Have China’s drylands become wetting in the past 50 years?

  • ZHANG Yu , 1, 2, 3 ,
  • ZHANG Yangjian 1, 2, 4 ,
  • CHENG Liang 5 ,
  • CONG Nan 1 ,
  • ZHENG Zhoutao 1 ,
  • HUANG Ke 1, 6 ,
  • ZHANG Jianshuang 1 ,
  • ZHU Yixuan 1, 2 ,
  • GAO Jie 1, 2 ,
  • SUN Yihan 1, 2
  • 1. Key Laboratory of Ecosystem Network Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China
  • 2. College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100190, China
  • 3. College of Life Science, Beijing Normal University, Beijing 100875, China
  • 4. CAS Center for Excellence in Tibetan Plateau Earth Science, Beijing 100101, China
  • 5. Beijing Applied Atmospheric Institute, Beijing 100029, China
  • 6. Department of Geosciences and Natural Resource Management, University of Copenhagen, Copenhagen 1350, Denmark
*Zhang Yangjian, PhD and Professor, E-mail:

Zhang Yu, PhD Candidate, specialized in global change ecology. E-mail:

Received date: 2021-11-12

  Accepted date: 2022-03-18

  Online published: 2023-01-16

Supported by

The Major Program of National Natural Science Foundation of China(41991234)

The National Science Fund for Distinguished Young Scholars(41725003)


© 2023


Recently, whether drylands of Northwest China (NW) have become wetting has been attracting surging attentions. By comparing the Standard Precipitation Evapotranspiration Indices (SPEI) derived from two different potential evapotranspiration estimates, i.e., the Thornthwaite algorithm (SPEI_th) and the Penman-Monteith equation (SPEI_pm), we try to resolve the controversy. The analysis indicated that air temperature has been warming significantly at a rate of 0.4°C decade−1 in the last five decades and the more arid areas are more prone to becoming warmer. Annual precipitation of the entire study area increased insignificantly by 3.6 mm decade−1 from 1970 to 2019 but NW presented significantly increasing trends. Further, the SPEI_th and SPEI_pm demonstrated similar wetting-drying-wetting trends (three phases) in China’s drylands during 1970-2019. The common periodical signals in the middle phase were identified both by SPEI_th and SPEI_pm wavelet analysis. Analysis with different temporal intervals can lead to divergent or even opposite results. The attribution analysis revealed that precipitation is the main climatic factor driving the drought trend transition. This study hints that the wetting trend’s direction and magnitude hinge on the targeted temporal periods and regions.

Cite this article

ZHANG Yu , ZHANG Yangjian , CHENG Liang , CONG Nan , ZHENG Zhoutao , HUANG Ke , ZHANG Jianshuang , ZHU Yixuan , GAO Jie , SUN Yihan . Have China’s drylands become wetting in the past 50 years?[J]. Journal of Geographical Sciences, 2023 , 33(1) : 99 -120 . DOI: 10.1007/s11442-022-2067-5

1 Introduction

Drylands refer to areas subject to permanent or seasonal water deficiency due to low precipitation and high potential evapotranspiration (PET) (Huang et al., 2017a, Lian et al., 2021), and they currently cover about 46% of global land surface and support over a third of the world population (IPCC, 2021). Drylands in China are the largest among all the developing countries (Yao et al., 2021) and they are faced with a variety of environmental and human subsistence issues (Fu et al., 2021). In resolving these issues, Chinese government has launched a set of re-vegetation programs (Feng et al., 2019), through which extensive vegetation greening over China’s drylands has been achieved (Chen et al., 2019). In selecting regions suitable for implementing the ecological restoration projects, determining the regional climate trends is a necessary step (Yao et al., 2021).
Dryland ecosystems are vulnerable to global changes. Under aggravated climate stresses, global dryland ecosystems suffer varying degrees of ecological degradation (Liu et al., 2019; Yao et al., 2019; Lobell et al., 2020), exhibited as systemic or abrupt status changes (Berdugo et al., 2020). With their extensive distributions and being exposed to regionally varied climates, dryland ecosystems must exhibit a non-uniform response. As a key region to achieve the global sustainable development goals (UN, 2015), clarifying climate change patterns are of utmost importance to understand ecosystem dynamics, also the basis to ensure ecological security in drylands. Although Northwest China has been widely reported to become wetting in the past decades, the trends contain high spatio-temporal uncertainties (Table 1).
Table 1 Summary of drying or wetting trends in China’s drylands
Location Period Data source Index Methods of regional
Related finding Reference
Northwest China 1950s-2000s Observation Temperature; Precipitation; Glacier melt water; Runoff; Lake level; Flood disasters; Vegetation cover; Dust storm days Geographic boundary A climate mutation to warm wet (1987 abrupt) Shi et al. (2003)
North China 1948-2008 CPC
Aridity Index (AI) Climate
Dryland extent has extended Li et al. (2015)
China 1961-2012 CRU SPEI_pm; PDSI_pm; sc_PDSI_pm; PDSI_th; sc_PDSI_th Köppen
Arid region and Tibetan Plateau (TP):↑; Semiarid regions:↓ Chen
et al. (2017)
Arid regions of Central Asia 1950-2015 CRU SPEI_pm; PDSI_th; PDSI_pm; sc_PDSI_th; sc_PDSI_pm Geographic and climatic
Xinjiang region and Hexi Corridor:↑ Hu et al. (2018)
China 1961-2016 CMA AI Geographic and climatic
Northwest China and TP:↑; North and Northeast China:↓ Liu et al. (2018)
Xinjiang, China 1961-2015 CMA SPEI_th Geographic boundary 1961-1996:↑;
Yao et al. (2018)
Eurasia and Xinjiang, China 1961-2019 CRU Temperature; Precipitation; SPI; PDSI Topographic characteristics and geographic boundary Wetting trend not a regional episode for Xinjiang Wang
et al. (2020)
Xinjiang, China 1963-2018 CRU SPEI_pm Geographic boundary Overall trend:↑ (1988 abrupt);
Southwest and Southeast part:↓
Song et al. (2021)

Note: ↑ (wetting trend); ↓ (drying trend).

In linking the wetting or drying climates to ecological process, it is better to consider a comprehensive set of climatic factors, rather than individual one (Cheng et al., 2016). Drought is generally categorized into five types: meteorological drought, agricultural drought, hydrological drought, socioeconomic drought, and ecological drought (Slette et al., 2019; West et al., 2019; Lian et al., 2021). Among the five types, meteorological drought is the basis of understanding other types (Wilhite et al., 1987). As a landmark of meteorological drought (represented as drought in this study) indices, the Palmer Drought Severity Index (PDSI) (Palmer et al., 1965) incorporates water supply and demand into a hydrologic system and is more capable of reflecting local moisture conditions (Trenberth et al., 2013). Afterwards the multi-scalar drought indices were developed, such as the Standard Precipitation Index (SPI) (Mckee et al., 1993) and the Standard Precipitation Evapotranspiration Index (SPEI) (Vicente-Serrano et al., 2010). The advantages of both SPI and SPEI lie in their ability to indicate drought conditions at different temporal intervals (e.g., 1, 3, or 12 months) (Wang et al., 2021). The World Meteorological Organization (WMO) recommends SPI as the standard drought indices. However, SPI can not reflect drought conditions caused by global warming (Berdugo et al., 2020; Ding et al., 2021). SPEI is calculated by incorporating both PET and precipitation (Vicente-Serrano et al., 2010). Compared to SPI and PDSI, SPEI is more relevant to soil moisture (Xu et al., 2021). For the same SPEI, largely different or even opposite trends have been found when PET was estimated by different methods (Yao et al., 2019; Aadhar et al., 2020; Li et al., 2020). The typical examples are the widely used Thornthwaite algorithm (Thornthwaite et al., 1948) and Penman-Monteith (PM) equation (Allen et al., 1998) (Peng et al., 2019; Anderegg et al., 2020; Chen et al., 2020; Hu et al., 2020; Xu et al., 2020, 2021; Jiao et al., 2021). It is well known that PET estimated value by PM equation is more consistent with pan evaporation than by Thornthwaite algorithm (Chen et al., 2005) because of its more comprehensive representation on physical processes. Climatic variables incorporated in PM equation include air temperature, relative humidity, wind speed, and sunshine duration (Allen et al., 1998). Some of these variables are not well available on a large scale (Hoffmann et al., 2020). Consequently, many studies still utilize the traditional Thornthwaite algorithm for SPEI due to a limited data availability (Chen et al., 2017; Xu et al., 2020; Xu et al., 2021). In recognition of the limitations of each PET calculation, robust conclusions can better be reached by integrating a variety of indices. Their consistent results suggest that actual changes are significant enough to override biases among these indices (Chen et al., 2017).
Wetting could be consequences of enhanced precipitation, or weakened evaporation under decreased temperature (Hu et al., 2019; Wang et al., 2020). Drying could be brought about by less precipitation or strengthened evaporation under warming (Cook et al., 2014; Dilinuer et al., 2021). Disentangling the major climate factor causing the wetting or drying is the basis to project future trend, also the foundation to implement appropriate measures to adapt.
The objectives of this study are to explore temperature, precipitation, and drought indices change trends in drylands of China. Two commonly used drought indices (SPEI), i.e., based on the Thornthwaite algorithm (SPEI_th) and the Penman-Monteith equation (SPEI_pm), were utilized. Herein, we try to (1) investigate changes in temperature and precipitation from 1970 to 2019; (2) compare the drying or wetting trends between calculated SPEI_th and SPEI_pm; (3) clarify the periodic characteristics of drought indices in the past 50 years; (4) identify the climatic variables controlling SPEI_th and SPEI_pm trends. The research findings of this study can shed further light on drying or wetting trends over China’s drylands.

2 Study area

The dryland extent was defined by the Aridity Index (AI) < 0.65 (AI = Precipitation / PET, with annual average) based on climate data of Climatic Research Unit (CRU) datasets (http://data.ceda.ac.uk/badc/cru/data) from 1901 to 2017. Then it was further divided into four subtypes of hyper-arid (AI<0.05), arid (0.05≤AI<0.2), semi-arid (0.2≤AI<0.5) and dry sub-humid (0.5≤AI<0.65) regions (Huang et al., 2017b) (Figure 1a).
Figure 1 China’s drylands (a) and geographic boundary (b). In this study area, 206 meteorological stations (green dots) in China’s drylands (1970-2019) were used. No.1, Northwest China; No.2, North China; No.3, Northeast China; over drylands regions
To refine the analysis, the entire study area was divided into Northwest (NW), North (NC), and Northeast China (NE) drylands (Figure 1b) based on geographic divisions: (1) NW featuring mountain-basin-desert terrain; (2) NC featuring wide range cropland; (3) NE featuring cold winter and limited thermal resources.

3 Data and methodology

3.1 Data retrieving

Meteorological data, comprising the observed monthly mean, maximum, and minimum air temperature (Tmean, Tmax, and Tmin), precipitation, relative air humidity (RH), wind speed, and sunshine duration, were obtained from China Meteorological Administration (CMA) (http://data.cma.cn). In total, 206 meteorological stations (Figure 1) data covering the period of 1970-2019 were utilized. Some stations over NW started to provide data since the 1960s. Thus, the target period in this study is from 1970 to 2019. Temporal and spatial consistency control on the meteorological data were conducted (Figure S1). Only stations with continuous observations were utilized. To ensure that each grid box contains at least one meteorological station with continuous observations, 96 grid boxes were generated with a spatial resolution of 2° by 2° (Zhang et al., 2016). For grid boxes containing more than one station, their mean values were utilized.

3.2 Calculations of drought indices

The drying and wetting trends were assessed based on two commonly used SPEI, i.e. Thornthwaite algorithm (SPEI_th) and PM equation (SPEI_pm).

3.2.1 PET calculation:

Two PET estimation methods, i.e. Thornthwaite algorithm (PET_th, Thornthwaite et al., 1948) and Penman-Monteith (PM) equation (PET_pm) (Allen et al., 1998), were used in this study.
$PET\_th=16K{{\left( \frac{10{{T}_{mean}}}{I} \right)}^{m}}$
$m=6.75\times {{10}^{7}}{{I}^{3}}-7.71\times {{10}^{5}}{{I}^{2}}+1.79\times {{10}^{2}}I+0.492$
$I=12{{\left( \frac{{{T}_{mean}}}{5} \right)}^{1.514}}$
where K is a correction factor as a function of latitude and month; Tmean, Tmax, and Tmin are monthly mean, maximum, and minimum air temperature (℃), respectively; I is a heat index; m is a coefficient depending on I.
$PET\_pm=\frac{0.408\text{ }\!\!\Delta\!\!\text{ }({{R}_{n}}-G)}{\text{ }\!\!\Delta\!\!\text{ }+\gamma (1+0.34{{u}_{2}})}+\frac{\gamma \frac{900}{{{T}_{mean}}+273}{{u}_{2}}({{e}_{sat}}-{{e}_{a}})}{\text{ }\!\!\Delta\!\!\text{ }+\gamma (1+0.34{{u}_{2}})}$
where Δ is the slope of the saturated vapor pressure (kPa ℃−1); Rn is net radiation (MJ m−2 day−1); G is soil heat flux density (MJ m−2 y−1); γ is psychrometric constant (kPa ℃−1); u2 is wind speed at a 2 m height (m s−1); esat and ea are saturated and actual vapor pressures (kPa), respectively.
Rn was calculated by:
${{R}_{ns}}=(1-\alpha ){{R}_{s}}$
${{R}_{s}}=\left[ {{a}_{s}}+{{b}_{s}}\left( \frac{n}{N} \right) \right]{{R}_{a}}$
${{R}_{nl}}=\sigma \left( \frac{T_{\max,k}^{4}+T_{\text{min,}k}^{4}}{2} \right)\left( 0.34-0.14\sqrt{{{e}_{a}}} \right)\left( 1.35\frac{{{R}_{s}}}{{{R}_{s0}}}-0.35 \right)$
where Rns, Rnl, Rs, Ra, and Rso are net shortwave, net longwave, shortwave, extraterrestrial, and clear sky radiations (MJ m−2 month−1), respectively; α is albedo (0.23); as and bs are regression constants; n is sunshine duration (hr); N is maximum possible sunshine durations (hr); σ is Stefan-Boltzmann constant (4.903 × 10−9 MJ K−4 m−2 d−1); Tmax,k and Tmin,k are maximum and minimum absolute air temperatures (K), respectively.
The monthly G was estimated by:
where the subscripts K and K−1 represent the order of each month.

3.2.2 A simplified water balance for the analyzed month

where Dj value is the difference between monthly precipitation Pj and potential evapotranspiration PETj; The accumulated difference Di,l value in a given month j and year i depends on the chosen temporal scale of k. For example, the 12-month temporal duration is calculated by:
$X_{i,j}^{k}=\underset{l=13-k+j}{\overset{12}{\mathop{\sum }}}\,{{D}_{i}}_{-1,l}+\underset{l=1}{\overset{j}{\mathop{\sum }}}\,{{D}_{i,l}}, j<k$
$X_{i,j}^{k}=\underset{l=j-k+1}{\overset{j}{\mathop{\sum }}}\,{{D}_{i,l}}, \text{ }\!\!~\!\!\text{ }j\ge k$
where Di,l is the P-PET difference in the first month of the year i.

3.2.3 Normalize the water balance to obtain the SPEI index series

A three-parameter log-logistic distribution was used, with the probability density function (PDF) and probability distribution function expressed as:
$f(x)=\frac{\beta }{\alpha }{{\left( \frac{x-\gamma }{\alpha } \right)}^{\beta -1}}{{\left[ 1+{{\left( \frac{x-\gamma }{\alpha } \right)}^{\beta }} \right]}^{-2}}$
$F(x)={{\left[ 1+{{\left( \frac{\alpha }{x-\gamma } \right)}^{\beta }} \right]}^{-1}}$
where α, β, and γ represent scale, shape, and origin parameters, respectively. SPEI can be obtained as standardized values of F(x) (Vicente-Serrano et al., 2010):
where $W=\sqrt{-2\ln (P)}$ for p≤0.5, p=1–F(x). If p>0.5, p is replaced by 1–p and the constants are C0=2.515517, C1=0.802853, C2=0.010328, d1=1.432788, d2=0.189269 and d3=0.001308.

3.3 Analysis

The dynamics of Tmean and precipitation, and drought indices (SPEI_th and SPEI_pm) were indicated by their linear trends. The definition of SPEI stipulates that if the trend value of SPEI is below 0, it denotes a drying trend and vice versa. The piecewise linear regression was used to quantify the magnitudes in the linear trends of precipitation, SPEI_th, and SPEI_pm (Toms et al., 2003). Considering potential non-linear changes in drought indices, continuous wavelet transform (CWT) method was used to explore periodic features in drought indices. In addition, wavelet transform coherence (WTC) describes the relationship strength between the two nonstationary time series in the time and frequency domains, and it was performed to detect the regions in time and frequency space where the two annual time series (drought indices and input climatic variables) covary. In this study, the phase difference in the two series was denoted as θ, indicating the relative change in the lead or lag state, which reflects the positive or negative coherence of the two series. When $\theta \in \left[ -\frac{\text{ }\!\!\pi\!\!\text{ }}{\text{2}},\ \frac{\text{ }\!\!\pi\!\!\text{ }}{\text{2}} \right]$, the two series keep in the same change pattern, i.e., increasing or decreasing simultaneously allowing for fine phase differences. When$\theta \in \left[ -\text{ }\!\!\pi\!\!\text{ },\ -\frac{\text{ }\!\!\pi\!\!\text{ }}{\text{2}} \right)$ or $\left( \frac{\text{ }\!\!\pi\!\!\text{ }}{2},\text{ }\!\!\pi\!\!\text{ } \right]$, the two series are treated as inverse changes. On the other hand, by measuring mean WTC and percent area of significant coherence (PASC) relative to the wavelet scale-location domain, the ability of climatic variables in explaining variations of the drought indices was assessed (Su et al., 2019). A greater mean WTC with larger PASC indicates that more SPEI_th or SPEI_pm variations are explained by a particular climatic variable. More details about CWT and WTC methods were provided by Grinsted et al. (2004). The code of wavelet analysis was obtained from MATLAB Toolbox (http://grinsted.github.io/wavelet-coherence/).

4 Results

4.1 Trends of temperature and precipitation

The 206 meteorological station average showed that air temperature has been warming at a rate of 0.4℃ decade−1 (p < 0.05) over China’s drylands in the last five decades (Figure 2a), approximately 2.2 times the global average (WMO, 2020). Annual precipitation increased insignificantly by 3.6 mm decade−1 (p = 0.30) in the same period (Figure 2b). The piecewise linear regression revealed that precipitation increased slightly prior to the late 1990s (slope = 0.38 mm yr−1, p < 0.05), but increased significantly (slope = 3.89 mm yr−1, p < 0.05) in the later period (Figure 2b).
Figure 2 Time series of the annual mean air temperature (a) and precipitation (b). All green dots are averaged for 206 meteorological stations in China’s drylands (1970-2019). The green shaded area indicates the 95% confidence interval. The blue dotted lines indicate changes in slopes of precipitation.
The significant positive trend in air temperature from 1970 to 2019 exhibits high spatial heterogeneity (Figure 3a). Among the three regions, warming rate was the highest in NC (0.40℃ decade−1) and NW (0.37℃ decade−1) but lowest in NE (0.33℃ decade−1). In NW, the hyper-arid region warmed at a rate higher than the semi-arid region in Xinjiang, and the highest warming rate occurred at Fuyun (hyper-arid region) in northern Xinjiang. In NC, warming rate for the arid regions of Inner Mongolia was higher than the dry sub-humid regions of North China Plain. These results suggest that more arid areas are more prone to becoming warmer in the past 50 years.
Figure 3 Long-term trends during 1970-2019 over China’s drylands in annual mean air temperature (a) and precipitation (b-d). Comparison of precipitation trends over China’s drylands between pre-turning point (1970-1998) (c) and post-turning point (1999-2019) (d). The black dots indicate the trends are statistically significant at the 95% significance level.
Spatial trends in annual precipitation from 1970 to 2019 also demonstrate great spatial variabilities (Figure 3b). Among the three regions, the highest positive trend occurred in NW. Meanwhile, most of Xinjiang (slope = 0.79 mm yr−1, p < 0.05), and headwaters of Yellow River (slope = 1.69 mm yr−1, p < 0.05) and Yangtze River (slope = 2.27 mm yr−1, p < 0.05) over NW presented significantly positive trends. However, precipitation decreased in NC, except for the region of Loess Plateau (LP). Figures 3c and 3d display the spatial trends of precipitation before and after the late 1990s. It was observed that precipitation increased slightly before the late 1990s (0.41 mm yr−1, NW; 0.88 mm yr−1, NC; 1.65 mm yr−1, NE), but increased markedly afterwards (2.36 mm yr−1, NW; 4.93 mm yr−1, NC; 7.19 mm yr−1, NE) with stronger trends. Precipitation increased significantly in NC and NE after the late 1990s. A high contrast precipitation trend was identified in NW and NC between prior and posterior the late 1990s, especially in the headwaters of Yellow River and Yangtze River, and LP (Figures 3c and 3d).

4.2 Trends of SPEI_th and SPEI_pm

To indicate the overall change trends in drought indices over China’s drylands, temporal changes of SPEI_th and SPEI_pm with multi-temporal intervals (1-, 3-, and 12-months) were plotted (Figure 4). The shorter temporal intervals (1- and 3-months) of SPEI_th and SPEI_pm detected more drought events (frequency and severity). In identifying long-term drying and wetting trends, the longer temporal interval (12-months) is more appropriate. The 12-month SPEI_th presented significant negative (drying) trend (slope = -0.0164 yr−1, p < 0.05) against positive (wetting) trend (slope = 0.0030 yr−1, p < 0.05) indicated by the 12-month SPEI_pm in the last 50 years (Figure 5).
Figure 4 Multi-scale (1-, 3-, and 12-month) SPEI time series over China’s drylands (1970-2019). Time series of two different SPEI estimates forced with Thornthwaite algorithm (denoted as SPEI_th) and with Penman-Monteith equation (denoted as SPEI_pm), respectively.
Figure 5 Comparisons of 12-month SPEI_th (a) and SPEI_pm (b) over China’s drylands from 1970 to 2019. Time series of two different SPEI estimates forced with Thornthwaite algorithm (denoted as SPEI_th) and with Penman-Monteith equation (denoted as SPEI_pm). The blue shaded areas indicate the 95% confidence interval. The red dotted lines indicate the changes in the slopes of SPEI.
The 12-month SPEI_th and SPEI_pm identified three phases of changes in drought (Figure 4). In the earlier phase, drought condition (drying or wetting trend) indicated by SPEI_pm exhibited a more apparent wetting trend. In the middle phase, the variations in SPEI_th and SPEI_pm were nearly flat. But they presented an aggravated situation of drought afterwards, and SPEI_pm showed greater fluctuations (Figure 4). Further, three phases of drought in the last five decades were detected by piecewise linear regression (Figure 5). SPEI_th and SPEI_pm demonstrated almost consistent wetting-drying-wetting trends through the three phases (Figure 5).
The 12-month SPEI_th presented a significant negative trend in most areas, especially in the arid NW (Figure 6a). In comparison, most parts of NW showed significant wetting trends according to the 12-month SPEI_pm (Figure 6b). The consistency between SPEI_th and SPEI_pm was high in the relatively humid regions (i.e., the semi-arid and dry sub-humid regions of NC and NE) (Figure 1). A large proportion of NC presented consistent drying trends according to the two indices in the last five decades, except for LP. The results indicate that different estimation methods turned out distinct SPEI trends in the arid NW during 1970-2019 (Figures 3a and 3b).
Figure 6 Long-term trends for 1970-2019 over China’s drylands according to 12-month SPEI_th (a) and SPEI_pm (b). Two different SPEI estimates forced with Thornthwaite algorithm (denoted as SPEI_th) and with Penman-Monteith equation (denoted as SPEI_pm), respectively. The black dots indicate the trends are statistically significant at the 95% significance level.
Spatial trends in 12-month SPEI_th and SPEI_pm for the first and third phases were exhibited (Figure 7). The middle phase was not considered due to its short duration (Figure 5). Distinct results were found between the SPEI_th and SPEI_pm before the late 1990s (Figures 7a and 7c). However, they demonstrated an almost consistent wetting trend in the NW (i.e., headwaters of the Yellow River and Yangtze River, and LP), NC, and NE in the later period (Figures 7b and 7d).
Figure 7 Comparison of 12-month SPEI_th (a-b) and SPEI_pm (c-d) trends over China’s drylands between pre-turning point and post-turning point periods. SPEI_th trend of 1970-1995 (a), SPEI_th trend of 1999-2019 (b); SPEI_pm trend of 1970-1993 (c), SPEI_pm trend of 2000-2019 (d). Two different SPEI estimates were forced with Thornthwaite algorithm (denoted as SPEI_th) and with Penman-Monteith equation (denoted as SPEI_pm), respectively. The black dots indicate the trends are statistically significant at a 95% significance level.

4.3 Periods of SPEI_th and SPEI_pm

Global dryland ecosystems normally experience greater changes in response to environmental changes, e.g., droughts, CO2 fertilization, invasions of exotic species, grazing, and disturbances (Burrell et al., 2020). All these processes induce non-linear effects on dryland ecosystem-climate interactions. Considering the potential non-linear changes in 12-month SPEI_th and SPEI_pm, continuous wavelet transform (CWT) method was used to explore their cyclical pattern (Figure 8). The thick black contours represent the 95% significance level against yellow noise. The thin black contours indicate the wavelet spectrum’s cone of influence (COI), and the pale regions outside COI denote no periodic characteristics. The common periodical signals were presented by SPEI_th and SPEI_pm wavelet patterns.
Figure 8 Continuous wavelet transforms (CWTs) for 12-month SPEI_th (a) and SPEI_pm (b) over China’s drylands from 1970 to 2019. The two different SPEI estimates are forced by Thornthwaite algorithm (denoted as SPEI_th) and Penman-Monteith equation (denoted as SPEI_pm), respectively. The period is measured in years. Each subplot shows the CWT in a region. NW = Northwest China’s drylands; NC = North China’s drylands; NE = Northeast China’s drylands; Total = China’s drylands. The color denotes the strength of the wavelet power. Tick contours denote 95% significance levels against yellow noise. Pale regions indicate the cone of influence where edge effects might bias the results.
Specifically, in NW, the periodical signal of SPEI_th was weaker than that of SPEI_pm during 1986-1997, but all cyclical durations were shorter than 5 years. SPEI_th and SPEI_pm had strong periodical signals (4 to 8 years during 1997-2012) in NC. In NE, they exhibited 2- to 8-year primary periodical signals during 1992-2010, and the significant periodical signal of SPEI_th was slightly weaker than that of SPEI_pm. For the entire study area, SPEI_th (2 years during 1975-1979 and 1990-1998, 5 years during 2003-2005) and SPEI_pm (2 years during 1992-1998, 5 years and 8 years during 2001-2009) both had significant periodical signals. Hence, in the middle phase (1996-1998, SPEI_th; 1994-1999, SPEI_pm), significant oscillations (around 2-8 years) existed in SPEI_th and SPEI_pm over China’s drylands during 1970-2019. The steep downward trends in the middle phase, as demonstrated by SPEI_th and SPEI_pm, lead to almost consistent rapid drying trends in that period (Figure 5).

4.4 Climatic variables controlling SPEI_th and SPEI_pm variabilities

The strong coherences (the yellow area inside the COI) and the high percent area of significant coherences (PASCs) were identified in the study area (Figures S2 and S3).
Effects of all input climatic variables on variations in 12-month SPEI_th and SPEI_pm were summarized (Figure 9). The mean WTCs between drought indices and precipitation, surpassed 0.50 for all regions, especially in NC (surpassed 0.86) and NE (larger than 0.79) (Figure 9a). It reflects high covariance between drought indices and precipitation. Hence, precipitation was the dominant climatic variable for SPEI_th and SPEI_pm over China’s drylands. Separately, Tmax was second to SPEI_th, and the mean WTCs ranged from 0.39 to 0.49. The highest coherence (0.49) between SPEI_th and Tmax occurred in NW, which is comparable to precipitation (0.50). For SPEI_pm series, RH (0.50) and sunshine duration (0.47) were next in importance, with higher coherence over China’s drylands. But differences of WTCs between SPEI_th and SPEI_pm had the most apparent response to precipitation in NW (22%), where precipitation showed significantly positive trends (Figure 3b).
Figure 9 Mean wavelet transform coherences (WTCs) between drought indices (SPEI_th and SPEI_pm) and input climatic variables (a); the percent area of significant coherences (PASCs) for the WTCs between drought indices and input climatic variables (b). NW = the Northwest China’s drylands; NC = the North China’s drylands; NE = the Northeast China’s drylands; Total = the China’s drylands; Pre = precipitation; Tmean = mean air temperature; Tmax = mean max air temperature; Tmin = mean minimum air temperature; RH = relative air humid; Ws = wind speed; Sd = sunshine duration.
Regarding PASCs for WTCs (Figure 9b), precipitation made greater contributions than other climatic variables to variations in SPEI_th and SPEI_pm, especially in NC (over 83%) and NE (larger than 75%). Separately, for SPEI_th, Tmax stood out as the most influential factor in the two regions (NC, NE), followed by precipitation. For SPEI_pm, PASCs of precipitation ranged from 64% to 84%, with a mean value of 72%. And sunshine duration and RH were in the top two positions (i.e., NW, NC; and NE), respectively. But the greatest differences of PASCs between SPEI_th and SPEI_pm varied with precipitation in NW (50%).

5 Discussion

This study comprehensively assessed the drying or wetting trends in China’s drylands by adopting results of several common drought indices. Compared with each individual climatic factor, drought indices can more comprehensively reflect moisture conditions due to their capability of simultaneously considering water supply from precipitation and water demand by PET. Overall, we found that China’s drylands experienced a wetting trend in the last two decades, and more arid parts are more prone to becoming warmer. However, high spatial heterogeneities are imbedded in the overall trend, and different drought indices even turned out opposite trends in some parts.

5.1 More arid areas are more prone to becoming warmer

For the several typical regions of China’s drylands, including NW and NC, their arid parts are more prone to becoming warmer than their relatively humid counterparts (Figure 1).
Similar findings have also been reported for other parts of the globe (Ji et al., 2014; Guan et al., 2015; Huang et al., 2017b). Anthropogenic emitted CO2 and other greenhouse gas are recognized as the main causes of global warming (IPCC, 2021). The phenomena of enhanced warming over more arid areas can be explained by thermodynamic mechanism driven by soil moisture and vegetation growth (Figure 10). Taking the hyper-arid and semi-arid regions as an example, the former (Figure 10a) normally possesses low vegetation coverage and soil moisture content, whose limited evaporation and transpiration rates would lead to a low latent heat flux (LH) compared with the latter (Figure 10b). Another contributing factor can be land surface albedo. The dry condition and low vegetation coverage of the hyper-arid regions leads to their higher albedo (Chapin et al., 2011). The high albedo reduces energy absorption and reflects a higher proportion of energy back to atmosphere, together with higher upward net longwave radiation (LW). To release heat from radiation by sensible heat flux (SH), surface temperatures over more arid areas would increase sharply to create a wider land-air temperature gradient, consequently causing a higher SH and a higher warming rate compared with the relatively humid region (Huang et al., 2017b). For the relatively humid regions, extra heat expended on evapotranspiration (Neelin et al., 2003) reduces the warming magnitude.
Figure 10 The thermodynamic mechanism of enhanced warming-more arid areas (SW = shortwave radiation; LW = longwave radiation; LH = latent heat flux; SH= sensible heat flux; G = ground heat flux)
The specific heat capacity of the more arid land surface is usually lower than that of the semi-arid counterparts, and the same amount of heat brings in enhanced warming over the former. In addition, more anthropogenic aerosols are emitted in the relatively humid region than in the extremely arid drylands (Huang et al., 2017b). Anthropogenic aerosols can directly cool land surface by scattering or absorbing sunlight. Furthermore, cloud conditions can be different between super-arid and less arid drylands. Optically thick low clouds are efficient reflectors of the incoming sunlight, and they occur more in relatively humid lands. Warming effect caused by LW can be limited due to low cloud tops, which leads to atmospheric cooling and humidifying. However, there are fewer clouds over the hyper-arid region due to low air humidity. Above all, the physical processes of enhanced warming over super-arid drylands are vital for interpreting warming patterns over global drylands.

5.2 Wetting trends in China’s drylands

The drought indices all reveal an overall wetting trend in China’s drylands in the last two decades, with embedded high spatial heterogeneities.
The NW is located at the junction region of westerly wind and Asian monsoon. The cold and dry westerly winds prevail in winter. The warm and moist vapor carried by Asian monsoon can barely feed NW in summer, consequently forming the aridity-dominated climate over NW (Yu et al., 2003; Cheng et al., 2006). In line with our findings, previous related studies also reported rising temperature and enhanced precipitation in NW of China’s drylands (Cheng et al., 2006; Zhao et al., 2011; Li et al., 2012; Qin et al., 2021). The weakened Siberian High is the primary reason causing the warming in NW (Li et al., 2012). The strengthened Western Pacific Subtropical High (WPSH) and North America Subtropical High (NASH) after the mid-1980s play a key role in enhancing precipitation in NW (Li et al., 2016). Although precipitation showed a positive trend in the last five decades in most areas of NW, its nature of arid climate remains unchanged (Wang et al., 2020). In addition, expanded agricultural irrigation and urban development can significantly increase land surface evaporation and provide additional moisture for precipitation in NW. Accelerated glacier ablation and rising water levels in rivers and lakes caused by climate warming and aerosol emissions may further enhance regional climate changes (Ren et al., 2016). Since the 1960s, rising temperature, enhanced precipitation, surging river runoff, accelerated glacier ablation, rising water level in rivers and lakes, and vegetation destruction have been widely reported (Shi et al., 2003; Shi et al., 2007; Peng et al., 2017). Overall, the complex driving forces causing the enhanced precipitation in NW and moisture source attribution entail being disentangled from atmospheric circulation in the future studies.
Compared with NW, precipitation decreased in NC during 1970-2019, except for the area of LP (Figure 3b). Due to a dominant contribution of precipitation to SPEI_th and SPEI_pm, NC exhibited overall drying trends in the past five decades, which are consistent with other related studies (Chen et al., 2015; Huang et al., 2017a; Liu et al., 2018). Interestingly, precipitation increased over NC for the first and third phases (Figures 3c and 3d), with an almost consistent wetting trend after the late 1990s according to the two indices (Figures 7b and 7d). Hence, the statistically decreased precipitation in NC over the past 50 years is mainly caused by the sharp precipitation decreasing in the middle phase of the study period. The climate in NC is governed by strong monsoon, which carries moistures primarily falling in summer. Accompanied by a marked weakening of the East Asian summer monsoon during the 1970s (Wang et al., 2002), atmospheric circulation and precipitation patterns have undergone significant changes (Han et al., 2007). Moisture in NC mainly originates from the Bay of Bengal and the South China Sea, the Pacific, Eurasia, the Indian Ocean, and eastern China (Jiang et al., 2017). Weakened water vapor sources from these regions causes declined precipitation in NC. However, LP over NC exhibited an overall wetting trend in the past five decades. Compared with other regions, the middle phase flux brings about an overall non-significant precipitation trend on the LP. Besides monsoon and atmospheric circulation effects, regional vegetation changes caused by artificial ecology engineering projects in LP can further add to the complexity of precipitation flux in LP.
Precipitation is the primary factor regulating climate drying or wetting trends. Besides precipitation, temperature, RH, wind speed, and sunshine duration also play vital roles in contributing to the trends. Among the ensemble of climatic factors, precipitation makes the greatest contribution to drought index designated drought conditions, such as by SPEI_th or SPEI_pm. Under strengthened precipitation, SPEI_th and SPEI_pm presented almost consistent wetting trends in NW (i.e., headwaters of the Yellow River and Yangtze River, and LP), NC, and NE in the last two decades. However, where the water sources are originated from and how other climatic factors interact with precipitation in causing the wetting trends still await further investigations.

5.3 Opposite trends when PET was estimated by different methods for SPEI

The commonly used drought indices of SPEI_th and SPEI_pm have exhibited similar performances in monitoring global drought conditions (Begueria et al., 2014). However, Cook et al. (2014) found that increased PET was the main cause of regional drought, and PET can lead to the transition from wetting to drying trend in areas with slightly increased precipitation. This study identified opposite drought trends when PET was estimated by different methods for SPEI, especially in the arid NW (Figure 6).
Admittedly, the Thornthwaite algorithm for SPEI can amplify the drying trends due to its higher sensitivity to temperature in the last five decades (Figure S4), when precipitation increased marginally from 1970 to 2019 (Figure 2) and the arid NW becomes warmer (Figure 3a). The 12-month SPEI_th presented a significant drying trend in most areas during 1970-2019, especially in the arid NW (Figure 6a). The drought conditions are the joint consequences between precipitation sourced moisture and PET consumed parts. The contribution of PET to drying or wetting trend can be as important as precipitation (Sun et al., 2016). It results in water availability ranging from the higher to the lower, as compared to SPEI_pm (Figure S5). SPEI_pm performs better based on observed variations in streamflow and soil moisture over China (Chen et al., 2015). Hence, there emerges a great discrepancy in the arid NW during 1970-2019 between the Thornthwaite algorithm and Penman-Monteith equation (Figures 5 and 6), which is consistent with Zhou et al.’s (2020). Briefly, SPEI_pm is more suitable for monitoring drying or wetting trends over China’s drylands than SPEI_th, particularly over the arid regions. Furthermore, reliable and expanded field observations are entailed to reflect the variety of drought conditions, including atmospheric, agricultural, hydrological, socioeconomic, and ecologic drought.

5.4 Uncertainties and limits

There are a few points awaiting further attentions. It is worth mentioning that easily accessible data sets (Slette et al., 2019), such as SPEIbase (https://spei.csic.es/database.html), can lead to mischaracterization of drying or wetting trends. SPEIbase data was only validated by observational data from German meteorological stations (Zang et al., 2020). Adequately validated data for each aimed region are in strong need. The meteorological data utilized in the present work has no field validation issue. The sparse and incomplete surface observations in NW, coupled with mountain-basin structure and mountain-desert environment, can undermine the research conclusions, particularly regarding precipitation (Yao et al., 2022). Given their sporadic station distributions in NW, each grid box containing at least one station has to be at a spatial resolution of 2° × 2° (Zhang et al., 2016). The coarse spatial resolution must bring about some uncertainties. Thus, improved spatial resolution data are urgently needed in the future studies (Wei et al., 2021).
SPEI is suitable for both short- and long-term drought monitoring, but selecting a proper temporal interval is critical. By comparing 1-month and 3-, 6-, 12-, 18-, and 24-month SPEI, it was found that 12-month temporal interval is more appropriate for exploring the long-term drought trends. But China’s drylands accommodate diverse topography and land cover types, which nourish complex regional climates. Appropriate temporal intervals need to be considered for each specific region when focusing on a different spatial scale. For example, Jiang et al. (2020) found that alpine meadow of the TP responded predominantly to 1-12 months SPEI, while response of desert grassland of the northern TP is more related to >12 months SPEI. In addition, forests have longer temporal growing durations than grasslands due to their high growth stature and deep fine root systems. The 3-6 months SPEI is probably more appropriate for irrigated cropland mainly distributed in NW and NC, because anthropogenic water supply is mostly concentrated during the 3-6 months of each growing season. Significantly, the diverse drying and wetting trends in China’s drylands need to be considered over appropriate temporal intervals in the future studies.
The drying and wetting trends based on SPEI are associated directly with climatic variables. Other factors can act on the drying or wetting trends significantly (Su et al., 2019), including general atmospheric circulation and anthropogenic activities (Burrell et al., 2020) related land greening and urbanization (Zeng et al., 2018; Wang et al., 2020). The installation of large-scale wind and solar power generation facilities could reduce surface wind speed (Li et al., 2018), which would indirectly tune climatic factors related to the input data of SPEI.

6 Conclusion

This study concludes that the 12-month SPEI_th and SPEI_pm demonstrated almost consistent wetting-drying-wetting trends during 1970-2019 for China’s drylands. The significant wetting trends over China’s drylands were presented by SPEI_th and SPEI_pm in the last two decades. Both drought indices calculation methods and temporal interval considered can have a significant influence on the derived drying and wetting trends. The different intervals considered can lead to different or even opposite conclusion. Second, precipitation increased significantly in the last two decades. Precipitation was a major factor regulating the drying and wetting trends, which results in a significant wetting trend over China’s drylands in the last two decades. Third, the wetting trend’s direction and magnitude in China’s drylands hinge tightly on the targeted study areas.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.


Thanks to the National Meteorological Information Center of China Meteorological Administration for archiving the observed data. Great thanks to the editors and anonymous reviewers for providing valuable comments that significantly improved this paper.

Data availability statement

All observation data used in this study are available at http://data.cma.cn. The CRU data can be accessed at http://data.ceda.ac.uk/badc/cru/data.

Supplementary Materials

Figure S1 96 (2° × 2°) grid boxes (color: blue) were identified containing at least one station with continuous observations. In total, 206 meteorological stations (color: spruce green) were used.
Figure S2 Wavelet transform coherences (WTCs) between 12-month SPEI_th and input climatic variables (a. mean air temperature; b. mean maximum air temperature; c. mean minimum air temperature; d. precipitation) over China’s drylands from 1970 to 2019. The SPEI estimates forced with Thornthwaite algorithm (denoted as SPEI_th). The period is measured in years. Each subplot shows the WTC between a region and the individual meteorological variable. NW = Northwest China’s drylands; NC = Northern China’s drylands; NE = Northeastern China’s drylands; Total = China’s drylands. The color denotes the strength of coherence. Thick contours reflect 95% significance levels against yellow noise. Pale regions indicate the cone of influence where edge effects might bias the results. Small arrows show the relative phase relationship (in-phase, arrow point right; anti-phase, arrows point left).
Figure S3 Wavelet transform coherences (WTCs) between 12-month SPEI_pm and input climatic variables (a. mean air temperature; b. mean maximum air temperature; c. mean minimum air temperature; d. precipitation; e. relative air humid; f. wind speed; g. sunshine duration) over China’s drylands from 1970 to 2019. The SPEI estimates forced with Penman-Monteith equation (denoted as SPEI_pm). The period is measured in years. NW = Northwest China’s drylands; NC = North China’s drylands; NE = Northeast China’s drylands; Total = China’s drylands.
Figure S4 Changes in potential evapotranspiration (PET) anomaly over China’s drylands: 1970-2019. Time series of two PET anomaly calculated by two approaches: Thornthwaite algorithm (denoted as PET_th) and Penman-Monteith equation (denoted as PET_pm), against temperature anomaly.
Figure S5 Changes in water balance (precipitation minus potential evapotranspiration (P-PET)) anomaly over China’s drylands: 1970-2019. Time series of two PET anomaly calculated by two approaches: Thornthwaite algorithm (denoted as PET_th) and Penman-Monteith equation (denoted as PET_pm), against temperature anomaly.
Aadhar S, Mishra V, 2020. Increased drought risk in South Asia under warming climate: Implications of uncertainty in potential evapotranspiration estimates. Journal of Hydrometeorology, 21(12): 2979-2996.


Allen R G, Pereira L S, Raes D et al., 1998. Crop Evapotranspiration Guidelines for Computing Crop Water Requirements, FAO Irrigation & Drainage Paper 56, Roma: FAO, Food and Agriculture Organization of the United Nations.

An C, Feng Z, Barton L, 2006. Dry or humid? Mid-Holocene humidity changes in arid and semi-arid China. Quaternary Science Reviews, 25(3/4): 351-361.


Anderegg W R L, Trugman A T, Badgley G et al., 2020. Divergent forest sensitivity to repeated extreme droughts. Nature Climate Change, 10: 1091-1095.


Beguería S, Vicente-Serrano S M, Reig F et al., 2014. Standardized precipitation evapotranspiration index (SPEI) revisited: Parameter fitting, evapotranspiration models, tools, datasets and drought monitoring. International Journal of Climatology, 34(10): 3001-3023.


Berdugo M, Delgado-Baquerizo M, Soliveres S et al., 2020. Global ecosystem thresholds driven by aridity. Science, 367(6479): 787-790.


Burrell A L, Evans J P, De Kauwe et al., 2020. Anthropogenic climate change has driven over 5 million km2 of drylands towards desertification. Nature Communications, 11: 3853.


Chapin III F S, Chapin M C, Matson PA et al., 2011. Principles of Terrestrial Ecosystem Ecology. New York: Springer, 57pp.

Chen C, Park T, Wang X et al., 2019. China and India lead in greening of the world through land-use management. Nature Sustainability, 2: 122-129.


Chen D, Gao G, Xu C Y et al., 2005. Comparison of the Thornthwaite method and pan data with the standard Penman-Monteith estimates of reference evapotranspiration in China. Climate Research, 28: 123-132.


Chen H, Sun J, 2015. Changes in drought characteristics over China using the standardized precipitation evapotranspiration index. Journal of Climate, 28(13): 5430-5447.


Chen H, Wang S, Zhu J et al., 2020. Projected changes in abrupt shifts between dry and wet extremes over China through an ensemble of regional climate model simulations. Journal of Geophysical Research: Atmospheres, 125(23): e33894.

Chen T, Zhang H, Chen X et al., 2017. Robust drying and wetting trends found in regions over China based on Köppen climate classifications. Journal of Geophysical Research: Atmospheres, 122(8): 4228-4237.

Cheng G, Wang G, 2006. Changing trend of drought and drought disaster in Northwest China and countermeasures. Earth Science Frontiers, 13(1): 3-14. (in Chinese)

Cheng L, Hoerling M, Aghakouchak A et al., 2016. How has human-induced climate change affected California drought risk? Journal of Climate, 29(1): 111-120.


Cook B I, Smerdon J E, Seager R et al., 2014. Global warming and 21st century drying. Climate Dynamics, 43(11): 2607-2627.


Dilinuer T, Yao J, Chen J et al., 2021. Regional drying and wetting trends over Central Asia based on Köppen climate classification in 1961-2015. Advances in Climate Change Research, 12(3): 363-372.


Feng X, Fu B, Piao S et al., 2016. Revegetation in China’s Loess Plateau is approaching sustainable water resource limits. Nature Climate Change, 6: 1019-1022.


Fu B, Stafford-Smith M, Wang Y et al., 2021. The Global-DEP conceptual framework: Research on dryland ecosystems to promote sustainability. Current Opinion in Environmental Sustainability, 48: 17-28.


Grinsted A, Moore J C, Jevrejeva S, 2004. Application of the cross wavelet transform and wavelet coherence to geophysical time series. Nonlinear Process in Geophysics, 11: 561-566.


Han J, Wang H, 2007. Interdecadal variability of the East Asian summer monsoon in an AGCM. Advances in Atmospheric Sciences, 24: 808-818.


Hoffmann D, Gallant A J E, Arblaster J M, 2020. Uncertainties in drought from index and data selection. Journal of Geophysical Research: Atmospheres, 125(18): e31946.

Hu B, Wang L, Li X et al., 2020. Divergent changes in terrestrial water storage across global arid and humid basins. Geophysical Research Letters, 48(1): e91069.

Hu Z, Chen X, Chen D et al., 2018. “Dry gets drier, wet gets wetter”: A case study over the arid regions of central Asia. International Journal of Climatology, 39(2): 1072-1091.


Huang J, Li Y, Fu C et al., 2017a. Dryland climate change: recent progress and challenges. Reviews of Geophysics, 55(3): 719-778.


Huang J, Yu H, Dai A et al., 2017b. Drylands face potential threat under 2°C global warming target. Nature Climate Change, 7: 417-422.


Jiang P, Ding W, Yuan Y et al., 2020. Diverse response of vegetation growth to multi-time-scale drought under different soil textures in China’s pastoral areas. Journal of Environmental Management, 274(15): 110992.


Jiang Z, Jiang S, Shi Y et al., 2017. Impact of moisture source variation on decadal-scale changes of precipitation in North China from 1951 to 2010. Journal of Geophysical Research: Atmospheres, 122(2): 600-613.


Jiao W, Wang L, Smith W K et al., 2021. Observed increasing water constraint on vegetation growth over the last three decades. Nature Communications, 12(1): 3777.


Li B, Chen Y, Chen Z et al., 2016. Why does precipitation in northwest China show a significant increasing trend from 1960 to 2010? Atmospheric Research, 167: 275-284.


Li B, Chen Y, Shi X, 2012. Why does the temperature rise faster in the arid region of northwest China? Journal of Geophysical Research: Atmospheres, 117: D16115.

Li L, She D, Zheng H et al., 2020. Elucidating diverse drought characteristics from two meteorological drought indices (SPI and SPEI) in China. Journal of Hydrometeorology, 21(7): 1513-1530.


Li Y, Huang J, Ji M et al., 2015. Dryland expansion in northern China from 1948 to 2008. Advances in Atmospheric Sciences, 32(6): 870-876.


Lian X, Piao S, Chen A et al., 2021. Multifaceted characteristics of dryland aridity changes in a warming world. Nature Reviews Earth and Environment, 2: 232-250.


Liu L, Wang Y, You, N et al., 2018. Changes in aridity and its driving factors in China during 1961-2016. International Journal of Climatology, 39(1): 50-60.


Liu Y, Kumar M, Katul G G et al., 2019. Reduced resilience as an early warning signal of forest mortality. Nature Climate Change, 9: 880-885.


Lobell D B, Deines J M, Tommaso S Di, 2020. Changes in the drought sensitivity of US maize yields. Nature Food, 1: 729-735.

Masson-Delmotte V, Zhai P, Pirani A et al., 2021. Intergovernmental Panel on Climate Change (IPCC):The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press. https://www.ipcc.ch/srcc.

Mckee T B, Doesken N J, Kleist J, 1993. The relationship of drought frequency and duration to time scales. Anaheim:8th Conference on Applied Climatology.

Milton Abramowitz, Irene A Stegun, 1965. Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables (Dover Books on Mathematics). New York: Dover Publications Inc.

Neelin J D, Chou C, Su H, 2003. Tropical drought regions in global warming and El Niño teleconnections. Geophysical Research Letters, 30(24): D2275.

Palmer W C, 1965. Meteorological Drought. U.S. Weather Bur. Res. Pap. No.45.

Peng D, Zhou T, 2017. Why was the arid and semiarid northwest China getting wetter in the recent decades? Journal of Geophysical Research: Atmospheres, 122(17): 9060-9075.

Peng J, Wu C, Zhang X et al., 2019. Satellite detection of cumulative and lagged effects of drought on autumn leaf senescence over the Northern Hemisphere. Global Change Biology, 25(6): 2174-2188.


Qin J, Su B, Tao H et al., 2021. Spatio-temporal variations of dryness/wetness over Northwest China under different SSPs-RCPs. Atmospheric Research, 259: 105672.


Ren G, Yuan Y, Liu Y et al., 2016. Changes in precipitation over Northwest China. Arid Zone Research, 33: 1-19. (in Chinese)

Shi Y, Shen Y, Li D et al., 2003. Discussion on the present climate change from warm-dry to warm-wet in Northwest China. Quaternary Sciences, 23(2): 152-164.

Slette I J, Post A K, Awad M et al., 2019. How ecologists define drought, and why we should do better. Global Change Biology, 25(10): 3193-3200.


Song Y, M J, Zuo Q et al., 2021. Analysis on characteristics of dry-wet variation on multi-time scale in Xinjiang. Water Resources Protection, 37(2): 43-48. (in Chinese)

Su L, Miao C, Duan Q et al., 2019. Multiple-wavelet coherence of world’s large rivers with meteorological factors and ocean signals. Journal of Geophysical Research: Atmospheres, 124(9): 4932-4954.

Sun S, Chen H, Wang G et al., 2016. Shift in potential evapotranspiration and its implications for dryness/wetness over Southwest China. Journal of Geophysical Research: Atmospheres, 121(16): 9342-9355.

Tao H, Borth H, Fraedrich K et al., 2014. Drought and wetness variability in the Tarim River Basin and connection to large-scale atmospheric circulation. International Journal of Climatology, 34(8): 2678-2684.


Thornthwaite C W, 1948. An approach toward a rational classification of climate. Geographical Review, 38: 55-94.


Toms J D, Lesperance M L, 2003. Piecewise regression: A tool for identifying ecological thresholds. Ecology, 84(8): 2034-2041.


Trenberth K E, Dai A, van der Schrier G et al., 2014. Global warming and changes in drought. Nature Climate Change, 4: 17-22.


United Nations (UN), 2015. Transforming Our World: The 2030 Agenda for Sustainable Development (United Nations General Assembly. https://www.un.org/sustainabledevelopment/progress-report/

Vicente-Serrano S M, Beguería S, López-Moreno J I, 2010. A multiscalar drought index sensitive to global warming: The standardized precipitation evapotranspiration index. Journal of Climate, 23(7): 1696-1718.


Wang H, 2002. The instability of the East Asian summer monsoon-ENSO relations. Advances in Atmospheric Sciences, 19: 1-11.


Wang J, Feng J, Yan Z et al., 2020. An analysis of the urbanization contribution to observed terrestrial stilling in the Beijing-Tianjin-Hebei region of China. Environmental Research Letters, 15(3): D34062.

Wang Q, Zhai P, Qin D, 2020. New perspectives on ‘warming-wetting’ trend in Xinjiang, China. Advances in Climate Change Research, 11(3): 252-260.


Wang Y, Fu B, Liu Y et al., 2021. Response of vegetation to drought in the Tibetan Plateau: Elevation differentiation and the dominant factors. Agricultural and Forest Meteorology, 306(15): 108468.


Wei L, Jiang S, Ren L et al., 2021. Evaluation of seventeen satellite-, reanalysis-, and gauge-based precipitation products for drought monitoring across mainland of China. Atmospheric Research, 263(1): 105813.


West H, Quinn N, Horswell M, 2019. Remote sensing for drought monitoring & impact assessment: Progress, past challenges and future opportunities. Remote Sensing of Environment, 232: 111291.


Wilhite D A, Glantz M H, 1987. Understanding the drought phenomenon: The role of definitions. Water International, 10(3): 111-120.


World Meteorological Organization, 2020. WMO statement on the status of the global climate in 2019. In: Organización Mundial de la Meteorología, 1248: 44pp.

Xu H, Wang X, Zhao C, 2021. Drought sensitivity of vegetation photosynthesis along the aridity gradient in northern China. International Journal of Applied Earth Observation and Geoinformation, 102: 102418.


Xu H, Wang X, Zhao C et al., 2020. Assessing the response of vegetation photosynthesis to meteorological drought across northern China. Land Degradation and Development, 32(1): 20-34.


Yao J, Chen Y, Guan X et al., 2022. Recent climate and hydrological changes in a mountain-basin system in Xinjiang, China. Earth-Science Reviews, 226: 103957.


Yao N, Li Y, Dong Q et al., 2019. Influence of the accuracy of reference crop evapotranspiration on drought monitoring using standardized precipitation evapotranspiration index in mainland of China. Land Degradation and Development, 31(2): 266-282.


Yao Y, Liu Y, Wang Y et al., 2021. Greater increases in China’s dryland ecosystem vulnerability in drier conditions than in wetter conditions. Journal of Environmental Management, 291: 112689.


Yu Y, Wang J, Li Q, 2003. Spatial and temporal distribution of water vapor and its variation trend in atmosphere over Northwest China. Journal of Glaciology and Geocryology, 25(2): 149-156. (in Chinese)

Zang C, Buras A, Esquivel-Muelbert A et al., 2020. Standardized drought indices in ecological research: Why one size does not fit all. Global Change Biology, 26(2): 322-324.


Zeng Z, Piao S, Li L Z X et al., 2018. Global terrestrial stilling: Does Earth’s greening play a role? Environmental Research Letters, 13: 124013.


Zhang J, Sun F, Xu J et al., 2016. Dependence of trends in and sensitivity of drought over China (1961-2013) on potential evaporation model. Geophysical Research Letters, 43(1): 206-213.


Zhao H, Yao J, Li X et al., 2020. The characteristic of climate change in Xinjiang during 1961-2015. Acta Scientiarum Naturalium Universitatis Sunyatseni, 59(5): 126-133. (in Chinese)

Zhao W, 2020. Extreme weather and climate events in China under changing climate. National Science Review, 7(5): 938-943.


Zhou J, Wang Y, Su, B et al., 2020. Choice of potential evapotranspiration formulas influences drought assessment: A case study in China. Atmospheric Research, 242: 104979.