schellekens, J., de Jeu, R.A.M., Bruijnzeel, Lu, Y., Zhang, L., Feng, X., Zeng, Y., Fu, B., Yao, X., Li, J., Wu, B., 2015. Recent ecological fl L.A., 2011. Global evaluation of four AVHRR-NDVI data sets: intercomparison and transitions in China: greening, browning, and in uential factors. Sci. Rep. 5, 8732. assessment against landsat imagery. Remote Sens. Environ. 115, 2547–2563. Ma, Z., Fu, C., 2006. Some evidence of drying trend over northern China from 1951 to – Begueria, S., Vicente-Serrano, S.M., Angulo-Martinez, M., 2010. A multiscalar global 2004. Chin. Sci. Bull. 51, 2913 2925. drought dataset: the SPEIbase: a new gridded product for the analysis of drought Marshall, M., Okuto, E., Kang, Y., Opiyo, E., Ahmed, M., 2016. Global assessment of variability and impacts. Bull. Am. Meteorol. Soc. 91, 1351–1354. vegetation index and phenology lab (VIP) and global inventory modeling and map- – Carlson, T.N., Ripley, D.A., 1997. On the relation between NDVI, fractional vegetation ping studies (GIMMS) version 3 products. Biogeosciences 13, 625 639. ff cover, and leaf area index. Remote Sens. Environ. 62, 241–252. McAuli e, J.R., Hamerlynck, E.P., 2010. Perennial plant mortality in the Sonoran and Choat, B., Jansen, S., Brodribb, T.J., Cochard, H., Delzon, S., Bhaskar, R., Bucci, S.J., Mojave deserts in response to severe, multi-year drought. J. Arid Environ. 74, – Feild, T.S., Gleason, S.M., Hacke, U.G., Jacobsen, A.L., Lens, F., Maherali, H., 885 896. Martinez-Vilalta, J., Mayr, S., Mencuccini, M., Mitchell, P.J., Nardini, A., Pittermann, McDowell, N., Pockman, W.T., Allen, C.D., Breshears, D.D., Cobb, N., Kolb, T., Plaut, J., J., Pratt, R.B., Sperry, J.S., Westoby, M., Wright, I.J., Zanne, A.E., 2012. Global Sperry, J., West, A., Williams, D.G., Yepez, E.A., 2008. Mechanisms of plant survival convergence in the vulnerability of forests to drought. Nature 491, 752–755. and mortality during drought: why do some plants survive while others succumb to – Churkina, G., Running, S.W., 1998. Contrasting climatic controls on the estimated pro- drought? New Phytol. 178, 719 739. ductivity of global terrestrial biomes. Ecosystems 1, 206–215. McDowell, N.G., Coops, N.C., Beck, P.S.A., Chambers, J.Q., Gangodagamage, C., Hicke, Copeland, S.M., On, S.P.H., Latimer, A.M., Damschen, E.I., Eskelinen, A.M., Fernandez- J.A., Huang, C., Kennedy, R., Krofcheck, D.J., Litvak, M., Meddens, A.J.H., Muss, J., Going, B., Spasojevic, M.J., Anacker, B.L., Thorne, J.H., 2016. Ecological effects of Negrón-Juarez, R., Peng, C., Schwantes, A.M., Swenson, J.J., Vernon, L.J., Williams, extreme drought on Californian herbaceous plant communities. Ecol. Monogr. 86, A.P., Xu, C., Zhao, M., Running, S.W., Allen, C.D., 2015. Global satellite monitoring of – 295–311. climate-induced vegetation disturbances. Trends Plant Sci. 20, 114 123. Craine, J.M., Ocheltree, T.W., Nippert, J.B., Towne, E.G., Skibbe, A.M., Kembel, S.W., Metzger, M.J., Bunce, R.G.H., Jongman, R.H.G., Mücher, C.A., Watkins, J.W., 2005. A fi Fargione, J.E., 2013. Global diversity of drought tolerance and grassland climate- climatic strati cation of the environment of Europe. Glob. Ecol. Biogeogr. 14, – change resilience. Nat. Clim. Change 3, 63–67. 549 563. Davidson, E.A., Verchot, L.V., Cattanio, J.H., Ackerman, I.L., Carvalho, J.E.M., 2000. Miao, L., Moore, J.C., Zeng, F., Lei, J., Ding, J., He, B., Cui, X., 2015. Footprint of research fi – Effects of soil water content on soil respiration in forests and cattle pastures of eastern in deserti cation management in China. Land Degrad. Dev. 26, 450 457. – Amazonia. Biogeochemistry 48, 53–69. Mishra, A.K., Singh, V.P., 2010. A review of drought concepts. J. Hydrol. 391, 202 216. de Beurs, K.M., Henebry, G.M., Owsley, B.C., Sokolik, I., 2015. Using multiple remote Neilson, R.P., 1995. A model for predicting continental-scale vegetation distribution and – sensing perspectives to identify and attribute land surface dynamics in Central Asia water balance. Ecol. Appl. 5, 362 385. 2001-2013. Remote Sens. Environ. 170, 48–61. Piao, S., Mohammat, A., Fang, J., Cai, Q., Feng, J., 2006. NDVI-based increase in growth Deng, X.P., Shan, L., Zhang, H.P., Turner, N.C., 2006. Improving agricultural water use of temperate grasslands and its responses to climate changes in China. Glob. Environ. – efficiency in arid and semiarid areas of China. Agric. Water Manage. 80, 23–40. Change 16, 340 348. Fensholt, R., Proud, S.R., 2012. Evaluation of earth observation based global long term Pinzon, J.E., Tucker, C.J., 2014. A non-stationary 1981-2012 AVHRR NDVI3g time series. – vegetation trends-comparing GIMMS and MODIS global NDVI time series. Remote Remote Sens. 6, 6929 6960. ff Sens. Environ. 119, 131–147. Pivovaro , A.L., Pasquini, S.C., De Guzman, M.E., Alstad, K.P., Stemke, J.S., Santiago, Friedl, M.A., Sulla-Menashe, D., Tan, B., Schneider, A., Ramankutty, N., Sibley, A., Huang, L.S., 2016. Multiple strategies for drought survival among woody plant species. – X., 2010. MODIS collection 5 global land cover: algorithm refinements and char- Funct. Ecol. 30, 517 526. acterization of new datasets. Remote Sens. Environ. 114, 168–182. Reichstein, M., Bahn, M., Ciais, P., Frank, D., Mahecha, M.D., Seneviratne, S.I., Fu, G., Shen, Z., Zhang, X., 2018. Increased precipitation has stronger effects on plant Zscheischler, J., Beer, C., Buchmann, N., Frank, D.C., Papale, D., Rammig, A., Smith, production of an alpine meadow than does experimental warming in the Northern P., Thonicke, K., van der Velde, M., Vicca, S., Walz, A., Wattenbach, M., 2013. – Tibetan Plateau. Agric. For. Meteorol. 249, 11–21. Climate extremes and the carbon cycle. Nature 500, 287 295. Gong, Z., Zhao, S., Gu, J., 2017. Correlation analysis between vegetation coverage and Roman, D.T., Novick, K.A., Brzostek, E.R., Dragoni, D., Rahman, F., Phillips, R.P., 2015. climate drought conditions in North China during 2001-2013. J. Geogr. Sci. 27, The role of isohydric and anisohydric species in determining ecosystem-scale re- – 143–160. sponse to severe drought. Oecologia 179, 641 654. Gouveia, C.M., Trigo, R.M., Begueria, S., Vicente-Serrano, S.M., 2017. Drought impacts on Ruppert, J.C., Harmoney, K., Henkin, Z., Snyman, H.A., Sternberg, M., Willms, W., ’ vegetation activity in the Mediterranean region: an assessment using remote sensing Linstaedter, A., 2015. Quantifying drylands drought resistance and recovery: the data and multi-scale drought indicators. Glob. Planet. Change 151, 15–27. importance of drought intensity, dominant life history and grazing regime. Glob. – Hansen, M.C., Defries, R.S., Townshend, J., Sohlberg, R., 2000. Global land cover clas- Change Biol. 21, 1258 1270. ffi sification at 1km spatial resolution using a classification tree approach. Int. J. Remote She eld, J., Wood, E.F., 2008. Projected changes in drought occurrence under future Sens. 21, 1331–1364. global warming from multi-model, multi-scenario, IPCC AR4 simulations. Clim. Dyn. – Holben, B.N., 1986. Characteristics of maximum-value composite images from temporal 31, 79 105. AVHRR data. Int. J. Remote Sens. 7, 1417–1434. Smith, M.D., 2011. An ecological perspective on extreme climatic events: a synthetic fi – Hua, T., Wang, X., Zhang, C., Lang, L., Li, H., 2017. Responses of vegetation activity to de nition and framework to guide future research. J. Ecol. 99, 656 663. drought in northern China. Land Degrad. Dev. 28, 1913–1921. Smith, A.M.S., Kolden, C.A., Tinkham, W.T., Talhelm, A.F., Marshall, J.D., Hudak, A.T., Boschetti, L., Falkowski, M.J., Greenberg, J.A., Anderson, J.W., Kliskey, A., Alessa, L.,

12 H.-j. Xu et al. Agricultural and Forest Meteorology 262 (2018) 1–13

Keefe, R.F., Gosz, J.R., 2014. Remote sensing the vulnerability of vegetation in nat- drought: a key to enhancing drought mitigation and preparedness. Water Resour. ural terrestrial ecosystems. Remote Sens. Environ. 154, 322–337. Manag. 21, 763–774. Sun, B., Zhao, H., Wang, X., 2016. Effects of drought on net primary productivity: roles of Wu, X., Liu, H., Li, X., Ciais, P., Babst, F., Guo, W., Zhang, C., Magliulo, V., Pavelka, M., temperature, drought intensity, and duration. Chin. Geogr. Sci. 26, 270–282. Liu, S., Huang, Y., Wang, P., Shi, C., Ma, Y., 2018. Differentiating drought legacy Tanaka, T.Y., Chiba, M., 2006. A numerical study of the contributions of dust source effects on vegetation growth over the temperate Northern Hemisphere. Glob. Change regions to the global dust budget. Glob. Planet. Change 52, 88–104. Biol. 24, 504–516. Thornthwaite, C.W., 1948. An approach toward a rational classification of climate. Geogr. Xu, H., Wang, X., 2016. Effects of altered precipitation regimes on plant productivity in Rev. 38, 55–94. the arid region of northern China. Ecol. Inf. 31, 137–146. Tian, F., Fensholt, R., Verbesselt, J., Grogan, K., Horion, S., Wang, Y., 2015. Evaluating Xu, X., Piao, S., Wang, X., Chen, A., Ciais, P., Myneni, R.B., 2012. Spatio-temporal pat- temporal consistency of long-term global NDVI datasets for trend analysis. Remote terns of the area experiencing negative vegetation growth anomalies in China over Sens. Environ. 163, 326–340. the last three decades. Environ. Res. Lett. 7, 035701. Trenberth, K.E., Dai, A., van der Schrier, G., Jones, P.D., Barichivich, J., Briffa, K.R., Xu, H., Wang, X., Yang, T., 2017. Trend shifts in satellite-derived vegetation growth in Sheffield, J., 2014. Global warming and changes in drought. Nat. Clim. Change 4, Central Eurasia, 1982-2013. Sci. Total Environ. 579, 1658–1674. 17–22. Yang, Q., Li, M., Zheng, Z., Ma, Z., 2017. Regional applicability of seven meteorological Tueller, P.T., 1987. Remote sensing science applications in arid environments. Remote drought indices in China. Sci. China-Earth Sci. 60, 745–760. Sens. Environ. 23, 143–154. Yu, M., Li, Q., Hayes, M.J., Svoboda, M.D., Heim, R.R., 2014. Are droughts becoming Vicente-Serrano, S.M., Begueria, S., Lopez-Moreno, J.I., 2010. A multiscalar drought more frequent or severe in China based on the standardized precipitation evapo- index sensitive to global warming: the standardized precipitation evapotranspiration transpiration index: 1951-2010? Int. J. Climatol. 34, 545–558. index. J. Clim. 23, 1696–1718. Yuan, W., Liu, D., Dong, W., Liu, S., Zhou, G., Yu, G., Zhao, T., Feng, J., Ma, Z., Chen, J., Vicente-Serrano, S.M., Begueria, S., Lorenzo-Lacruz, J., Julio Camarero, J., Lopez- Chen, Y., Chen, S., Han, S., Huang, J., Li, L., Liu, H., Liu, S., Ma, M., Wang, Y., Xia, J., Moreno, J.I., Azorin-Molina, C., Revuelto, J., Moran-Tejeda, E., Sanchez-Lorenzo, A., Xu, W., Zhang, Q., Zhao, X., Zhao, L., 2014. Multiyear precipitation reduction 2012. Performance of drought indices for ecological, agricultural, and hydrological strongly decreases carbon uptake over northern China. J. Geophys. Res. 119, applications. Earth Interact. 16, 1–27. 881–896. Vicente-Serrano, S.M., Gouveia, C., Julio Camarero, J., Begueria, S., Trigo, R., Lopez- Zeiter, M., Schärrer, S., Zweifel, R., Newbery, D.M., Stampfli, A., 2016. Timing of extreme Moreno, J.I., Azorin-Molina, C., Pasho, E., Lorenzo-Lacruz, J., Revuelto, J., Moran- drought modifies reproductive output in semi-natural grassland. J. Veg. Sci. 27, Tejeda, E., Sanchez-Lorenzo, A., 2013. Response of vegetation to drought time-scales 238–248. across global land biomes. Proc. Natl. Acad. Sci. U. S. A. 110, 52–57. Zeng, F., Collatz, G.J., Pinzon, J.E., Ivanoff, A., 2013. Evaluating and quantifying the von Wehrden, H., Hanspach, J., Ronnenberg, K., Wesche, K., 2010. Inter-annual rainfall climate-driven interannual variability in global inventory modeling and mapping variability in Central Asia-a contribution to the discussion on the importance of en- studies (GIMMS) normalized difference vegetation index (NDVI3g) at global scales. vironmental stochasticity in drylands. J. Arid Environ. 74, 1212–1215. Remote Sens. 5, 3918–3950. Wang, L., Chen, W., 2014. A CMIP5 multimodel projection of future temperature, pre- Zhang, L., Xiao, J., Zhou, Y., Zheng, Y., Li, J., Xiao, H., 2016. Drought events and their cipitation, and climatological drought in China. Int. J. Climatol. 34, 2059–2078. effects on vegetation productivity in China. Ecosphere 7, e01591. Wang, T., Xue, X., Zhou, L., Guo, J., 2015. Combating aeolian desertification in northern Zhang, Q., Kong, D., Singh, V.P., Shi, P., 2017. Response of vegetation to different time- China. Land Degrad. Dev. 26, 118–132. scales drought across China: spatiotemporal patterns, causes and implications. Glob. Wang, J., Dong, J., Yi, Y., Lu, G., Oyler, J., Smith, W.K., Zhao, M., Liu, J., Running, S., Planet. Change 152, 1–11. 2017. Decreasing net primary production due to drought and slight decreases in solar Zhao, M., Running, S.W., 2010. Drought-induced reduction in global terrestrial net pri- radiation in China from 2000 to 2012. J. Geophys. Res. 122, 261–278. mary production from 2000 through 2009. Science 329, 940–943. West, A.G., Dawson, T.E., February, E.C., Midgley, G.F., Bond, W.J., Aston, T.L., 2012. Zheng, J., Bian, J., Ge, Q., Hao, Z., Yin, Y., Liao, Y., 2013. The climate regionalization in Diverse functional responses to drought in a Mediterranean-type shrubland in South China for 1981-2010. Chin. Sci. Bull. 58, 3088–3099. Africa. New Phytol. 195, 396–407. Zou, X., Zhai, P., 2004. Relationship between vegetation coverage and spring dust storms Wilhite, D.A., Svoboda, M.D., Hayes, M.J., 2007. Understanding the complex impacts of over northern China. J. Geophys. Res. 109, D03104.