Satellite Evidence for Exacerbated Degradation and Desertification Of

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

Satellite evidence for exacerbated degradation and desertification of grassland in Central Asia Geli Zhang1, Chandrashekhar Biradar3, Xiangming Xiao1,2†, Jinwei Dong1, Yuting Zhou1, Yuanwei Qin1, Yao Zhang1, Fang Liu4, Mingjun Ding5, and Richard J. Thomas6

1Department of Microbiology and Plant Biology, and Center for Spatial Analysis, University of Oklahoma, Norman, OK 73019, USA; 2Institute of Biodiversity Science, Fudan University, Shanghai, 200433, China; 3International Center for Agricultural Research in Dry Areas (ICARDA), Amman, 11195, Jordan; 4Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China; 5Key Lab of Poyang Lake Wetland and Watershed Research of Ministry of Education and School of Geography and Environment, Jiangxi Normal University, Nanchang, 330022, China; 6CGIAR Research Program on Dryland Systems, c/o ICARDA, Amman, 11195, Jordan Journal to be submitted: JGR, Journal of Geophysical Research: Earth Surface

Abstract: Grassland degradation and desertification is a complex process, including both state conversion (e.g., grasslands to deserts) and gradual within-state change (e.g., greenness dynamics). Existing studies generally did not separate the two components and analyzed them based on time series vegetation indices, which however cannot provide a clear and comprehensive picture for degradation and desertification. Here we proposed a desertification zone classification-based grassland degradation strategy to detect the degradation and desertification process of grassland in Central Asia. First, annual spatially explicit maps of grasslands and deserts were generated to track the conversion between grasslands and deserts. The results showed that 12.5 % of grasslands were converted to deserts from 2000 to 2014, with an increasing degradation and desertification northward in the latitude range of 43-48°N. Second, a fragile and unstable Transitional zone (TZ), as one of hotspots in this region, was identified in southern Kazakhstan based on desert frequency maps. Third, gradual vegetation dynamics based on Enhanced Vegetation Index (EVI) during the thermal growing season (EVITGS) were investigated using Linear Regression and Mann-Kendall approaches. The results indicated that grasslands generally experienced widespread degradation in Central Asia, with an additional hotspot identified in the northern Kazakhstan with significant decreasing of EVITGS. Finally, attribution analyses of grassland degradation and desertification were conducted by correlating vegetation dynamics with three different drought indices (Palmer Drought Severity Index (PDSI), Standardized Precipitation Index (SPI), and Drought Severity Index (DSI)), precipitation, and temperature. It showed that the degradation and desertification of grassland was exacerbated by droughts, and persistent drought was the main factor in Central Asia. This study provided essential information for taking practical actions to prevent the further grassland degradation and targeting right spots for better intervention to combat the land degradation in the region.

Key words: Central Asia, grassland degradation, desertification, MODIS, EVI, drought

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

Key points: 1. State conversion and gradual within-state change were considered for degradation and desertification of grassland 2. Grassland desertification expanded northward during 2000-2014, accounting for 12.5% of grassland 3. Two hotspots with evident grassland degradation and desertification were identified 4. Grassland degradation and desertification was greatly attributed to persistent droughts.

1 Introduction With climate warming, dryland areas are projected to increase dramatically to cover half of the global land surface by the end of this century, in all CMIP5 scenarios, resulting in increased land degradation and desertification (Huang et al. 2015; Sternberg et al. 2015). Expansion of dryland causes a series of environmental issues ranging from soil erosion and salinization, more frequent dust storms, decreased wildlife species, decreased carbon storing capacity, to local and regional climate change (D'Odorico et al. 2013). Some studies showed that the global warming and the increasing hydroclimatic intensity will lead to more extreme precipitation events and an increased frequency of dry spells (Easterling et al. 2000; Giorgi et al. 2014; Huang et al. 2015). The aridity is expected to be intensified, resulting in arid ecosystems becoming more sensitive and fragile, and more susceptible to land degradation and desertification (Kefi et al. 2007). Central Asia is a dryland hotspot dominated by moisture-limited grassland, semi-deserts and deserts ecosystems (Loboda et al. 2012). Due to the relatively infertile soil and sparse vegetation cover, the grassland in Central Asia is sensitive to climate variability and human activities (Huang et al. 2015; Lioubimtseva and Henebry 2009; Seddon et al. 2016). With the enhanced warming and land cover and land use change due to human activities in the past several decades, concerns on grassland degradation and desertification in Central Asia are raising increasingly (de Beurs et al. 2015; Klein et al. 2014; Kuemmerle et al. 2011; Li et al. 2013; Li et al. 2015; Wright et al. 2012; Xi and Sokolik 2015; Zhou et al. 2015). Desertification is considered as the loss of the ability of a landscape to provide ecosystem services, due to a change in soil properties, vegetation or climate (D'Odorico et al. 2013). Desertification is linked with a transition between the two stable states in bi-stable ecosystem dynamics, corresponding to a vegetated state and an un-vegetated state according to desertification theories (Figure 1) (D'Odorico et al. 2013; D'Odorico et al. 2007). The resilience of a bi-stable system is extremely limited when compared with systems having only one stable state (Holling 1973). Thus, if an external disturbance (e.g., aridity) induces a transition to the unvegetated (desertified) state, the removal of this perturbation would not necessarily allow the system to spontaneously return back to its initial status. If the disturbance is beyond a critical threshold (e.g., by reducing the vegetation cover), these systems move toward the alternative stable state of degraded land. Therefore, the observations of desertification, especially the identification of transition from grasslands to deserts, are critical for the environmental protection and management of this landscape.

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

Figure 1 Desertification theory of transition between two stable states (unvegetated and vegetated) in bistable ecosystem, referred to D'Odorico et al., 2013. Given the harsh environment in dryland areas and their vast geographic extent, satellite remote sensing has become a major approach for large scale grassland degradation and desertification detection (de Beurs et al. 2015; Wessels et al. 2012). Remotely sensed vegetation indices, as the signal of vegetation, such as Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Soil-Adjusted Vegetation Index (SAVI), and Tasseled Cap Greenness (TCG), have been widely used to track changes in the vegetated land surface in dryland areas (de Beurs et al. 2015; Eckert et al. 2015; Jeong et al. 2011; Sternberg et al. 2015; Tucker et al. 1991; Zhou et al. 2015). For terrestrial ecosystems, it has been hypothesized that vegetation growth status could be used as a signature of imminent transitions (Kefi et al. 2007; Scheffer et al. 2009). A recent study using GIMMS3g NDVI time-series data has reported that Central Asia experienced a significantly decreasing trend for 1992-2011 due to continued warming and socio-economic conditions (Zhou et al. 2015). De Beurs et al. (2015) found different variables (e.g., EVI, NDVI and TCG) can reveal very similar trends for land surface dynamics. These remote sensing indicators can reflect the greenness or vegetation dynamics; however, the existing studies have not paid enough attention to the identification of three states of desertification: unvegetated (deserts) state, vegetated (grassland) state, and transitional (mixed grassland and deserts, or sparsely vegetation) state. That is, the previous studies paid more attention to the “within-state” shifts of vegetation coverage, instead of the land cover conversion from grasslands to deserts. In addition, previous studies based on temporal analysis of remote sensing vegetation indices can be improved from several aspects. Firstly, most of existing efforts on the identification of grassland degradation and desertification focused on the trend analysis based on vegetation indices; however, this approach is unsuitable for the barren or sparsely vegetated regions with significant seasonal variability (de Beurs et al. 2015; Wessels et al. 2012) which take up large areas of Central Asia. One reason is that the vegetation index values in barren or sparsely vegetated regions are largely affected by the soil background. The barren or sparsely vegetated lands are the transition ecotone from grassland (vegetated) to desert (no-vegetated), and also are the most sensitive, fragile, and unstable region (Figure 1). This transition ecotone

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

greatly determines the direction of desertification (turning to vegetated or unvegetated state) under external perturbation (Figure 1). However, in order to avoid the disturbance from the barren or sparse vegetated lands, some studies just excluded the area with very low NDVI when analyzing vegetation change at regional scale (Piao et al. 2011a; Zhang et al. 2013; Zhou et al. 2001). Therefore, the division of “vegetated” zone, “unvegetated” zone, and transitional zone could be an effective approach for a more complete understanding of degradation and desertification. Secondly, the study objects in these vegetation change efforts were, to some degree, vague (de Beurs et al. 2015; Mohammat et al. 2013; Zhou et al. 2015), which would affect attribution analysis. Specifically, most of studies presented the changes in vegetation greenness and their trends in the entire Central Asia. When investigating factors of land degradation and desertification, previous studies did not carefully consider the regions under evident human- induced land cover type conversions (e.g., abandonment, urbanization), or were merely based on previous knowledge or certain land cover map in a given period (Gessner et al. 2013). In fact, vegetation browning caused by land cover changes (e.g., cropland abandonment, fallows, urbanization) would induce a feint of land degradation and desertification. Moreover, the land use change intensity in local regions of Central Asia is increasing due to increasing human disturbances (de Beurs et al. 2015; Zhou et al. 2015). Therefore, when exploring grassland degradation and desertification, the identification of specific vegetative states including grassland and desert is a critical component for examining ecological and environmental changes. In this study, we try to clarify the targeted land cover types (grasslands and deserts) for the land degradation and desertification observations. Thirdly, annual or seasonal NDVI was generally used in previous studies for vegetation change analysis in Central Asia (Mohammat et al. 2013; Piao et al. 2011b; Zhou et al. 2015). However, the study region is across a large latitudinal range from 35° to 57° N, displaying the transition from cold temperate zone, temperate zone to subtropical zone. Large differences in vegetation growth phases exist from north to south. Therefore, it is not suitable to define the vegetation growing season with the same period. A thermal growing season defined by using daily surface temperature (Zhang et al. 2015) could provide a better option for the tracking vegetation dynamics in a spatially and temporally consistent way. Based on the above concerns and limitations of previous studies, it is critical to understand the grassland degradation and desertification process and establish early warning signals of conversion from grassland to desert. The major objective of this study is to provide more previse/robust satellite evidence for the grassland degradation and desertification process in Central Asia by considering both the land-cover conversion between grassland and desert and gradual vegetation greenness changes. Specifically, we would like to answer four questions: 1) what is the annual dynamics of grassland, desert areas from 2000 to 2014? 2) What was the status of grassland degradation in last 15 years? 3) Where are the sensitive and fragile regions or hotspots of grassland degradation and desertification? 4) What is the main factor for grassland degradation and desertification in this region? The results are expected to serve decision makers in order to take early action to prevent further desertification in Central Asia.

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

2 Data and methods

2.1 Study area Central Asia, the core region of Asian continent, consists of Kazakhstan, Uzbekistan, Turkmenistan, Kyrgyzstan, and Tajikistan (Figure 2a). It stretches from Afghanistan and Iran in the south to Russia in the north, and from the Caspian Sea in the west to China in the east. The total area of Central Asia is about 4 million km2; the area of Kazakhstan is twice as large as the total area of the other four countries (de Beurs et al. 2015). The total population in this region is about 66 million in 2013-2014; the density of population is low and the distribution of population varies widely, due to the terrain and landscape (https://en.wikipedia.org/wiki/Central_Asia).

Figure 2 Location of study area. (a) Spatial distribution of land cover types in Central Asia based on MCD12Q1 IGBP classification scheme in 2010; (b) Specific targeted area of grassland and desert region in Central Asia. We combined IGBP-based open and closed shrublands into one class in (a). The terrain of Central Asia is high in the east but low in the west (Figure S1); it is mainly dominated by plains and hills, and mountainous areas are small, mostly distributed in Kyrgyzstan and Tajikistan. Central Asia includes a range of landscapes with a transition from steppes in the north to semi-deserts and deserts in the south (Gessner et al. 2013), due to the gradual weakening of the Arctic Ocean moisture from north to south. The grassland is largely located in the north of Kazakhstan, the sparsely grassland is distributed in the south of Kazakhstan, and semi-desert and desert are distributed in Uzbekistan and Turkmenistan. The grassland in this region forms central

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

part of the Eurasian steppe. The area of cropland is limited and small, and mostly located in valley areas (e.g., Amu Darya and Syr Darya) and north of Kazakhstan. Agro-pastoralism is a widespread land use, distributed in the region between the mountains and the flatter grasslands of Central Asia. Major rivers of Central Asia include the Amu Darya, the Syr Darya, Irtysh, the Hari River and the Murghab River. The Aral Sea and Lake Balkhash are the main lakes. Due to climate change and irrigation and industrial purposes, these water bodies have shrunk significantly in recent decades (Klein et al. 2012). Central Asia has continentally temperate desert and steppe climate, characterized by dry climate, scarce precipitation, intensive evaporation, and large fluctuation of temperatures in a day or in a year. The annual precipitation is normally below 300 mm. The mean temperature in summer ranges from 26-32 °C. The mean temperature in winter ranges from -20 °C in the north to 2°C in the south. The daily temperature difference reaches 20-30 °C.

2.2 Data 2.2.1 Remote sensing data The surface reflectance, land surface temperature (LST), and land cover data from the MODIS on board NASA’s Terra satellite were used in this study, which were obtained from the USGS EROS Data Center (https://lpdaac.usgs.gov/), including 8 tiles (H21/22/23V03, H21/22/23V04, and H22/23V05) for Central Asia. A detailed data preprocessing is described below. Vegetation Indices (VIs): The vegetation indices (VIs) were calculated using the 8-day composite Surface Reflectance Product (MOD09A1 C5) (Vermote and Vermeulen 1999; Zhou et al. 2015) from 2000-2014, including 1) EVI, which is considered to have higher robustness to atmospheric conditions and soil background relative to NDVI (Huete et al. 2002; Huete et al. 1997; Xiao et al. 2003), 2) land surface water index (LSWI), which is sensitive to equivalent water thickness (Maki et al. 2004; Xiao et al. 2002) because the imbedded SWIR band is sensitive to leaf water and soil moisture, 3) and normalized difference snow index (NDSI) which was used to identify snow pixels (Hall et al. 2002).

 EVI 2.5 nirred (1) nirredblue67.51  LSWI  nirswir (2) nirswir  NDSI  greennir (3)  greennir where ρblue, ρgreen, ρred, ρnir, and ρswir are the surface reflectance for the blue, green, red, near- infrared, and shortwave-infrared bands, respectively.

The bad observations, including clouds, cloud shadows, and snow, were excluded to form a long-term time series VI datasets, by using a two-step strategy. First, we used the data quality information from the quality control flag layer to extract the clouds and cloud shadows from each image. Second, we applied an additional restriction in which the pixels with a blue reflectance of ≥ 0.2 were also labeled as cloudy (Xiao et al. 2005) (the red round symbols in the

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

LSWI curves of Figure S2). The snow cover pixels were also excluded using the NDSI and the Near-Infrared (NIR) band (Hall et al. 2002), specifically NDSI>0.40 and NIR>0.11 (Xiao et al. 2005). All the pixels identified as cloud, cloud shadow, or snow covers were excluded. The gaps in time series VI data due to bad-quality observations were gap-filled using the linear interpolation approach (Xiao et al. 2005). Land Surface Temperature (LST): The Land Surface Temperature and Emissivity 8-Day L3 Global 1km product (MOD11A2) from 2000-2014 was used to calculate the nighttime land surface temperature (LST)-based thermal growing season in this study. The retrieval of LST has been improved by correcting noise, agreed well with in situ-based LST (Wan et al. 2002). The digital number values (DN) from MOD11A2 were converted to LST with centigrade unit values based on the following formula: LST (°C) = DN×0.02–273.15 (Wan 2008; Wan et al. 2002). The gaps in time series LST data were filled by using linear interpolation. Then the start and end dates of thermal growing season with nighttime LST above 0°C (TGS0) in continuous three 8- day intervals were calculated for each year from 2000-2014. The resultant maps of the start and end dates of TGS0 were resampled to 500 m using the nearest neighbor interpolation method to be spatially consistent with the VI maps from MOD09A1. Land Cover Type: The MODIS Land Cover Type Yearly L3 Global 500 m SIN Grid product (MCD12Q1) was used to identify the targeted area. This dataset provides five global land cover classifications schemes from 2001-2012, and IGBP-based global vegetation classification scheme was used in this study, which identifies 17 land cover classes globally (Friedl et al. 2010). According to the land cover data in 2010, Central Asia primarily consists of five land cover classes: Grasslands (64%), Barren (15%), Shrublands (10%), Croplands (7%), and Water (2%). 2.2.2 Climate data and drought indices datasets The air temperature and precipitation data were obtained from the University of East Anglia Climatic Research Unit (CRU) Time Series 3.2 dataset (Harris et al. 2014). The CRU dataset is derived from archives of climate station records, available at the British Atmospheric Data Centre website (http://badc.nerc.ac.uk/). This dataset is monthly climate observations at 0.5°×0.5° spatial resolution, covering the period from 1901-2013. Three drought indices datasets were used to illustrate the drying trend and relationship with grassland degradation and desertification in this study region, including MODIS-based Drought Severity Index (DSI), the meteorological-based Standardized Precipitation Index (SPI), and the Palmer Drought Severity Index (PDSI). The DSI data product is a remotely sensed global index using MODIS terrestrial evapotranspiration (ET)/potential ET and MODIS NDVI data as primary inputs (Mu et al. 2013). This dataset covers all vegetated land areas with 1 km spatial resolution at 8-day, monthly, and annual intervals over the period from 2000-2011. Annual DSI was used in this study, obtained from the Numerical Terra dynamic Simulation Group, University of Montana (ftp://ftp.ntsg.umt.edu/pub/MODIS/Mirror/DSI) (Mu et al. 2013). SPI data is commonly used to monitor drought and anomalous wet period, which is calculated based only on the long-term precipitation record for a desired period (McKee et al. 1993). This dataset includes the SPI at 3 months, half-year, and annual scales for global land surface from 1949- 2012 with 1° × 1° grids, obtained from Climate Data Guide (https://climatedataguide.ucar.edu/climate-data/standardized-precipitation-index-spi) (O'Loughlin et al. 2012). PDSI is the other reasonably successful index at quantifying long-term

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

drought for global land surface from 1850-2012 with 2.5° × 2.5° grids, which was also obtained from Climate Data Guide (https://climatedataguide.ucar.edu/climate-data/palmer-drought- severity-index-pdsi) (Dai et al. 2004; Dai 2011). Monthly PDSI is estimated by readily available temperature and precipitation data (Palmer 1965). 2.2.3 Livestock data The global distribution of livestock data were obtained from the Livestock of the World v2.0 dataset (Robinson et al. 2014). This dataset is publically available via the Livestock Geo-Wiki (http://www.livestock.geo-wiki.org/). The dataset provided modelled livestock densities of the world for the reference year 2005 with 1 km resolution, including cattle, chicken, ducks, goats, pigs, and sheep maps. Table 1. Summary of the Data Used in This Study Spatial Temporal Product Data used Period Reference Resolution Resolution VIs MOD09A1 500 m 8 day 2000-2014 (Vermote et al. 2002) LST MOD11A2 1000 m 8 day 2000-2014 (Wan et al. 2002) Land Cover MCD12Q1 500 m Yearly 2001-2012 (Friedl et al. 2010) MODIS DSI 0.05 ° Yearly 2000-2011 (Mu et al. 2013) ET/NDVI SPI P 1 ° Yearly 1949-2012 (O'Loughlin et al. 2012) PDSI T, P 2.5° Yearly 1850-2012 (Dai et al. 2004; Dai 2011) T, P CRU 3.2 0.5 ° monthly 1901-2013 (Harris et al. 2014) Note: T: Temperature, P: Precipitation.

2.3 Methods In order to understand and detect the trend and process of grassland degradation and desertification in Central Asia, a desertification zone classification-based grassland degradation strategy, which observes both state conversion (e.g., between grasslands and deserts) and gradual within-state change (e.g., greenness dynamics), was developed in this study. First, in order to exclude evident disturbance from land use changes (e.g., agricultural abandonment) in this region, a targeted area, only including grassland, barren or sparsely vegetated, and desert region, was clearly defined by the combination of land cover data and additional masks (see section 2.3.1 for detail). Secondly, the grassland and desert were classified, and conversion between grasslands and deserts was analyzed to determine the state conversions. Thirdly, the division of “Never or Occasionally Deserted” zone (NODZ), Transitional zone (TZ), and “Persistently Deserted” zone (PDZ) was executed by counting the frequency of annual desert from 2000-2014. Fourthly, vegetation variation and trend analyses in these three zones were conducted. Last, the causes of grassland degradation and desertification were investigated by correlation analysis with different climate variables and drought indices. Figure 3 provides a flow chart illustrating the methodology and workflow used in this study.

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

MOD09A1 MOD12A2 MCD12Q1 Reflectance data LST data 2001-2012 2000-2014 2000-2014

IGBP-based Time Series EVI and LST>=0 Thermal land cover LSWI (8-day intervals) Growing Season (TGS) AAnnnnuuaall MMaasskkss Cropland, forest, wetland Grassland, Barren or sparsely Water bodies Time Series VIs vegetated within TGS

Annual Grassland and Desert Maps from 2001-2012 SDEVI iinntteerrsseeccttiioonn Setting of study objective: Grassland and desert

DDeesseerrttiiffiiccaattiioonn FFrreeqquueennccyy aannaallyyssiiss AAnnnnuuaall LLaanndd CCoovveerr aannaallyyssiiss Never or Persistently Grassland Desert Occasionally Transitional Deserted Deserted Zone Zone SDEVI>=0.02 SDEVI<0.02 Zone (TZ) SDEVI>=0.02 SDEVI<0.02 (PDZ) (NODZ)

ffrreeqq <<==11 11<< ffrreeqq <<1144 ffrreeqq >>==1144

Vegetation Variation and Trend Analysis Temperature CRU 2000-2013

Attribution Analysis Precipitation DSI 2000-2011

Drought Index SPI 2000-2012

PDSI 2000-2012

Figure 3 Flow chart illustrating the data sources and methodology in this study.

2.3.1 Determination of targeted area by excluding croplands and forests Three steps are used to identify the spatially explicit extent of targeted area, including grassland, barren or sparsely vegetated, and desert. First, annual information of grasslands or barren or sparsely vegetated lands, based on the IGBP classification scheme from MCD12Q1 data, was extracted for each year. Second, we applied additional masks to eliminate potential errors in the classification process (Klein et al. 2012) and disturbances from other land cover types. An integrated mask of croplands, forests, and wetlands was generated using LSWI time series data. Cropland, forest, and wetlands areas tend to have a long period with LSWI>0 during LST0-based growing season (see Section 2.3.3 for definition of LST0-based growing season in detail) within

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

a year (Figure S2a, b, c). While for grassland in drylands, the period with LSWI>0 during LST0- based growing season is very short, or even does not occur (Figure S2d, e). The cropland, forest, and wetland areas are identified as those pixels having LSWI values of >0 with good observations over at least ten 8-day composites during LST0-based growing season within a year (Figure S3). Besides, there is also a need to separate water bodies from grassland and desert region. We assumed a pixel is covered by water if the maximum EVI with good observations is smaller or equal to zero during LST0-based growing season within a year (Figure S4). We produce an integrated mask of croplands, forests, and wetlands as well as a water body mask for each year. After applying these masks, annual refined grassland and sparsely vegetated land maps from 2001-2012 were generated. Last, the specific targeted area was ascertained by intersection of annual maps from 2001-2012 (Figure 2b). 2.3.2 Separation of Grasslands and Deserts Clear distinction between grassland and desert is essential for understanding how grassland degrades and converts into desert from a land cover change perspective. Generally, grasslands have higher greenness than deserts, at the same time, the seasonal variation of EVI in deserts is smaller and more stable than that in grasslands. In this study, standard deviation (SD) of EVI is used as an index to identify grassland and desert. Comparing the EVI curves of grassland and desert in Central Asia, the fluctuation in EVI curve of desert is more stable than that of grassland (Figure S2d, e, f). In other words, fluctuation in EVI curve will decrease with the decrease of vegetation coverage. We summarized the distributions of SD of the EVI (SDEVI) from random 10,000 pixels in 2010 for each IGBP-based vegetation type in Central Asia. Among all the vegetation types, barren or sparsely vegetated had the lowest SDEVI (Figure S5). We used 0.02 as the threshold to distinguish grasslands and deserts in the study area (Figure S6, S7). 2.3.3 Vegetation variation and trend analysis Before analyzing vegetation variation and trend for the targeted area of Central Asia, we defined the LST-based thermal growing season first for calculation of the mean EVI during the thermal growing season. Plants generally begin to grow when a stable temperature threshold is reached, so they will not suffer damage from cold temperatures. By comparing the observational vegetation phenology data and the temporal profile of nighttime LST (Figure S2), we determined the start and end date of TGS when the nighttime LST surpasses 0 °C in continuous three 8-day intervals (TGS0) (Figure S8).

The mean EVI during the LST0-based thermal growing season (EVITGS) for each year was used for the vegetation change analysis over the period 2000-2014. Two approaches, Linear Regression (LR) and Mann-Kendall (MK) test, were used for the statistical analysis. The LR calculates the slope of the linear least squares regression line fit to the inter-annual variation of the EVITGS values. The statistical significance of EVITGS change was mapped and assessed based on the two-tailed significance tests (Zhang et al. 2013). Mann-Kendall trend test is a nonparametric test used to identify a trend in EVITGS, even if there is a seasonal component in the series (Fensholt et al. 2012; Zhang et al. 2016). 2.3.4 Attribution analysis To explore main factor for grassland degradation and desertification in this region, we performed correlation analysis between EVITGS and three different drought indices (PDSI, SPI, DSI), precipitation, and temperature.

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

3 Results

3.1 Increasing conversion from grasslands to deserts from 2000-2014 Central Asia is primarily covered by grassland, mostly distributed in central Kazakhstan and adjacent with the desert band in the southern Kazakhstan (Figure S7). The changes in the SDEVITGS-indicated grassland area of Central Asia over the period 2000-2014 showed an evident decrease with a rate of -0.14×105 km2 (P=0.08) (Figure 4). Grassland decreased by 12.5 % from 2000 to 2014. Specially, the area of grassland in 2002 was the largest, then followed by 2007 and 2009, while that in 2008 was the lowest. The changes in SDEVITGS-indicated grassland area of Central Asia by latitude in past 15 years showed that the grassland area evidently decreased with a latitude range of 43-48°N (Figure 5). Besides, the first year with desert occurrence over the period 2000-2014 gradually delayed from south to north in Central Asia (Figure 5a), indicating the extension of desert northward, which brought the large decrease in grassland in Central Asia in these years.

Figure 4 Variation in standard deviation of EVI-indicated grassland area in targeted area of Central Asia over the period 2000-2014.

Figure 5 Geographical pattern of grassland desertification. (a) Spatial pattern of first occurrence of desert for each pixel over the period 2000-2014; (b) Distribution of desert area along latitude gradient in each year.

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

According to the desertification theory (D'Odorico et al. 2013), “vegetated” zone, transitional zone, and “unvegetated” zone were divided based on the frequency map of SDEVITGS-indicated desert in study area over the period 2000-2014 (Figure 6a). It included vegetated ‘Never or Occasionally Deserted Zone’ (NODZ for short) with frequency less or equal to 1, ‘Transitional Zone’ (TZ for short) with frequency larger than 1 and smaller than 14, and unvegetated ‘Persistently Desert Zone’ (PDZ for short) with frequency larger or equal to 14 (Figure 6b). From the desertification zone classification, this region was mainly dominated by NODZ, accounted for 46%; followed by TZ with 37%; and the area of PDZ was the smallest, only 17% (Figure 6b). NODZ, TZ, and PDZ are distributed in Central Asia from north to south (Figure 6), which showed the gradually decreased vegetation cover from north to south. This spatial pattern was consistent with that of the multi-year averaged EVITGS in study area (Figure S9). The distribution of NODZ and TZ presented continuous strips from west to east. From Figure 5, there was evidently decreasing in grassland area in TZ. TZ, the fragile and sensitive region, was identified as one of hotspots in Central Asia, which directly determined the direction of desertification.

Figure 6 (a) Spatial pattern of frequency of desert in Central Asia; (b) Spatial distribution of Never or Occasionally Deserted Zone (NODZ), Transition Zone (TZ), and Persistently Desert Zone (PDZ). The inset in (a) shows the distribution of area percent of different frequency of desert, and the inset in (b) shows the area percent of different zones.

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

3.2 Vegetation variation within-state change We conducted comparison analyses of interannual variations in EVITGS in different desertification classification zones mentioned above (Figure 7). The EVITGS of targeted area over the period 2000-2014 showed an evident decrease trend at a rate of -0.0014 (R2=0.45, P=0.006, n=15). There was consistent trend of EVITGS between in NODZ and TZ (R=0.85, P<0.001), which showed a significant decrease on the order of -0.0021 (R2=0.49, P=0.004, n=15) and -0.001 (R2=0.34, P=0.02, n=15), respectively. The EVITGS in PDZ showed a non-significant and fluctuating change (R2=0.12, P=0.20, n=15), which indicated that the definition of PDZ is reasonable.

Figure 7 Mean anomalies in averaged EVITGS in (a) targeted area, (b) NODZ, (c) TZ, and (d) PDZ of Central Asia over the period 2000-2014.

The LR and MK trends were used to analyze EVITGS in targeted area. In terms of the spatial pattern, the EVITGS exhibited a decreasing trend in almost all areas of targeted area (90%), and the decreasing trend was significant in 25% (36%) with a significance level of P<0.05 (P<0.1) based on the LR trend (Figure 8a, Table 2). The majority of significant decreases in EVITGS mainly occurred in the northern parts of targeted area with large swaths, which is the mainly region of NODZ, with 67% (64%) of significant decreases at significance level of P<0.05 (P<0.1) distributed in NODZ (Figure 8a, Table 2). The area with an increasing trend of EVITGS was very limited, only 10%, mainly distributed in the eastern part of Central Asia; and the increasing trend was significant in less than 1% with a significance level of P<0.1. Besides, the MK-based trend analysis also showed the similar spatial pattern of EVITGS change in this region (Figure 8b, Table 2). The area with significant decreasing in EVITGS in NODZ was identified as the other hotspot in Central Asia.

Figure 8 Spatial patterns of (a) Linear Regression (LR) trend and (b) Mann-Kendall (MK) trend of averaged EVITGS in targeted area of Central Asia over the period 2000-2014. The insets show the region with statistically significant at 0.05 level.

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

Table 2. Area percentages (%) of different significance levels of decreasing trend for averaged EVITGS with Linear Regression (LR) and Mann-Kendall (MK) methods in targeted area and NODZ of Central Asia over the period 2000-2014 LR MK Targeted area NODZ Targeted area NODZ Trend<0 and P<0.05 25 36 23 33 Trend<0 and P<0.1 36 50 33 46 Trend<0 90 94 89 94 Ratio between area of trend<0 and P<0.05 in NODZ 67 65 to that in targeted area Ratio between area of trend<0 and P<0.1 in NODZ 64 63 to that in targeted area Note: NODZ means Never or Occasionally Deserted Zone.

3.3 Effects of droughts on grassland degradation and desertification Despite different data sources and spatial resolutions, the three drought indices (PDSI, SPI, and DSI) were highly correlated (correlation coefficients between each other larger than 0.73, and all of P < 0.05), and showed similar decrease trend over the targeted area of Central Asia since 2000 (Figure 9a). There were significant decrease in PDSI and DSI with rates of -0.10 (P=0.04) and - 0.13 (P=0.009), respectively; the decrease trend of SPI was not significant (Figure 9a). These trends indicated that the drought intensified in this region from 2000 even though there was no great decrease in precipitation (slope=-1.68 mm/year, P=0.36) and no evident increase in temperature (P=0.73) from 2000-2013 (Figure 9b, c). It is noteworthy that the mean temperature anomaly during 1997-2013 was evidently larger than that during 1950-1996, showing the recent 17 years from 1997-2013 was the warmest period in the last 60 years (Figure 9c). At the spatial scale, the drying trends from three drought indices were widespread across this region, northern region showing a significant negative trend (P<0.05) (Figure 10a-c). Precipitation also showed consistent decreasing trend with SPI at spatial pattern; however, for temperature, it showed no significant increase in northwestern part and decease in eastern part (Figure 10d, e). The relationships of the spatial EVITGS with three drought indices (DSI, PDSI, and SPI), annual precipitation, and annual mean temperature were examined using statistical analysis. From the inter-annual variations, there were strong relationships between EVITGS and the three drought indices with P<0.01 (Figure 11a-c). Although EVITGS was significantly related with annual precipitation (P=0.01), the significant level was smaller than those with three drought indices (Figure 11d). Whereas, the EVITGS was not significantly related to annual mean temperature (P=0.42) (Figure 11e). Besides, by comparing the spatial pattern of LR trend of EVITGS and drought indices, precipitation, and temperature, it also showed that drought widely determined the vegetation growth (Figure S10).

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

Figure 9 Mean anomalies in (a) Drought Severity Index (DSI) from 2000-2011, Palmer Drought Severity Index (PDSI) from 1950-2012, and Standardized Precipitation Index (SPI) from 1950-2012, (b) rainfall from 1950-2013, and (c) mean temperature from 1950-2013 in targeted area of Central Asia.

Figure 10 Spatial patterns of Linear Regression trend of (a) Drought Severity Index (DSI) from 2000-2011, (b) Standardized Precipitation Index (SPI) from 2000-2012, (c) Palmer Drought Severity Index (PDSI) from 2000-2012, (d) annual precipitation from 2000-2013, and (e) annual mean temperature from 2000-2013. In each map, regions with plus symbol have a linear trend that is statistically significant at P<0.05.

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

Figure 11 Relationships of averaged EVITGS with (a) Drought Severity Index (DSI), (b) Standardized Precipitation Index (SPI), (c) Palmer Drought Severity Index (PDSI), (d) annual precipitation, and (e) annual mean temperature in targeted area of Central Asia.

4 Discussion In this study, we detected grassland degradation and desertification in Central Asia using a desertification zone classification-based grassland degradation strategy, which combined state conversion (conversion between grasslands and deserts) and gradual vegetation change (greenness trend or variation) analyses. To our knowledge, this is the first attempt to integrate these two approaches to show a clear picture of grassland degradation and desertification process. Previous studies (de Beurs et al. 2015; Mohammat et al. 2013; Zhou et al. 2015) based on trend analysis of vegetation indices overlooked the sparsely vegetation regions, which happened to be the fragilely transitional zone determining the direction of desertification. Our results showed that the significant browning in Central Asia based on trend analysis was mostly located in NODZ with relative good vegetation coverage (the area with significant decrease in NODZ accounting for 67% of that in our targeted area (Table 2)), which indicated the limits of trend analysis for the regions that are partly barren or sparsely vegetated. According to the desertification theory, our study tried to identify the NODZ, TZ, and PDZ. In vegetated NODZ, located in the north of Kazakhstan, it showed evidently browning based on trend analysis (grassland degradation), which was highly consistent with previous studies (de Beurs et al. 2015; Mohammat et al. 2013; Zhou et al. 2015). As for the vegetated region, the grassland in NODZ tends to return to its initial configuration when removing the disturbance (e.g., drought). If the perturbation increases, the vegetated NODZ would be likely to become TZ. TZ is the variably transitional ecotone between grasslands and deserts. The effects of external disturbance on this region were random, which means the removal of disturbance would allow this region to return to vegetated NODZ, or to unvegetated PDZ. The area of desert evidently increased in TZ in last 15 years (Figure 5b), showing TZ tended to turn to unvegetated PDZ in the context of persistent drought. On the whole, our study presented observational evidence for a widespread grassland degradation over the past 15 years combining of those two approaches, and offset the limits when detecting the degradation and browning of sparse vegetation by pure trend analysis (de

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

Beurs et al. 2015; Wessels et al. 2012). And it also helps to identify the concretely sensitive and fragile regions, which were the significant browning region based on trend analysis and the region with higher desert frequency in TZ. The explicit scope of targeted area is of importance to promote the understanding of process and causes of grassland degradation and desertification in Central Asia. In this study, the targeted area (grasslands and deserts) was identified based on the land cover data from MODIS and additional masks from 2000-2012 (see 2.3.1 for detail) to exclude croplands, forests, wetlands, and so on. If the other land cover types are not excluded, the causes of grassland degradation and desertification would be more complex. For example, de Beurs et al. (2015) considered that the vegetation browning in the northwest of Kazakhstan was attributed to a combination of drought, increasing fallow periods or ongoing agricultural abandonment. Zhou et al. (Zhou et al. 2015) demonstrated that NDVI was significantly related with precipitation in this region. In fact, in our study, cropland has been excluded in study region by the determination of targeted area, and effects of agricultural activities on vegetation browning cannot be considered. In this sub-region, both EVITGS and three drought indices showed evident decreasing since 2000 (Figure S11), and the relationships between EVITGS and three drought indices, annual precipitation, and annual mean temperature indicated that the drought may be the most significant cause for vegetation browning (P<0.001, Figure S12). For the whole targeted area in Central Asia, we also found that the observed EVITGS change was principally attributable to persistent drought (P<0.005, Figure 11a-c). Furthermore, the low density of livestock (Figure S13) and population in wide grassland and desert region (Figure 2a) also showed that anthropogenic activities were limited and negligible for grassland degradation and desertification in targeted area of Central Asia, which was consistent with previous studies (de Beurs et al. 2015; Zhou et al. 2015). Besides, the area of SDEVITGS-indicated grassland in 2008 was the lowest (Figure 4). It was very clear that the extreme drought in this year was the main factor for it through the variations of three drought indices and precipitation (Figure 9). Overall, the clear targeted area definition is significant for detecting the cause mechanisms of grassland degradation and desertification through removing the disturbance from the land use changes and human activities (e.g., agricultural abandonment, grassland cultivation, urbanization). Due to the difference in climate types from north to south in Central Asia, the assumption with same length of growing season on the whole region is unsuitable, such as from April to October used in previous studies (Mohammat et al. 2013; Zhou et al. 2015). The spatial distribution of start date of growing season (SOS) of vegetation in Central Asia varies greatly. For example, the SOS of grassland in north region (Figure S2d) and south region (Figure S2e) began around April and March, respectively. The spatial pattern of start date of LST0-based TGS map in 2010 in this region (Figure S8) also showed significant difference from north to south, such as the end of March or April in northern region, and January in southern region. If the universal and consistent length of growing season is considered in this region, such as April as the SOS of vegetation, the early and short-lived vegetation would be neglected (Figure S2e). Therefore, the identification of vegetation growing season in Central Asia needs be more careful, and the method utilized in this study also can be used in the other studies about the vegetation change at large scale. It is noteworthy that the degradation and desertification of grassland, the main land cover type in Central Asia, was analyzed, but not shrublands, croplands, and the other land cover types.

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

In the near future, the degradation and desertification of different land cover types should be further studied in order to ascertain a comprehensive picture of desertification in the region.

5 Conclusions Knowledge of the process of grassland degradation and desertification against the background of global warming is essential for taking early action to prevent the increase in desertification in central Asia. In this study, we carried out a comprehensive analysis and provided a clear picture for grassland degradation and desertification, including a state conversion between grasslands and deserts and within-state gradual change. This improved understanding on the grassland desertification also contributed to the attribution analysis. Our analysis provided three key insights into grassland degradation and desertification in Central Asia. First, the state conversion analysis showed that grassland degradation has increased from 2000 to 2014, and desertification was gradually expanding northward. Secondly, our results clearly identified sensitive and fragile regions, one was the significant browning region in the north of Kazakhstan, and the other was the region with higher desert frequency of TZ in central of Kazakhstan. Thus these ‘hot spots’ of degradation require more attention in order to prevent the further aggravation of regional desertification. Last, the grassland degradation and desertification was largely linked with persistent droughts in Central Asia. Given the projected increase in aridity and warming (Huang et al. 2015), grassland degradation and desertification is expected to exacerbate in Central Asia in the future. The identification of sensitive and fragile regions in this study can help decision makers to effectively mitigate desertification in Central Asia. The desertification zone classification-based grassland degradation strategy, by integrating abrupt state conversion information and gradual vegetation dynamics, can also be used to detect grassland degradation and desertification in the other dryland regions.

Acknowledgments This study was supported by research grants from the NASA Land Use and Land Cover Change program (NNX14AD78G) and the National Institutes of Health NIAID (1R01AI101028- 01-A2) and International Center for Agricultural Research in Dry Areas (ICARDA, Russian Government Funds). We thank XXX for their comments in the earlier versions. C. Biradar and R.J. Thomas acknowledge the support of the CGIAR Dryland Systems program.

References D'Odorico, P., Bhattachan, A., Davis, K.F., Ravi, S., & Runyan, C.W. (2013). Global desertification: Drivers and feedbacks. Advances in Water Resources, 51, 326-344 D'Odorico, P., Caylor, K., Okin, G.S., & Scanlon, T.M. (2007). On soil moisture-vegetation feedbacks and their possible effects on the dynamics of dryland ecosystems. Journal of Geophysical Research-Biogeosciences, 112 Dai, A., Trenberth, K.E., & Qian, T.T. (2004). A global dataset of Palmer Drought Severity Index for 1870-2002: Relationship with soil moisture and effects of surface warming. Journal of Hydrometeorology, 5, 1117-1130 Dai, A.G. (2011). Characteristics and trends in various forms of the Palmer Drought Severity Index during 1900- 2008. Journal of Geophysical Research-Atmospheres, 116 de Beurs, K.M., Henebry, G.M., Owsley, B.C., & Sokolik, I. (2015). Using multiple remote sensing perspectives to identify and attribute land surface dynamics in Central Asia 2001–2013. Remote Sensing of Environment, 170, 48-61 Easterling, D.R., Meehl, G.A., Parmesan, C., Changnon, S.A., Karl, T.R., & Mearns, L.O. (2000). Climate extremes: observations, modeling, and impacts. Science, 289, 2068-2074 Eckert, S., Husler, F., Liniger, H., & Hodel, E. (2015). Trend analysis of MODIS NDVI time series for detecting land degradation and regeneration in Mongolia. Journal of Arid Environments, 113, 16-28

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

Fensholt, R., Langanke, T., Rasmussen, K., Reenberg, A., Prince, S.D., Tucker, C., Scholes, R.J., Le, Q.B., Bondeau, A., Eastman, R., Epstein, H., Gaughan, A.E., Hellden, U., Mbow, C., Olsson, L., Paruelo, J., Schweitzer, C., Seaquist, J., & Wessels, K. (2012). Greenness in semi-arid areas across the globe 1981-2007 - an Earth Observing Satellite based analysis of trends and drivers. Remote Sensing of Environment, 121, 144-158 Friedl, M.A., Sulla-Menashe, D., Tan, B., Schneider, A., Ramankutty, N., Sibley, A., & Huang, X.M. (2010). MODIS Collection 5 global land cover: Algorithm refinements and characterization of new datasets. Remote Sensing of Environment, 114, 168-182 Gessner, U., Naeimi, V., Klein, I., Kuenzer, C., Klein, D., & Dech, S. (2013). The relationship between precipitation anomalies and satellite-derived vegetation activity in Central Asia. Global and Planetary Change, 110, 74-87 Giorgi, F., Coppola, E., & Raffaele, F. (2014). A consistent picture of the hydroclimatic response to global warming from multiple indices: Models and observations. Journal of Geophysical Research-Atmospheres, 119, 11695-11708 Hall, D.K., Riggs, G.A., Salomonson, V.V., DiGirolamo, N.E., & Bayr, K.J. (2002). MODIS snow-cover products. Remote Sensing of Environment, 83, 181-194 Harris, I., Jones, P.D., Osborn, T.J., & Lister, D.H. (2014). Updated high-resolution grids of monthly climatic observations - the CRU TS3.10 Dataset. International Journal of Climatology, 34, 623-642 Holling, C.S. (1973). Resilience and Stability of Ecological Systems. Annual Review of Ecology and Systematics, 4, 1-23 Huang, J., Yu, H., Guan, X., Wang, G., & Guo, R. (2015). Accelerated dryland expansion under climate change. Nature Climate Change Huete, A., Didan, K., Miura, T., Rodriguez, E.P., Gao, X., & Ferreira, L.G. (2002). Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sensing of Environment, 83, 195-213 Huete, A.R., Liu, H.Q., Batchily, K., & vanLeeuwen, W. (1997). A comparison of vegetation indices global set of TM images for EOS-MODIS. Remote Sensing of Environment, 59, 440-451 Jeong, S.J., Ho, C.H., Brown, M.E., Kug, J.S., & Piao, S.L. (2011). Browning in desert boundaries in Asia in recent decades. Journal of Geophysical Research-Atmospheres, 116 Kefi, S., Rietkerk, M., Alados, C.L., Pueyo, Y., Papanastasis, V.P., ElAich, A., & de Ruiter, P.C. (2007). Spatial vegetation patterns and imminent desertification in Mediterranean arid ecosystems. Nature, 449, 213-U215 Klein, I., Gessner, U., & Kuenzer, C. (2012). Regional land cover mapping and change detection in Central Asia using MODIS time-series. Applied Geography, 35, 219-234 Klein, I., Gessner, U., & Künzer, C. (2014). Generation of Up to Date Land Cover Maps for Central Asia. In L. Mueller, A. Saparov, & G. Lischeid (Eds.), Novel Measurement and Assessment Tools for Monitoring and Management of Land and Water Resources in Agricultural Landscapes of Central Asia (pp. 329-346): Springer International Publishing Kuemmerle, T., Olofsson, P., Chaskovskyy, O., Baumann, M., Ostapowicz, K., Woodcock, C.E., Houghton, R.A., Hostert, P., Keeton, W.S., & Radeloff, V.C. (2011). Post-Soviet farmland abandonment, forest recovery, and carbon sequestration in western Ukraine. Global Change Biology, 17, 1335-1349 Li, J.Y., Yang, X.C., Jin, Y.X., Yang, Z., Huang, W.G., Zhao, L.N., Gao, T., Yu, H.D., Ma, H.L., Qin, Z.H., & Xu, B. (2013). Monitoring and analysis of grassland desertification dynamics using Landsat images in Ningxia, China. Remote Sensing of Environment, 138, 19-26 Li, Z., Chen, Y.N., Li, W.H., Deng, H.J., & Fang, G.H. (2015). Potential impacts of climate change on vegetation dynamics in Central Asia. Journal of Geophysical Research-Atmospheres, 120, 12345-12356 Lioubimtseva, E., & Henebry, G.M. (2009). Climate and environmental change in arid Central Asia: Impacts, vulnerability, and adaptations. Journal of Arid Environments, 73, 963-977 Loboda, T.V., Giglio, L., Boschetti, L., & Justice, C.O. (2012). Regional fire monitoring and characterization using global NASA MODIS fire products in dry lands of Central Asia. Frontiers of Earth Science, 6, 196-205 Maki, M., Ishiahra, M., & Tamura, M. (2004). Estimation of leaf water status to monitor the risk of forest fires by using remotely sensed data. Remote Sensing of Environment, 90, 441-450 McKee, T.B., Doesken, N.J., & Kleist, J. (1993). The relationship of drought frequency and duration to time scales. In, Proceedings of the 8th Conference on Applied Climatology (pp. 179-183): American Meteorological Society Boston, MA Mohammat, A., Wang, X., Xu, X., Peng, L., Yang, Y., Zhang, X., Myneni, R.B., & Piao, S. (2013). Drought and spring cooling induced recent decrease in vegetation growth in Inner Asia. Agricultural and Forest Meteorology, 178–179, 21-30 Mu, Q.Z., Zhao, M.S., Kimball, J.S., McDowell, N.G., & Running, S.W. (2013). A Remotely Sensed Global Terrestrial Drought Severity Index. Bulletin of the American Meteorological Society, 94, 83-98

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

O'Loughlin, J., Witmer, F.D.W., Linke, A.M., Laing, A., Gettelman, A., & Dudhia, J. (2012). Climate variability and conflict risk in East Africa, 1990-2009. Proc Natl Acad Sci U S A, 109, 18344-18349 Palmer, W.C. (1965). Meteorological drought. US Department of Commerce, Weather Bureau Washington, DC Piao, S.L., Cui, M.D., Chen, A.P., Wang, X.H., Ciais, P., Liu, J., & Tang, Y.H. (2011a). Altitude and temperature dependence of change in the spring vegetation green-up date from 1982 to 2006 in the Qinghai-Xizang Plateau. Agricultural and Forest Meteorology, 151, 1599-1608 Piao, S.L., Wang, X.H., Ciais, P., Zhu, B., Wang, T., & Liu, J. (2011b). Changes in satellite-derived vegetation growth trend in temperate and boreal Eurasia from 1982 to 2006. Global Change Biology, 17, 3228-3239 Robinson, T.P., Wint, G.R.W., Conchedda, G., Van Boeckel, T.P., Ercoli, V., Palamara, E., Cinardi, G., D'Aietti, L., Hay, S.I., & Gilbert, M. (2014). Mapping the Global Distribution of Livestock. Plos One, 9 Scheffer, M., Bascompte, J., Brock, W.A., Brovkin, V., Carpenter, S.R., Dakos, V., Held, H., van Nes, E.H., Rietkerk, M., & Sugihara, G. (2009). Early-warning signals for critical transitions. Nature, 461, 53-59 Seddon, A.W.R., Macias-Fauria, M., Long, P.R., Benz, D., & Willis, K.J. (2016). Sensitivity of global terrestrial ecosystems to climate variability. Nature Sternberg, T., Rueff, H., & Middleton, N. (2015). Contraction of the Gobi Desert, 2000-2012. Remote Sensing, 7, 1346-1358 Tucker, C.J., Dregne, H.E., & Newcomb, W.W. (1991). Expansion and Contraction of the Sahara Desert from 1980 to 1990. Science, 253, 299-301 Vermote, E., & Vermeulen, A. (1999). Atmospheric correction algorithm: spectral reflectances (MOD09). ATBD version, 4 Vermote, E.F., El Saleous, N.Z., & Justice, C.O. (2002). Atmospheric correction of MODIS data in the visible to middle infrared: first results. Remote Sensing of Environment, 83, 97-111 Wan, Z.M. (2008). New refinements and validation of the MODIS Land-Surface Temperature/Emissivity products. Remote Sensing of Environment, 112, 59-74 Wan, Z.M., Zhang, Y.L., Zhang, Q.C., & Li, Z.L. (2002). Validation of the land-surface temperature products retrieved from Terra Moderate Resolution Imaging Spectroradiometer data. Remote Sensing of Environment, 83, 163-180 Wessels, K.J., van den Bergh, F., & Scholes, R.J. (2012). Limits to detectability of land degradation by trend analysis of vegetation index data. Remote Sensing of Environment, 125, 10-22 Wright, C.K., de Beurs, K.M., & Henebry, G.M. (2012). Combined analysis of land cover change and NDVI trends in the Northern Eurasian grain belt. Frontiers of Earth Science, 6, 177-187 Xi, X., & Sokolik, I.N. (2015). Dust interannual variability and trend in Central Asia from 2000 to 2014 and their climatic linkages. Journal of Geophysical Research-Atmospheres, 120 Xiao, X., Boles, S., Frolking, S., Salas, W., Moore, B., Li, C., He, L., & Zhao, R. (2002). Landscape-scale characterization of cropland in China using Vegetation and landsat TM images. International Journal of Remote Sensing, 23, 3579-3594 Xiao, X.M., Boles, S., Liu, J.Y., Zhuang, D.F., Frolking, S., Li, C.S., Salas, W., & Moore, B. (2005). Mapping paddy rice agriculture in southern China using multi-temporal MODIS images. Remote Sensing of Environment, 95, 480-492 Xiao, X.M., Braswell, B., Zhang, Q.Y., Boles, S., Frolking, S., & Moore, B. (2003). Sensitivity of vegetation indices to atmospheric aerosols: continental-scale observations in Northern Asia. Remote Sensing of Environment, 84, 385- 392 Zhang, G.L., Xiao, X.M., Dong, J.W., Kou, W.L., Jin, C., Qin, Y.W., Zhou, Y.T., Wang, J., Menarguez, M.A., & Biradar, C. (2015). Mapping paddy rice planting areas through time series analysis of MODIS land surface temperature and vegetation index data. ISPRS Journal of Photogrammetry and Remote Sensing, 106, 157-171 Zhang, G.L., Zhang, Y.J., Dong, J.W., & Xiao, X.M. (2013). Green-up dates in the Tibetan Plateau have continuously advanced from 1982 to 2011. Proc Natl Acad Sci U S A, 110, 4309-4314 Zhang, Y., Peng, C.H., Li, W.Z., Tian, L.X., Zhu, Q.Q., Chen, H., Fang, X.Q., Zhang, G.L., Liu, G.M., Mu, X.M., Li, Z.B., Li, S.Q., Yang, Y.Z., Wang, J., & Xiao, X.M. (2016). Multiple afforestation programs accelerate the greenness in the 'Three North' region of China from 1982 to 2013. Ecological Indicators, 61, 404-412 Zhou, L.M., Tucker, C.J., Kaufmann, R.K., Slayback, D., Shabanov, N.V., & Myneni, R.B. (2001). Variations in northern vegetation activity inferred from satellite data of vegetation index during 1981 to 1999. Journal of Geophysical Research-Atmospheres, 106, 20069-20083 Zhou, Y., Zhang, L., Fensholt, R., Wang, K., Vitkovskaya, I., & Tian, F. (2015). Climate Contributions to Vegetation Variations in Central Asian Drylands: Pre- and Post-USSR Collapse. Remote Sensing, 7, 2449-2470

Confidential manuscript submitted to Journal of Geophysical Research: Earth Surface

Figure legend Figure 1 Desertification theory of transition between two stable states (unvegetated and vegetated) in bistable ecosystem, referred to D'Odorico et al., 2013. Figure 2 Location of study area. (a) Spatial distribution of land cover types in Central Asia based on MCD12Q1 IGBP classification scheme in 2010; (b) Specific targeted area of grassland and desert region in Central Asia. We combined IGBP-based open and closed shrublands into one class in (a). Figure 3 Flow chart illustrating the data sources and methodology in this study. Figure 4 Variation in standard deviation of EVI-indicated grassland area in targeted area of Central Asia over the period 2000-2014. Figure 5 Geographical pattern of grassland desertification. (a) Spatial pattern of first occurrence of desert for each pixel over the period 2000-2014; (b) Distribution of desert area along latitude gradient in each year. Figure 6 (a) Spatial pattern of frequency of desert in Central Asia; (b) Spatial distribution of Never or Occasionally Deserted Zone (NODZ), Transition Zone (TZ), and Persistently Desert Zone (PDZ). The inset in (a) shows the distribution of area percent of different frequency of desert, and the inset in (b) shows the area percent of different zones.

Figure 7 Mean anomalies in averaged EVITGS in (a) targeted area, (b) NODZ, (c) TZ, and (d) PDZ of Central Asia over the period 2000-2014. Figure 8 Spatial patterns of (a) Linear Regression (LR) trend and (b) Mann-Kendall (MK) trend of averaged EVITGS in targeted area of Central Asia over the period 2000-2014. The insets show the region with statistically significant at 0.05 level. Figure 9 Mean anomalies in (a) Drought Severity Index (DSI) from 2000-2011, Palmer Drought Severity Index (PDSI) from 1950-2012, and Standardized Precipitation Index (SPI) from 1950- 2012, (b) rainfall from 1950-2013, and (c) mean temperature from 1950-2013 in targeted area of Central Asia. Figure 10 Spatial patterns of Linear Regression trend of (a) Drought Severity Index (DSI) from 2000-2011, (b) Standardized Precipitation Index (SPI) from 2000-2012, (c) Palmer Drought Severity Index (PDSI) from 2000-2012, (d) annual precipitation from 2000-2013, and (e) annual mean temperature from 2000-2013. In each map, regions with plus symbol have a linear trend that is statistically significant at P<0.05.