Response to hydrometeorological changes and evaluation of habitat quality in the Dongting Lake basin, China

Necessary hydrological processes within watersheds are one of the most important factors in maintaining the rich diversity of ecosystems, and similarly, hydrological processes within rivers play an essential role in protecting ecological functioning (Sabater et al. 2018; Keith et al. 2022). However, over the last century, extreme climate change and large-scale human-industrial advances have had significant impacts on many rivers and even watersheds globally (Wang et al. 2020). Climate variability and human activities affect the conditions of species in both terrestrial and aquatic ecosystems, not only directly by altering the habitat conditions of species but also indirectly by influencing ecosystem functioning through the biological chain (Zolfagharpour et al. 2022). Therefore, the study of hydrological processes in rivers will be important for the maintenance of water ecology and ecosystem diversity in watersheds (Zhang et al. 2023).

The Brisbane Declaration of 2007 defines environmental flows as the quantity, timing, and quality of water needed to sustain freshwater ecosystems and the livelihoods of the plants and animals that depend on them. Existing studies have shown that river flow regimes play an important role in maintaining the biodiversity, ecological integrity, and restoration of ecosystem services of rivers and the surrounding region, which is more evident in wetlands (Mao et al. 2023). Therefore, many local governments and water management agencies have invested considerable effort in environmental flow releases, such as manipulating dams, reservoirs, or other water management facilities to maintain annual or seasonal flow requirements (Cid et al. 2022; Thompson et al. 2022). Climate change has always been an important factor affecting surface runoff, especially in high-temperature and rainfall zones. In large catchment areas, the hydrological impact of surface runoff has long been a hot research-topic, with researchers worldwide examining the varying factors of surface water from different perspectives (Oliveira & Quaresma 2017). For example, Klotz et al. (2022) simulated changes in surface runoff under different scenarios using hydrological models, providing a visual comparison of the varying effects of various factors. Machine learning has also demonstrated superior performance in simulating natural runoff and restoring runoff changes under natural climatic factors, thereby facilitating comparative analysis of runoff variations (Ayzel & Heistermann 2021). Singh et al. (2023) developed an integrated multi-model framework based on various machine-learning models to reconstruct gridded runoff in Europe over the past 500 years. Previous studies have delved into the mechanisms underlying the interplay between climate and river flow dynamics, yet there remains a gap in understanding how subsequent alterations in watershed habitats correspond to these climate–flow dynamics. Hence, there is a pressing need for further research pertaining to watershed habitats to address the management requirements associated with such changes (Lei & Middleton 2021).

Under the dual impact of environmental changes (surface water, climate, etc.) and human activities, the quality of surface habitats is facing unprecedented challenges (Zhang et al. 2018; Wang et al. 2023). In response to these unknown risks, efforts have been made to assess the ecological and habitat quality of the surface environment through various approaches (Sepulveda et al. 2023). Yuan et al. (2021) evaluated the ecology of the Dongting Lake (DTL) basin using the remote sensing-based ecological index (RSEI) and identified potential influencing factors of ecological changes within the basin. Meanwhile, model-based methods have gained popularity in recent years due to their ease of operation, strong visualization capabilities, and comprehensive theoretical frameworks. The InVEST model is one such example that has been widely adopted (Bastos et al. 2023). Lei et al. (2022) utilized the InVEST model to assess the habitat quality of Hainan Island and analyzed the impact and trend changes of human engineering activities on the island's habitat quality. The DTL basin in China is one of the 200 globally prioritized ecological protection areas and a crucial stopover and wintering ground for migratory birds (Geng et al. 2021a, 2021b; Zhu et al. 2021). Studies by Qi et al. (2023) on the net primary productivity of vegetation in the DTL basin have revealed that some regions within the basin are facing ecological degradation crises. To address these ecological issues, a series of ecological restoration projects have been implemented in the basin. However, due to a lack of regional studies on hydrological and ecological differences, there are significant variations in the effectiveness of ecological restoration across different regions (Wang et al. 2021). Therefore, it is necessary to assess the habitat quality of the DTL basin and investigate the ecological and hydrological responses of its tributary basins. This is crucial for ecological protection and restoration efforts within the basin and holds significant implications for global biodiversity conservation. By conducting such assessments and studies, we can gain insights into the current state of the basin's ecology, identify key areas for restoration, and develop targeted strategies to enhance the overall health and resilience of the DTL basin ecosystem (Cao et al. 2023; Luo et al. 2023).

The aim of this paper is to link the changes in the main rivers in DTL with climate change in the basin to analyze the changes in habitat quality in the basin, and finally to compare and analyze the relationship between the flow regime of the rivers and climate change based on previous studies. To this end, we did the following: (1) analyzed the environmental flow distribution characteristics of the four major rivers in the DTL basin, (2) calculated the distribution of meteorological inputs in the basin based on the RClimDex model, and explored the potential linkages with the four rivers, (3) attributed the changes in runoff from the four rivers to the four rivers on a monthly scale using the long short-term memory (LSTM) model, and (4) assessed the habitat quality of DTL basin based on the InVEST model. The results of the study can provide a reference and theoretical basis for surface water resource management and habitat protection in the DTL basin.

For the hydrological data, the measured daily flows of the Yangtze River, the Four Rivers (Li, Yuan, Zi, and Xiang rivers), and the representative hydrological stations of DTL were collected from 1990 to 2020, and were provided by the Yangtze River Water Conservancy Commission (Table 1). Chenglingji (CLJ) Hydrological Station is located at the outlet section of DTL, about 4 km from the confluence of the Yangtze River, and controls the changes in the hydrological situation in DTL. Shimen (SM), Taoyuan (TY), Taojiang (TJ), and Xiangtan (XT) hydrological stations are the last hydrological stations before their respective rivers flow into DTL, representing the hydrological situation of the confluence in the DTL basin. By analyzing the five representative hydrological stations, the evolution of the hydrological situation in the DTL basin can be visualized and accurately reflected.

Bilateral tests are used here. At the α-significant level, if ⁠, the sequence negates the original hypothesis of no trend. The sequence has an increasing or decreasing trend; conversely, the sequence has no increasing or decreasing trend.

For the M-K mutation test, the original time-series statistic UF is calculated at the same time as its inverse sequence column UB (UF = UB), and if an intersection of the two UF and UB curves is observed in the given confidence-level α, it is considered that a mutation occurs there.

The environmental flow components (EFCs) are the flow and its processes required to maintain the river's ecosystem and are described by the dry flow, the extra-dry flow, the high flow pulse, the minor flood, and the major flood, a total of 34 ecologically significant indicators in five groups (Table 2). The threshold delineation of flow events for relevant environmental flows is detailed in Peres & Cancelliere (2016).

The RClimDex model is a source of Fortran programs written in R for different indicator calculations. First, for data quality checking, the model checks for invalid data in the dataset, including cases of negative precipitation, minimum temperature greater than or equal to the maximum temperature, and so on. Then a series of statistical calculations are performed to determine 27 meteorological indicators including 11 extreme precipitation indices and 16 extreme temperature indices. In this study, we selected eight extreme precipitation indices (http://etccdi.pacificclimate.org/) and four representative extreme temperature indices as representatives for calculation and analysis (Table 3). More detailed information about meteorological indices can be found in the study by dos Santos et al. (2011).

Gray correlation analysis was first proposed by Professor Deng Julong, an expert in cybernetics, and a series of gray system-related theories were established after that. Gray correlation analysis is one of the important methods in gray system theory, which is a multi-factor analysis method, and research focuses on the main correlation relationship between the factors (Dong et al. 2018). The calculation of gray correlation is analyzed as follows.

The LSTM is a type of deep-learning neural network that has been used by many scholars in recent years for interpolation and prediction of hydrometeorological data. In this study, meteorological data from various watersheds were utilized as inputs to generate five meteorological–flow models. These models were then used to reconstruct the flow processes of the Four Rivers and the DTL region solely under the influence of climate change. This reconstruction allowed for the assessment of the impact of external stressors on the runoff within the watershed. For the development of the watershed models, measured data from 1961 to 1990 were utilized for model calibration, while measured data from 1990 to the year of abrupt change were used to validate the model's accuracy. The performance of the LSTM model was comprehensively evaluated using the Nash–Sutcliffe efficiency (NSE), root mean square error (RMSE), and the coefficient of determination (R²).

The model run should be set up with parameters that are selected considering the InVEST model user's guide and determined considering the results of research in various ecological fields. For threat sources and their weights and the maximum influence distance of threat factors, the parameter settings are shown in Table 4; for the degree of sensitivity of various habitats to threat factors, the parameter settings are shown in Table 5. The higher the sensitivity of natural habitat quality to the threat factors, the greater the influence of this threat source on habitat quality (Guan et al. 2022).

In Figure 6(b)–6(e), it can be found that the high-energy pulses in the four water basins cover most of the time, but in the post-surge period of XT and TJ, it can be found that the frequency of small- and medium-sized floods in the Xiang and Zi rivers is increased, which reflects the influence of the extreme climatic conditions in the DTL basin. The extremely low dry water flow of both TJ and SM reflects the seriousness of the water ecological safety hazard in the basin. In the distribution of the environmental flows of the Four Rivers, it can be seen that the peak values of the environmental flows of the Four Rivers were generally reduced in 2003 and 2004, especially for the small floods, reflecting the reduced probability of floods in the DTL basin as a humid and rainy area.

In general, compared with the period before the mutation, the frequency and magnitude of extra-dry flow events and large and small floods were reduced after the mutation, especially the extra-dry flow events, which seldom occurred after the mutation, and the peak flow of small flood events decreased, especially in the DTL basin where the high-energy pulses took up most of the time intervals, and the extra-dry flow and large floods almost did not occur. The construction of water conservancy projects has played a great role in protecting water security in the basin.

Figure 7 shows the regional spatial distribution of temperature indices in the DTL basin, and it can be found that the minimum temperature indices are mostly distributed in the southeast and west, where the mountain ranges are isolated to form an obvious low-temperature holding zone, specifically the east–west trending Nanling Mountain Range that stretches across the border between Hunan and Guangdong, which blocks the cold air from the north from coming down to the south to form a cold-air layer in the sky above Hunan, and the unique mountain range distribution has an impact on the distribution of the minimum temperature. By comparing the distribution of the maximum temperature index, it can be found that in the northeast and southwest of the zone, the climate of high-temperature weather stays higher, under the influence of the mountain range, resulting in the southern zone of the tropical monsoon being blocked. In the northeast of DTL, the TX10p and TX90p indices are significantly higher in the high-temperature areas around the wetlands of DTL than in the other areas due to the plains and fewer hills and mountains.

Figure 8 shows the regional spatial distribution of extreme rainfall indices, and it can be noticed that the high rainfall indices are mostly distributed in the west as well as in the northern belt and in the south, while there is a significant decrease in rainfall closer to the east. In Figure 8(g) of the annual total rainfall index PRCPTOT, it can be found that the high rainfall areas in the basin are less distributed in the southwestern and northern zones, which are in the condition of blocking by the mountain range, and the rainfall events in the whole year are significantly reduced compared with the eastern region. In the Rx1day and Rx5day distributions, the annual maximum daily precipitation and the annual maximum five consecutive days of precipitation in the northern region are significantly smaller than those in the other regions, and the intensity of rainfall over a short period of time represents a magnitude of heavy rainfall that will be much lower than that in the other regions. In the R10mm distribution map (Figure 8(d)), it can only be seen that the rainfall zone in the basin occurs in the eastern part of the basin, and the number of rainfall days in a year is significantly higher than that in the other basins of the Four Rivers due to the coverage of the Xiang River basin in this area. Specifically in the distribution of the Four Rivers watersheds of DTL, it can be found that the rainfall in the Xiang River area is significantly higher than that in the other watersheds. The extreme daily rainfall intensity in the Yuan and Li rivers is significantly higher than that in the other areas, and they are in the same rainstorm area as the Yangtze River, so the impacts on the main stream of the Yangtze River will be much greater than those on the Zi and Xiang Rivers in a combined rainstorm event.

Figure 9(b)–9(e) shows the distribution of the correlation of the four sub-basins to the meteorological indices in the basin, and it can be found that most of the correlations between the environmental flow indices and the meteorological indices in the Four Rivers are very similar, and the correlation of the meteorological indices in the Xlowl Dur and the Sfldl Peak are all blank, which may be due to the frequency of large floods and small floods as well as the occurrence of extreme dry water events in the four sub-basins. This may be due to the frequency of major and minor floods as well as extreme dry water events, on the one hand, as the Four Rivers basin is a humid area with sufficient rainfall, and on the other hand, to the difference between the high flow rate in the Four Rivers as Class II and Class IV rivers compared with the Yangtze River. The response to the correlation to the meteorological indices is obvious, even in the river since the sudden change of the river to no more flood events, and the correlation with the meteorological indicators naturally cannot be analyzed. In the subsequent comparison of the four sub-basins, it was found that for the XT and TY hydrographic stations the correlation of the meteorological index of their respective basins is significantly higher than for the TY and SM hydrographic stations analyzed.

Based on the spatial pattern of habitat quality in the DTL basin (Figure 12), it was observed that habitat quality varied considerably spatially, with a wider distribution of higher habitats (0.8–1) in general. Calculations showed that the mean values of different habitat quality in 1980, 1990, 2000, 2005, 2010, and 2020 were 0.8261, 0.82, 0.8188, 0.8183, 0.8172, and 0.8106, respectively, and that the habitat quality in the DTL basin showed a gradually decreasing trend of change, with an overall obvious decline. In terms of the different habitat-quality grades in the DTL basin, the ‘low’ grade showed an increasing trend from 1980 to 2020, with an increase of 1.51%. The ‘medium’ grade shows a decreasing trend, with a decrease of 1.53%, indicating that the rapid economic development under the influence of human activities has led to a clear downward trend in habitat quality. The changes in the ‘relatively high’ and ‘high’ grades were smaller, with the area increasing by 0.1% and decreasing by 0.08%, respectively. In the DTL basin, the ‘high’ grade of habitat quality accounts for more than 60% of the total area, which can indicate that the overall performance of habitat quality in the DTL basin is better, but from 1980 to 2020, the quality of habitat gradually declined, and there is a risk of habitat deterioration in the DTL basin.

In this study, by examining the changes in flow processes recorded at various hydrological stations in the DTL basin, the results show that human activities play a dominant role most of the time. However, the influence played by climatic aspects cannot be ignored; the distribution of the correlation between the RClimDex index and the environmental flow found that the influence of rainfall was significantly higher than that of the temperature index, and rainfall was affected by the geomorphology, and part of the rainfall converged to DTL through the river to form a larger catchment area, and this part of the rainfall played a more pronounced role in evapotranspiration compared with the temperature (Oliveira & Quaresma 2017). Within watersheds with connected lakes, watershed hydrological changes significantly influence the ecological succession of the lakes. Also, within the Poyang Lake watershed in the lower section of the Yangtze River, Li et al. (2015) demonstrated that changes in hydrological regimes within the watershed significantly influence lake connectivity and ecological conditions, and drive the dynamic migration of vegetation and water quality changes in the lake area. Zhang et al. (2021) and others demonstrated that hydrological changes within a watershed lead to changes in ecological status on a larger scale through confluence, and Khatun et al. (2021) accurately identified wetland ecological responses to changes in hydrological regimes while exploring scenarios in which water quality changes and biological responses are leading to declines in the abundance of water bodies in connected lake wetlands. Changes in the hydrological regimes of rivers connected to lakes will significantly alter the ecological status of lakes, and these ecological changes will be more severe if they occur in complex lake wetlands or floodplain wetlands (Lai et al. 2014).

In recent years, human activities such as urban expansion, lake reclamation for farmland, and industrial pollution have significantly altered the original state of ecological structures, leading to degradation in habitat quality and disruption of ecological balance (Calderon et al. 2019). As evident from the habitat quality results (Figure 11), high-grade habitat quality is primarily distributed in areas far from the DTL water body, where there are numerous mountainous and hilly regions with relatively high forest coverage. These regions benefit from various ecological protection policies for flora and fauna (Yuan et al. 2021). Conversely, human activities centered around the DTL perimeter have led to a decline in ecological conditions in nearby areas, resulting in an inverse relationship between habitat quality and population density (Hu et al. 2015). These spatial differences in habitat degradation suggest that habitats closer to urban and industrial areas are more likely to experience degradation. The habitat quality along the riverbanks within the basin is moderate (0.4–0.6), and there has been a significant increase in low-grade habitat quality near the riverbanks after 2005. This trend is consistent with the high population density along the banks of the DTL basin and the decrease in habitat quality caused by human activities, aligning with research findings in other regions (Mengist et al. 2021).

Within large catchment areas, climate change is a major factor affecting habitats, particularly in terms of rainfall (Astagneau et al. 2022). Precipitation is a direct manifestation of climate change, directly influencing changes in surface water and ecological water demand in the region, with river catchments playing an equally important role. With the rapid development of industry and population increase, the utilization of water resources in the DTL basin is increasing, the contradiction between supply and demand of water resources is becoming more and more serious, and the trade-off relationship between water resources and environmental protection needs to be further studied (Balata et al. 2022). As of 2019, 14,096 reservoirs have been built in the DTL basin, with a total capacity area of 51.412 billion m³, which significantly regulates the hydrological patterns of rivers in the DTL basin and causes serious ecological problems (Zheng et al. 2023). For this reason, the coordination of ecological quality and social development in the region has widely attracted the attention of researchers and managers worldwide (Mao et al. 2019). Due to the dependence of numerous ecological patterns in the watershed on water resources, the ecosystems connected through the vast water system will be significantly affected by the water resources pattern and fluctuate continuously, thus putting more demands on the regulation of water resources.

Uncertainty is a widespread phenomenon in the process of hydrological modeling. The research on climate-driven hydrological changes presented in this article is based on the LSTM deep-learning model. Both model parameters and the quality of input data are significant sources of uncertainty in simulation results (Cui & Yu 2023). To ensure the quality of input data, all data used in the model originate from measured observations, and outlier removal as well as mean normalization have been applied. To minimize the impact of model uncertainty, we employed a manual trial-and-error method to train the optimal hyperparameters. This approach aimed to achieve both minimal model fluctuations and optimal accuracy. Additionally, we utilized a mean-based approach to process the output results. Under these constraints, the flow reconstruction results for all hydrological stations during the baseline period exhibited excellent agreement with the measured values (Table 7). The NSE for both the calibration and validation datasets was higher than 0.75, and the differences in fitting metrics were below 5%. These results indicate that the model's performance during the variation period can be used to evaluate the impact of climate change on hydrology (Frame et al. 2023). Overall, despite the inherent uncertainty in hydrological modeling, the approach taken in this study, including data preprocessing, hyperparameter optimization, and output processing, has effectively minimized the uncertainty and produced reliable results for assessing climate-driven hydrological changes.

This study was conducted to gain insight into the linkages between climate conditions and the hydrology of tributaries in the DTL basin and to analyze the evolution of habitat quality in the basin. We analyzed the evolutionary characteristics of climate indicators and environmental flow indicators of tributaries in the DTL basin, introduced the LSTM model to quantify the contribution of climate change and human activities to runoff changes on short and long time-scales, and introduced the InVEST model to assess the habitat-change status in the basin during 1980–2020. The main conclusions are summarized below:

• (1) The tributaries in the DTL basin showed a smaller range of environmental flow characteristics after the abrupt change, and the frequency of extreme flow and hydrograph time decreased.

• (2) Within each tributary, the rainfall index showed a high correlation with the environmental flow index, and the temperature index showed a low correlation except for DTL. The correlation of climate indices was strongest in the Xiang River basin and weakest in the Yuan River basin.

• (3) The attribution analysis showed that the impacts of human activities on the annual scale were generally above 80%, with the highest being 89% in the Yuan River. On the intra-annual scale, human impacts were mostly dominant, but on the Xiang River, the impacts of climatic factors were more dominant.

• (4) Higher habitat-quality distribution accounted for the highest percentage, in all years accounting for more than 60% and mainly distributed away from the water body of DTL. Habitat quality is lower in highly populated areas and around DTL.

In the future, further consideration can be given to policy orientation, species invasion and industrial pollution damage effects, and the correlation between them can be analyzed for a more comprehensive assessment.

This study was supported by the Science and Technology Program of Guizhou Province (KT202008) and the Basic Research Project of Key Scientific Research Projects of Colleges and Universities of Henan Province (23ZX012).

X.B. conceived the study and wrote the first draft, the data collection and analysis was done by X.B. and W.Y., the methodology was analyzed with the help of Y.M. and M.Z., literature by W.G. and H.W. supervised the paper, and all authors provided comments and assistance on the first few versions of the manuscript. All authors read and approved the final manuscript.

The scope of the work does not require any approval of the ethics committee. The ethical approval is not applicable for the present work.

The scope of the work does not require participation of any candidate. The consent to participate does not apply for the present work.

Consent of all the authors was taken before the communication of the manuscript in its present form.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.