Monitoring of Extreme Agricultural Drought of the Past 20 Years in Southwest China Using GLDAS Soil Moisture

Abstract

Drought can cause severe agricultural economic losses and hinder social and economic development. To manage drought, the process of drought events needs to be described with the help of an effective drought indicator. As a comprehensive variable, soil moisture is an essential indicator for describing agricultural drought. In this work, the extreme drought events in southwest China were analysed by the Global Land Data Assimilation System (GLDAS) root zone soil moisture data set. To define the drought quantitatively, a Standardized Soil Moisture Drought Index (SSMI) was calculated using the soil moisture data, then used to get the duration, frequency, and severity of drought events in southwest China. The results showed that the frequency and intensity of drought in southwest China had an apparent upward trend before 2014 and an apparent downward trend since 2014. Moreover, there are apparent differences in the frequency and intensity of drought in various regions of southwest China. Yunnan Province is prone to spring drought events. Guangxi Province and Guizhou Province are prone to spring, autumn and winter droughts, and the intensity of autumn and winter droughts is significantly higher than that of spring droughts. The Sichuan-Chongqing border area is prone to summer drought. We found that the monthly variation of soil moisture in different provinces in southwest China is consistent, but the seasonal variation of drought is different. Meanwhile, the performance of the SSMI was compared to the commonly used drought indices, the Standardized Precipitation Evapotranspiration Index (SPEI) and the Palmer Drought Severity Index (PDSI). The results showed that the SSMI is more sensitive to drought than both SPEI and PDSI in southwest China. The results also demonstrate that GLDAS soil moisture data can be used to study drought at a small regional scale.

1. Introduction

Drought is one of the most destructive and urgent natural disasters in the world [1], as it causes insufficient soil moisture and disrupts crop water balance and reduces yield. Among wholly natural disasters, drought is one of the most serious that endangers agriculture and animal husbandry production [2]. In the context of global warming, the frequency and duration of drought events also show an upward trend. From 1950 to 2008, the frequency of drought events in arid regions around the world increased by about 1.74% every ten years [3]. Increasingly, facts prove that the impact of drought is extremely extensive and far-reaching [4]. For example, drought in 2012 caused severe agricultural disasters throughout North America, leading to a sharp increase in food prices [5]. From 2001 to 2013, the intensity and frequency of droughts in the north-eastern and southern regions of China showed a clear trend of intensification [6]. In the summer of 2011, a drought in the upper and middle reaches of the Yangtze River in China affected 30 million people and caused $2.4 billion in damage [7]. According to statistics, about 80 million people in the world have been threatened by drought since the 1970s, and the number of direct deaths due to drought will exceed 1 million [8].

Drought has a high frequency, long duration, a wide range of impacts, and causes economic and environmental consequences. For those reasons, much effort has been devoted to developing techniques for drought analysis and monitoring. Most research on the identification and analysis of drought events use the drought index. Drought indices are the most widely used, but subjectivity in defining drought has made it very difficult to establish a unique and universal drought index [9]. According to the World Meteorological Organization data, there are 55 drought indices that are commonly used [1], for example, the Standardized Precipitation Evapotranspiration Index (SPEI), the Standardized Precipitation Index (SPI) and the Palmer Drought Severity Index (PDSI). Vicente Serrano [10] et al. evaluated the performance of different drought indices and found that the SPEI and the SPI have excellent performance and are very sensitive to drought events. Liu [11] et al. used the PDSI, SPI and the SPEI to evaluate and monitor drought events in the North China Plain. The study found that the SPI and the SPEI have similar performance in monitoring drought, while the PDSI has obvious lag. Ajaz [12] et al. developed a new drought index (Soil Moisture Evapotranspiration Index) to monitor and evaluate the impact of drought on agriculture in Oklahoma. This drought index can effectively monitor agricultural drought in Oklahoma. Dutra [13] et al. used ECMWF Reanalysis—40 years (ERA-40) data to calculate the SPI, PDSI and the Normalized Soil Moisture (NSM) in Iberia. The results found that the NSM can have a good correlation with the SPI and the PDSI, and the NSM can monitor regional drought very well. Pena [14] et al. used the PDSI, SPEI, and the SPI to assess the impact of drought on tree growth and other conditions. The results found that compared with the PDSI, the SPI, SPEI, etc. have more advantages in monitoring tree growth and estimating net primary productivity. Wei [15] et al. used weather station data to calculate the PDSI, the Precipitation Anomaly (PA), and the Surface Wetness Index (SWI) in the northeast and analysed the relationship and differences between the three. The study found that the SPEI can accurately describe the nature and intensity of drought. Most of these drought indexes are only based on rainfall data, because drought is directly caused by insufficient rainfall, and rainfall data is easier to obtain and process than other data [16]. However, it also has been reported that many droughts are caused by reduced rainfall and other factors that affect water balance conditions. The drought index also has certain drawbacks. For example, the PDSI index does not consider the impact of human activities on the water balance, and there are many input data, which are difficult to obtain [3]. The SPI has a strong simulation dependence on precipitation data and its probability distribution, which can easily cause misjudgement of drought conditions [17]. Studies have found that using precipitation alone to measure regional drought conditions is not appropriate. Drought will first affect the agricultural sector, and soil moisture is more important than precipitation for agricultural drought. Sufficient soil moisture is necessary for crops to grow in different periods [18], and soil moisture governs the size and change of water and energy flux at the landing site-atmosphere interface, and controls plant growth and biology. Crop production is severely affected by soil moisture [19]; accurate soil moisture data is, therefore, vital for monitoring and forecasting agricultural drought.

At present, there are three main methods for acquiring soil moisture data: site observation, remote sensing satellite observation, and model simulation methods. Early forms of obtaining soil moisture are mainly observed through site data [20,21,22,23,24]. The soil moisture data followed by the site data can be considered accurate and widely used in studying drought events. However, monitoring stations cannot provide long-term series of large-scale soil moisture data currently, and the distribution of soil moisture monitoring stations in China is uneven. The number of stations in southwest China is significantly less than in the northeast, Huaihe, and other regions [20]. Satellite remote sensing technology itself has the advantage of large-scale simultaneous observation. With the continuous development of remote sensing satellite technology in the past 20 years, related research based on remote sensing observation of soil moisture has gradually increased [25,26,27,28]. Remote sensing satellite technology can obtain large-scale soil moisture data sets. A multi-band, multi-temporal, and multi-polarized soil moisture observation mode has been formed [29]. However, the current remote sensing technology can only obtain the soil moisture in the shallow soil layer (1–5 cm) [26]. In recent years, soil moisture models have developed rapidly. The model can combine the advantages of remote sensing data with site data. The soil moisture data achieved through the model has large-scale, long-term, temporal, and spatial consistency [30]. In drought research, soil moisture data calculated by the model has also been widely used [31,32,33,34,35].

In the past ten years, tremendous drought disasters have repeatedly occurred in southwest China. Frequent drought disasters have caused enormous economic losses in the southwest China region and even threatened the safety of the drinking water supply. Many scholars have carried out research. Feng et al. used precipitation data from stations in southwest China and data from the National Environmental Forecast Center to establish a drought detection equation and analysed the causes of drought events in southwest China [36]. Wang et al. calculated the relative humidity index (M) using data from weather stations in southwest China. They used the relative humidity index as an indicator to reflect the evolution characteristics of seasonal drought in southwest China from the perspective of the frequency and intensity of drought [37]. Zhang et al. and others used the Drought Stress Index (DSI) retrieved by the Moderate Resolution Imaging Spectrometer (MODIS) to monitor the duration, intensity, and spatial resolution of extreme droughts in southwest China from 2009 to 2010. The results show that southwest China experienced severe drought from November 2019 to March 2010, and the affected area of crops accounted for 74% of the total area of the study area [38]. Li et al. and others established a remote sensing monitoring and evaluation method for the drought and its impact in southwest China in the spring of 2010 using the multispectral and thermal infrared data of domestic environmental disaster mitigation stars and the US medium-resolution MODIS [39]. By calculating the Vegetation Condition Index (VCI) and Temperature Condition Index (TCI), indices that characterize the drought, and through the linear combination model of these two indices, it accurately reflects the comprehensive information such as the range of the drought occurrence area and the degree of drought. Yi et al. used Gravity Recovery and Climate Experiment Follow-On (GRACE-FO) gravity satellite data and GLDAS hydrological model data to study the temporal and spatial changes of water reserves in southwest China and combined precipitation data for analysis [40].

In the study of drought events in southwest China, soil moisture data mainly come from site observation and satellite remote sensing measurement. Research on drought events in southwest China using soil moisture data obtained through model simulations is relatively rare. The Global Land Data Assimilation System (GLDAS) publishes a global 25 km soil moisture datasets widely used in drought research world-widely. Therefore, this study selected the 25 km monthly soil moisture datasets (0–10 cm) published by GLDAS-Noah as the basis for studying drought in southwest China. The purpose of this experiment is to analyse the potential of GLDAS soil moisture data to identify and characterize drought events with spatio-temporal continuous data (southwest China). We used SSMI to monitor drought in southwest China. The SSMI can measure the severity of global agricultural drought events. The structures of this paper are as follows: In the second part, the work introduces the data used in the research and the calculation process of SSMI. In the third part, we used soil moisture data and the SSMI to study drought events throughout southwest China and we compared the SPEI, PDSI, and the SSMI for drought monitoring in southwest China. The fourth part discusses our research. The final section provides a summary.

2. Materials and Methods

2.1. The Global Land Data Assimilation System (GLDAS) and Soil Moisture Data

The Global Land Data Assimilation System (GLDAS) simulates satellite and ground observation data products using advanced land surface data assimilation systems and surface modelling to generate accurate surface flux data [41]. The GLDAS uses various high-resolution satellite remote sensing data, including high-resolution LAI, soil, and elevation.

The GLDAS analyses observation-based precipitation and downward shortwave radiation data using an optimal model of the atmospheric assimilation system [41,42,43,44].

The GLDAS-2.1 data products have been simulated since 1 January 2000, providing medium and high-resolution land surface data sets [45]. The GLDAS-2.1 datasets provide spatial resolutions of 0.25°, 1°, and temporal resolutions of 3 h, 1 day, and 1 month. The GLDAS data mainly includes various meteorological data, including rainfall, wind speed, air temperature, evapotranspiration, etc., and soil moisture, soil temperature, and other data. The GLDAS-2.1 soil moisture data are divided into 4 layers; the deepest can reach 1 m and more. A study shows that GLDAS data products have strong spatiotemporal continuity and high product accuracy [30,34,46,47]. In this work, we selected the underground 0–10 cm soil moisture data to study in southwest China.

2.2. The Global 0.5° Gridded SPEI Data and Monthly Gridded Global PDSI (Palmer Drought Severity Index)

The SPEI datasets (https://spei.csic.es/, accessed on 1 March 2020) used in this work come from the global SPEI database, which currently provide SPEI data from 2000 to 2019 with a spatial resolution of 0.5° and a temporal resolution of one month [48,49]. The time scale of the SPEI datasets ranges from 1 to 48 months: we chose the one-month SPEI to compare with our results. Below we give a brief introduction to the calculation of SPEI:

$$P=1-F\left(X\right)=1-{\left[1+{\left(\frac{\alpha }{x-\gamma }\right)}^{\beta }\right]}^{-1}$$

(1)

When cumulative probability $P\le 0.5$, $W=\sqrt{-2 \ln \left(P\right)}$:

$$SPEI=W-\frac{C_0+C_{1}W+C_2{W}^2}{1+d_{1}W+d_2{W}^2+d_3{W}^3}$$

(2)

When cumulative probability $P>0.5$, $W=\sqrt{-2 \ln \left(1-P\right)}$:

$$SPEI=-W+\frac{C_0+C_{1}W+C_2{W}^2}{1+d_{1}W+d_2{W}^2+d_3{W}^3}$$

(3)

where the constant $C_0=2.5155$, $C_1=0.8029$, $C_2=0.0103$, $d_1=1.4328$, $d_2=0.1893$, $d_3=0.0013$, where $\alpha$, $\beta$ and $\gamma$ are the scale parameter, shape parameter and starting parameter, respectively, It can be calculated by the method of linear moments.

The Palmer Drought Severity Index (PDSI) was devised by Palmer (1965) [50]. The PDSI dataset used in this work is provided by the Dai datasets (https://psl.noaa.gov/data/gridded/, accessed on 1 March 2020). The PDSI products are up to 2018, with a temporal resolution of one month and a spatial resolution of 0.5°.

Table 1. Classification of drought by SPEI and PDSI drought index.

Category	SPEI	PDSI
Extreme drought	SPEI ≤ −2.0	PDSI ≤ −4.0
Severe drought	−2.0 < SPEI ≤ −1.5	−4.0 < PDSI ≤ −3.0
Moderate drought	−1.5 < SPEI ≤ −1.0	−3.0 < PDSI ≤ −2.0
Mild drought	−1.0 < SPEI ≤ −0.5	−2.0 < PDSI ≤ −1.0
No drought	SPEI ≥ −0.5	PDSI ≥ −1.0

is the classification of drought intensity by SPEI and PDSI. The PDSI index is used to characterize the water deficit in which the actual water supply in a certain area is continuously less than the suitable supply of the local climate for a period. Its main calculation formula is as follows:

$$\widehat{P}=\widehat{ET}+\widehat{R}+\widehat{RO}-\widehat{L}$$

(4)

Among them, $\widehat{P}$ is the climatic suitable precipitation in a certain month, $\widehat{ET}$, $\widehat{R},$ And $\widehat{L}$ are the evapotranspiration, soil water replenishment, yield, and soil water loss under suitable climatic conditions, respectively. After the $\widehat{P}$ of each month is calculated, the water deficit of each month can be calculated:

$$d=P-\widehat{P}$$

(5)

Moisture Anomaly Index $Z$:

$$Z=dK$$

(6)

$$K=\frac{\overline{PE}+\overline{R}}{\overline{P}+\overline{L}}$$

(7)

$$X_i=0.897\ast X_{i-1}+\frac{Z_i}{3}$$

(8)

$X_i$ and $X_{i-1}$ are the PDSI values of the current month and the previous month, respectively, and $Z_i$ is the index value of the current month.

2.3. Study Region

The study areas selected for this work include Yunnan Province, Sichuan Province, Guizhou Province, Guangxi Province, and Chongqing City of China (Figure 1) [51]. The study area is between $91°{21}^{\prime }\mathrm{E}$–$112°{04}^{\prime }\mathrm{E}$ and $20°{54}^{\prime }\mathrm{E}$–$34°{19}^{\prime }\mathrm{N}$, with an altitude of 100–5500 m and an area of about 1.38 million km², accounting for 14.2% of China’s total area. The average population density of the southwest China region is 183 persons per square kilometres, the urbanization rate is 33.5%, and the GDP accounts for about 10.7% of China; the cultivated land area is 184,973 km², the effective irrigation area is 62153 km², and the annual grain output is about 86.73 million tons, accounting for 16.4% of China [52]. The dependence on agriculture in southwest China is vital, so the drought undeniably impacts regional drinking water safety and farmer livelihoods.

The southwest region is in the upper reaches of the Yangtze River and the Pearl River. The landforms are very complex, mainly mountainous, and hilly areas. The study area is mainly composed of the Sichuan Basin, the Yunnan-Guizhou Plateau, and Guangxi Province and Guangdong’s hilly parts. The landform of the southwest China is mainly mountainous karst landform, and the soil texture is sandy soil, and sandy loam. The terrain of the entire southwest China area is intricate and complicated. This feature makes the surface runoff rapid and the land cannot quickly accumulate a large amount of water [53].

The depth difference of the river valley will cause great difficulty in groundwater development. There are noticeable regional differences in climate in southwest China. The Sichuan Basin belongs to the northern subtropical monsoon humid climate, with a temperate climate, high humidity, and relatively flat terrain [54]. The Yunnan-Guizhou-Guangxi belongs to the low-latitude plateau central-south subtropical monsoon climate, and the area is relatively flat, suitable for the development of agriculture. Some tropical seasonal rain forest climate zones are distributed in a small part of the southern end. The combination of low latitude and high latitude in the southwest has increased the magnitude of climate change. The abrupt changes of different mountains have caused a significant impact on some areas. Complex climate change exacerbates the water cycle, increasing the frequency and intensity of extreme weather and extreme hydrological events (such as floods and droughts). Different topography also affects the atmospheric water cycle, making regions prone to droughts [55].

2.4. SSMI-Based Identification of Drought Events

Due to the wide range of spatial effects of drought, the diversification of time distribution, and the different needs of people for water supply systems, it is difficult to propose a single definition of drought [56]. The American Meteorological Society (AMS 1997) divided drought into four categories: meteorological drought or climatic drought, agricultural drought, hydrological drought, and socioeconomic drought [57]. In order to accurately identify and quantify various types of droughts, researchers have proposed a variety of drought indices specifically used to describe drought events, including the SPEI [58], the SPI [59], the PDSI [60] etc. Among the four types of drought, agricultural drought is characterized by soil moisture content and plant growth status [61], which refers to the lack of water in the soil due to long-term no rain during the agricultural growing season. The standardized soil moisture index (SSMI) [62], crop water deficit index anomaly (CWDIa), soil moisture index (SMI), and vegetation supply water index (VSWI) [63] are commonly used internationally as indicators to describe agricultural drought. The CWDIa is related to crop water loss in a certain period time. Although the CWDIa can reflect the degree of crop water satisfaction, the water storage capacity is generally large in dry climate areas, it has not considered the role of irrigation [64]. The SMI is defined as the percentage of actual soil moisture in the field capacity, reflecting the relative dryness of the soil. The SMI can be used to detect droughts in specific soil types in small-scale areas, but it has limitations [65]. Although the VSWI monitoring method is intuitive, it is easily affected by cloud and vegetation conditions. Especially in southwest China, where vegetation is lush and foggy weather occurs all the year-round, this makes remote sensing monitoring of drought index with more significant uncertainty.

The SSMI determines the severity of drought by calculating the ratio between the deviation of soil moisture from the multi-year average soil moisture at the grid points and the standard deviation of the multi-year soil moisture. The SSMI is a dimensionless drought index that monitors drought intensity by calculating the deviation of soil moisture. The SSMI is one of the most straightforward indices developed and validated in different studies and is a powerful tool for monitoring agricultural drought [63,66,67,68]. The multi-year soil moisture average value is for each pixel, and the multi-year soil moisture average value in different months is obtained by calculating the pixel-by-pixel average value of the long-term series in different months. The calculation formula of the SSMI is as follows:

$$SSM{I}_{i,j}=\frac{{\theta }_{i,j}-\mu {\theta }_i}{\sigma {\theta }_i}$$

(9)

where $SSM{I}_{i,j}$ is the $SSMI$ for year $j$ and month $i$, ${\theta }_{i,j}$ is the mean soil moisture for year $j$ and month $i$, $\mu {\theta }_i$ is the mean of the long-term series of monthly soil moisture. $\sigma {\theta }_i$ is the standard deviation of the long-time series of soil moisture at the monthly scale. The greater the absolute value of the negative SSMI, the more severe the drought in the area, and the relative rather than absolute soil moisture loss can be used to compare the severity of drought in different periods and regions [62].

Using GLDAS soil moisture data to calculate the SSMI in southwest China we could identify and quantify the drought events in the southwest China. According to the statistical results, combined with the China Meteorological Administration’s statistical data, the southwest China’s statistical yearbook and the SSMI classification standard [69], we found that when the SSMI was less than −0.5, the identified temporal and spatial variations of drought were consistent with the drought development process documented in the literature and government materials. The grades are divided in detail, as shown in

Table 2. Soil moisture classification based on the SSMI.

Category	SSMI
Extreme drought	≤−2.0
Severe drought	−2.0 to −1.5
Moderate drought	−1.5 to −1.0
Mild drought	−1.0 to −0.5
No drought	≥−0.5

:

3. Results

3.1. Identification of Drought Events in Southwest China

In 2010, a severe drought occurred in southwest China (Yunnan Province, Guizhou Province, Guangxi Province, Sichuan Province, and Chongqing City). Ren Fumin, the National Climate Centre chief expert, said that this is the most severe drought in southwest China since meteorological data has been available. Figure 2 is a schematic diagram of the SSMI from October 2009 to June 2010 in southwest China. Combined with the drought classification criteria in Table 2, we made detailed statistics on drought events in southwest China in 2010. The results are shown in

Table 3. Statistics for drought area in southwest China from October to December 2009.

Year	Area of Arid Area in October (km2)	Area of Arid Area in November (km2)	Area of Arid Area in December (km2)
2009	515,625	774,375	712,500

and

Table 4. Statistics for drought area in southwest China from January to March 2010.

Year	Area of Arid Area in January (km2)	Area of Arid Area in February (km2)	Area of Arid Area in March (km2)
2010	818,125	981,875	941,250

. According to Figure 2, it reveals that drought had already occurred in southwest China, and it was mainly concentrated in the border areas of Yunnan Province, Guizhou, and Guangxi in October 2009. In November, the drought in southwest China expanded and spread to the whole of Chongqing, and the degree of the drought was further aggravated. Through calculation, we found that the dry area in southwest China reached 774.375 × ${10}^3 {\mathrm{km}}^2$. In December, the severity of drought in southwest China eased, and the average soil moisture was around 0.297 ${\mathrm{m}}^3/{\mathrm{m}}^3$, which was mainly due to the rainfall in many areas in December. In January 2010, the main affected areas of drought in southwestern China were Yunnan Province, Guizhou Province, Chongqing City, and eastern Sichuan. The severity of the drought was further aggravated, and the average soil moisture was around 0.245 ${\mathrm{m}}^3/{\mathrm{m}}^3$, drought area reached 818.125 × ${10}^3 {\mathrm{km}}^2$. In February, the severity and scope of drought in the southwest region reached their peak with the drought-affected area almost covering the entire region. The most arid provinces, Chongqing, and Guizhou, reached extreme drought conditions; the average soil moisture in southwest China was around 0.219 ${\mathrm{m}}^3/{\mathrm{m}}^3$, and the drought area reached 981.875 × ${10}^3 {\mathrm{km}}^2$. In March, the affected area of drought in southwest China decreased, but the degree of drought did not weaken. The affected area of drought shifted to the eastern part of southwest China, mainly including Guangxi Province, Guizhou Province, Chongqing City, and the eastern areas of Sichuan and Yunnan; the average soil moisture in southwest China was around 0.248 ${\mathrm{m}}^3/{\mathrm{m}}^3$, and arid area reached 941.250 × ${10}^3 {\mathrm{km}}^2$. In April 2010, the intensity and coverage of the drought decreased significantly, and there was a slight drought in some areas. In May and June, the drought in southwest China slowed down. The main affected areas included Yunnan and Guizhou. The intensity of the drought gradually weakened, and it was mainly mild drought.

Figure 3 shows mean soil moisture and the 95% confidence interval in southwest China from 2000 to 2020, obtained from the long-term series for each month. From Figure 3, the monthly average soil moisture in southwest China has apparent fluctuations. From January to April, the average soil moisture has an apparent downward trend, mainly due to the unusual spring rainfall. It is rare and only accounts for less than 30% of the annual rainfall, and the temperature rose very fast in spring, which accelerated the evapotranspiration and consumed soil moisture. In addition, the 95% confidence interval of soil moisture fluctuated more from January to April, indicating that the soil moisture fluctuated more in spring, and drought was prone to occur. After entering May, the average soil moisture rose rapidly and peaked in July-August when the rainfall was the highest. After September, the rainfall decreased, and the soil moisture showed a decreasing trend, but the decreasing trend was relative.

3.2. Identification of Drought Events in the Border Area of Sichuan and Chongqing

Using the GLDAS-NOAH root zone soil moisture and the SSMI to obtain drought data for the Sichuan-Chongqing border area, we found that the drought events were concentrated in summer and autumn. Figure 4 is a schematic diagram of the SSMI of the drought months in the Sichuan-Chongqing junction from 2000 to 2020. According to the data, a severe drought event occurred in the Sichuan-Chongqing region in 2006. Based on the SSMI, we quantified the drought events in this area for the past 20 years (

Table 5. Statistics of drought area in the Sichuan-Chongqing border area from July to September 2006.

Year	Area of Arid Area in July (km2)	Area of Arid Area in August (km2)	Area of Arid Area in September (km2)
2006	199,375	231,250	202,500

and

Table 6. Statistics of drought area in the Sichuan-Chongqing border area from October to December 2006.

Year	Area of Arid Area in October (km2)	Area of Arid Area in November (km2)	Area of Arid Area in December (km2)
2006	198,125	148,125	186,250

).

By analysing Figure 4, we found that drought events occurred in the summer at the junction of Sichuan and Chongqing in 2006. In July 2006, drought gradually covered the entire Sichuan-Chongqing junction area. At this time, the average soil moisture in this area was around 0.310 ${\mathrm{m}}^3/{\mathrm{m}}^3$. The drought developed to extreme conditions in August with a drought area reaching 231.250 × ${10}^3 {\mathrm{km}}^2$; the average soil moisture was around 0.205 ${\mathrm{m}}^3/{\mathrm{m}}^3$. In September, the drought eased somewhat. At this time, the average soil moisture in the area was about 0.275 ${\mathrm{m}}^3/{\mathrm{m}}^3$, and the arid area decreased to 202.500 × ${10}^3 {\mathrm{km}}^2$. From October to December 2006, the drought was mainly concentrated in the Chongqing area and north-eastern Sichuan. The intensity had not decreased, but the drought’s impact had gradually narrowed to 186.250 × ${10}^3 {\mathrm{km}}^2$.

Figure 5 is a schematic diagram of the monthly average soil moisture and 95% confidence interval in the Sichuan-Chongqing border area from 2000 to 2020. As shown in Figure 5, we analysed the average soil moisture in the area for the past 20 years. We found that the soil moisture in the Sichuan-Chongqing communication area showed a downward trend from January to March and a fluctuating change after a decline from July to August. The 95% confidence interval of soil moisture in the Sichuan-Chongqing area is large from July to October, which indicates that soil moisture fluctuates wildly in summer and autumn (July to October) and is prone to summer and autumn drought events. The above experimental results are consistent with relevant academic research and government statistical records.

3.3. Identification of Drought Events in Yunnan

Soil moisture data for the Yunnan Province in the recent 20 years were analysed based on the SSMI values. The results show that Yunnan Province is prone to severe drought in spring, and extreme spring drought events occurred in 2005, 2012, 2013, 2014 and 2019. Figure 6 is a schematic diagram of the spring SSMI values during the drought years in Yunnan Province from 2000 to 2020. The analysis shows that the frequency and intensity of drought in spring in Yunnan Province had an upward trend. We made a quantitative analysis of drought events in Yunnan Province by using the drought program combined with the drought classification criteria in Table 1, see

Table 7. Statistics of the drought area in Yunnan Province from February to April for 2000 to 2020.

Year	Area of Arid Area in February (km2)	Area of Arid Area in March (km2)	Area of Arid Area in April (km2)
2005	282,500	0	49,375
2012	315,625	298,125	206,250
2013	216,250	281,875	330,000
2014	160,000	75,000	297,500
2019	56,250	101,250	205,000

.

The results (Table 7, Figure 6) show that a severe drought event occurred in Yunnan Province at the end of the spring in 2005 when drought area reached 282.500 × ${10}^3 {\mathrm{km}}^2$, accounting for about 71% of the province’s total area. A severe drought also occurred in Yunnan Province in the spring of 2013 when drought area reached 318.625 × ${10}^3 {\mathrm{km}}^2$ in February, the average soil moisture was only 0.209 ${\mathrm{m}}^3/{\mathrm{m}}^3$, and it accounted for about 81% of the province’s area, i.e., nearly the entire Yunnan province. In 2014, severe drought events also occurred in Yunnan in winter and spring. In April, the drought covered almost the entire Yunnan region, and the drought area reached 297.500 × ${10}^3 {\mathrm{km}}^2$. In the spring of 2019, a drought event again occurred in Yunnan, but its intensity was weaker than that in 2013. The minimum soil moisture during the drought reached about 0.292 ${\mathrm{m}}^3/{\mathrm{m}}^3$. The drought coverage mainly includes the central and southern parts of Yunnan Province.

We used GLDAS soil moisture data to calculate the Trend of the average soil moisture change in Yunnan Province in the past 20 years (Figure 7). This showed that the average soil moisture had a significant downward trend from January to March, and the soil moisture increases significantly from April to August, due to the increase in rainfall in Yunnan in the summer. The rate of decline is less than in spring. The 95% confidence interval a from March to May (spring) in Yunnan Province was marked, indicating that the soil moisture fluctuated most in in spring and was prone to drought events. The overall results are consistent with relevant literature and government records. For example, in the 2012 Flood and Drought Disaster Bulletin, it was recorded that severe drought events occurred in winter and spring in Yunnan Province, causing hundreds of thousands of people to have difficulty obtaining drinking water [70,71,72,73,74,75].

3.4. Monthly Statistics Related to Drought in Southwest China

Through the drought identification program, this work has carried out detailed data analyses of drought events in various regions in southwest China. In this work, the drought identification program was used to calculate the cumulative number of different types of drought pixels in different months in the entire southwest China region, Yunnan Province, Guangxi Province, and other regions from 2000 to 2020 (Figure 8).

From 2000 to 2020, the accumulated number of drought pixels in spring and winter in southwest China is the largest and in March can reach up to 6500 (Figure 8). Although the cumulative number of drought pixels in autumn is lower than that in spring and winter, the cumulative number of severe and extreme drought pixels in summer and autumn is higher than in other seasons; severe types of drought events are prone to occur in the southwest China in autumn (September–November). From this point of view, southwest China is more prone to drought events in spring and winter. The frequency of drought in summer is significantly lower than in other seasons; although it is low in autumn, the intensity is high.

Figure 9 shows the trend of the cumulative drought pixel area in different months in various regions of southwest China. The number of accumulative drought pixels in winter and spring in Yunnan Province were significantly more than other seasons, and the number of severe and extreme drought pixels in winter and spring were also more than other seasons. In the border region of Sichuan and Chongqing, the number and intensity of accumulative drought pixels in spring and summer were significantly higher than those in other seasons. The Sichuan-Chongqing region was prone to severe drought events in spring and summer. The cumulative numbers of drought pixels in spring and winter in Guangxi Province were higher than other seasons, but the cumulative number of severe and extreme droughts in summer and autumn was higher than that in spring and winter, which indicates that Guangxi Province is very strongly prone to drought in spring and winter. As for the Sichuan-Chongqing region, Guizhou Province had significantly higher cumulative drought pixels and intensity than other seasons in spring and summer. The cumulative number of drought pixels in August reached 1183, of which 307 were severe and extreme droughts [43,70,71,72,73,74,75,76,77,78,79].

3.5. Comparison and Trends

In this work, the agricultural drought index SSMI was calculated using the soil moisture data of the GLDAS root zone map to monitor drought in southwest China from 2000 to 2020. Figure 10 shows the changing trends of monthly average for SSMI, SPEI, and PDSI values in southwest China from 2000 to 2018. The frequency and intensity of droughts from 2003 to 2014 had an obvious upward trend. After 2014, the frequency and severity of droughts began to decrease, and southwest China entered a “humid period”. The trend was the same for SSMI, PDSI, and SPEI values. The detailed trends showed that the overall intensity and frequency of droughts in southwest China were slightly lower before 2009. The duration of droughts was 3–4 months, and the recovery of droughts was faster. After 2009, the intensity and duration of drought showed an obvious upward trend. From October 2009 to April 2010, a severe drought lasting seven months occurred. The drought eased in 2011, but in 2011, from the second half of the year to August 2012, a severe drought event occurred again, and the drought intensity was high. In 2013 and 2014, severe spring drought events occurred again in southwest China, and the drought intensity was higher than that before 2009, lasting about four months. In general, through the SPEI, SSMI and PDSI data, it is observed that the frequency and intensity of droughts before 2014 had an obvious increasing trend that decreased significantly after 2014.

Through comparison, we can find that the overall temporal trends of the SSMI, SPEI, and PDSI data for southwest China are similar, and they can all be monitored for drought. For further comparison, we took the 2010 mega-drought event in southwest China as an example to compare the accuracy of the SPEI, PDSI, and SSMI data in monitoring the affected area. Figure 11 shows the spatial trend of the SPEI, PDSI and SSMI data in spring 2010. According to the records of the 2010 Bulletin of flood and drought disasters in China, “precipitation continued to occur in the Southwest China in March 2010, and the drought in most parts of Chong-Qing, Sichuan Basin, and Guangxi Province gradually eased, and most of Guizhou Province, western Yunnan, and a small part of Guangxi Province experienced drought. Southwest China entered the rainy season in May, and the drought was basically relieved”. From Figure 11 and the government report, we can observe that the SSMI is more sensitive to drought in the southwest China.

4. Discussion

This work calculated the SSMI drought index by GLDAS root zone soil moisture data to monitor the development process, frequency, and intensity of agricultural drought in southwest China from 2000 to 2020. The main purpose of this work is to monitor agricultural drought in the past 20 years and verify that the GLDAS soil moisture data can be applied to the monitoring of drought in small areas. Compared with previous studies on drought in southwest China, the novelty of this paper is that we used soil moisture data from a land surface model and calculated a new agricultural drought index SSMI. We quantitatively evaluated China based on the SSMI results showing the intensity and duration of drought in various regions of southwest China and the temporal and spatial changes of drought. By comparing the changes of different types of droughts in different regions in different months, we revealed the seasonal characteristics of droughts in various regions of southwest China. Finally, the article compared the results of the SSMI, SPEI and PDSI indices and combined the government and related literature records for verification. We found that the SSMI, SPEI and PDSI could capture the historical drought in the southwest China. However, the SSMI was more sensitive to drought monitoring and could describe the occurrence and changes more accurately. Based on the quantitative assessment of the SSMI on drought in southwest China, we can carry out specific work on the prevention of future drought. For example, our research found that Yunnan Province has been prone to drought in spring and winter for the past 20 years; therefore, before the winter and spring seasons come, specific drought prevention can be carried out through water conservancy projects. These works are the new finding with the help of SSMI.

Our research also has certain deficiencies and limitations. First, the spatial resolution of the soil moisture data used to calculate the drought index is low, making it difficult to accurately monitor the details of the drought development process and bring about the precise quantitative evaluation of drought. Although the calculation of the agricultural drought index SSMI is simple, it is a single element (soil moisture) drought index, which is only suitable for monitoring agricultural drought. As only soil moisture is used for calculation, SSMI cannot reflect the multi-scale characteristics of drought events, which will cause specific errors in the results of drought monitoring. Simultaneously, SSMI requires long-term and continuous spatiotemporal soil moisture data. If the regional soil moisture data is missing or the period is too short, this will lead to apparent deviations in the SSMI results, because the SSMI - monitors drought by calculating the degree of deviation of soil moisture. Therefore, the SSMI is not suitable for the situation of missing regional soil data and short time series.

Drought events have occurred frequently in southwest China in the past 20 years. Many studies have pointed out that the increase in the frequency and intensity of drought may be caused by climate change and related human activities [38]. Global warming has caused abnormal atmospheric circulation and increased surface temperature, which has led to a decrease in regional rainfall and an increase in ground evapotranspiration, which has led to frequent droughts [1]. Frequent human activities have led to increased water demand, and unreasonable logging has exacerbated drought development [80]. The frequent occurrence of drought events will cause the loss of grain production and even threaten the safety of the drinking water supply. Therefore, further analysis of the mechanism of drought and improvement of the accuracy of monitoring for the early warning signs of drought are the next focus. This work used the soil moisture obtained by the land surface process model to study the drought in the southwestern region and confirmed the feasibility of the land surface process model to obtain soil moisture data for the study of drought in small areas. Through the quantitative research on the severity, duration and frequency of drought, the occurrence of past drought can be studied in detail. For the next step, we will study the mechanism of drought events in southwest China in conjunction with the land surface process model. Combined with the relevant results of drought severity-duration-frequency occurrence, the relevant quantitative indicators for future drought monitoring and prediction are set.

5. Conclusions

This work used the monthly data set of surface soil moisture (0–10 cm) provided by GLDAS 2.1-Noah from 2000 to 2020, with a spatial resolution of 0.25° × 0.25°, and the SSMI to monitor drought in southwest China. The SSMI was used to measure the severity of agricultural drought and can reveal a stable drought process and a reasonable start and end time for drought. We used the SSMI to collect data on the intensity, frequency, and scope of drought in southwest China revealing an apparent upward trend before 2014 and an apparent downward trend after 2014. Moreover, there are apparent differences in the frequency and intensity of drought in various regions of southwest China. Yunnan Province is prone to drought in spring. The intensity and frequency of drought events are higher than in other seasons. The border area between Sichuan Province and Chongqing City is prone to severe drought events in summer. The provinces of Guangxi Province and Guizhou Province are prone to drought events in spring, autumn and winter, and the frequency of drought in autumn and winter is lower than that in spring. However, the drought intensity and image range are significantly higher than in spring once it occurs. By comparing the results of drought monitoring by the SSMI, SPEI, and PDSI indices in the southwestern region, the SSMI was shown to more accurately monitor the development process of agricultural drought. At the same time, we found that the monthly variation of soil moisture in different provinces in Southwest China is consistent, but the seasonal variation of drought is different. We believe that this phenomenon is mainly caused by different meteorological changes.