A comprehensive drought monitoring method integrating multi-source data

Study area

The Yellow River Basin is located 32–42°N and 96–119°E, a drainage area of 795,000 km², as shown in Fig. 1 (Lu et al., 2021). The Yellow River is the second-longest river in China with a length of 5,464 km. It originates from the Tibetan Plateau, discharges into the Bohai Sea, and flows through nine provinces (Qinghai, Sichuan, Gansu, Ningxia, Inner Mongolia, Shaanxi, Shanxi, Henan, and Shandong) (Aoa et al., 2020). There are many mountains in the Yellow River Basin, and the elevation difference between east and west is large. From west to east are found the Qinghai-Tibet Plateau, the Loess Plateau, and the Huang-Huai-Hai Plain. It is rainy in summer and autumn, the annual precipitation is 123–1,021 mm, and the average annual precipitation is about 378 mm (Huang, Wang & Chen, 2022). The precipitation from June to September accounts for about 70% of the year, and the annual evaporation is 700–1,800 mm. There are four climatic regions: arid, semi-arid, semi-humid, and humid. The Yellow River irrigates 15% of the cultivated land, supports 12% of the country’s population, and drives 14% of the country’s gross domestic product (Wang et al., 2012). As such, it is known as the Mother River of China. However, in recent years, the challenges regarding balancing the supply and demand of water resources in the Yellow River Basin have become prominent, and drought problems have become serious.

Data

Remote Sensing data: The MODIS NDVI data (MOD13A3) with 1 km spatial resolution was downloaded from https://ladsweb.modaps.eosdis.nasa.gov for 2002 to 2015.

Drought event data: The historical drought events from 2006 to 2015 were obtained from the China Flood and Drought Bulletin published by the Ministry of Water Resources of the People’s Republic of China (http://www.mwr.gov.cn/sj/tjgb/zgshzhgb). The Flood and Drought Bulletin records the occurrence and development of drought, as well as intensity, scope, and associated disasters.

Meteorological Data: The monthly total precipitation data and monthly average temperature data for 2002–2015 were provided by the Meteorological Data Sharing Service System of China Meteorological Administration (http://data.cma.cn/).

scPDSI dataset: The scPDSI dataset (self-calibrating Palmer drought severity index based on Penman-Monteith) came from the Climate Research Unit (CRU) of the University of East Anglia (http://www.cru.uea.ac.uk/data); the spatial resolution is 0.5° × 0.5° and the period from 1901 to 2017.

GRACE data: This paper uses Gravity Recovery and Climate Experiment (GRACE) Level-2 gravity field model data provided by the Center for Space Research (CSR) of the University of Texas (https://www.csr.utexas.edu). Removed from the data are the effects of solid tides, sea tides, polar tides, non-tidal atmosphere and ocean mass generated by the Earth’s rotation, and the gravity disturbance caused by extraterrestrial stars. Finally, details of the datasets used in this study for drought indexes were given in Table 1.

TWSA changes from GRACE

GRACE estimates of mass variations over land reflect primarily related to changes in TWS (including snow/ice, soil moisture, and groundwater), plus other unmodeled signals, such as those related to postglacial rebound (PGR) or tectonics (Han et al., 2019; Jyolsna, Kambhammettu & Gorugantula, 2021). CSR GRACE data are regularized, spherical, harmonic coefficients that exclude the effects of solid tides, sea tides, solid polar tides, marine polar tides, non-tidal atmospheres and seas, and gravity disturbances generated by planets like the sun and moon (Yang et al., 2015b). The SLR coefficient replaces the C₂₀ term and the first-order term is rectified. The 60-order coefficient is reduced in this article to prevent the impact of high-order noise and odd-even-order correlation error. Gaussian filtering and the P3M15 reverse boundary extension were employed for processing. The mass variations were then converted to TWSA in units of equivalent water height (EWH) (Scanlon et al., 2016). The WGS 1984 datum were used to project the 1° × 1° grid data.

(1) $$\begin{array}{ll}TWSA& =\frac{a{\rho }_{ave}}{3{\rho }_{water}}{\sum }_{l=0}^{\mathrm{\infty }}{\sum }_{m=0}^l{P}_{lm}\left(\cos \theta \right)\frac{2l+1}{1+k_l}\\ & \cdot \left(\mathrm{\Delta }{C}_{lm}\cos \left(m\lambda \right)+\mathrm{\Delta }{S}_{lm}\sin \left(m\lambda \right)\right)lm{P}_{lm}\left(\cos \theta \right)k_l\end{array}$$

In the formula, $TWSA$ is the monthly water reserve change, $a$ is the equatorial radius; ${\rho }_{ave}$ is the mean density of the Earth; $l$, $m$ are the order and number of the Earth’s gravity field, $P_{lm}\left(\cos \theta \right)$ is the fully specified connective Lejeune function; $k_l$ is the load Lejeune number corresponding to order $l$; $\mathrm{\Delta }{C}_{lm}$, $\mathrm{\Delta }{S}_{lm}$ are the normalized spherical harmonic coefficient variables; $\theta$ is the calculation point residual is the longitude of the calculation point; $\lambda$ is the longitude of the calculation point.

The single drought index calculation

SWSI is a hydrological drought index, which is used to evaluate the loss state of water storage change (Araghinejad, 2011). Its formula is:

(2) $$\mathrm{S}\mathrm{W}\mathrm{S}\mathrm{I}=\frac{\mathrm{T}\mathrm{W}\mathrm{S}{\mathrm{A}}_{\mathrm{i},\mathrm{j}}-\mathrm{T}\mathrm{W}\mathrm{S}{\mathrm{A}}_{\mathrm{j},\mathrm{m}\mathrm{e}\mathrm{n}}}{\mathrm{T}\mathrm{W}\mathrm{S}{\mathrm{A}}_{\mathrm{j},\mathrm{s}\mathrm{t}\mathrm{d}}}$$

In the formula, $TWS{A}_{i,j}$ is the monthly water reserve change, $TWS{A}_{j,mean}$ is the monthly water reserve change mean, and $TWS{A}_{j,std}$ is the monthly water reserve change standard deviation.

For vegetation regions, the weather-related Normalized Difference Vegetation Index (NDVI) changes are smaller than the ecosystem-related ones and the drought impacts on vegetation cannot be easily detected from NDVI data directly (Quiring & Ganesh, 2010). Thus, Kogan (1995) developed a vegetation condition index to control local differences in ecosystem productivity, which is used to reflect the state of vegetation in different growth periods. It is defined as:

(3) $$VCI=\frac{NDV{I}_i-NDV{I}_{min}}{NDV{I}_{max}-NDV{I}_{min}}$$where NDVI_i is the monthly normalized vegetation index, NDVI_max and NDVI_min are monthly normalized vegetation index maximums and minimums over the study period. VCI ranges from 0 to 1 corresponding to the changes in vegetation condition from unfavorable to optimal.

The SPEI is a regional meteorological drought monitoring index proposed by Vicente-Serrano (Vicente-Serrano, Beguería & López-Moreno, 2010). The index represents regional drought conditions by calculating the deviation degree of the difference between precipitation and potential evapotranspiration from the average state (Beguería et al., 2014). Considering the influence of precipitation and temperature on drought, the calculation is:

(4) $$SPEI=\{\begin{array}{c}w-\frac{c_0-c_1\times w+c_2\times w^2}{1+d_1\times w+d_2\times w^2+d_3\times w^3}\hfill \\ \begin{array}{c}\begin{array}{c}w=\sqrt{-2\ln \left(p\right)}\begin{array}{cc},\hfill & p\le 0.5\hfill \end{array}\hfill \\ -\left(w-\frac{c_0-c_1\times w+c_2\times w^2}{1+d_1\times w+d_2\times w^2+d_3\times w^3}\right)\hfill \end{array}\hfill \\ w=\sqrt{-2\ln \left(1-p\right)}\begin{array}{cc},\hfill & p>0.5\hfill \end{array}\hfill \end{array}\hfill \end{array}$$

where the constant terms c₀, c₁, c₂, d₁, d₂, and d₃ are 2.515517, 0.802853, 0.010328, 1.432788, 0.189269, and 0.001308; w is the cumulative probability-weighted moment; the cumulative probability P is the normalized result of the different data series of precipitation and potential evapotranspiration, where the calculation of potential evapotranspiration adopts the Penman formula.

Weight of the drought index calculation

The corresponding grid values were extracted from the meteorological stations, and 11,952 groups of grid values from April 2002 to December 2015 were used as weight samples. The weights of four indexes were calculated based on the principle of entropy weight-AHP weight method. Finally, the weights of each index in the comprehensive index were calculated by combining the two weights (Ni et al., 2009; Xie, Gu & Lin, 2014). The following calculation is used:

(5) ${\omega }_i=\lambda \omega {}_i^{\mathrm{\prime }}+\left(1-\lambda \right){\omega }_i^{\mathrm{\prime }\mathrm{\prime }}$

where ${\omega }_i$ is the comprehensive weight of the first evaluation index and $\lambda$ is the preference coefficient ( $\lambda$ = 0.5).

Aggregation

To create a monthly comprehensive drought index in the Yellow River Basin, a single drought index is integrated using the weight calculated by the formula (4). The equation is:

(6) $$YCDI={\omega }_{1}SPEI+{\omega }_{2}scPDSI+{\omega }_{3}VCI+{\omega }_{4}SWSI$$

where ${\omega }_i$ is the weight corresponding to a single drought index.

Drought frequency calculation

According to the classification of YCDI (Table 2), the frequencies of meteorological drought days for each level were calculated by the following equation:

(7) $$P_i=\frac{n_i}{N}$$

where N is the total months (165 months), n_i is the total number of different droughts, and i is the drought level (extreme drought, severe drought, moderate drought, mild drought) (Felicia, Omid & Amir, 2021).

Run theory

The primary priority of run-length theory is to uncover the statistical law of continuous occurrence of random events and calculate the probability distribution and return period of continuous durations quantitatively (Moyé & Kapadia, 1995). It is a useful theoretical method for figuring out the basic law of drought occurrence. Based on the YCDI calculation results, three characteristic variables of drought factors were separated using the run theory method from the YCDI sequence: drought frequency, drought duration, and drought intensity. During the drought process, drought intensity (s) refers to the accumulation of the difference between the drought index value and the drought threshold, and drought duration (t) refers to the time from the start to the end of the drought. Drought duration and intensity are two terms that can be used to describe and reflect the severity of a drought (Tayyebeh et al., 2020; Wang et al., 2020).

Mann-kendall and moving T test

The Mann-Kendall nonparametric statistical test is recommended by the International Meteorological Organization (WMO) for time series trend analysis of environmental data and is a tool for testing the trend of drought time series. The Mann-Kendall nonparametric test was used to analyze the significance of YCDI changes in the Yellow River basin, and when |Z| > 1.96, the significance test with confidence level α = 0.05 was passed. For the specific calculation of Mann-Kendall can be referred to Yue & Wang (2004). A Moving T test was also introduced in this study to improve the accuracy of the mutation test and eliminate the erroneous mutation sites created by the M-K test. For the specific calculation of Moving T test can be referred to Joses et al. (2019).

Verification of the comprehensive drought index in the yellow river basin

Based on the calculation of the principal component analysis-entropy weight method, the weights of drought indexes are shown in Table 2.

According to the weight of each drought index, the comprehensive drought index model of the Yellow River Basin is:

(8) $YCDI=0.2705SPEI+0.245scPDSI+0.2155VCI+0.269SWSI$

Based on the above formula and the observation data of 79 meteorological stations in the Yellow River Basin, we selected 13,035 YCDI values as our samples. The SPEI and other comprehensive drought index (Tayyebeh et al., 2020; Vicente-Serrano, Beguería & López-Moreno, 2010; Yu et al., 2020) were classified according to the national standard of Meteorological Drought Grades, combined with relevant research results and the frequency of drought occurrence in different grades in the Yellow River Basin. According to the requirements of 2% special drought, 5% severe drought, 10% medium drought, 15% light drought, and 68% no drought, five grades are shown in Table 3.

According to the 2009 China Flood and Drought Bulletin, a drought occurred in China’s winter wheat region in December 2008 and lasted until February 2009, with severe drought areas persisting until March. From October 2008 to March 2009, the YCDI monitored the drought development. The drought began in the Ningmeng reach, eastern Gansu, and western Shaanxi in October 2008. The drought spread to the southeast in November and December, with the range reaching its peak in January 2009. The drought was alleviated from February to March 2009 because of increased precipitation in the basin Fig. 2.

The China Flood and Drought Bulletin and Meteorological Drought Yearbook were used as standards to verify the monitoring accuracy of YCDI. The results showed that the accuracy of YCDI in drought monitoring could reach ~87%. (Table 4). In addition, we also verified the drought level and drought intensity. First, we compared the drought levels monitored by YCDI with the officially reported drought levels, and the agreement between them reached ~69% (Table 4). Second, we used the area of crop extinction in major provinces in the Yellow River basin from 2006 to 2015 as a metric of drought intensity to verify the accuracy of the YCDI drought intensity. Table 5 shows the coefficients of correlation between crop extinction area and drought intensity for each province in the Yellow River Basin (Note: Drought did not occur in Shandong and Qinghai provinces). By comparison, we find that the YCDI is very applicable for monitoring drought intensity in the Yellow River Basin.

Spatial-temporal analysis of drought in the Yellow River Basin

Drought trends

The corresponding Moving T test (Fig. 3A) exceeds the significance level line in 2004, 2006, 2009 and 2013. As shown in Fig. 3B, the mutation results of M-K of YCDI had UF > 0, an increasing trend in 2002 to 2006 and 2012 to 2015, and the UF and UB curves had intersection points in 2003, 2009 and 2015. the mutation time points of YCDI based on M-K test were 2003, 2009, 2015. Taken together, YCDI underwent a mutation in 2009. The reason for the abrupt change is that the annual precipitation in northern China in 2009 was the lowest in nearly 30 years, and that in some areas was the lowest in nearly 50 years. From June 23 to 27, 2009, there was widespread high temperature weather in the north of China, and this high temperature weather affected a wide range and intensity, with local maximum temperatures reaching 40 °C to 43 °C in southern Hebei, northern Henan, Shandong, etc. (Jiang et al., 2012).

The trends of drought occurrence in the Yellow River Basin also show obvious spatial patterns (Fig. 4). The YCDI of Henan, Shandong, and southwestern Shanxi in the lower reaches of the basin shows the most obvious decreasing trends, and the maximum decreasing rate reached 0.097/yr. The YCDI of central Inner Mongolia, southern Ningxia, central Shaanxi, eastern Gansu, northern Xining City in Qinghai Province, and the Yellow River coastal area of Shanxi Province showed a weak downward trend, indicating that the drought situation in the above areas is more serious. The YCDI in the southeast of Qinghai, the southwest of Gansu, and the north of Ningxia and Sichuan increased slightly, and the YCDI in the southeast of Qinghai increased most significantly with a maximum increase of 0.096/yr.

Drought frequency variation

As shown in Fig. 5, the frequency of extreme drought in the Yellow River Basin was 0–4.83% (Fig. 5A), mainly distributed in northern Ningxia, central Inner Mongolia, the junction of Shanxi and Henan, and the Gansu and Sichuan areas. The frequency of severe drought was 0.65–10.64% (Fig. 5B), mainly distributed in Ningxia, southern Inner Mongolia, central Shaanxi, southwestern Shanxi, and parts of Henan and Gansu. The frequency of moderate drought was 5.31–15.75% (Fig. 5C), mainly distributed in Ningxia Hui Autonomous Region, eastern Gansu, and the Yellow River in Inner Mongolia. The frequency of mild drought was 7.26–24.58% (Fig. 5D), mainly distributed in central Shaanxi Province, eastern Gansu Province, northern Shanxi Province, northern Ningxia, and central Inner Mongolia. The frequency of different grades of drought in the Yellow River Basin during the research duration was: mild drought (15.84%), moderate drought (12.52%), severe drought (4.03%), and extreme drought (0.97%), respectively.

Differences in response to typical drought occurrences between different drought indexes

From November 2008 to February 2009, severe droughts occurred in most parts of North China, eastern Northwest China, Huang-Huai, Jiang-Huai, Jianghan, and other northern winter wheat areas (Jiang et al., 2012). Compared with the same period in previous years, the precipitation decreased by 3–5%. The SPEI, scPDSI, VCI, SWSI, and YCDI drought index values corresponding to the Huining meteorological station were selected to explore the response differences of different drought indexes to drought events (Fig. 6). It can be seen that the trend of the YCDI is synchronous with SPEI and scPDSI, and it is highly sensitive to drought in the early stages. In the subsequent development, the SPEI index is weaker than YCDI and scPDSI in capturing persistent drought. The main reason is that the leading factor in the SPEI index is precipitation, and the evapotranspiration anomaly caused by temperature changes is a secondary factor. When precipitation occurs, the SPEI index will reduce the degree of drought evaluation and cannot achieve complete monitoring. The scPDSI index is slightly higher than other drought indexes because the scPDSI index not only includes precipitation and potential evapotranspiration but also considers the maximum effective water holding capacity of the soil. In arid areas or continuous drought events, the lower maximum effective water holding capacity will cause the scPDSI drought index to overestimate the drought grade.

The peak value of the SWSI lags after the YCDI, SPEI, and scPDSI. The reason is that the important supply source of land water reserves in the Yellow River Basin is precipitation, and the conversion of precipitation to water reserves has a lag, which is consistent with the time required for meteorological drought to develop into hydrological drought. Compared with other indexes, the SWSI index is not sensitive to severe drought, which may be limited by GRACE data quality accuracy. The VCI index shows strong annual variation characteristics and has a poor correlation with other indexes. It is not possible to fully describe drought characteristics solely based on vegetation growth state. Second, precipitation is not the most important factor in natural vegetation growth; instead, runoff and groundwater are more important. As a result, the VCI index will understate the severity of the drought. In conclusion, the YCDI developed in this paper combines the other four drought indexes, allowing for more accurate monitoring of drought occurrence, duration, and ending.

By selecting YCDI, scPDSI, SPEI, SWSI, and VCI indexes and precipitation data from August 2008 to May 2009, the advantages of the YCDI in monitoring drought events were further verified (Fig. 7). From August 2008 to May of the following year, the precipitation at Huining station showed a strong temporal difference. From August, precipitation began to decrease. Except for VCI and SWSI, the other three indexes showed a one-month lag to the precipitation, and the drought signal was captured in September. The sensitivity of the SPEI index was the highest, followed by scPDSI, and YCDI was the weakest. In the subsequent development, YCDI was capable of monitoring drought, while the SPEI index was greatly affected by precipitation and maintained the same trend as precipitation. The scPDSI index maintained a downward trend from September to February, which overestimated the degree of drought. After January, each index increased, with scPDSI and SPEI indexes having the most significant increases. The VCI index began to increase in March, and the SWSI index had the largest increase in March. Because precipitation needs a certain period to recharge runoff and groundwater, the monitoring results have a certain delay. The comparison shows that the YCDI has a better continuous monitoring ability for drought events.

The spatial change of drought in the Yellow River Basin as monitored by various drought indexes from October 2008 to March 2009 is shown in Fig. 8. The differences can be attributed to both the index compositions and climate change. The scPDSI is the most analogous to the YCDI, however, when compared to other indexes that monitored drought development, changes are not as clear. The key reason for this is that scPDSI generates parameters using historical data from each station and more objectively examines the regional variability of parameters (Shrestha et al., 2020). The SWSI constructed based on TWSA lags slightly behind other drought indexes for about two months. The monitoring results of SWSI in the eastern portion of the basin from December 2008 to March the following year were not consistent with those of other drought indexes. Soil moisture in North China is influenced by both natural and human influences. The exploitation of groundwater provides for 70% of total water resource use in the North China plain and irrigation water accounts for 81.7% of agricultural water. In the North China plain, 79.7% of groundwater is used directly for agricultural irrigation (Yang et al., 2017). Irrigation will invariably have an impact on shallow soil water and vegetation growth. As a result, VCI, scPDSI, and SPEI are unable to adequately depict drought in North China.

Limited by the number of meteorological stations, YCDI also failed to effectively monitor drought in North China. Related studies have shown that the applicability of SPEI in humid areas is high, but it is relatively poor in arid and semi-arid areas because SPEI will overestimate the contribution of temperature anomalies to drought (Liu, Kang & Wang, 2016). VCI reflects drought from the growth of vegetation, but other factors, such as pests and diseases, floods, improper fertilization, and lack of irrigation, will also lead to poor vegetation growth (Jibril et al., 2022). As a consequence, VCI only represents comprehensive information on vegetation growth and cannot specifically reflect drought information, particularly in agricultural producing areas, as is shown in Fig. 8. The comparison reveals that the YCDI incorporates the benefits of each indicator and that drought monitoring results are consistent with official reports.

Factors influencing drought in the Yellow River Basin

The trend and frequency of drought in the Yellow River Basin have significant regional variation. From the spatial distribution and frequency of YCDI, we can see that the eastern and central parts of the Yellow River basin belong to drought-prone areas (Figs. 3 and 4). This conclusion is consistent with the findings of Huang et al. (2020) and Peng et al. (2011). In addition, we found that the frequency of droughts in the Yellow River basin gradually increases from south to north, which is also verified with the results of She, Xia & Du (2012) and Xu, Yang & Yang (2014). Therefore, the YCDI proposed in this study is scientifically reliable and has strong applicability compared with the results of existing studies on drought in the Yellow River basin.

Under the combined effect of climate warming and human activities, the whole Yellow River basin is moving toward a warming and drying trend, and extreme climate events are becoming more frequent. As a result, this has led to an increased risk of drought in the Yellow River basin, especially in the middle and lower reaches of the basin. Over the past 70 years, residential water use in the Yellow River basin has jumped from 20% to 80% of runoff, with surface water resource utilization well above the 40% runoff health alert (Piao et al., 2010). However, due to the large east-west span of the Yellow River basin and significant climatic differentiation, the causes of disasters also differ significantly. The drought in the lower Yellow River basin is mainly influenced by human activities. This region is one of the most important food resources in China, and crop irrigation leads to a continuous decrease in effective soil water holding capacity, which corresponds to the trend of YCDI. In contrast, the high-frequency drought in the Ningxia-Inner Mongolia region in the upper part of the basin is caused by both human activities and climate change. The Ningxia-Inner Mongolia section of the upper Yellow River has a temperate continental climate with low precipitation (200–300 mm/year) and high evaporation, where grassland is the primary vegetation type (Wu et al., 2018). Due to the considerable variations in regulating climate, soil, and water conservation between single vegetation ecosystems and complex community ecosystems, as well as the loss of limited precipitation in the form of runoff, effective water control is not achievable. Secondly, energy bases with high water consumption, such as Ningxia Eastern Energy Base and Hohhot-Baotou-Erdos Economic Circle. Regional droughts are becoming more frequent and their intensity is growing as the regional economy continues to expand and urbanization accelerates (Ma et al., 2019).

The limitations of YCDI for drought monitoring

To construct a comprehensive drought index, it is necessary to integrate meteorological, hydrological, and vegetation factors at different temporal and spatial scales. Different drought indexes have varied monitoring results for drought events, which depend on the calculation method and underlying data (Feng et al., 2019). The four single indexes used in this paper describe drought from aspects of meteorology, vegetation, and hydrology, but the ability of the VCI index to monitor drought is limited in the winter dormancy period of vegetation. The resolutions of the four single indexes were inconsistent. After the unified resolution, the YCDI resolution is 1° × 1° which can be used for effective monitoring of drought on a large scale but is still not sensitive to drought on small scales.

Because of global climate change and land-water cycle disruptions, extreme hydrological events are more common (Levizzani & Cattani, 2019). Flash droughts are very short (e.g., a few days or weeks) and are high magnitude events, and are often accompanied by a heatwave, soil moisture loss, and evapotranspiration increase, thus having a serious impact on crop yield and water supply (Wang & Yuan, 2018; Zhang et al., 2022). The YCDI on a monthly scale has poor adaptability to unexpected drought, and it cannot match the eight-day or even daily scale drought monitoring requirements. In the future, our focus will be on developing a set of short-term drought monitoring indexes.

Compared with other research results, the main contribution of this paper is the introduction of the SWSI. Its advantage is that drought can be considered comprehensively from the water stored in each layer of land and it considers the impact of human activities on hydrological flux. In the Loess area of the middle reaches of the Yellow River Basin, the complex correlation between surface water and groundwater makes it difficult for most hydrological models and observation methods to capture the accurate water change in the region. TWSA is an important indicator of regional water resources changes, reflecting the changes in surface water, groundwater and river and lake water (Tapley et al., 2019), the SWSI has the advantages of uniform distribution and scale. In general, SWSI has great application potential and can reflect the temporal and spatial variation of drought in time and space, especially for large-scale areas with hydrological and meteorological data (Han et al., 2021b). However, the limitation of the SWSI is that the accuracy and stability of its monitoring results are affected by the quality of GRACE data. The TWSA time series is relatively short, resulting in its multi-year average not being representative. Because of the periodic variations of TWSA, some scholars believe that the time series of TWSA over 30 years is preferable (Deng et al., 2020). These factors make the use of SWSI for drought monitoring and analysis still uncertain, but the GRACE gravity satellite data remains the most effective technical means for monitoring land water reserves (Nie et al., 2018). With the operation of GRACE-Follow on, the future construction of a comprehensive drought index based on a long-term series of the SWSI will improve the monitoring accuracy.