A comprehensive drought monitoring method integrating multi-source data

Droughts are the most expensive natural disasters on the planet. As a result of climate change and human activities, the incidence and impact of drought have grown in China. Timely and effective monitoring of drought is crucial for water resource management, drought mitigation, and national food security. In this study, we constructed a comprehensive drought index (YCDI) suitable for the Yellow River Basin using principal component analysis and the entropy weight-AHP method, which integrated a standardized precipitation evapotranspiration index (SPEI), self-calibrating Palmer drought severity index (scPDSI), vegetation condition index (VCI), and standardized water storage index (SWSI). SWSI is calculated by the terrestrial water storage anomaly (TWSA), which can more comprehensively reflect the impact of surface water resources on drought (as compared with soil moisture-based indexes). The study results showed that: (1) compared with single drought index, YCDI has stronger ability to monitor drought process. In terms of time scale and drought degree, the monitoring results based on YCDI were similar with data presented in the China Flood and Drought Bulletin and Meteorological Drought Yearbook, reaching ~87% and ~69%, respectively. The correlation between drought intensity and crop harvest area was 0.56. (2) By the combined analysis of the Mann-Kendall test and Moving T test, it was found that the abrupt change of YCDI index at the time of 2009, mainly due to the precipitation in 2009 reached the lowest value in the past 30 years in northern China and extreme high temperature weather. (3) The YCDI of Henan and Shandong provinces in the middle and lower reaches of the basin decreased more significantly, with the maximum value reaching 0.097/yr, while the index in the upper reaches showed an increasing trend with the maximum rate of 0.096/yr. (4) The frequency of mild drought, moderate drought, severe drought and extreme drought in the Yellow River basin during the study period was 15.84%, 12.52%, 4.03% and 0.97%, respectively. Among them, the highest frequency of droughts occurred in Ningxia, Inner Mongolia and central Shaanxi provinces. Drought causation in the Yellow River basin is more influenced by human activities than climate change in the middle and lower reaches, while climate change is the main factor in the upper reaches. Overall, YCDI is a reliable indicator for monitoring the spatial and temporal evolution of drought in the Yellow River basin, and it can be used for monitoring soil moisture changes and vegetation dynamics, which can provide scientific guidance for regional drought governance.


INTRODUCTION
Climate change and human activities have resulted in the frequent occurrence of hydrological extremes characterized by drought (Huang et al., 2019;Xiao et al., 2016). Drought is a naturally recurring hazard, different from other natural hazards since it often has a slow onset and it is difficult to forecast. The frequency, intensity, and duration of extreme drought events have increased since the early 2000s from global warming and during El Niño periods. As such, the living environment for human beings is compromised and the stability of the economy and society are threatened (Yue et al., 2015). China's climate is warming and drying, and the frequency and duration of droughts in the north are predicted to increase to all-time highs (Ayantobo, Li & Song, 2019;Dai et al., 2020;Shakeel et al., 2019;Trenberth et al., 2013;Zhu et al., 2018). Thus, timely and accurate monitoring of drought occurrence, severity, and spatial range are of significance.
According to different scientific perspectives, four types of droughts can be identified: meteorological, agricultural, hydrological, and socioeconomic drought (Leng, Tang & Rayburg, 2015). Meteorological drought is defined as the phenomenon of a water shortage caused by the imbalance between evaporation and precipitation, and the water loss is greater than the water gain in a certain period (Zhang et al., 2022). Agricultural drought is defined as the water deficit caused by the shortage of soil moisture during the growth period of crops, which affects the normal growth and development of plants (Zhou et al., 2021). Hydrological drought is caused by the long-term shortage of precipitation, resulting in a reduction of river flows, lake water levels, and reservoir impoundments (Loon & Anne, 2015). Socio-economic drought is defined as an abnormal water shortage caused by the imbalance between the supply and demand of water resources in natural systems and human socio-economic systems (Zhao et al., 2019). The social demand for water can be divided into industrial water demand, agricultural water demand, and water demand for living and service industries. The above four types of droughts are closely intertwined.
A drought index can quantify the complex mechanisms of drought from the perspectives of duration, strength, and severity (Mishra & Singh, 2010). However, due to the differences in drought characteristics around the world, scholars from various countries have developed numerous drought indexes and several drought information systems (e.g., measuring severity or magnitude) (Liu et al., 2020;Pulwarty & Sivakumar, 2014). At one time, researchers primarily defined drought as a decrease in precipitation over a certain period (such as with a Z index or the precipitation anomaly percentage) (Rahmati et al., 2020;Yu et al., 2020). The focus now is increasingly on water supply and demand differences, which links water deficit with duration and further supplements the concept of drought (such as Standardized Precipitation Index (SPI) and SPEI) (Erfurt et al., 2020;Yimam et al., 2021). The Palmer Drought Index, proposed in 1965, has been widely used in meteorology, hydrology, water resources, climate, agriculture, and other fields (Palmer, 1965). All drought indexes, however, are constrained by data sources and their regional application is limited by locations of particular weather stations.
The construction and calculation of drought indexes is a critical foundation for studying regional drought quantitatively. The meteorological drought index, the hydrological drought index, the agricultural drought index, and the socio-economic drought index are the most commonly used drought indexes in the international community. They are based on the relative deficit of water in the water cycle, that is, the corresponding form of drought and the classification of drought (Andrew et al., 2022). The drought indicator based on insufficient precipitation is the most common since it is easy to understand, simple, and intuitive. However, it has a long response time, a low sensitivity, and a poor portrayal of drought severity, such as precipitation anomaly percentage, SPI, drought Z index, and so on (Hurtado et al., 2021;Wu et al., 2001). Although a measure based on surface water deficits such as runoff distance level percentage and natural water area distance level percentage can more accurately depict the actual drought conditions, it lacks real-time and comprehensive observation data (Jeng et al., 2005). The index based on soil water deficit include soil relative humidity and soil effective water storage. These indexes can better reflect agricultural drought conditions and related drought losses, but the raw data is difficult to obtain and the accuracy is low. The percentages of groundwater level spacing based on groundwater shortages can better reflect actual drought situations, although observation data is lacking. The standardized precipitation evapotranspiration index, the integrated meteorological drought index (He et al., 2015), the integrated hydrological drought index (Yang et al., 2015a), and the Palmer drought index based on overall water deficit are constructed from several factors in the hydrological cycle, which can explain the drought phenomenon and level from the perspective of water conversion process (Palmer, 1965). In general, none of the current drought indexes have strong physical meaning. Therefore, developing comprehensive drought indexes is an important challenge for drought monitoring.
The establishment of a comprehensive drought index is of great importance for drought monitoring. In recent years, many researchers have attempted to create a comprehensive drought index that integrates multiple drought-related hydrometeorological variables. Considering the needs of different agencies for drought monitoring, integrated drought indexes are often preferred (Huang et al., 2015;Huang et al., 2019;Zhang et al., 2019). Comprehensive drought indexes take into account meteorological, hydrological, and agricultural drought factors. The methods used to combine drought indexes include weight combination, multivariate joint distribution, and machine learning (Andrew et al., 2022;Han et al., 2021a). The raw data for establishing integrated drought indexes is also broadly divided into two categories: one using basic data such as meteorology, vegetation, and soil, and the other using existing drought indexes (Shrestha et al., 2020;Wang et al., 2021).
By referring to other research results, we found that SPEI, scPDSI, VCI, and SWSI all have their own advantages for different types of drought monitoring. The SPEI integrates precipitation and evapotranspiration, allowing it to reflect not just the distribution and trend of regional wet and dry, but also variations in drought at different scales, which can reflect various sorts of drought conditions (Yang et al., 2017). The scPDSI calculation takes into account the entire water balancing process, including precipitation, soil water content, runoff, and potential evapotranspiration, and can more correctly describe drought conditions in a given region (Zhang et al., 2022). The VCI can better depict water stress status and not only monitor and track regional drought but also explain spatial and temporal changes in vegetation, rendering it the optimum data for large-scale remote sensing drought monitoring, and its reliability has been proved by a large amount of data (Zhang et al., 2017). The SWSI is calculated based on information on changes in surface water, soil moisture and groundwater storage, and thus is of great use in monitoring and assessing global and regional droughts (Tapley et al., 2019). As a result, we selected four typical drought indexes to create a comprehensive drought index for the Yellow River basin in this research. A combined weight scheme of subjective and objective weights was used, based on entropy weight-Analytic Hierarchy Process (AHP), to determine reliable weight values of multiple indexes.
Understanding the temporal and spatial variation of drought in the Yellow River Basin in the context of climate change is critical for water resource management and agricultural production in inland China (Piao et al., 2010). The goal of this study was to introduce a general drought definition. This general definition can improve drought monitoring, help decision-makers better allocate water resources, and reduce the impact of drought on different economic and social sectors. The specific objectives of this study are to: (1) Construct the YCDI of the Yellow River Basin using the entropy weight-AHP weight method, based on SPEI, scPDSI, VCI and SWSI; (2) validate the comprehensive drought index (YCDI) of the Yellow River Basin using typical drought events with the China Flood and Drought Bulletin and Meteorological Drought Yearbook; (3) use trend analysis, mutation analysis, and drought frequency analysis to investigate the temporal and spatial variation of drought in the Yellow River Basin; (4) explain the causes of drought in the Yellow River Basin and assess the impact of human activities and climate change on drought disasters.

Study area
The Yellow River Basin is located 32-42 N and 96-119 E, a drainage area of 795,000 km 2 , 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.  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 C 20 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.
In the formula, TWSA is the monthly water reserve change, a is the equatorial radius; q ave is the mean density of the Earth; l, m are the order and number of the Earth's gravity field, P lm cos h ð Þ is the fully specified connective Lejeune function; k l is the load Lejeune number corresponding to order l; DC lm , DS lm are the normalized spherical harmonic coefficient variables; h is the calculation point residual is the longitude of the calculation point; k 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: In the formula, TWSA i;j is the monthly water reserve change, TWSA j;mean is the monthly water reserve change mean, and TWSA 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: 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: where the constant terms c 0 , c 1 , c 2 , d 1 , d 2 , and d 3 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: where x i is the comprehensive weight of the first evaluation index and k is the preference coefficient (k = 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: where x 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: 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

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: 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   (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.

Drought trends
The corresponding Moving T test (Fig. 3A)    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)  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) andXu, 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.

CONCLUSIONS
Drought indexes that rely on a single variable are unable to effectively reflect drought conditions. As a result, developing an integrated indicator for drought evaluation is critical.
In this study, a comprehensive drought monitoring index YCDI, with SWSI, scPDI, VCI, and SPEI as raw data, were constructed. Compared with typical drought events, the applicability of the YCDI was verified, and on this basis, the temporal and spatial distribution of drought in the basin were analyzed. The main conclusions are: 1) YCDI has stronger ability to monitor drought process. In terms of time scale and drought degree, the monitoring results based on YCDI were similar with data presented in the China Flood and Drought Bulletin and Meteorological Drought Yearbook, reaching~87% and 69%, respectively. The correlation between drought intensity and crop harvest area was 0.56.
2) By the combined analysis of the Mann-Kendall test and moving T test, it was found that the abrupt change of YCDI index at the time of 2009.
3) The YCDI of the basin showed a weak downward trend in the past 14 years, that is, the drought risk has increased. The drought-prone areas are mainly distributed in the upper reaches of the basin, such as Gansu, Ningxia, and Inner Mongolia; the Guanzhong Plain in the middle reaches; and in the lower reaches, Henan and Shandong. The area west of the Lanzhou section maintained a state of no or light drought.
4) The YCDI of Henan and Shandong in the lower reaches of the basin decreased at a maximum rate of 0.097/yr, and the index of Qinghai in the upper reaches increased at a maximum rate of 0.096/yr.

Data Availability
The following information was supplied regarding data availability: The monthly total precipitation data and monthly average temperature data for 2002-2015 are available at the Meteorological Data Sharing Service System of China Meteorological Administration (http://data.cma.cn/).