Evaluation of the impact of anthropogenic storage on the hydrological drought propagation in two contrasting semi-arid river basins

The IPCC reports (Seneviratne et al. 2021) indicate the high occurrences of severe weather extremes including drought events around the world. In the recent and early years of the twenty-first century, global society has witnessed several instances of severe floods, droughts, and heatwaves (Wanders et al. 2015; Mishra et al. 2019; Spinoni et al. 2019). Drought is a recurring, gradual natural calamity that has a substantial impact on environmental deterioration, agricultural productivity, and water resource management (Udmale et al. 2015). The characterization of drought is considered to be complex in nature as it takes different forms like meteorological, agricultural, hydrological, and socio-economic droughts Mishra & Singh (2010). Though the research community has made great progress in drought monitoring, characterization, and prediction research one major theme where there is more scope for research is in drought propagation from one form to another (Bhardwaj et al. 2020; Huang et al. 2021). Knowledge of drought propagation is significant as it can give an early warning of transitions that can be influenced by anthropogenic activities (e.g., dam and reservoir regulations) along with climatic factors (Huang et al. 2017; Xu et al. 2019; Zhang et al. 2021).

Over recent decades, various drought monitoring indices have been characterized by researchers for meteorological drought (MD) assessment, e.g., Aridity Index (AI; Martonne 1925), Rainfall Anomaly Index (RAI; van Rooy (1965), Palmer Drought Severity Index (PDSI; Palmer (1965), Deciles Gibbs & Maher (1967), Standardized Precipitation Index (SPI; McKee et al. (1993), China Z Index (CZI; Edwards & McKee 1997), and many more. Hydrological drought (HD) is related more to surface and subsurface water scarcity (Mishra & Singh 2010). Dracup et al. (1980) made a clear distinction between low flows and streamflow drought, stating that low flow has a daily or weekly time-step, whereas drought events are studied on monthly or seasonal scale. In the past, many researchers attempted to define HD by developing indices such as the Standardized Reservoir Supply Index (SRSI) Gusyev et al. (2015), the Standardized Streamflow Index (SSI) Modarres (2007), the Standardized Water Level Index (SWLI) Bhuiyan (2004), the Streamflow Drought Index (SDI) (Nalbantis & Tsakiris (2009)), and the Surface Water Supply Index (SWSI) Shafer & Dezman (1982). Among these, SPI and SSI have been extensively utilized for MD and HD characterizations, respectively, due to their various strengths like: (i) simplicity, (ii) flexibility, and (iii) suitability at different time and space scales even in data-deficient situations (Zhao et al. 2014; Jehanzaib et al. 2020).

Many researchers have studied the linkage between MD and HD by investigating the propagation from one drought type to another (Huang et al. 2017; Bhardwaj et al. 2020; Li et al. 2020; Ding et al. 2021). MD triggers HD with some time-lag, and this period is called drought propagation time (DPT) (Apurv et al. 2017). Most of the studies on drought propagation are focused on the propagation time between the subtypes of droughts (Van Loon & Laaha 2015; Ding et al. 2021; Shi et al. 2022a, 2022b). Some researchers (Fang et al. 2020; Li et al. 2020) have used terms like ‘propagation threshold’, ‘lag time’, and ‘response time’ interchangeably to explain the linear relation of drought propagation. Previous studies in the literature have used different approaches for the quantification of the DPT by building the relationship between MD and HD. Other approaches adopted are: (i) correlation coefficient (CC)-based analysis (Xing et al. 2021), (ii) linear relationship (Edossa et al. 2010), (iii) non-linear relationship (Wu et al. 2021), (iv) propagation probability (Shin et al. 2018; Sattar et al. 2019), and (v) causal analysis (Shi et al. 2022a). Drought propagation from MD to HD varies with different climatic conditions and seasons; hence, the linkage between drought types depends on the climate of the region (Ding et al. 2021). Wu et al. (2018) studied the effect of reservoirs on drought occurrence downstream of the Jinjiang River basin of China through the establishment of a non-linear relationship between SPI and SSI. Wang et al. (2019) used the SWAT model in the Luanhe River basin for studying the influence of the reservoir on drought propagation. Another study by Xing et al. (2021) investigated the Pearson correlation relation between MD and HD to assess the effect of the reservoir on the Daling basin of China.

The Chirchik River basin, situated in Central Asia (Uzbekistan), and the Tapi River basin, located in South Asia (India), exhibit contrasting yet complex geomorphological characteristics. The Chirchik basin is marked by diverse rock compositions from different geological eras and ongoing tectonic activity, resulting in rugged terrains with mountain ranges, valleys, and plateaus. In contrast, the Tapi basin features a mix of igneous, metamorphic, and sedimentary rocks, with deep valleys and gorges carved by rivers, along with plateaus and plains shaped by erosion and weathering processes. Price et al. (2011) emphasized that the interplay between land use and geomorphic features of the river basin can significantly impact the development of severe drought conditions. Therefore, conducting a comparative study, as detailed in this research, could yield novel insights with meaningful implications for environmental flow management in both river basins. Apart from the differences in the geomorphology of river basins, in the Chirchik basin, snowmelt in spring and summer contributes significantly to river flow, whereas in the Tapi basin, rainfall from the monsoon season primarily drives flow formation. Exploring these variations is intriguing as it throws light on the varied impacts of drought propagation on different river- basin dynamics and hydrological patterns. Some research has stressed the importance of evaluating the diverse drivers and attributes of droughts, particularly in snow-dominated catchments and monsoonal climates, due to the distinct processes involved in the drought propagation (Van Loon 2015; Wang et al. 2024).

Many of the above-mentioned drought studies are specific to a river basin or region but there are a few comparative studies across river basins from the contrasting climatic conditions. The future projected drought characteristics for the Central Asian rivers are different from the South Asian rivers due to various aspects like: shifts in the peak of river discharge, snow accumulation shortages, and flow deficiency in summer months (Didovets et al. 2021). The climate in most Central Asian river basins is strongly continental arid to semi-arid with hot cloud-free summers. However, there are some river basins in South Asia with similar arid/semi-arid climatic conditions, where the HD responses are different because of the monsoonal effect. Understanding drought propagation across diverse agro-climatic settings, including South Asia and Central Asia, is important for several motivating factors such as: (i) differences in seasonality, climate aridity, and timing of the precipitation (Apurv et al. 2017), (ii) differences in regulated and natural water systems, and (iii) differences in the soil types and agricultural practices (Ding et al. 2021). Some studies have highlighted that drought formation is extremely complex and the physical factors driving the drought conditions are different in both South Asia and Central Asia (Zhang et al. 2021, 2022; Saha et al. 2023). Anthropogenic factors serve as the primary drivers of long-term changes in drought conditions across both South and Central Asian regions (Roodari et al. 2021; Shah et al. 2024). Particularly in downstream areas, these factors can influence the management and operation of reservoirs, contingent upon crop requirements and other regional water demands. Moreover, the practices employed in reservoir management can also affect the attributes of drought, leading to additional environmental and sustainability ramifications. Therefore, investigating these aspects through comparative studies holds significance. To the authors' knowledge, there have been no studies that have conducted a comparative analysis of drought propagation in South Asia and Central Asia. This motivated the selection of the Tapi basin with the Ukai reservoir in India and the Chirchik Basin with the Charvak reservoir in Uzbekistan as the case-study locations. Thus, the motivation of this study is two-fold. The first is to understand the differences in the propagation from MD to HD in the selected arid river basins of South Asia and Central Asia. Secondly, to understand how differently the drought propagation is altered under the influence of the reservoir for the selected basins.

This study is meant to quantify the correlation and propagation characteristics between MD and HD in two semi-arid river basins in India and Uzbekistan. The specific objectives of this study are:

• (i) to characterize MD and HD using the multiscale SPI and SSI, respectively, for both the selected basins;

• (ii) to study the propagation from MD to HD using the maximum value of Spearman rank correlation coefficient for the natural and reservoir-influenced parts of the selected river basins;

• (iii) to analyze the variation in the predictive skill of MD to identify the HD with and without the reservoir-influenced regions of the two river basins having different hydroclimatic settings.

This paper is organized as follows: Section 2 describes the study area, data, and methods for characterizing drought propagation from MD to HD. The results are presented in Section 3, followed by discussion and conclusions in Section 4.

The Chirchik River basin is located in the northwestern range of the Tian Shan Mountain system of Uzbekistan. The Chirchik River is formed by the confluence of the Chatkal and Pskem rivers in the Charvak reservoir. The total area of the basin is 14,240 km² and the total river length is 174 km. The Chirchik River originates in the mountains of Western Tian Shan, the height of which reaches marks in the basin close to 4,500 m above mean sea level. The Chirchik River is fed by snowmelt, precipitation, and glaciers. Atmospheric processes which are characteristic of Central Asia are of huge importance for climate formation in the Chirchik basin. The whole territory is influenced by the western transfer of air masses. A significant role is also played by the cyclones, moist western air-masses, and cold northern air-masses, causing precipitation and decrease in air temperature. Table 1 summarizes all the hydro meteorological characteristics of the Tapi and Chirchik basins.

The Tapi River basin in India largely flows on the plain region and is dominated by agricultural land use, which is almost 67% of the area of the basin; thus, spatiotemporal variation of drought in this region is vital for local government and policy managers (Ramkar & Yadav 2018). The downstream streamflow is very low in the Tapi basin during the non-monsoon period as it is difficult to maintain environmental flow downstream of the dam. Likewise, the climatic conditions across the Chirchik River basin of Uzbekistan vary spatially for different topographical features and are highly influenced by the Arctic's cold and the westerly warm and humid air-masses (Mamadjanova et al. 2018).

The main aim of this study is to compare the drought propagation results for the Tapi and Chirchik River basins of India and Uzbekistan, respectively; hence, the global monthly precipitation time series data is downloaded from the Climate Research Unit (CRU v4.04: https://crudata.uea.ac.uk/cru/data/hrg/) having a spatial resolution of 0.5° × 0.5°. Reservoir storage and streamflow data for the Tapi basin is acquired from the India-WRIS (https://indiawris.gov.in/wris/#/RiverMonitoring) for the stations (Sarangkheda and Ghala: upstream and downstream of the Ukai reservoir, respectively) having continuous streamflow records from 1980 to 2004. Missing streamflow records are imputed using the linear interpolation approach. Overlapping streamflow time-series data is collected from Uzhydromet for stations (Chatkal-Khudoydodsay and Pskem-Mulala) of the Chirchik River basin. The precipitation and streamflow data span the period from 1980 to 2004, which is the common timeframe for both basins based on the data availability.

The present study is focused on computing the DPT from MD to HD along with studying the influence of climate–catchment characteristics of the selected basins. We also studied the variable ability of MD (SPI) to capture the HD (SSI) for the Tapi and Chirchik basins having different rainfall seasonalities and terrestrial water storages. A detailed description of the research process, including a brief overview of the methodology followed, is described below.

SPI is calculated using monthly rainfall data at different aggregation timescales (1, 2, 3, 4, … ,12) following the procedure given by McKee et al. (1993). Because of the ease of use and flexibility of calculating on multiple timescales, SPI has become a popular index for drought modeling in the field of hydrology (Hayes et al. 2011). The strength of SPI is that it can be calculated on multiple timescales, which further can be used to abstract information on the different drought types such as agricultural and HD. Agricultural drought conditions are well represented by the precipitation anomalies on a shorter timescale such as 1–6 months. However, HD conditions are more related to SPI at the timescale of 6–24 months (Svoboda et al. 2012). As per McKee et al. (1993), MD has been classified into four dryness levels based on the SPI values (see Table 2).

Table 2

SPI/SSI ranges showing different drought severity classes

Sr. No. .	SPI/SSI range .	Drought category .
1	0 to −0.99	Mild drought
2	−1.0 to −1.49	Moderate drought
3	−1.50 to −1.99	Severe drought
4		Extreme drought

Sr. No. .	SPI/SSI range .	Drought category .
1	0 to −0.99	Mild drought
2	−1.0 to −1.49	Moderate drought
3	−1.50 to −1.99	Severe drought
4		Extreme drought

View Large

SPI and SSI time-series datasets are utilized further to study the effect of the basins' climatic and hydrological conditions on drought occurrence and propagation using the Spearman-rank-correlation-based lag-time analysis. We calculated the SPI for MD using monthly precipitation data for the timescales of 1, 2, 3, … , 12 months, whereas for HD (streamflow drought) we calculated the SSI using monthly streamflow data from 1980 to 2004. The cross-correlation based on Spearman rank correlation was done between one-month SSI (SSI-1) and lagged values of SPI starting from one to 12 months. SPI timescales having a maximum value of Spearman rank correlation coefficient with SSI-1 is treated as a DPT (Barker et al. 2016; Wu et al. 2018). The detailed workflow for the computation of the DPT using multiscale SPI and SSI-1 is shown in Figure 2. In the current work, DPT is computed for each month, starting from January to December, and the variation in the DPT values is compared for both the basins under consideration to see the effect of different hydroclimatic and topographical characteristics.

The hit score and FAR are calculated using Equations (1) and (2), respectively. A perfect predicting capability of the MD index corresponds to a greater hit-score value and a smaller FAR value.

Tapi is a monsoon-dominated basin, whereas Chirchik has the presence of snowfall and glacier melt. The Chirchik basin has two major periods, one is the accumulation period (October–March) and the other is the melting period (April–September). However, Tapi has a single runoff period overlapping with the duration of monsoon precipitation (June–September). Drought characterization and propagation are analyzed in the natural and reservoir-influenced parts of both the basins and varying responses for these basins are reported. Further, the POD of HD from MD is quantified using hit score and FAR matrices, which clearly shows how the presence of the dam influences the downstream HD in both the river basins. The detailed results of the analysis are described in the following sections.

According to the observations (see Supplementary Figure S2), SPI-n (1, 2, 3, … , 12) has demonstrated similarities for the natural and reservoir-influenced portions of the catchment for the Tapi basin. As shown in the heatmap (see Supplementary Figure S2), the annual average (average value of 12 months for each year) SPI range at multiple timescales for both the basins is different, as for Tapi, it varies from −1.67 to 0.64 (upstream of the reservoir) and from −1.84 to 0.61 (downstream of the reservoir). However, for the Chirchik basin, SPI at multiple timescales varies from −1.46 to 1.53 (upstream of the reservoir) and from −1.47 to 1.54 (downstream of the reservoir). From Supplementary Figure S2, it seems like both upstream and downstream SPI have similarities in the data, but this is because the precipitation grid point selected overlaps with the streamflow gauging station, which is nearby for both upstream and downstream parts of the reservoirs. Supplementary Figures S5 and S6 (long-term annual average of SPI at multiple timescales) depict the slight variation in the upstream and downstream data, which also supports the inference of higher drought intensity for the Tapi basin (SPI values are on the negative side) as compared with the Chirchik basin. Long-term drought episodes (10, 11, and 12 months) are noticed for the years 1992, 2001, and 2002 for the Tapi basin both in the upstream and downstream parts of the catchment (see Supplementary Figure S2). The long-term droughts observed during the period from 2001 to 2004 can be justified by the El Niño effect as 2002 and 2004 are recorded as El Niño years as reported by the IMD report (Shewale & Kumar 2005) on drought incidences in India, whereas the Chirchik basin (both natural and reservoir-influenced parts) faced more severe, short and long-term MD events for the five years from 1982 to 1986. A similar observation is reported in the study done by Guo et al. (2018) on the Central Asia region, highlighting the long-term MD events of higher intensity during the timeframe of 1982–1992.

For the downstream part of the Tapi basin, due to reservoir regulation, no such definite pattern was observed in the HD in response to the MD due to weak CC between SPI and SSI. The highest DPT was observed for June (11 months) which is similar to the upstream part, i.e., highest DPT of nine months. For both the upstream and downstream parts of the basins, DPT starts declining due to the influence of the strong rainfall–runoff relation as the dominant presence of the monsoon rain and corresponding runoff.

For the downstream part of the Tapi basin, lower CC values for the shorter SPI timescales are observed because of the reservoir influence; rainfall in the upstream part of the basin is not reflected as runoff into the downstream part is influenced by release from the reservoir (see Figure 6 for the CC heatmap). Similar results were reported by Xing et al. (2021) in the drought propagation study done on the Darling River basin of China which reported the reservoir influence on the DPT and CC values. The study showed that the presence of a reservoir has a significant impact on the correlation between SPI and SSI, resulting in notably lower values, particularly during the wet season.

DPT observation for the downstream part of the Chirchik basin shows lower values during the melting period due to the presence of runoff; however, higher values are observed during the accumulation period. For the upstream part of the Charvak reservoir, DPT during the melting period ranges from four months (July) to nine months (September), however, for the downstream part of the basin, it ranges from five months (June–July) to 12 months (April). See Figure 7(a) and 7(b) for the detailed variation of the DPT in the upstream and downstream parts of the Chirchik basin during melting (April–September) and accumulation periods (October–March). Correlation analysis results for the Chirchik basin are shown in Figure 7, which shows the increasing trend of CC values for SPI-1 to SPI-12. Maximum CC (MCC) for the natural (0.89) and reservoir-influenced parts (0.59) of the Chirchik basin is observed at five- and 12-month timescales for July and March, respectively. However, the weakened strength of CC is evident downstream of the reservoir as compared with the natural part of the basin (see Figure 7 for the CC heatmap).

The DPT results for the Chirchik basin have a contrasting nature as compared with the Tapi, as Chirchik has the effect of snow, which is absent in the case of Tapi. This storage effect for the Chirchik basin, due to the snow dominance in the natural part and the additional influence of the reservoir for the downstream part, has resulted in MCC values occurring at higher SPI timescales during the accumulation period (see Figure 7). A similar storage effect and MCC dynamics were observed for the downstream part of the Tapi (Ghala) showing prolonged DPT values (see Figure 6).

In the later part of the study, the POD of HD by MD is quantified using the hit-score matrix and the probability of false alarm by FAR. From the results (see Table 4), it is evident that the natural part of the Tapi basin (Sarangkheda station) is experiencing a higher POD of HD (0.4) from MD events as the greater number of MD events will be reflected in streamflow stress due to the absence of human disturbances. A similar observation is reported for the upstream part of the Chirchik basin, with a POD value of 0.8, whereas for the downstream part, lower POD values, i.e., 0.2 and 0.4 for the Tapi and Chirchik basins, respectively, are observed because of the reservoir influence. Differences in the POD values of the Tapi and Chirchik basins can be attributed to the differences in the rainfall and streamflow seasonalities and water abstraction activities as streamflow is highly sensitive to these natural and manmade aspects (Somisetty et al. 2022).

Propagation from MD to HD is critical in terms of surface-water management and preparedness. Many previous researchers have studied the different factors influencing the DPT such as water extraction, reservoir operation, and human-induced land-use–land-cover change at a single-basin scale (Huang et al. 2017; Guo et al. 2020), however, few studies have also compared the variation across basins all over the globe (Fuentes et al. 2022; Shi et al. 2022a, 2022b). The current analysis is focused on studying drought propagation under the influence of reservoirs for two climatically different river basins, one in India (Tapi) and another in Uzbekistan (Chirchik) to underline the influence of climatic conditions and reservoirs on DPT.

The analysis of the monthly SSI time series for the Tapi basin revealed that upstream areas of the basin exhibit fewer signs of HD, whereas the portion of the basin influenced by the reservoir has frequent occurrences of short- as well as long-term droughts with higher duration (see Figure 3(a)). A similar observation was reported by Wang et al. (2021) for the Huaihe River basin of China, highlighting enhanced drought duration and severity due to the presence of a reservoir. For the Chirchik basin, it is evident that short-term drought episodes (see Figure 4(b)) occur more frequently during the course of the study period. This may be because the Chirchik basin is a water-scarce basin and receives major input fluxes in the form of snow. Also, the Chirchik basin has faced more intense HD both upstream and downstream of the basin as compared with the Tapi. A study done by Brunner et al. (2023) on HD in the snow-dominated European Alps has reported identical findings, showing how snowmelt deficit can lead to more frequent HD events.

In the current analysis, the drought propagation process was studied and inference on DPT was drawn on the basis of maximum values of CC. A similar approach has been adopted by other researchers (Huang et al. 2017; Ding et al. 2021) in the past to study the drought propagation mechanism. From the results of drought propagation analysis, it is observed that climate and precipitation mechanisms play a crucial role in deciding response time or DPT. From the results, it is clear that the storage effect induced either by a natural process (snow accumulation) or anthropogenic actions (reservoir operation) affects the DPT for both Tapi and Chirchik basins (see Figures 6 and 7). Our results are in line with the findings reported in many previous studies (van Langen et al. 2021; Wang et al. 2021; Xing et al. 2021; Ma et al. 2022), highlighting the natural and manmade storage effect on altering the HD propagation pattern. The study revealed that the presence of reservoirs in both the Tapi and Chirchik River basins influences river connectivity in the downstream regions and disturbs the natural hydrological process and it is evident through influences on drought propagation as explained in Xing et al. (2021). The probable reasons for the differences in the propagation results in different months for the downstream part of both Ukai (Tapi basin) and Charvak (Chirchik basin) reservoirs are because of the differences in the reservoir operation rules in both river basins and the presence of snowmelt in the Chirchik River basin and their responses are different under drought conditions (Ma et al. 2019; Xing et al. 2021).

This study innovatively assesses reservoir impacts on drought propagation in the Tapi basin with the Ukai reservoir in India and the Chirchik basin with the Charvak reservoir in Uzbekistan. Utilizing the SPI and SSI, it identifies correlations between MD and HD. Through hit-score-based evaluations, the research demonstrates how reservoir operations influence the drought propagation patterns. The present study analyzes the monthly lag-time values to contrast monsoon and snow-melting influences on drought propagation dynamics in both the river basins considering the effect of reservoirs as well. The study also highlights the reservoir-influenced trends in propagation predictive skill, offering insights for water resource management and drought warnings in arid regions. Furthermore, the findings of this study can yield insights applicable to these river basins, facilitating the establishment of early drought-forecasting capabilities. This, in turn, provides invaluable guidance for reservoir management practices across varied environmental contexts.

The following particular conclusions are drawn from the current analysis:

• (i) The Tapi and Chirchik River basins have different climatic conditions and hence, show a major difference in MD characterization. In contrast to Chirchik, where water-stress conditions were predominant, Tapi has fewer instances of drought and water scarcity.

• (ii) Long-term HD are observed to have greater duration duration in the downstream part of the basin under the influence of the reservoir for Tapi; however, for the Chirchik basin, this influence is nullified due to the incoming natural streams in the downstream part.

• (iii) Drought propagation inference drawn from MCC values, for the reservoir-influenced part, has shown that DPT decreases during the monsoon (June–September) period for the Tapi basin and melting period (April–September) for the Chirchik basin due to the stronger rainfall–runoff relation at shorter timescales.

• (iv) The natural mechanism of drought propagation is altered in the downstream of both the Tapi and Chirchik basins; however, snow accumulation has the dominant influence on the DPT for the Chirchik basin (both upstream and downstream).

• (v) POD matrix (hit score) has given evidence for altered drought propagation mechanism under the reservoir influence for both the Tapi and Chirchik basins.

This study primarily focuses on the modified relationship between SPI and SSI in the presence of human disturbance (reservoir operation) or natural disturbance (snow storage) and one potential limitation is that it has ignored the influence of climate change on drought propagation. It demonstrates the challenges in accurately estimating DPT using the MCC-based approach in such scenarios. The current analysis focuses just on one human-caused disturbance (reservoir) and one natural disturbance (snow). However, additional modifications such as changes in land-use–land-cover (Omer et al. 2020) and large-scale climatic factors like El Niño–Southern Oscillation (Das et al. 2022) can be examined more extensively in future studies.

We would like to thank and acknowledge the support of the Department of Science & Technology (DST), Government of India for the project – INT/UZBEK/P-11 sanctioned under the Indo-Uzbek Joint Research Programme.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.