Next Article in Journal
Optimizing WRF Configurations for Improved Precipitation Forecasting in West Africa: Sensitivity to Cumulus and PBL Schemes in a Senegal Case Study
Previous Article in Journal
The Perceptions of Rural Australians Concerning the Health Impacts of Extreme Weather Events: A Scoping Review
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Reconstruction of Water Storage Variability in the Aral Sea Region

by
Nikita Murzintcev
1,
Sahibjamal Nietullaeva
2,
Timur Berdimbetov
1,
Buddhi Pushpawela
3,*,
Asiya Tureniyazova
1,
Sherly Shelton
4,
Bakbergen Aytmuratov
1,
Khusen Gafforov
5,6,
Kanat Parakhatov
7,
Alimjan Erdashov
1,
Abdul-Aziz Makhamatdinov
8 and
Timur Allamuratov
9
1
Department of Computer Engineering, Nukus State Technical University, Nukus 130100, Uzbekistan
2
Department of Data Transmission Systems and Networks, Nukus State Technical University, Nukus 130100, Uzbekistan
3
The Department of Physics and Astronomy, The University of Alabama in Huntsville, Huntsville, AL 35899, USA
4
The Department of Plants, Soils, and Climate, Utah State University, Logan, UT 84322, USA
5
Scientific Research Institute of Irrigation and Water Problems, Tashkent 100187, Uzbekistan
6
Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi 830011, China
7
Information Systems and Technologies, Nukus State Technical University, Nukus 130100, Uzbekistan
8
Department of Artificial Intelligence and Cybersecurity, Nukus State Technical University, Nukus 130100, Uzbekistan
9
Department of Television Technologies, Nukus State Technical University, Nukus 130100, Uzbekistan
*
Author to whom correspondence should be addressed.
Climate 2025, 13(9), 182; https://doi.org/10.3390/cli13090182
Submission received: 7 June 2025 / Revised: 6 August 2025 / Accepted: 22 August 2025 / Published: 29 August 2025

Abstract

The Gravity Recovery and Climate Experiment (GRACE) mission, operational from 2002 to 2017, provided critical measurements of Earth’s gravity field anomalies which have been extensively used to study groundwater and terrestrial water storage (TWS) dynamics. In this research, we utilize GRACE data to identify, model, and analyze potential climate parameters contributing to the reconstruction of TWS variability in the Aral Sea Basin region (ASB). We assess the impact of climate change and anthropogenic nature management on TWS change using a quantitative method. Our analysis reveals a significant decline in the TWS at a rate of 0.44 cm year−1 during the 2005–2009 period, primarily attributed to the prevailing drought conditions in the region. Notably, the estimated impact of anthropogenic influence on TWS during the same period of −1.39 cm year−1 is higher than the influence of climatic variables, indicating that anthropogenic activity was the dominant factor in water resource depletion. In contrast, we observed an increase in TWS at a rate of 0.82 cm year−1 during the 2013–2017 period, which can be attributed to the implementation of more effective water resource management practices in the ASB.

1. Introduction

The Aral Sea (AS) region is an endorheic basin formed by two major rivers, the Amu Darya and Syr Darya, originating from the Pamirs and Tianshan Mountains [1,2]. The lake has been shrinking at an alarming rate since the 1960s. By 1997, its surface area had reduced to 10% of its original size, and fluctuated between 10% and 25% thereafter [3,4]. This shrinkage is primarily attributed to agricultural activities, as river water is extensively consumed upstream for irrigation, leaving little inflow to sustain the lake [5,6].
Previous studies have analyzed the effects of both natural and anthropogenic factors on changes in terrestrial water storage anomalies (TWSAs) [7,8,9,10,11]. In particular, Yi et al. [10] identified a linear correlation between precipitation anomalies and water storage fluxes, estimating the annual anthropogenic impact on TWS change in Asia as 18,187 ± 38 km3 per year [10]. Similarly, Felfelani et al. [8] quantified human influence by comparing two models–one incorporating anthropogenic forcing and the other excluding it [8]. However, these studies face limitations due to uncertainties in global conditions and coarse temporal resolution. For instance, Yi et al. [10] assumed a constant anthropogenic contribution and zero climate-related effect throughout the entire period from 1979 to 2015 [10]; with data only available at annual intervals. Felfelani et al. [8] also noted that fluctuations of the TWSA trend may arise from dataset errors, which can be as significant as the variability across different hydrological models [8]. In addition, global groundwater and land surface models can often exhibit substantial shifts in the amplitude of TWSA estimates [12,13]. Zhong et al. [11] extended this line of research by analyzing both the climate TWSA values and their long-term trends over the last decade, alongside the anthropogenic contribution, using GRACE-derived TWSA and precipitation data spanning four decades [11].
GRACE-derived TWS data have proven valuable for understanding long-term hydrological dynamics and persistent climate anomalies [7,8]. A relatively new method involves analyzing changes in the terrestrial water storage deficit index (TSDI) based on GRACE TWS, and only a few researchers have employed this approach to detect changes in drought. Cao et al. [14] successfully used GRACE TWS data to analyze TSDI changes observed in the Northwestern China region during 2003–2012 [14]. Also, Nemati et al. [15] applied this method to monitor drought processes observed in Iranian territory from 2002 to 2015 [15]. However, these methods remain relatively new and have not been previously applied to monitor TWS dynamics and the drought process in the Aral Sea Basin.
In this study, we reconstructed terrestrial water storage (TWS) variability in the Aral Sea Basin (ASB) from 1989 to 2020 and estimated the human impact on the groundwater balance. GRACE satellite data were further utilized to model groundwater changes and to analyze possible climate parameters influencing the variability of TWS in the ASB. The proposed method enables the separation of climate-related and anthropogenic contributions to changes in water resources. This approach provides clear insights into observed transformations in water conservation and resource management practices. In addition, we analyzed the deficit of TWS in the region using a combination of GRACE-derived TWS data and precipitation records.

2. Materials

2.1. Study Area

We selected a sub-region of the Aral Sea Basin (ASB) where groundwater levels exhibit an increasing trend. In this area, a positive groundwater trend of 1–3 cm per year or more is observed during the growing season [1]. The study region (Figure 1) is primarily located in the southern part of the ASB territory and encompasses large irrigated agricultural lands [16,17,18,19], which are the main consumers of water from the Amu Darya and Syr Darya rivers [20]. The research region spans approximately 550,000 km2, extending about 980 km north to south and 560 km east to west (see Figure 1). For the purpose of analysis, the area is subdivided into three sub-regions—upper (UASB), middle (MASB), and lower (LASB)—based on the dominant hydrological regime and in alignment with established conventions in the scientific literature [16]. According to the ECA CCI data (2002), the largest share of cropland area is concentrated in the lower ASB, accounting for 64.1% of the total cropland in the basin. In contrast, the upper and middle parts of the ASB regions comprise 14.5% and 24.4%, respectively [16,21]. Analysis of CRU climate data reveals that the lower and middle ASB regions experience relatively higher temperatures and potential evapotranspiration (PET), along with lower precipitation during the growing season, compared to the upper ASB sub-region [17,20,22].

2.2. Datasets

To train the hydrological model, we use GRACE-derived terrestrial water storage (TWS) as the target (the prediction) variable and total precipitation as the primary input variable. The quality of these datasets is therefore crucial to our analysis and is discussed in detail later in this section. The GRCTellus Land dataset represents mass change estimates derived from spherical harmonic solutions produced by the GRACE instruments operated by CSR, JPL, and GFZ. It has a spatial resolution of 1 degree in both latitude and longitude [12,23,24], and covers the period from 2002 to 2017. However, the GRACE time series is notably irregular, with temporal intervals ranging from 11 to 99 days [25]. Our method of processing this uneven time series is presented in Section 3.4. While Humphrey et al. [26] utilized GPL mascon solutions, Zhong et al. [11] compared different GRACE products, ultimately recommending CSR data for medium- and small-sized basins. We also employ the CSR dataset and apply gain factors as described in previous research [12].
Total precipitation data are obtained from the ERA5 reanalysis dataset, a global climate product developed by the European Center for Medium-Range Weather Forecasts (ECMWF). ERA5 provides data from 1979 to the present at an hourly interval. For our purpose, we aggregate this data into daily total precipitation and resample it from 0.25° to a 1° spatial resolution to match the GRACE dataset [27].

3. Methodology

3.1. Climate-Driven Water Storage Variability

Our research builds upon the foundational work introduced in Humphrey et al. [28] and further developed in their subsequent studies [11,26]. In their approach, Humphrey et al. proposed predicting TWS using climate-driven variables, with a linear polynomial model. In this model, a combination of input parameters, specifically temperature and precipitation, is fitted against the GRACE-derived observations as follows:
T W S = a 0 + a 1 × f i l t e r P R E , τ + a 2 × T M P + ε
A key innovation in Humphrey et al. [28] is the application of a decay filter, denoted as f i l t e r P R E , τ , to the total precipitation time series. This filtering approach, described in detail in Section 3.5, accounts for the delayed hydrological response to precipitation, particularly by distributing the influence of intense rainfall events occurring near the end of GRACE’s irregular observation intervals into subsequent months. While we adopt the same general modeling framework, our methodology differs in the decomposition of the time series (see Section 3.4) and in the constraints applied to the GRACE data. Nevertheless, the core processing approach, including the decay filter, remains consistent with that of Humphrey et al. [28]. Further details are provided in Section 3.5.

3.2. Model Identification

To estimate free parameters a 0 ,   a 1 , a 2 , a n d   τ in the model defined by Equation (1), Humphrey et al. [26] employed Markov chain Monte Carlo (MCMC) [29] techniques to identify the optimal parameter set (Figure 2) [26,30]. However, the decay filter operation is computationally expensive. Since the model is linear in form, we adopt a more efficient approach by directly calibrating the filter’s decay parameter τ for each grid cell (Section 3.5), followed by fitting the mode accordingly. In the original formulation, Humphrey et al. [28] introduced temperature in the model as “a proxy for evapotranspiration”. However, subsequent studies, including Zhong et al. [11], observed that including temperature does not significantly improve model performance [11,30,31]. As noted in the previous section, in the ASB area, particularly during the growing season, there is a weak correlation between GRACE TWS anomalies and both temperature and PET derived from CRU data [1,21,32]. Consequently, our model reconstructs climate-driven TWS variations using only precipitation data. Finally, this leads to the following simplified reconstruction mode:
T W S r e c = a 0 + a 1 × f i l t e r P R E , τ  
where T W S r e c denotes the reconstruction of variations in GRACE TWS and P R E is the total precipitation. The linear coefficients a i ' s are calibrated against the GRACE mascon data. To implement this model, the total precipitation time series must be smoothed using a decay filter, with τ representing the decay steepness. All climate variables are aggregated to monthly means to match the temporal resolution of the GRACE dataset. Before model fitting, the GRACE time series and input variables are regularized by removing seasonal and linear trend components (see Section 3.4), as these are influenced primarily by large-scale climate processes. This ensures the model focuses on reconstructing detrended TWS anomalies (TWSAs), capturing regional-scale variations driven by precipitation.

3.3. Estimation of Climate and Human TWS Trends

Figure 3 illustrates the long-term climate-driven TWS changes over the period 1989–2017. According to the reconstruction methodology outlined in Humphrey et al. [26], the global linear trend of TWS over the reconstruction period (January 1989 to December 2017) is assumed to be zero.
By comparing two time-series, one reconstructed from precipitation data (TWSREC) and the other derived from the GRACE mission (TWSGRACE), we define the human-induced component of terrestrial water storage (TWSHI) as the difference between the observed and the predicted values:
T W S H I = T W S G R A C E T W S R E C    
This residual component, TWSHI, captures the anthropogenic impacts on water storage, including groundwater extraction, irrigation, and reservoir operations, which are not accounted for by climate-driven precipitation alone. Due to the high level of assumptions and simplified nature of the model, this analysis is best performed on trends extracted from decomposed time series. As described in Section 3.4, we first remove both the seasonal cycle and the global linear trend. The remaining components, comprising interannual and sub-annual variability, are then used to estimate regional trends in terrestrial water storage [11]. This approach allows for a more accurate assessment of localized, climate-driven and anthropogenic influences by isolating long-term signals from short-term fluctuations.
T r e n d ( H I ) = T r e n d ( G R A C E l o n g + s h o r t ) T r e n d ( R E C l o n g + s h o r t )  

3.4. Time Series Decomposition

Climatic and hydrological time series reflect the combined influence of multiple processes, such as global climate trend, long-term (inter-annual) groundwater depletion, seasonal regular harmonics, and short-term anomalies driven by atmospheric processes. To facilitate the analysis and comparison of time series with different origins and characteristics, it is common practice to decompose their signals into distinct components, ranging from low- to high-frequency. This decomposition allows for a clearer understanding of the dominant temporal patterns and enables the separation of natural variability from anthropogenic impacts.
T S = T S t r e n d + T S l o n g + T S s e a s o n a l + T S s h o r t  
While there are several well-established methods for decomposing regularly spaced time series, only a limited number are suitable for processing unevenly spaced data [33]. One such method is STL [34], which was originally designed to handle missing values and is moderately effective on uneven time series. Humphrey et al. [28] proposed a modified STL approach specifically tailored to handle GRACE data irregularities. In our study, we employ a custom implementation of the iterative algorithm described in Eckner [33], which is particularly well-suited for reconstructing high-quality seasonal components and converting unevenly spaced data in the series to a regular time grid. This facilitates subsequent data processing and improves the reliability of the trend analysis. This method was chosen because GRACE observations are irregularly spaced in time, posing a significant challenge for consistent temporal analysis. Converting these observations into a regular monthly time series is essential for comparing them with climate and anthropogenic impact data. Eckner’s method offers a statistically robust framework specifically designed to handle non-equidistant time series. Unlike simpler interpolation techniques, it effectively preserves the intrinsic variability of the original data while minimizing artifacts. As a result, it enhances the temporal consistency and analytical reliability of the total water storage anomaly (TWSA) time series, which is crucial for identifying long-term trends and seasonal patterns.

3.5. Exponential Decay Filter

Following common practice in hydrological modeling, Humphrey et al. [28] assumed that the influence of a precipitation event on water storage observed at time t i decays exponentially over time. To account for this delayed response, they proposed a weighting function that distributed the impact of precipitation over an extended period:
w t ; t i = 1 τ × e t t i τ  
where t denotes the discrete step of time (in days) elapsed since the precipitation event. For a strict mathematical definition of the method, refer to Humphrey et al. [28]. In their subsequent work, Humphrey et al. [26] stated that “applying an exponential decay filter to daily precipitation time series before averaging them to monthly resolution greatly improved their correlation with sub-seasonal TWS anomalies” ( G R A C E s h o r t ). However, this statement is somewhat misleading, as the authors explicitly calibrated the decay filter to maximize the correlation with GRACE observations. The task of determining the optimal value of decay parameter τ can be defined as a one-dimensional optimization problem: maximizing the correlation between G R A C E s h o r t and P R E m o n t h l y . This optimization function always has a single global maximum, allowing it to be efficiently solved using numerical approximation techniques, such as the method proposed by Brent [35] (for example, implemented in the ‘optimize’ function in R).

3.6. Analysis Based on Drought Index TSDI and SPI

The Standard Precipitation Index (SPI), introduced by McKee in 1993, is one of the most widely used and effective tools for monitoring drought variability and intensity over different time scales [36,37]. In this section, we apply the SPI to assess temporal changes in drought conditions across the study region. The classification of drought severity based on SPI values is summarized in Table 1.
The total storage deficit index (TSDI) is a highly effective method for identifying the wet and dry years within a given time period [14,38,39]. To evaluate changes in the TSDI, the first step is to compute the cumulative terrestrial storage deficit (TSD), which captures deviations in water storage from long-term climatological norms. The TSD is calculated monthly using Equation (7), and it quantifies how terrestrial water storage anomalies (TSAs) derived from GRACE data deviate from typical seasonal conditions. Specifically:
T S D i j = T S A i j M T S A j M a x T S A j M i n T S A j × 100 %  
where T S D i j represents the terrestrial storage deficit for the i -th year and j -th month. T S A i j denotes the terrestrial water storage anomaly in the same time period, as derived from GRACE data. M e a n T S A j refers to the mean of TWS anomalies in all j -th months over the study period. M a x T S A j and M i n T S A j are the maximum and minimum observed TWS anomalies in the j -th month, respectively. The TSD values are normalized to range between −100 and +100, where negative values indicate dry (arid) conditions and positive values represent wet (humid) conditions. Once the monthly TSDI values are calculated, the TSDI is derived by applying a recursive formulation that links current and past TSD values to capture drought persistence and intensity over time. This is given by Equation (8):
T S D I i = p × T S D I i 1 + q × T S D i    
Here, T S D I i represents the terrestrial storage deficit index at the i-th time step. In order to calculate the coefficients p and q , we construct a cumulative plot of the TSD values over time, with the cumulative TSD on the ordinate (y-axis) and time on the abscissa (x-axis). A linear regression is then applied to this plot, and the best-fitting line is used to determine the slope p and intercept q, which describe the long-term trend and baseline behavior of terrestrial water storage deficits over time. The coefficients are calculated as follows:
p = m m + b q = C m + b    
Here, m and b represent the slope and intercept derived from the linear trend of the cumulative TSD curve. The parameter C represents the intensity of the dryness, defined as the slope of the best-fit line applied to the cumulative TSD during the identified drought period. According to Palmer’s drought severity classification [40,41], droughts are categorized into four classes, and the appropriate value of C is selected accordingly (Table 2). It is important to note that there is no universally accepted standard for defining and characterizing drought severity (a monograph on drought), as discussed in the existing literature [14]. In our study, the value of the parameter C was estimated using the technique described in (Palmer, 1965) [40], which combines the cumulative TSD trends with the SPI [36]. The SPI was computed from precipitation data within and around the study area [14,15].

4. Results

4.1. Climate-Driven and Human-Induced Contributions to Water Storage Variations in the ASB

The method proposed by Humphrey and his co-authors was originally developed and tested using data from China. It enables the reconstruction of total water storage anomalies (TWSAs) and their analysis through time-series methods, which are widely used in climate change research. When applied to the Aral Sea basin, the method produced consistent and interpretable results. It also facilitated the estimation of human impact on groundwater storage dynamics, with findings that aligned with the independent studies on drought processes in the same region. The proposed modifications and optimizations in our study enhanced the robustness of the method and supported its applicability across diverse geographic and hydrological contexts.

4.2. Inter-Annual Change of TWSA

To better illustrate the year-to-year variability of TWSAs, monthly climate-driven and human-induced TWSA values were averaged to obtain annual means (Table 3). Both climate-driven and human-induced annual TWSA values exhibited a general increasing trend between 2002 and 2005 across all sub-regions. However, in the MASB and the LASB, the human-induced TWSAs remained consistently at negative during the period 2002–2003, indicating intensified anthropogenic stress in those areas. The maximum climate-driven TWSA occurred in different years for each sub-region: the UASB recorded its peak climate-driven TWSA in 2010 (1.36 cm), the MASB in 2005 (0.97 cm), and the LASB in 2016 (1.04 cm). Conversely, all sub-regions experienced their lowest climate-driven TWSA in 2008. In addition, a consistent decline in climate-driven TWSA was noted from 2006 to 2009, with negative values across all sub-regions, underscoring a regional climate-related water deficit.
In both GRACE-derived and human-induced TWSA changes, the maximum positive value was recorded in 2005 across all sub-regions. The minimum TWSA occurred in the UASB in 2008, with GRACE and human-induced values of −6.29 cm and −4.25 cm, respectively. In the MASB and the LASB, the lowest values were observed in 2002. Over the entire study period, the mean annual human-induced TWSA change was slightly positive in the UASB (0.20 cm) and the MASB (0.04 cm), while the LASB recorded a small negative mean value (−0.03 cm). Despite this, the human-induced TWSA trend in the LASB shifted positively in the later years (2015–2017), indicating a reduction in anthropogenic water stress. A similar positive shift was observed in the MASB over a longer period, from 2003 to 2017. Notably, the LASB also recorded the highest single-year positive human-induced TWSA value during the entire study period.
Table 4 presents the change in terrestrial water fluxes (TWFs), defined as the difference in annual TWS change between two consecutive years [11]. High positive climate-driven TWF changes were observed in the UASB between 2009 and 2010 (1.99 cm), while the MASB and the LASB recorded high values between 2009 and 2010 and between 2016 and 2015, respectively. For human-driven TWS, substantial positive changes were recorded successively between 2004 and 2005 across all sub-regions, with TWF values of 4.39 cm in UASB, 2.18 cm in the MASB, and 3.23 cm in the LASB.
Conversely, extreme negative climate-driven TWF values were observed between 2010 and 2011 in both the UASB and the LASB. The minimum GRACE and human-driven TWF values were observed between 2012 and 2013 in the UASB. Over the entire study period (2002–2017), the average TWF based on GRACE, climate-driven, and human-induced components showed a net positive trend across all sub-regions. Notably, in the MASB, no change was recorded in the human-induced TWF between 2013 and 2014. Similarly, the LASB showed a minimal difference of 0.04 cm between climate-driven and human-induced TWF for the 2012–2013 period, during which GRACE-derived TWF in the LASB was recorded as zero.

4.3. Trend of Climate-Driven and Human-Induced TWS

The identified model is relatively simple, relying solely on precipitation data. Therefore, the absolute values of the reconstructed TWS should not be considered fully credible. However, the model remains useful for analyzing the linear trends and temporal dynamics in groundwater change across different periods. Within the study timeframe, we identified two distinct periods during which climate-driven and human-induced TWSAs exhibited unusual patterns of development.
The first notable interval spans were from January 2005 to February 2009 (Figure 4a–c), corresponding to an extended drought period in the region. During this time, the climate-driven TWS variables were expected to decrease across the study area, with a mean trend 0.44 cm per year. Sub-region-specific trends showed varying rates of decrease: −0.84 cm per year in the UASB, −0.12 cm per year in MASB and −0.36 cm per year in the LASB. In contrast, the human-induced TWS trend during this period demonstrates a high negative trend compared to the climate-driven component, especially in the UASB. The human-induced TWS trend in this region was −2.52 cm per year, while the MASB and the LASB recorded −0.96 cm per year and −0.72 cm per year, respectively. Across the entire study area, the mean trend of human-induced TWS was −1.39 cm per year, indicating substantial anthropogenic pressure on groundwater resources during the drought.
The second interval, from January 2013 to December 2017, can be characterized as a relatively favorable period. During this period, both the predicted and observed thicknesses of the groundwater exhibited increasing trends across all sub-regions. Across the region, the UASB recorded a high climate-driven TWS trend (0.59 cm per year), while the LASB showed a lower trend value (0.34 cm per year). Notably, in the UASB, the climate-driven TWS trend (0.59 cm per year) closely matched the total observed TWS trend (0.58 cm per year), suggesting a dominant climatic influence. In contrast, the LASB exhibited the lowest human-induced TWS trend (0.14 cm per year), whereas the MASB had a moderately higher value of 0.29 cm per year.
A comparison between the two identified intervals reveals a clear shift in the nature of anthropogenic influence. During the 2005–2009 period, human activities significantly contributed to groundwater depletion. In contrast, the 2013–2017 interval reflects more sustainable natural resource management practices, with positive trends in both climate-driven and human-induced TWS. However, it remains unclear whether the observed groundwater decline in the earlier period resulted from direct extraction (e.g., pumping) or from insufficient irrigation water supply due to broader hydrological stress. The lack of detailed data on irrigation infrastructure and the hydraulic connectivity between surface and groundwater systems limits our ability to draw definitive conclusions. As previously noted, the study period reflects two contrasting scenarios: a predominantly negative anthropogenic impact on TWS during the first interval and a more favorable human contribution in the second. In the following section, we investigate the influence of water stress and climate variability on regional drought dynamics using GRACE-derived TWS data.

4.4. Determine Parameters of Terrestrial Storage Deficit Index

The TSD change was calculated using Equation (7) (Figure 5). Between April 2008 and September 2009, the mean TSD value was −26.43%, remaining consistently negative for 18 consecutive months. The lowest negative TSD value during this period was −49.21%, indicating a prolonged and severe drought. Therefore, this period is defined as the extreme dry period. Based on the TSD changes observed from April-2008 to September-2009, the necessary parameters for Equations (8) and (9) were computed.
Based on the cumulative TSD curve (Figure 6), the critical parameters m , and b needed to calculate the p and q parameters in Equation (9) were determined. The best-fit line was obtained from the cumulative TSD plot during the dry period, with a slope of approximately −3. Accordingly, the parameter C was set to 3 . Using the values C = 3 , m = 29.15 and b = 0.122 , p   and q in Equation (9) were calculated as p = 0.004   and q = 0.104 , respectively. Substituting the p and q values in Equation (10) yields the following recursive formula for the TSDI in the basin was obtained using Equation (9):
T S D I i = 0.004 × T S D I i 1 + 0.104 × T S D i  
For the initial month, the starting value T S D I 0 is set to 2% of T S D 1 following the recommendation in [1]. Furthermore, in cases where GRACE data are missing for a given month, it is recommended that T S D I i for that month be estimated as 2% of the previous month’s TSD value, T S D i 1 . Finally, the drought status of the study area was evaluated by calculating TSDI values and interpreting them using the classification criteria provided in Table 5 [42].

4.5. Temporal Drought Characteristics Based on the TDSI and SPI

Figure 7 shows the monthly TSDI and monthly SPI changes observed between 2002 and 2017 across the sub-regions. In the UASB, the mean TDSI value was 0.15, with the highest value of 4.78 recorded in May 2005 and the lowest of −4.71 in June 2008 (Figure 7a). In the MASB, the maximum and minimum TDSI values were observed at the beginning of the study period (Figure 7b), with the peak in May-2005 (5.11) and the lowest point in October-2002 (−4.76). In this region, mean TDSI was negative at −0.12. Among the sub-regions, the LASB appears to be the driest (Figure 7c), as it consistently recorded lower TSDI values compared to the UASB and the MASB. Specifically, the highest TSDI value was 4.13 (May 2005), while the lowest was −5.16 (October 2002). For the SPI, the highest positive value (2.29) was recorded in the UASB in November 2005. The lowest SPI value (−3.03) was observed in the MASB in April 2007. The mean SPI values also support the classification of the LASB as the driest region, with a mean SPI of −0.13, compared to 0.06 in the UASB and −0.11 in the MASB.
According to the drought classification criteria, a continuous TDSI value below −1 for a period of three months or more is considered a drought period [15]. Based on this threshold, we analyzed TDSI variations by segmenting the data into continuous drought intervals across the study area. Between 2002 and 2017, four distinct drought episodes—both long and short—were identified in the UASB (Table 6). The longest drought in the UASB occurred between May 2007 and June 2009, lasting 26 months, with a cumulative TDSI slope of −2.85. During this interval, the minimum and mean TDSI values were −4.78 and −2.49, respectively. The corresponding SPI values also indicate that a moderate drought affected the region during this time. Another prolonged moderate drought was observed in the UASB between February 2013 and October 2014, with a cumulative TDSI slope of −1.57. In the MASB, five consecutive drought periods were recorded throughout the study period. The longest of these was a mild drought lasting 22 months, from September 2007 to June 2009, with a cumulative TDSI slope of −2.22. During this interval, the mean TDSI and SPI values were −1.85 and −0.12, respectively. Additionally, another mild drought occurred between January 2013 and September 2014, lasting 11 months, with a cumulative TDSI of −1.58. Both mean and minimum values of the TDSI and the SPI during this period confirm the presence of drought conditions.
In the LASB, the most frequent and continuous drought periods were observed. Between 2002 and 2017, seven distinct drought events occurred across various time intervals. Notably, two mild drought episodes were recorded: one between February 2008 and August 2009, and another between February 2011 and November 2011, lasting 12 and 10 months, respectively. The cumulative TDSI slopes during these periods were −0.98 and −1.31. Corresponding mean and minimum SPI values also reflected negative indices, confirming drought conditions. It is important to highlight that during the first two years of the study (2002–2003), long-term continuous droughts of varying severity were observed across all regions—particularly in the LASB. In this sub-region, a 19-month continuous severe drought occurred with a cumulative TDSI slope of −3.26. During the same years, the UASB and the MASB experienced relatively shorter, yet severe, droughts lasting 12 months, with cumulative TDSI slopes of −3.16 and −4.34, respectively. Between 2002 and 2003, the mean TDSI and SPI values were closely aligned in the LASB, both indicating significant drought. However, in the UASB and the MASB, the mean SPI values showed only slight negative or near-neutral conditions, suggesting that the co-drought classification was strongest in the LASB. Additionally, in the LASB, a seven-month mild drought occurred from February 2013 to August 2013. This episode had a cumulative TDSI slope of −0.61 and a minimum TDSI value of −1.22. Both the mean and minimum SPI values during this time confirmed the presence of drought conditions in the region.
Drought frequency was analyzed using both TDSI and SPI values. Over the full 188-month study period, TDSI identified 77 drought months in the UASB and the LASB, and 87 months in the MASB. According to SPI, 87 drought months were observed in the LASB, 85 in the UASB, and 82 in the MASB. A year-by-year analysis reveals that the driest period was from 2007 to 2009, during which negative mean TDSI values were recorded across all regions. Mean annual SPI values also confirm that this interval marked a widespread drought. In contrast, the period following 2009 is characterized as relatively wet, as all regions recorded only positive annual TDSI values, particularly in 2010 when TDSI peaked. A pronounced seasonal difference is also evident. In the UASB, the mean TDSI was −0.16 during the vegetation (growing) season and 0.19 during the non-vegetation (non-growing) season. Similarly, in the MASB, mean TDSI values were −0.13 in the growing season and 0.21 in the non-growing season. In the LASB, both seasonal TDSI values were positive: 0.06 in the vegetation season and 0.36 in the non-vegetation season. Seasonal SPI trends support these findings, indicating drier conditions during the growing season and wetter conditions in the non-growing season, consistent with the TDSI observations. When comparing the two drought periods identified in Section 4.5—namely, January 2005 to February 2009 and June 2013 to June 2016—the data show that the first interval experienced significant drought, while the latter exhibited a transition to wetter conditions in all three sub-regions. Mean SPI values also confirm this pattern, showing mild drought during the first period and a relatively wet phase during the second.

5. Discussion

In this study, we quantitatively assessed the relative contributions of climatic and anthropogenic factors to terrestrial water storage (TWS) changes in the Aral Sea Basin (ASB), focusing on cropland areas within the Upper, Middle, and Lower ASB (UASB, MASB, and LASB) sub-regions [16,17,20,32,43,44]. GRACE-derived TWS observations for the period 2002–2015 exhibited negative correlations with all climate parameters except precipitation [17]. Following the methodology of previous research [26], we used ERA-Interim precipitation data as the sole climate variable to isolate climate-driven TWS changes. The delineation of the study area was guided by patterns of groundwater change [1].
Our analysis indicated that the monthly climate-driven TWS exhibited a declining trend of −0.012 cm per year from 2002 to 2017, while human-induced TWS changes revealed a modest increasing trend of 0.016 cm per year in recent years. Interannual climate-driven TWS anomalies were persistently negative during 2011–2017, averaging −0.67 cm in the UASB and −0.42 cm in the MASB and the LASB. Notably, all three sub-regions experienced strong negative TWS anomalies in 2008, whereas the highest annual positive anomaly was recorded in UASB in 2010 at 1.36 cm.
Seasonal analysis of monthly climate-driven TWS changes (Table 7) revealed positive anomalies during the non-growing season and negative anomalies during the growing season. The highest positive monthly TWS anomaly was observed in the UASB in January, while the MASB and the LASB recorded their peaks in December. Across all regions, climate-driven TWS values were consistently negative in the summer months (June–August), with the MASB showing the lowest anomaly of −0.72 cm in July.
Previous studies have documented a long-term decline in precipitation across the ASB region [1,45,46], primarily attributed to increasing evapotranspiration rates. In line with these findings, the observed negative trends in both precipitation and climate-driven TWS during the study period are likely driven by enhanced evapotranspiration. As noted in earlier studies, the daily potential evapotranspiration (PET) trend during the vegetation season increased by approximately 0.027 mm per day from 2002 to 2017, and concurrently, temperature trends in the region also exhibited a consistent upward trajectory [1,20,21].
As noted earlier, the monthly human-induced TWS trend showed an overall increase during the study period. The corresponding annual human-induced TWS trend was also positive, averaging 0.191 cm per year. Between 2002 and 2017, the highest positive trend was recorded in the LASB (0.096 cm per year), followed by the MASB (0.072 cm per year) and the UASB (0.025 cm per year). The contribution of human activity to TWS changes was negative during the early years of the study and became increasingly positive in later years. A comparative analysis of pre- and post-2010 periods highlighted this shift: in the UASB, the mean monthly human-induced TWS change was −0.19 cm before 2010, rising to 0.78 cm after 2010. Similarly, the MASB experienced an increase from −0.24 cm to 0.49 cm across the same periods. The LASB showed a comparable pattern of change.
Inter-annual human-induced TWS changes generally displayed positive values during the growing season, in contrast to climate-driven TWS, which exhibited negative changes during the same period. However, negative human-induced TWS values were observed during May–August in the MASB and the LASB, and from May–July in the UASB. Outside these months, all regions recorded positive human-induced TWS values. The lowest monthly value (−0.49 cm) was observed in the LASB in May. Notably, in July, both the UASB and the MASB recorded identical negative values (−0.31 cm per year) for seasonal human-induced TWS. On a seasonal basis, the human-induced TWS trend was negative during the vegetation period (−0.31 cm per year) and positive during the non-vegetation period (0.17 cm per year). From this, we can draw a straightforward conclusion. During the study period, water demand and pressure were significantly higher during the growing season, particularly in the agriculturally intensive regions of the ASB. Below, we provided a more detailed explanation of the mechanisms behind both the negative and positive human impacts on groundwater storage.
In 2010, the irrigated area in the ASB reached 8.2 million km2, and 92 km3 of river water was withdrawn, representing approximately 84% of the region’s annual water volume [45]. As discussed, both cropland and urbanized areas in the region expanded steadily from 1995 to 2015. According to ECA CCI data, cropland area increased by 0.12% between 2003 and 2015, reaching 256.87 × 103 km2 by 2015 [16,17]. During the same period, urban areas experienced a dramatic increase of 55.6%, totaling 6.2 × 103 km2 by 2015 [16,17]. As previously noted, the human-induced TWS trend was negative during the earlier period (January 2005 to February 2009) and became positive in the later period (January 2013 to June 2016). These shifts in human-induced water pressure aligned with observed land-use changes. Specifically, cropland area increased by 0.28% in the earlier period, but decreased by 0.09% in the latter period. While urban expansion occurred in both periods, its growth rate slowed significantly in the later period, decreasing by 12.46% compared to the earlier interval [16,17].
Another major contributor to water loss in the region is the outdated irrigation infrastructure. Much of the irrigation equipment is obsolete and has not been replaced or modernized in a timely manner [47]. In addition, the lack of maintenance and repair of both large and small water supply canals has led to inefficient water distribution. As a result, significant volumes of river water have been lost through evaporation, particularly during the summer months [46]. As noted in the previous section, the failure to maintain and service drainage systems has also contributed to land degradation in the region [16].
In recent years, the World Bank, along with countries within the Aral Sea Basin, has initiated several projects aimed at mitigating the Aral Sea crisis [45,48]. These efforts have begun to yield measurable results. For instance, between 2010 and 2014, the surface area of the Small Aral Sea increased modestly from 3.41 km2 to 3.43 km2 [1,4,44,49,50,51].
Drought conditions in the region were assessed using the Terrestrial Drought Severity Index (TDSI), derived from GRACE data, and the Standardized Precipitation Index (SPI), based on ERA-Interim precipitation data. The annual drought analysis indicated that 2005 and 2009 were particularly dry years, with the TDSI recording its lowest values of −2.93 and −2.51, respectively. SPI data further confirmed that drought conditions had been prevalent in recent years. The post-2009 period, however, marked a relative shift to wetter conditions, especially in 2012 when the TDSI reached its peak positive value.
To better understand these trends, we compared changes in the TDSI with river water volume data from the Tuyemuyun hydrological station, located in the lower reaches of the Amu Darya River [21]. Between 2003 and 2009, river flow decreased by 17.8%, while it increased by 16.7% during 2010–2015. The average flow between 2003 and 2015 was 719.59 km2, with the lowest recorded in 2008 at 343.1 km2 [52].
Previous studies on drought variability in Central Asia [53,54] reported drought occurrences across various timeframes using indices such as SPI, SPEI, and PDSI, often derived from MODIS data [53,55]. In contrast, our analysis is the first to employ a GRACE-derived TDSI to monitor drought processes in the Central Asian region. We also found that groundwater storage increased in central ASB, likely due to higher infiltration from surface water accumulation in reservoirs and losses from unlined irrigation canals. While storing surface water in aquifers or reservoirs can buffer against climatic variability and mitigate evaporation losses, it does not address anthropogenic stress on water systems. Such storage does not contribute to the direct replenishment of the Aral Sea.
Moreover, mismanagement of groundwater and reservoirs can jeopardize agricultural sustainability and exacerbate socio-political tensions. Therefore, it is essential that countries within the ASB region collaborate to develop optimal and equitable water-sharing solutions. Modernizing irrigation systems would substantially reduce water losses and create the foundation for more sustainable water resource management.
Given that the regional economy heavily relies on cotton and grain production, an accurate understanding of available water resources in this arid environment is vital. Water losses in areas adjacent to the Aral Sea continue to strain regional infrastructure, and a growing population may intensify these challenges. Ongoing monitoring of climate change impacts and water dynamics in the ASB is critical. Our findings affirm that the GRACE satellite mission is an effective tool for this purpose. Encouragingly, our analysis shows a slight but measurable decline in anthropogenic water stress in recent years, coinciding with the onset of a wetter period. Continued improvements in water management practices may sustain this trend in the future.

6. Conclusions

In this study, we quantitatively analyzed the contributions of climate anthropogenic factors to changes in terrestrial water storage (TWS) using the data-driven reconstruction method. GRACE satellite data and ERA precipitation data were used in the analysis. Following the method described in Humphrey [28] we reconstructed the TWS variations across the ASB. The proposed optimization in the model identification process presents potential benefits for future research on hydrological systems affected by both natural variability and human interventions.
Our analysis demonstrates that, in certain years, water consumption exceeded the region’s natural replenishment capacity. Given the fragile ecological situation in the ASB region, such levels of anthropogenic activity pose serious risks to water sustainability. The inter-annual variations in climate-driven and human-induced TWS changes during the 2002–2017 period indicate that the anthropogenic impacts were the dominant factor in water storage decline. Drought events in the region were found to result from both anthropogenic stress and climatic variability, specifically, increased water demand and reduced precipitation. In this study, we also monitored drought conditions using two parameters based on GRACE TWS data. Periods in which the human-induced TWS trend showed a decline corresponded with increased values of the TSDI, indicating drought conditions. Based on these results, we infer that anthropogenic drought occurred in the region from 2005 to 2009, while conditions improved after 2013, with the region experiencing a transition toward wetter conditions. After the appropriate calibration of the groundwater level observation data in the control wells according to the GRACE time series, our method can be applied for operational monitoring of water overuse. Such an approach could help in preventing further disruption to the hydrological regime sustaining the Aral Sea and may support efforts aimed at stabilizing and potentially restoring its water levels.

Author Contributions

Conceptualization, N.M., T.B. and B.P.; methodology, T.B.; software, N.M.; validation, N.M., S.N. and A.T.; formal analysis, T.B. and S.S.; investigation, B.A. and K.G.; resources, A.E., A.-A.M. and T.A.; data curation, S.N. and K.P.; writing—original draft preparation, N.M. and T.B.; writing—review and editing, B.P.; visualization, T.B.; supervision, B.P.; project administration, T.B. and B.P.; funding acquisition, B.P. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

Data used in this manuscript can be requested for research purposes from the co-author, timur.berdimbetov@tea.ac.cn, and have also been uploaded to a publicly accessible GitHub repository, which can be freely accessed at https://github.com/nikita-moor/aral-sea-2025 (accessed on 27 August 2025).

Acknowledgments

We are grateful to the Nukus State Technical University, Utah State University, and the University of Alabama in Huntsville for providing the opportunity to work on this study.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Berdimbetov, T.T.; Ma, Z.-G.; Liang, C.; Ilyas, S. Impact of Climate Factors and Human Activities on Water Resources in the Aral Sea Basin. Hydrology 2020, 7, 30. [Google Scholar] [CrossRef]
  2. Cretaux, J.-F.; Kostianoy, A.; Bergé-Nguyen, M.; Kouraev, A. Present-day water balance of the Aral Sea seen from satellite. In Remote Sensing of the Asian Seas; Springer: Charm, Switzerland, 2019; pp. 523–539. [Google Scholar]
  3. Micklin, P. The Aral Sea Disaster. Annu. Rev. Earth Planet. Sci. 2007, 35, 47–72. [Google Scholar] [CrossRef]
  4. Lioubimtseva, E. Impact of Climate Change on the Aral Sea and Its Basin. In The Aral Sea; Springer: Berlin/Heidelberg, Germany, 2014; pp. 405–427. [Google Scholar]
  5. Li, Q.; Li, X.; Ran, Y.; Feng, M.; Nian, Y.; Tan, M.; Chen, X. Investigate the relationships between the Aral Sea shrinkage and the expansion of cropland and reservoir in its drainage basins between 2000 and 2020. Int. J. Digit. Earth 2020, 14, 661–677. [Google Scholar] [CrossRef]
  6. Klein, I.; Gessner, U.; Kuenzer, C. Regional land cover mapping and change detection in Central Asia using MODIS time-series. Appl. Geogr. 2012, 35, 219–234. [Google Scholar] [CrossRef]
  7. Huang, Y.; Salama, M.S.; Krol, M.S.; Su, Z.; Hoekstra, A.Y.; Zeng, Y.; Zhou, Y. Estimation of human-induced changes in terrestrial water storage through integration of GRACE satellite detection and hydrological modeling: A case study of the Yangtze River basin. Water Resour. Res. 2015, 51, 8494–8516. [Google Scholar] [CrossRef]
  8. Felfelani, F.; Wada, Y.; Longuevergne, L.; Pokhrel, Y.N. Natural and human-induced terrestrial water storage change: A global analysis using hydrological models and GRACE. J. Hydrol. 2017, 553, 105–118. [Google Scholar] [CrossRef]
  9. Rodell, M.; Famiglietti, J.S.; Wiese, D.N.; Reager, J.T.; Beaudoing, H.K.; Landerer, F.W.; Lo, M.H. Emerging trends in global freshwater availability. Nature 2018, 557, 651–659. [Google Scholar] [CrossRef] [PubMed]
  10. Yi, S.; Sun, W.; Feng, W.; Chen, J. Anthropogenic and climate-driven water depletion in Asia. Geophys. Res. Lett. 2016, 43, 9061–9069. [Google Scholar] [CrossRef]
  11. Zhong, Y.; Feng, W.; Humphrey, V.; Zhong, M. Human-Induced and Climate-Driven Contributions to Water Storage Variations in the Haihe River Basin, China. Remote Sens. 2019, 11, 3050. [Google Scholar] [CrossRef]
  12. Landerer, F.; Swenson, S. Accuracy of scaled GRACE terrestrial water storage estimates. Water Resour. Res. 2012, 48, 4531. [Google Scholar] [CrossRef]
  13. Scanlon, B.R.; Zhang, Z.; Save, H.; Sun, A.Y.; Muller Schmied, H.; van Beek, L.P.H.; Wiese, D.N.; Wada, Y.; Long, D.; Reedy, R.C.; et al. Global models underestimate large decadal declining and rising water storage trends relative to GRACE satellite data. Proc. Natl. Acad. Sci. USA 2018, 115, E1080–E1089. [Google Scholar] [CrossRef]
  14. Cao, Y.; Nan, Z.; Cheng, G. GRACE Gravity Satellite Observations of Terrestrial Water Storage Changes for Drought Characterization in the Arid Land of Northwestern China. Remote Sens. 2015, 7, 1021–1047. [Google Scholar] [CrossRef]
  15. Nemati, A.; Ghoreishi Najafabadi, S.H.; Joodaki, G.; Mousavi Nadoushani, S.S. Monitoring and Analysis of Drought Using Gravity Recovery and Climate Experiment (GRACE). Hydrology 2019, 6, 75. [Google Scholar] [CrossRef]
  16. Berdimbetov, T.; Ma, Z.-G.; Shelton, S.; Ilyas, S.; Nietullaeva, S. Identifying Land Degradation and its Driving Factors in the Aral Sea Basin From 1982 to 2015. Front. Earth Sci. 2021, 9, 690000. [Google Scholar] [CrossRef]
  17. Berdimbetov, T.; Shelton, S.; Pushpawela, B.; Rathnayake, U.; Koshim, A.G.; Yegizbayeva, A.; Nietullaeva, S. Use of intensity analysis and transfer matrix to characterize land conversion in the Aral Sea Basin under changing climate. Model. Earth Syst. Environ. 2024, 10, 4717–4729. [Google Scholar] [CrossRef]
  18. Jiang, L.; Bao, A.; Jiapaer, G.; Guo, H.; Zheng, G.; Gafforov, K.; Kurban, A.; De Maeyer, P. Monitoring land sensitivity to desertification in Central Asia: Convergence or divergence? Sci. Total Environ. 2019, 658, 669–683. [Google Scholar] [CrossRef]
  19. Jiang, L.; Jiapaer, G.; Bao, A.; Li, Y.; Guo, H.; Zheng, G.; Chen, T.; De Maeyer, P. Assessing land degradation and quantifying its drivers in the Amudarya River delta. Ecol. Indic. 2019, 107, 105595. [Google Scholar] [CrossRef]
  20. Timur, B. Spatio-Temporal Variations of Climate Variables and Extreme Indices over the Aral Sea Basin during 1960–2017. Trends Sci. 2023, 20, 5664. [Google Scholar] [CrossRef]
  21. Berdimbetov, T.; Pushpawela, B.; Murzintcev, N.; Nietullaeva, S.; Gafforov, K.; Tureniyazova, A.; Madetov, D. Unraveling the Intricate Links between the Dwindling Aral Sea and Climate Variability during 2002–2017. Climate 2024, 12, 105. [Google Scholar] [CrossRef]
  22. Guo, H.; Bao, A.; Chen, T.; Zheng, G.; Wang, Y.; Jiang, L.; De Maeyer, P. Assessment of CMIP6 in simulating precipitation over arid Central Asia. Atmos. Res. 2021, 252, 105451. [Google Scholar] [CrossRef]
  23. Swenson, S.; Wahr, J. Post-processing removal of correlated errors in GRACE data. Geophys. Res. Lett. 2006, 33, L08402. [Google Scholar] [CrossRef]
  24. Swenson, S.; Wahr, J.; Milly, P.C.D. Estimated accuracies of regional water storage variations inferred from the Gravity Recovery and Climate Experiment (GRACE). Water Resour. Res. 2003, 39, 1223. [Google Scholar] [CrossRef]
  25. Alghafli, K.; Ali, A.M.; Shi, X.; Sloan, W.; Obeid, A.A.A.; Shamsudduha, M. Evaluation of runoff estimation from GRACE coupled with different meteorological gridded products over the Upper Blue Nile Basin. J. Hydrol. Reg. Stud. 2023, 50, 101545. [Google Scholar] [CrossRef]
  26. Humphrey, V.; Gudmundsson, L.; Seneviratne, S.I. A global reconstruction of climate-driven subdecadal water storage variability. Geophys. Res. Lett. 2017, 44, 2300–2309. [Google Scholar] [CrossRef]
  27. Ahmed, J.S.; Buizza, R.; Dell’Acqua, M.; Demissie, T.; Pè, M.E. Evaluation of ERA5 and CHIRPS rainfall estimates against observations across Ethiopia. Meteorol. Atmos. Phys. 2024, 136, 17. [Google Scholar] [CrossRef]
  28. Humphrey, V.; Gudmundsson, L.; Seneviratne, S.I. Assessing Global Water Storage Variability from GRACE: Trends, Seasonal Cycle, Subseasonal Anomalies and Extremes. Surv. Geophys. 2016, 37, 357–395. [Google Scholar] [CrossRef]
  29. Chib, S. Chapter 57—Markov Chain Monte Carlo Methods: Computation and Inference. In Handbook of Econometrics; Heckman, J.J., Leamer, E., Eds.; Elsevier: Amsterdam, The Netherlands, 2001; Volume 5, pp. 3569–3649. [Google Scholar]
  30. Harris, I.; Osborn, T.J.; Jones, P.; Lister, D. Version 4 of the CRU TS monthly high-resolution gridded multivariate climate dataset. Sci. Data 2020, 7, 109. [Google Scholar] [CrossRef]
  31. Fan, Y.; Dool, H. A global monthly land surface air temperature analysis for 1948–present. J. Geophys. Res. 2008, 113, D01103. [Google Scholar] [CrossRef]
  32. Berdimbetov, T.; Ilyas, S.; Ma, Z.; Bilal, M.; Nietullaeva, S. Climatic Change and Human Activities Link to Vegetation Dynamics in the Aral Sea Basin Using NDVI. Earth Syst. Environ. 2021, 5, 303–318. [Google Scholar] [CrossRef]
  33. Eckner, A. A Note on Trend and Seasonality Estimation for Unevenly Spaced Time Series. Available online: http://www.eckner.com/papers/ (accessed on 21 August 2025).
  34. Robert, B.C.; William, S.C.; Jean, E.M.; Irma, T. STL: A seasonal-trend decomposition procedure based on loess. J. Official Stat. 1990, 6, 3–73. [Google Scholar]
  35. Brent, R. Algorithms for Minimization Without Derivatives; Prentice Hall: Englewood Cliffs, FL, USA, 2002; Volume 19. [Google Scholar] [CrossRef]
  36. McKee, T.B.; Doesken, N.J.; Kleist, J. The relationship of drought frequency and duration of time scales. In Proceedings of the Eighth Conference on Applied Climatology, Anaheim, CA, USA, 17–22 January 1993; pp. 179–186. [Google Scholar]
  37. Cheng, Y.-S.; Lu, J.-R.; Yeh, H.-F. Multi-Index Drought Analysis in Choushui River Alluvial Fan, Taiwan. Environments 2024, 11, 233. [Google Scholar] [CrossRef]
  38. Agboma, C.O.; Yirdaw, S.Z.; Snelgrove, K.R. Intercomparison of the total storage deficit index (TSDI) over two Canadian Prairie catchments. J. Hydrol. 2009, 374, 351–359. [Google Scholar] [CrossRef]
  39. Yirdaw, S.Z.; Snelgrove, K.R.; Agboma, C.O. GRACE satellite observations of terrestrial moisture changes for drought characterization in the Canadian Prairie. J. Hydrol. 2008, 356, 84–92. [Google Scholar] [CrossRef]
  40. Palmer, W. Meteorological Drought; Research Paper, No. 45; Weather Bureau: Washington, DC, USA, 1965; pp. 1–65. [Google Scholar]
  41. Diodato, N.; De Guenni, L.B.; Garcia, M.; Bellocchi, G. Decadal Oscillation in the Predictability of Palmer Drought Severity Index in California. Climate 2019, 7, 6. [Google Scholar] [CrossRef]
  42. Narasimhan, B.; Srinivasan, R. Development and evaluation of Soil Moisture Deficit Index (SMDI) and Evapotranspiration Deficit Index (ETDI) for agricultural drought monitoring. Agric. For. Meteorol. 2005, 133, 69–88. [Google Scholar] [CrossRef]
  43. Yegizbayeva, A.; Ilyas, S.; Berdimbetov, T. Drought Characterisation of Syrdarya River Basin in Central Asia Using Reconnaissance Drought Index. In Proceedings of the IGARSS 2022–2022 IEEE International Geoscience and Remote Sensing Symposium, Kuala Lumpur, Malaysia, 17–22 July 2022; pp. 6356–6359. [Google Scholar]
  44. Foldvary, L.; Tureniyazova, A.; Berdimbetov, T.; Abdurahmanov, I. Relationship between climate and land cover change in Aral Sea Basin. E3S Web Conf. 2023, 386, 06003. [Google Scholar] [CrossRef]
  45. Philip, M.; Aladin, N.V.; Plotnikov, I. The Aral Sea. The Devastation and Partial Rehabilitation of a Great Lake; Micklin, P., Ed.; Springer: Berlin/Heidelberg, Germany, 2014; p. 453. [Google Scholar]
  46. Becker, R.; Zmijewski, K. Estimating the Effects of Anthropogenic Modification on Water Balance in the Aral Sea Watershed Using GRACE: 2003–12. Earth Interact. 2014, 18, 1–16. [Google Scholar] [CrossRef]
  47. Robinson, S. Land Degradation in Central Asia: Evidence, Perception and Policy. In The End of Desertification? Springer Earth System Sciences; Springer: Berlin/Heidelberg, Germany, 2016; pp. 451–490. [Google Scholar]
  48. Shabad, T.; Bond, A.R.; Bradshaw, M.J.; Cushman, J.F.; Harris, C.D.; Hausladen, G.J.; Lewis, R.A.; Lydolph, P.E.; Micklin, P.P.; North, R.N.; et al. Panel on the Soviet Union in the Year 2000. Sover. Geogr. 2013, 28, 388–433. [Google Scholar] [CrossRef]
  49. Liu, H.; Chen, Y.; Ye, Z.; Li, Y.; Zhang, Q. Recent Lake Area Changes in Central Asia. Sci. Rep. 2019, 9, 16277. [Google Scholar] [CrossRef]
  50. Lioubimtseva, E.; Cole, R. Uncertainties of Climate Change in Arid Environments of Central Asia. Rev. Fish. Sci. Aquac. 2007, 14, 29–49. [Google Scholar] [CrossRef]
  51. Lioubimtseva, E.; Cole, R.; Adams, J.M.; Kapustin, G. Impacts of climate and land-cover changes in arid lands of Central Asia. J. Arid Environ. 2005, 62, 285–308. [Google Scholar] [CrossRef]
  52. CAWater-Info. Portal of Knowledge for Water and Environmental Issues in Central Asia. Available online: http://www.cawater-info.net/ (accessed on 21 August 2025).
  53. Aitekeyeva, N.; Li, X.; Guo, H.; Wu, W.; Shirazi, Z.; Ilyas, S.; Yegizbayeva, A.; Hategekimana, Y. Drought Risk Assessment in Cultivated Areas of Central Asia Using MODIS Time-Series Data. Water 2020, 12, 1738. [Google Scholar] [CrossRef]
  54. Hu, Z.; Chen, X.; Chen, D.; Li, J.; Wang, S.; Zhou, Q.; Yin, G.; Guo, M. “Dry gets drier, wet gets wetter”: A case study over the arid regions of central Asia. Int. J. Climatol. 2019, 39, 1072–1091. [Google Scholar] [CrossRef]
  55. Guo, H.; Bao, A.; Liu, T.; Jiapaer, G.; Ndayisaba, F.; Jiang, L.; Kurban, A.; De Maeyer, P. Spatial and temporal characteristics of droughts in Central Asia during 1966–2015. Sci. Total Environ. 2018, 624, 1523–1538. [Google Scholar] [CrossRef] [PubMed]
Figure 1. The Aral Sea basin and the research area. The thick black line delineates the hydrological basin, while blue lines indicate the Amu Darya and Syr Darya rivers, as well as the remnants of the Aral Sea. The red line outlines the study area, which is subdivided into three regions: upper (UASB), middle (MASB), and lower (LASB). A fine mesh corresponding to the 1° GRACE grid is overlaid for spatial reference.
Figure 1. The Aral Sea basin and the research area. The thick black line delineates the hydrological basin, while blue lines indicate the Amu Darya and Syr Darya rivers, as well as the remnants of the Aral Sea. The red line outlines the study area, which is subdivided into three regions: upper (UASB), middle (MASB), and lower (LASB). A fine mesh corresponding to the 1° GRACE grid is overlaid for spatial reference.
Climate 13 00182 g001
Figure 2. General workflow for hydrological model identification by (Humphrey et al., 2017 [26]).
Figure 2. General workflow for hydrological model identification by (Humphrey et al., 2017 [26]).
Climate 13 00182 g002
Figure 3. Reconstructed, climate-driven total water storage (TWS, in cm) time series are shown for the period from 1989 to 2002 (black line) and from 2002 to 2017 (red line). The dash red line represents the trend of climate-driven TWS between May-2002 and December-2017. The mean value of reconstructed, climate-driven TWS is derived from the whole study area.
Figure 3. Reconstructed, climate-driven total water storage (TWS, in cm) time series are shown for the period from 1989 to 2002 (black line) and from 2002 to 2017 (red line). The dash red line represents the trend of climate-driven TWS between May-2002 and December-2017. The mean value of reconstructed, climate-driven TWS is derived from the whole study area.
Climate 13 00182 g003
Figure 4. Time series of GRACE total water storage (TWS, magenta bar, unit: cm), climate-driven TWS (blue line, unit: cm), and human-induced TWS (black line, unit: cm) from 2002 to 2017 for each region: (a) UASB, (b) MASB, and (c) LASB. The red dashed line represents the trend of climate-driven TWS, while the violet dashed line indicates the trend in human-induced TWS.
Figure 4. Time series of GRACE total water storage (TWS, magenta bar, unit: cm), climate-driven TWS (blue line, unit: cm), and human-induced TWS (black line, unit: cm) from 2002 to 2017 for each region: (a) UASB, (b) MASB, and (c) LASB. The red dashed line represents the trend of climate-driven TWS, while the violet dashed line indicates the trend in human-induced TWS.
Climate 13 00182 g004
Figure 5. Time evolution of total storage deficit (TSD, %) for the period 2002–2017 from the whole study period. Two dashed blue lines represent the extreme dry period.
Figure 5. Time evolution of total storage deficit (TSD, %) for the period 2002–2017 from the whole study period. Two dashed blue lines represent the extreme dry period.
Climate 13 00182 g005
Figure 6. Cumulative graph of Terrestrial Storage Deficit (TSD) and division of this graph area into four equal parts.
Figure 6. Cumulative graph of Terrestrial Storage Deficit (TSD) and division of this graph area into four equal parts.
Climate 13 00182 g006
Figure 7. Drought changes in the (a) UASB, (b) MASB and (c) LASB between 2002 and 2017.
Figure 7. Drought changes in the (a) UASB, (b) MASB and (c) LASB between 2002 and 2017.
Climate 13 00182 g007
Table 1. Classification of the drought index by SPI.
Table 1. Classification of the drought index by SPI.
ClassCoefficient C
Wet0 or more
Near normal0~−0.5
Mild drought−0.5~−1
Moderate drought−1~−1.5
Severe drought−1.5~−2
Extreme drought−2~less
Table 2. Classification of drought severity based on the dryness intensity coefficient C, adapted from Palmer.
Table 2. Classification of drought severity based on the dryness intensity coefficient C, adapted from Palmer.
Class Coefficient   C
Mild drought−1
Moderate drought−2
Severe drought−3
Extreme drought−4
Table 3. Inter-annual variations in total water storage anomalies (TWSAs; cm) across the upper (UASB), middle (MASB), and lower (LASB) sub-regions of the Aral Sea Basin from 2002 to 2017, including both climate-driven and human-induced components. Negative TWS values (−) indicate dry years.
Table 3. Inter-annual variations in total water storage anomalies (TWSAs; cm) across the upper (UASB), middle (MASB), and lower (LASB) sub-regions of the Aral Sea Basin from 2002 to 2017, including both climate-driven and human-induced components. Negative TWS values (−) indicate dry years.
YearUASBMASBLASB
GRACE
TWSA
Climate TWSAHuman TWSAGRACE
TWSA
Climate TWSAHuman TWSAGRACE
TWSA
Climate TWSAHuman TWSA
2002−3.590.16−3.75−2.230.10−2.33−3.990.52−4.51
20030.620.78−0.16−0.250.69−0.94−2.140.78−2.92
20040.950.670.270.450.68−0.23−0.120.54−0.65
20055.300.654.662.920.971.952.720.152.58
20061.36−0.431.790.40−0.080.481.12−0.161.28
2007−1.86−1.00−0.86−0.14−0.190.051.37−0.191.56
2008−6.29−2.04−4.25−2.12−0.88−1.25−0.60−0.900.30
2009−2.27−0.62−1.64−0.54−0.01−0.53−0.29−0.380.09
20104.151.362.791.210.201.021.10−0.111.20
20110.75−0.281.03−0.73−0.62−0.11−0.92−0.68−0.24
20122.59−0.032.620.82−0.211.02−0.32−0.540.22
2013−1.62−0.90−0.72−0.48−0.29−0.19−0.32−0.500.18
2014−2.88−1.38−1.50−0.94−0.75−0.19−0.59−0.600.01
20150.08−0.540.620.61−0.180.79−0.13−0.04−0.09
2016−0.010.41−0.420.21−0.130.341.121.040.08
20172.011.110.900.730.030.700.830.400.43
Table 4. Inter-annual variations of terrestrial water fluxes (TWF; difference between TWS of consecutive two years) from GRACE, climate-driven TWF, and human-induced TWF for the period from 2002 to 2017.
Table 4. Inter-annual variations of terrestrial water fluxes (TWF; difference between TWS of consecutive two years) from GRACE, climate-driven TWF, and human-induced TWF for the period from 2002 to 2017.
YearUASBMASBLASB
GRACE
TWF
Climate TWFHuman TWFGRACE
TWF
Climate TWFHuman TWFGRACE
TWF
Climate TWFHuman TWF
2002−3.590.16−3.75−2.230.10−2.33−3.990.52−4.51
20034.210.613.591.980.591.391.850.261.59
20040.33−0.100.430.71−0.010.722.03−0.242.26
20054.36−0.034.392.470.292.182.84−0.393.23
2006−3.94−1.07−2.87−2.53−1.05−1.47−1.61−0.31−1.30
2007−3.22−0.57−2.65−0.53−0.10−0.430.26−0.030.28
2008−4.43−1.04−3.41−1.98−0.69−1.29−1.97−0.72−1.25
20094.021.412.611.580.860.720.320.52−0.21
20106.421.994.431.750.211.551.380.271.11
2011−3.40−1.65−1.76−1.94−0.82−1.12−2.02−0.57−1.45
20121.840.251.591.550.421.130.600.140.46
2013−4.21−0.87−3.34−1.30−0.08−1.220.000.04−0.04
2014−1.26−0.48−0.78−0.45−0.460.00−0.27−0.09−0.18
20152.960.842.121.540.570.970.460.56−0.10
2016−0.100.95−1.05−0.400.04−0.441.251.080.17
20172.030.701.320.520.160.36−0.29−0.640.35
Table 5. The classification of the area condition by using TSDI.
Table 5. The classification of the area condition by using TSDI.
ClassCoefficient
Wet1~more
Near normal−1~1
Mild drought−2~−1
Moderate drought−3~−2
Severe drought−4~−3
Extreme drought−4~less
Table 6. Duration and severity of droughts in the study area.
Table 6. Duration and severity of droughts in the study area.
RegionTimeDrought Duration in MonthsMin TDSIMean TDSIMin SPIMean SPISlope of the Ccumulative TSDI
UASB2002-05 to 2003-0212−3.16−1.71−1.630.11−1.95
2007-05 to 2009-0626−4.78−2.49−2.54−0.17−2.85
2011-04 to 2011-085−1.22−0.58−2.41−0.75−0.67
2013-02 to 2014-1021−3.29−1.49−1.460.07−1.57
MASB2002-05 to 2003-0412−4.34−1.96−1.990.14−2.33
2006-06 to 2006-105−0.68−0.43−1.26−0.13−0.53
2007-09 to 2009-0622−3.48−1.85−1.01−0.02−2.22
2011-02 to 2011-109−2.22−1.28−2.23−0.41−1.55
2013-01 to 2014-0911−2.27−1.56−1.68−0.11−1.58
LASB2002-05 to 2003-1119−5.16−2.99−1.92−0.16−3.26
2004-05 to 2004-084−1.54−0.51−1.510.76−0.52
2006-08 to 2006-093−0.51−0.46−1.03−0.87−0.01
2008-09 to 2009-0812−1.31−0.92−1.71−0.13−0.98
2011-02 to 2011-1110−1.86−1.14−1.84−0.31−1.31
2013-02 to 2013-087−1.22−0.64−1.41−0.19−0.61
2014-04 to 2014-096−2.12−1.48−1.240.23−0.25
2015-02 to 2015-076−1.62−0.96−1.59−0.13−1.01
Table 7. Mean monthly climate-driven and human-induced total water storage (TWS, cm) change between 2002 and 2017.
Table 7. Mean monthly climate-driven and human-induced total water storage (TWS, cm) change between 2002 and 2017.
MonthUASBMASBLASB
Climate TWSHuman TWSClimate TWSHuman TWSClimate TWSHuman TWS
January0.260.210.230.320.180.25
February0.210.400.220.190.110.18
March0.170.210.130.12−0.100.12
April0.230.110.160.17−0.110.21
May−0.07−0.04−0.40−0.050.14−0.49
June−0.11−0.13−0.70−0.140.13−0.35
July−0.20−0.18−0.72−0.18−0.32−0.27
August−0.220.07−0.71−0.18−0.38−0.34
September−0.210.46−0.660.21−0.360.24
October−0.200.86−0.240.39−0.480.28
November0.190.500.160.18−0.150.13
December0.120.890.250.440.250.55
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Murzintcev, N.; Nietullaeva, S.; Berdimbetov, T.; Pushpawela, B.; Tureniyazova, A.; Shelton, S.; Aytmuratov, B.; Gafforov, K.; Parakhatov, K.; Erdashov, A.; et al. Reconstruction of Water Storage Variability in the Aral Sea Region. Climate 2025, 13, 182. https://doi.org/10.3390/cli13090182

AMA Style

Murzintcev N, Nietullaeva S, Berdimbetov T, Pushpawela B, Tureniyazova A, Shelton S, Aytmuratov B, Gafforov K, Parakhatov K, Erdashov A, et al. Reconstruction of Water Storage Variability in the Aral Sea Region. Climate. 2025; 13(9):182. https://doi.org/10.3390/cli13090182

Chicago/Turabian Style

Murzintcev, Nikita, Sahibjamal Nietullaeva, Timur Berdimbetov, Buddhi Pushpawela, Asiya Tureniyazova, Sherly Shelton, Bakbergen Aytmuratov, Khusen Gafforov, Kanat Parakhatov, Alimjan Erdashov, and et al. 2025. "Reconstruction of Water Storage Variability in the Aral Sea Region" Climate 13, no. 9: 182. https://doi.org/10.3390/cli13090182

APA Style

Murzintcev, N., Nietullaeva, S., Berdimbetov, T., Pushpawela, B., Tureniyazova, A., Shelton, S., Aytmuratov, B., Gafforov, K., Parakhatov, K., Erdashov, A., Makhamatdinov, A.-A., & Allamuratov, T. (2025). Reconstruction of Water Storage Variability in the Aral Sea Region. Climate, 13(9), 182. https://doi.org/10.3390/cli13090182

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop