Linking Critical Zone With Watershed Science: The Example of the Heihe River Basin

Abstract

Watersheds are the basic unit of Earth's terrestrial systems and are therefore ideal units for the study of critical zones (CZs). However, grand challenges remain regarding the observation, modeling, and management of CZs and watershed systems. We synthesize the progress and breakthroughs associated with watershed science and CZ research in the Heihe River basin (HRB), a large-scale endorheic river basin with unique mountain cryosphere-oasis-desert landscapes and prominent human-nature competition for water resources. The HRB observation system consists of mountain cryosphere, agricultural oasis, and natural oasis observatories and is promoted by large-scale comprehensive experiments to achieve multiscale observations. A watershed system model that couples ecohydrological models with socioeconomic models is developed to investigate the complex interactions among water, ecology, and socioeconomics in the HRB. The model is embedded in a decision support system to bridge science and decision-making and to better serve river basin sustainability. Significant progress and breakthroughs have been made in CZ and watershed process research (e.g., cryospheric hydrological processes, ecological and hydrological interactions, and surface-groundwater interactions) in the HRB. Nevertheless, observation and modeling of geochemical and geomorphological processes in the CZ have not been well addressed in integrated watershed studies of the HRB. In the future, new observation technologies, agent-based models, machine learning, and data assimilation will benefit both watershed science and CZ science and help to address complexity and dynamics in the CZ at the river basin scale. Overall, the HRB has successfully demonstrated how an experimental river basin can link CZ science with watershed science.

Linking Critical Zone Science With Watershed Science

Our planet is currently under unprecedented pressure from climate change and human activities; these pressures are pushing planetary boundaries and potentially introducing irreversible and threshold-abrupt changes to Earth systems (Rockström et al., 2009, 2021). These pressures are expected to increase in the coming decades (Banwart et al., 2013), threatening the stability of Earth's life support systems and its capacity to support our future generations (C. Zheng et al., 2021). Maintaining planetary stability and resilience while ensuring sustainable human development are among the greatest challenges facing scientists and politicians (Rockström et al., 2021; Steffen et al., 2020), and these goals require a deep integration of knowledge, techniques and expertise derived from multiple fields to build a holistic and unified understanding of the Earth system (Sullivan et al., 2017).

Building on multidisciplinary and interdisciplinary techniques, many new efforts have emerged over the past two decades to link traditional disciplines that were often studied in isolation to improve our understanding of the structure, functioning and evolution of the Earth system and its responses to climate change and anthropogenic disturbances and to promote the sustainable development of the Earth system (Cheng & Li, 2015; Steffen et al., 2020). Critical zone (CZ) science (Baldocchi et al., 2001; Lin, 2010; Sullivan et al., 2017) and watershed science (Cheng & Li, 2015; Cheng et al., 2014; He & James, 2021; Montanari et al., 2015) are two stark manifestations of such efforts, and they both represent integrated research pathways in Earth system science. Despite the significant advances made in CZ and watershed science in recent decades, grand challenges regarding the observation, modeling, and management of CZ and watershed systems remain. The major challenge regarding observations lies in the surface-subsurface integration of multiscale and multitechnology observation systems that can capture the strong heterogeneities and associated uncertainties of the Earth system (Li et al., 2013). With respect to modeling, the main challenge involves determining how to couple geochemical cycles with human behaviors and other slow and rapid natural processes to obtain integrated human-nature system simulations (Li, Zhang, et al., 2021; Lund, 2015). The greatest management challenge involves synthesizing multisource information, including observations, model simulations, expert knowledge, and public opinions, to better serve sustainable development. CZ science and watershed science have different characteristics that, if integrated and linked, we believe can provide the best opportunity for addressing the challenges mentioned above and for understanding the Earth system in ways that cannot be obtained using either of them independently.

In recognition of the fact that the Earth's surface sustains nearly all terrestrial life, CZ science defines a new unit of research, that is, the CZ, which spans vertically from the top of the vegetation layer to the bottom of the aquifer layer (National Research Council, 2001). The CZ concept has stimulated a paradigm shift in Earth science and provided a unifying framework for the holistic study of terrestrial physical, chemical and biological processes at the interface between Earth's surface and shallow subsurface (Lin, 2010). However, the horizontal CZ boundary is not clear and can vary among different land use types, watersheds, and climatic and geological zones (Brantley et al., 2007; Zhang, Song, & Wu, 2021). Watershed science, on the other hand, considers horizontally quasi-closed systems (i.e., watersheds or basins) as the unit of research; watersheds are considered the best unit for practicing Earth system science because watersheds exhibit all of the complexities of the Earth system and are the basic unit for the hydrological cycle and water resource management (Cheng & Li, 2015). However, the vertical boundaries of the watershed are not explicitly defined in existing watershed science. In this regard, linking CZ science with watershed science would allow the definitions of the boundaries of the basic Earth system science research unit, both horizontally and vertically, to be more clearly defined.

CZ science relies on multidisciplinary and interdisciplinary approaches to study the system of coupled and cross-scale chemical, biological, physical, and geological processes within CZs to support life on the Earth's surface (Brantley et al., 2007), while watershed science emphasizes a systematic approach to the holistic and integrated study of watersheds (including the entire water-soil-air-plant-human nexus) (Cheng & Li, 2015). CZ science explicitly emphasizes the study of biogeochemical cycles, whereas watershed science does not due to the tremendous challenges of modeling integrated hydrological-biogeochemical processes (Li, Sullivan, et al., 2021); these challenges include spatial- and temporal-scale mismatches and computational cost constraints (Hubbard et al., 2018, 2020). Moreover, CZ science focuses on both rapid cyclic processes (e.g., river flows and vegetation growth) and long-term cumulative changes (e.g., bedrock weathering, ecosystem succession, and tectonic movements) (Guo & Lin, 2016), which have not attracted extensive attention from watershed science research. Watershed science highlights the coupling characteristics of humans and nature (Cheng & Li, 2015; Cheng at al., 2014), as we are currently living in the Anthropocene (Richter & Mobley, 2009; Sivapalan et al., 2014), an epoch characterized by intense anthropogenic influence on Earth systems, whereas CZ science has done less to capture the effects of anthropogenic influences or disturbances on CZ processes, especially in its early stages (Minor et al., 2019). Although human impacts have recently received much attention from CZ science, human activities are typically treated as extrinsic drivers, without considering the bidirectional interaction between human activities and CZ processes. In this regard, linking CZ science with watershed science would yield a better understanding and interpretation of all natural and human processes, as well as their interactions in CZs and watersheds, hereby helping us to better describe, predict, and manage the Earth system.

In this paper, we aim to provide a synthesized report of the progress and breakthroughs associated with watershed science and CZ research in the Heihe River basin (HRB), a typical endorheic basin located in northwest China. The HRB is a large-scale endorheic river basin (more than 140,000 km²) with unique mountain cryosphere-oasis-desert landscapes and prominent human-nature competition for water resource. The ecosystems in the HRB experienced deterioration during the second half of the twentieth century but have been greatly restored after the implementation of an ecological water diversion project (EWDP) in 2000. The HRB has also been used as an experimental river basin to carry out integrated studies of the water–ecosystem–economy system in pursuing sustainable development in endorheic basins. We review this progress and give particular emphasis to the critical role of the HRB observation system and watershed system model in advancing our understanding of watershed and CZ-spanning processes at multiple spatiotemporal scales (including cryospheric processes, surface-groundwater processes, and vegetation-hydrology and carbon-water interactions). We also propose the advancement of watershed science to better predict and manage CZ dynamics and the development of CZ observatories and modeling platforms to capture the full spectrum of CZ processes, especially the underrepresented geochemical processes and human behaviors.

HRB Observation System

Earth system science cannot survive without Earth system observations, as this field requires diverse measurements collected across a wide range of spatial and temporal scales; thus, the need to support the establishment of integrated and multiscale observing systems resonates in the CZ science and watershed (or hydrology) science communities (Cheng & Li, 2015; Goodrich et al., 2021; Kirchner, 2006; Sullivan et al., 2017). In recent decades, several watershed/basin-scale observation systems, such as the CZ Observatories of the United States (Brantley et al., 2017), the Terrestrial Environmental Observatories of Germany (Bogena, Montzka, et al., 2018; Zacharias et al., 2011), the Danish Hydrological Observatory (Jensen & Refsgaard, 2018), the Changing Cold Regions Network of Canada (Debeer et al., 2015), the Qinghai Lake Basin CZ Observatory on the Qinghai Tibet Plateau (Li,Yang, et al., 2018), and the HRB observation system in Northwest China (Li et al., 2013; Liu, Li, et al., 2018), have been established. Despite the differences in the specific design of the programs, these observation systems share a common feature in their multifaceted, multidisciplinary approach to observing the Earth's surface by taking advantage of both ground-based and remote sensing technologies, as well as several other new approaches (e.g., wireless sensor technologies and the Internet of Things) (Banwart et al., 2013; Bogena, White, et al., 2018). These observation systems provide CZ and watershed scientists with comprehensive data sets that can be applied for process understanding, hypothesis testing, model development, and remote sensing technique validations.

The HRB observation system was established and developed under the major research plan of the National Natural Science Foundation of China (NSFC), that is, the integrated research on the ecohydrological process in the HRB (referred to as the “Heihe Plan”). This observation system was designed from an interdisciplinary watershed science perspective to capture the spatial and temporal variations in land surface, hydrological, ecological, and near-surface atmospheric processes to serve the understanding of watersheds as a whole and the development and validation of integrated watershed models. The HRB observation system has been recording data for over 20 years and has been greatly promoted in large-scale comprehensive experiments, that is, the Watershed Allied Telemetry Experimental Research (WATER) (Li et al., 2009) and the Heihe Watershed Allied Telemetry Experimental Research (HiWATER) (Li et al., 2013).

Three representative observation areas have been established in the basin based on the different characteristics of the upper, middle, and lower reaches of the HRB and the sensitivity to climate change and human activities. These areas are the mountain cryosphere observatory in the upstream area, the agricultural oasis observatory in the midstream area, and the natural oasis observatory in the downstream area (Figure 1). These regions are coordinated with long-term operational monitoring and research stations to cover the whole HRB and are allied with airborne and satellite remote sensing to achieve multiscale observations ranging from in situ, pixel/grid, and subbasin/irrigation district observations to observations taken at the entire river-basin scale (Figure 2). In this multiscale observation system, remote sensing plays an important role. For example, airborne remote sensing, either manned or unmanned, is an experimental tool used to bridge the scale gaps between the in situ, pixel/grid, and subbasin scales. Therefore, upscaling methods have been applied to in situ observations, particularly those obtained from sensor networks (Jin et al., 2017; Liu, Xu, et al., 2016). Additionally, various river basin-scale ecohydrological remote sensing products at high spatial and temporal resolutions have been operationally produced and used in ecohydrological models and data assimilation systems (Li et al., 2022). Therefore, remote sensing has become an indispensable joint force in this watershed observation system.

Figure 1. Overview of the Heihe River basin (HRB) observation system. The upper left panel shows the location of the HRB.

Figure 2. Schematic diagram of the multiscale observation system in the Heihe River basin. RS: remote sensing, AMS: automatic meteorological stations, EC: eddy covariance, WSN: wireless sensor network, LAS: large-aperture scintillators, COSMOS: cosmic-ray soil moisture observing system.

Mountain Cryosphere Observatory

The mountain cryosphere observatory conducts observations at three scales: the whole upstream area of the mainstream of the Heihe River, the subbasin (the Babao River basin), and some representative watersheds (i.e., the Hulugou, Dayekou, and Binggou watersheds) (Figure 1). As shown in Figure 3, the mountain cryosphere observatory mainly includes basic hydrometeorological and flux networks and observations of glacial/permafrost/snow hydrology and sediment processes as well as forest hydrology processes.

Figure 3. Schematic diagram of the mountain cryosphere observatory in the Heihe River basin (only the main observation instruments active on multiple scales are listed here). EC: eddy covariance, LAS: large-aperture scintillators, GPR: ground-penetrating radar, SSG-2: snow scales, a device that measures the snow water equivalent of the snowpack, GMON: GammaMONitor, the instrument used to measure snow water equivalent, NMR: nuclear magnetic resonance, RS: remote sensing, UAV: unmanned aerial vehicle, WSN: wireless sensor network, and COSMOS: cosmic-ray soil moisture observing system.

In the whole upstream HRB, ten hydrometeorological-flux observation stations are deployed; in particular, two supersites carry out comprehensive observations of snow and soil freeze–thaw processes. In the Babao River basin, a subbasin of the eastern branch of the upper mainstream of the Heihe River, a wireless sensor network, has been implemented (Jin et al., 2014). More than 40 sensor network nodes are used to capture the soil moisture and temperature in the active layer (i.e., the top layer of ground subject to annual thawing and freezing), land surface temperatures, snow depths, and precipitation. The Hulugou, Dayekou, and Binggou watersheds are used as typical experimental watersheds, each with an area of tens of square kilometers, with different focuses on cold-region hydrology, forest ecohydrology, and remote sensing product and ecohydrological model validation. Some research stations have been built to conduct long-term monitoring (Che et al., 2019; R. Chen et al., 2018) in conjunction with remote sensing technologies.

One of the purposes of the mountain cryosphere observatory is to capture the vertical zonal ecohydrological features that vary with elevation. From high to low elevations, different landscapes and corresponding ecohydrological processes are measured with state-of-the-art instruments and observational technologies. For glaciers, a combination of graduated snow sticks, ground-based light detection and ranging (LiDAR), and ground-penetrating radar (GPR) are used to observe the glacial ablation and mass balance (Wang & Pu, 2009; Wu et al., 2010; Xu, Li, et al., 2019). For snow, the Snow Pack Analyzing System (SPA, Sommer, and Austria), snow scales SSG-2, GammaMONitor, eddy covariance (EC), and FlowCap are used to observe the physical snow accumulation properties and snow mass–energy exchange processes (e.g., wind-blown snow) at the pixel scale (Che et al., 2019). The cosmic-ray soil moisture observing system (COSMOS) is used to observe the snow depth and soil moisture at the pixel scale (Bogena et al., 2022). Additionally, snowmelt runoff processes are studied by fine measurements of isotopes and hydrochemical tracers. Ground-based nuclear magnetic resonance (NMR) (Kass et al., 2017) is used to detect subsurface water in permafrost at 24 measuring points and 7 measuring lines. Relevant water chemistry and isotope analyses are used to investigate the flow processes of permafrost groundwater and its transformation mechanisms with surface water. Forest hydrological processes, including precipitation, forest interception, stemflow, fall-through, and the Evapotranspiration (ET) of the understory vegetation layers, are observed in the Guantan forest field mainly by EC, evaporation dishes, sap flow monitors (thermal dissipation probes, TDPs) (Fuchs et al., 2017), rain gauges, and interception gauges that are installed according to established canopy gap fractions (Li et al., 2009). Additionally, airborne and ground-based LiDAR are used to retrieve canopy structure parameters.

Agricultural Oasis Observatory

The agricultural oasis observatory is characterized by multiscale monitoring experiments and control experiments performed to support the understanding of ecohydrological processes in agroecosystems and their interactions with agricultural activities. This observatory is mainly developed in three typical irrigation districts, including the Yingke, Daman, and Pingchuan districts, in the midstream area of the HRB. These irrigation district-scale observatories were selected to represent the agricultural oasis-riverine-wetland-desert complex system in this work. The irrigation district is mainly planted with corn, wheat, vegetables, and fruits that are irrigated mainly by the water resources from the mountainous upstream area.

Evapotranspiration is the key component of the water cycle in agroecosystems; however, the upscaling of ET is very difficult due to the strong kinetic and thermal heterogeneities of the land surface (Li et al., 2013). Thus, the Multi-Scale Observation Experiment on Evapotranspiration over heterogeneous land surfaces experiments as a component of HiWATER was conducted in the Yingke-Daman irrigation district (Liu, Li, et al., 2018; Xu et al., 2013). One superstation (Figure 4) cooperates with a dense flux observation matrix, large-aperture scintillators (LAS), a field isotope laboratory, and a soil temperature, moisture, and leaf area index (LAI) wireless sensor network that have been deployed in coordination with ground-enhanced observations, such as atmospheric soundings performed using radiosonde balloons and wind profiler radar. Additionally, several airborne remote sensing missions, including a multiangular thermal infrared imager, multispectral imager, LiDAR, and microwave radiometer, were implemented in 2012 (Li et al., 2013, 2017; Liu, Li, et al., 2018) to support the retrieval of land surface temperatures, surface reflectance, albedo, fraction of radiation absorbed by vegetation, LAI, canopy height, chlorophyll content, soil moisture, and fractional vegetation cover. The Heihe remote sensing experimental research station of the Chinese Academy of Sciences is located in this region and mainly conducts multisource remote sensing satellite calibrations and validations, develops remote sensing data assimilation systems, and finally produces ecohydrological remote sensing products that are representative of the whole basin.

Figure 4. Instrumental configuration of the grid-scale observation system in the agricultural oasis zone of the Heihe River basin. EC: eddy covariance, LAS: large-aperture scintillators, AMS: automatic meteorological stations, COSMOS: cosmic-ray soil moisture observing system, STM WSN: wireless sensor network for soil temperature and moisture, LAI: leaf area index.

In the Pingchuan irrigation district, strip samples along a transect are laid out to monitor the gradient changes in the soil moisture, soil temperature, frost depth, and groundwater level along the farmland–shelterbelt forest–desert transition zone. Both neutron moisture meter and time domain reflectometry techniques are used to investigate soil moisture in this region. A total of 17 groundwater level monitoring wells were installed. Soil hydraulic properties, such as the soil texture and saturated hydraulic conductivity, and the root distribution are also surveyed, and irrigation events are recorded (Yi et al., 2015). These observations are used to understand water exchanges, soil freezing and thawing processes, and the surface-groundwater exchange process among farmlands, shelterbelt forests, and desert landscape units. The Linze inland river basin comprehensive research station is located in this region and is equipped with agricultural water and fertilizer control experimental fields in which comprehensive observations of farmland ecosystems in desert-oasis ecotones are conducted (Li, Liu, et al., 2019).

Natural Oasis Observatory

The natural oasis observatory is characterized by multiscale oasis-desert ecosystem observations designed to support the understanding of ecohydrological processes in an extremely arid environment and to assess the ecological recovery of the terminal lakes and riparian ecosystems in this region. The climate in the area is extremely arid, and the natural oasis ecosystem has extremely fragile and sparse vegetation dominated by Populus euphratica, Tamarix, and Haloxylon ammodendron on both sides of the river. The observatory mainly covers the typical riparian forest area in the core oasis of Ejin Banner in the downstream region of the HRB (Figure 1).

Accurate estimates of the ecological water demand and ecological water consumption for the natural oasis ecosystem are the key to optimizing and assessing the integrated watershed management of the HRB, particularly the EWDP implemented since 2000. ET is a basic variable for these estimates and assessments, but monitoring ET at the regional scale is challenging. This challenge is related to the complex scale effect and the energy-water exchanges between the oasis and the desert (X. Li et al., 2013). Therefore, a multiscale flux observation system is used to capture the complex energy-water exchanges between the oasis and the desert and within the oasis (Figure 5). The system consists of LASs, ECs distributed in the poplar, tamarisk, and mixed poplar/tamarisk forests and in the agricultural fields, bare lands and deserts at the periphery of the oasis, a wireless sensor network for monitoring soil moisture, and multiple TDPs used to monitor sap flow at the single-tree or shrub scale. A series of instruments, such as Li-6400 (Li-Cor Inc), are used to investigate the physiology and photosynthesis parameters of the major desert plants. The stable isotopes technique has been used to investigate the water sources for different ages of Populus euphratica (Liu et al., 2015). Ground surveys and LiDAR measurements and airborne hyperspectral imagers and thermal imagers are utilized to retrieve the structural parameters representing the main vegetation types. Additionally, the water surface evaporation of terminal lakes is also measured by an EC tower to support water balance estimation at terminal lakes. Groundwater table monitoring is conducted along the river streams to understand changes in the groundwater. The abovementioned observations are used in the eco-hydrological models to understand the multiscale eco-hydrological processes to eventually contribute to the sustainable management of the ecosystems in the downstream areas of the HRB.

Figure 5. Schematic diagram of the observation of energy-water exchanges for the oasis-desert system in the downstream Heihe River basin. The arrows indicate the directions of air flow circulation within the oasis and between the oasis and the desert.

Watershed System Model, Data Information System, and Decision Support System in the HRB

The integrated framework that links the HRB observation system, the data information system, the watershed system model, and the scientific model-based decision support systems (DSSs) for sustainable development assessments is illustrated in Figure 6. The watershed system model couples ecohydrology with the socioeconomy (Li, Zhang, et al., 2021). Specifically, an integrated ecohydrological model was first developed by loosely coupling a geomorphology-based ecohydrological model (GBEHM) designed for the mountainous cryosphere (Gao et al., 2016; Yang, Gao, et al., 2015) with a hydrological-ecological integrated watershed-scale flow (HEIFLOW) model designed for oasis and desert areas (Han et al., 2021; Tian et al., 2018; Zheng, Tian, et al., 2020). The integrated ecohydrological model was then coupled with the socioeconomic model through two interface models including a land-use model and a water resource model. Techniques such as the surrogate modeling, the offline file transfer approach, and code-level integration were used to couple some of the models mentioned above (Li, Zhang, et al., 2021).

Figure 6. Framework designed for the integrated study of the Heihe River basin.

The HRB observation system provides a large number of multidisciplinary and multiscale data sets. The HRB data information system was developed for collecting, managing, and sharing these data sets (X. Li et al., 2010). In this system, the HRB observation data sets are collected through an automatic submission system and are subjected to rigorous data quality control in a variety of ways, including the adoption of uniform observation specifications, data standardization, and data quality evaluation (Guo & Liu, 2015; Guo et al., 2021; Wang et al., 2014). The metadata were reviewed and revised in several rounds and released at the National Tibetan Plateau Data Center (TPDC) (http://data.tpdc.ac.cn/en/special/heihe/) to provide data services. All the HRB data sets are free and open to the public. The TPDC provides a cloud-based platform with integrated online data collection, quality control, analysis and visualization capabilities to maximize the efficiency of data sharing (Pan, Guo, et al., 2021). The data topics of the HRB are further divided into HiWATER (Li et al., 2013, 2022), WATER (Li et al., 2009), surface process observation network (Liu, Li, et al., 2018), and other HRB data based on the different experimental stages of data acquisition for the convenience of data users.

The growing amount of ecohydrological observational data supports the development of models in the HRB. In the most recent phase, a watershed system model that couples GBEHM and HEIFLOW with socioeconomic models to better represent the integrated human-nature system in the HRB was developed and applied to simulate a variety of ecohydrological processes at different spatial scales and to investigate the complex interactions among water, ecology, and socioeconomics in the HRB. A comprehensive review of the development and application of the watershed system model in the HRB can be found in our recent publication (Li, Zhang, et al., 2021).

The watershed system model and its components are still under development. The new developments focus on introducing more human processes in the watershed system. (a) The GBEHM has been improved so that the model is fully adaptable to remote sensing inputs and has been successfully used in watersheds in cold regions, such as the river basins of the Tibetan Plateau (Shi et al., 2020, 2022; Zheng, Yang, et al., 2020). (b) The subgrid simulations have been improved to better represent sparse vegetation in arid regions and heterogeneous irrigation in agricultural areas (Han et al., 2021; Zhang, Tian, et al., 2021). (c) An agent-based model (ABM) representing farmers' water use has been coupled with models to fully integrate water policies, water uses and hydrological processes to enhance the microscale description of human-water interactions (Du et al., 2020, 2022; Yuan et al., 2021). (d) New modules have been developed in existing models to simulate high-resolution irrigation activities, water-saving techniques and water management policies (Zheng, Tian, et al., 2020).

To serve watershed management and sustainability assessments, a river basin sustainable development DSS was developed by taking scientific models (i.e., the HRB watershed system model) as the backbone and using the surrogate modeling technique (Ge et al., 2013, 2018, 2022). Surrogate models refer to the replacement of complex physics-based models with data-driven relationships between multiple explanatory variables and model outputs (Razavi et al., 2012). This approach can greatly improve the computational efficiency of watershed system models without sacrificing the simulation accuracy (Li, Zhang, et al., 2021). A great number of climate and socioeconomic scenarios were developed by localizing shared socioeconomic pathways using land-use, socioecological, and regional climate models (Ge et al., 2022; Hu et al., 2022; Pan et al., 2012; Xiong & Yan, 2013). Based on these scenarios, the DSS has been successfully applied to assess sustainable development in the HRB and analyze trade-offs between sustainable development goal indicators (Ge et al., 2022). The existing model development work in the HRB has provided a solid foundation and impetus for future efforts along these directions.

New Insights and Findings on the Watershed and CZ Processes of the HRB

Through systematic observations and model simulations, we can improve our understanding of major watershed and CZ processes in the typical endorheic basin (i.e., the HRB), including cryospheric hydrological processes, ecological and hydrological interactions in arid and cold regions, and complex surface-groundwater interactions. We can also address the complex and puzzling issues that have remained unresolved for a long time, including the multiscale water balance closure issue, the surface energy disclosure problem, and water use efficiency (WUE). These new findings and insights are important for understanding the watershed and CZ processes of cold and arid regions in endorheic river basins and promoting sustainable river basin development.

Cryospheric Hydrological Processes

Cryospheric processes exert significant impacts on the hydrological processes in the alpine region of the HRB. We found that the cryospheric hydrological processes in this region are experiencing rapid changes that could lead to cascading effects in the watershed system. Small mountainous glaciers have shrunk in recent decades mainly due to climate warming and are likely to disappear by the mid-21st century (R. Chen et al., 2018; Cheng et al., 2020). By the end of 2040, more than 45% of the glacial area will be lost, according to projections performed under established climate change scenarios (Yang et al., 2020). In the HRB, glacial melt contributes less than 5% to the total streamflow (R. Chen et al., 2018; Gao et al., 2018). Nevertheless, glaciers are able to supply water resources and regulate river flow by reducing peak flow and replenishing insufficient flows; thus, their continuous shrinking requires attention.

We found that snowmelt can be retained in soil voids, even on snow-free days, and can therefore continuously contribute to hydrological processes throughout the year (Li, Li, et al., 2019). At high elevations in the upstream HRB region, wind-blown snow is evident and can enhance snow accumulation due to the unsaturated moisture near the snow surface induced by the vertical diffusion of water vapor (Dai & Huang, 2014; Huang & Shi, 2017). Snowmelt and glacial meltwater infiltrate into the subsurface and can feed river streamflow in the form of springs across high-elevation subbasins (Z. Li et al., 2014). In addition, snowmelt and glacial melt from the Qilian Mountains are important contributors to groundwater in the middle and downstream areas of the HRB (Zhao et al., 2018).

Regarding permafrost, both the permafrost and seasonally frozen ground in the upstream HRB area have been subjected to warming, exhibiting clear ground temperature increases (Cao et al., 2018). Based on long-term hydrological simulations, approximately 8.8% of the permafrost was found to have degraded into seasonally frozen ground in the upstream HRB from 1971 to 2013 (Gao et al., 2018; Qin et al., 2016); this permafrost was mainly distributed in areas between 3,500 and 3,900 m above sea level. Permafrost was also found to play a critical role in controlling groundwater-streamflow interactions and groundwater flow in the upstream area of the HRB. In the piedmont plain, groundwater is recharged by lateral inflows from the permafrost zone and stream infiltration, while discharges to streams occur as baseflow (Ma et al., 2017, 2021). Rivers and soils freeze during the cold season, resulting in an increased groundwater head and a decreased hydraulic gradient between groundwater and rivers, thereby reducing the amount of groundwater discharged to rivers (Ma et al., 2017). Over the upstream region of the HRB, permafrost degradation increased the infiltration rate of the soil, resulting in the weakening or even loss of the water-blocking effect of the permafrost layer, thus increasing groundwater recharge and winter runoff (Gao et al., 2018). The permafrost area is projected to decrease by 23% from 2011 to 2060; the maximum freezing depth of seasonal permafrost will decrease by 5.4 cm/decade; and the active layer depth of permafrost will increase by 6.1 cm/decade under the representative concentration pathway 4.5 emission scenario (Wang et al., 2018).

Ecohydrological Processes in the HRB

With the continuous development of the HRB observation system and models, our understanding of ecohydrological processes in alpine forests, agricultural lands, and riparian forested areas has improved considerably. In the upstream mountainous area of the HRB, forest-area precipitation has shown an increasing trend in recent decades, but this increase is not sufficient to increase water production in the Qinghai spruce forests due to increased transpiration (Chang et al., 2017). In these Qinghai spruce forests, the well-established exponential decay relationship between the total precipitation and interception percentage following canopy saturation is supported by observations obtained by an automatic throughfall-collecting system; however, the interception percentage change is almost independent of the total precipitation change until the forest reaches canopy saturation (Peng et al., 2014). The canopy structure morphology exerts a complex effect on the stemflow of alpine shrubs in the Qilian Mountains, and the height and the projected area of the canopy are important factors that influence stemflow under a constant rainfall intensity (Liu, Chen, & Song, 2011).

In the midstream agricultural oasis of the HRB, significant changes have been observed in hydrological conditions due to agricultural development and the implementation of the EWDP, which aims to ensure a minimum flow from the midstream reach to the downstream reach (Cheng et al., 2014). Studies have indicated that three major water-exchange mechanisms occur among farmland, forests, and deserts: (a) water in surface soils is transferred from irrigated to nonirrigated areas due to the soil water potential gradient; (b) groundwater is transferred from irrigated to nonirrigated lands due to the groundwater level gradient; and (c) the soil water in farmlands is partially used by forests through their extended root systems (Yi et al., 2015). The irrigation depth far exceeds the water uptake depth of crops, leading to a low irrigation efficiency (Yang, Wen, & Sun, 2015). Only 53% of the total applied irrigation water was effectively utilized through evapotranspiration (Xu, Jiang, et al., 2019), according to the simulations of the agro-hydrological model. Moreover, numerical experiments performed with an agro-hydrological model (SWAP-EPIC) have indicated that more than 15% of irrigation water can be saved without negatively affecting crop yields by improving water conveyance systems and irrigation scheduling and by optimizing water allocation and cropping patterns (Jiang et al., 2015, 2016). Vegetables have been found to have the highest water use efficiencies (∼3.0 kg m⁻³), while spring wheat reportedly has the lowest WUE (∼1.5 kg m⁻³) (Li et al., 2016). The water consumed by crops has increased significantly in recent decades, mainly due to the expansion of farmland and changes in the planting structure, rather than climate change (Zou et al., 2017).

In the downstream area of the HRB, the groundwater table controls the changes and compositions of plant communities (Yu & Wang, 2012). The vegetation in this region has recovered considerably in the last two decades due to the EWDP. However, a streamflow threshold of 11.5 × 10⁸ m³/year exists for the downstream region, beyond which the contribution of the streamflow to vegetation recovery is trivial in the short term (Sun et al., 2018). Moreover, the most appropriate restoration area of East Juyan Lake (a terminal lake) has been identified to be 42 km², and at this scale, the lake has the highest water storage efficiency with relatively low evaporation loss (M. Zhang et al., 2019). In general, plants in this region have deep root systems that ensure their survival even when the groundwater table is at a considerable depth (Y. Zhang et al., 2011). Desert vegetation has been found to utilize water vapor under unsaturated conditions using isotope and fluorescence tracing techniques (Wang et al., 2016). The leaves of desert vegetation can absorb water from the atmosphere to relieve drought stress (Wang et al., 2016). Desert vegetation responds to water stress mainly in two ways: (a) by maintaining a low water potential by dissipating water and increasing the water potential gradient difference between the soil and plants (the open-source approach) and (b) by reducing water dissipation by regulating stomatal conductance (the throttling approach) (C. Zhang et al., 2017). Desert vegetation has a strong dependence on groundwater, and different species of trees, shrubs and herbs in desert environments absorb water at different depths (Ding et al., 2017; Li, Tong, et al., 2019).

Groundwater System

In endorheic basins that spatially concur with arid climates, groundwater critically influences the interrelated ecological and hydrological processes (Wheater et al., 2010). However, our understanding of groundwater systems and their roles in ecohydrological systems remains imperfect, as observing and modeling complex groundwater processes in arid regions is a great challenge. The HRB is characterized by frequent river-groundwater flux exchanges; and groundwater plays a critical role in maintaining the health of ecosystems in the basin (Yao et al., 2018). Groundwater is, to a large extent, recharged by surface water from upstream areas and is then discharged into rivers in downstream areas.

In the upstream area of the HRB, approximately 19% of precipitation reaches the groundwater table as groundwater recharge, of which approximately 65% discharges to mountain streams as baseflow, while the remaining 35% directly discharges to downstream aquifers (Yao et al., 2017). The aquifer system in the pluvial fans in the piedmont region of the upstream Qilian Mountains has a thick unsaturated zone and serves as a natural reservoir that stores water from stream leakage (Yao et al., 2018). The groundwater discharges from the aquifer system to the main channel of the HRB in the form of springs when the river incises aquifers and the river stage is below the groundwater head.

In the midstream area of the HRB, groundwater is an important source of water for irrigation. The discharge of groundwater to the total streamflow has been estimated to be as high as 28% in this region by using river surface temperature data obtained from airborne thermal infrared remote sensing technology (Liu, Liu, et al., 2016; Liu & Zheng, 2015). The overexploitation of groundwater for irrigation has led to declines in the groundwater level of up to 2 m in some groundwater-fed irrigated areas in the midstream HRB (Liu, Jiang, et al., 2018; Niu et al., 2019). The upper boundary of the groundwater discharge area in the midstream HRB has shifted downstream as more groundwater has been pumped out due to the restrictions of the EWDP on surface water use and the expansion of irrigated cropland (Li, Cheng, Ge, et al., 2018). The groundwater level in the midstream area of the HRB is projected to decline in the future due to the enhanced ET resulting from the warming climate, the decreased lateral groundwater inflow, and the increased usage of groundwater for agricultural irrigation (Cheng et al., 2020).

In the downstream area of the HRB, groundwater is a major factor affecting the distribution and health of riparian forests. In this area, the groundwater inflowing to the downstream region was sourced from the midstream area and the surrounding lower mountainous areas. Moreover, it was recently determined that both the Badain Jaran Desert and the downstream area of the HRB belong to the Yingen-Ejinan Banner Mesozoic basin (Wang & Zhou, 2018; P. Wang et al., 2013); this was found based on new evidence from systematic surveys and numerical modeling.

Multiscale Water Balance

Closure of the water balance at different spatial scales is essential for understanding the hydrological cycle and water resource management at the basin scale (Jensen & Refsgaard, 2018). Multiple integrated observations obtained by the HRB observation system were used together with the watershed system model (Li, Zhang, et al., 2021) to close the water balances at the basin, subbasin, landscape, channel, and irrigation district scales in the HRB and to analyze the associated changes in the hydrological cycle (Li, Cheng, Ge, et al., 2018). The hydrological cycle in the HRB has distinct characteristics: (a) vertical zonality is obvious in the upstream mountainous area, and cryospheric hydrological processes play an important role in runoff generation; (b) intensive interactions occur between surface water and groundwater in the midstream oasis area, and human activities such as irrigation significantly affect these hydrological processes; and (c) the potential ET is very large in this region, but precipitation is scarce, and most of the available water is consumed by ecosystems in the extremely arid downstream area.

The streamflow from the upstream mountainous area of the HRB exhibited low interannual variabilities due to the significant contribution of meltwater from snow, frozen ground, and glaciers. The climate in Northwest China has shifted from “warm-dry” to “warm-wet,” and the transition is predicted to last for a long time into the future (Shi et al., 2003). Due to the increased precipitation, snowmelt and glacial meltwater under the warm-wet climate, the streamflow from the upstream mountainous area of the HRB has shown a significant increasing trend over the past few decades. This increase in streamflow has appeared to benefit agricultural development and ecological conservation in the middle and lower reaches of the basin. However, hydrological projections indicate that streamflow from the upstream mountainous area of the HRB are likely to decrease in the future due to a higher ET under a warmer climate and greater groundwater infiltration resulting from the shifts of permafrost to seasonal frozen ground (Li, Zhang, et al., 2021; Y. Wang et al., 2018; L. Zhang et al., 2015). The possible reduction in streamflow from the upstream mountainous area may increase the agricultural and ecological risks of the HRB in the future.

The midstream area is the major agricultural production zone of the HRB and is characterized by intensive human activity footprints (i.e., river water diversions and groundwater withdrawals for irrigation). Despite the increasing inflows from the upstream area, the groundwater storage has shown an apparent downward trend in the midstream area of the HRB, mainly due to groundwater overuse resulting from the expansion of croplands and the restriction of surface water usage since 2000, when the EWDP was implemented. In the future, the precipitation is projected to have an insignificant trend, while the temperature will increase significantly in the midstream area of the HRB, which will increase the ET by up to 0.49 × 10⁸ m³ from 2021 to 2050 under the PCP8.5 emission scenario (Zou et al., 2020).

The downstream region is characterized by an extremely arid environment, and the vast majority of the land area is covered by desert. Most of the available water (>70%) is used to sustain natural oases and is lost to the atmosphere via ET in desert regions (Xu et al., 2020, 2021). The ecosystems in the downstream region of the HRB and the terminal lake areas have been substantially restored due to the implementation of the EWDP since 2000. Nevertheless, the trade-offs among water, ecosystems, and agricultural production in the HRB are stronger now than ever before, especially in the midstream areas. For example, expanding irrigated cropland can increase agricultural production, but it can lower the groundwater table and affect the growth of vegetation in downstream areas. In contrast, if our goal is to restore groundwater storage in midstream areas, irrigation withdrawals should be reduced; however, this is detrimental to agricultural production. The trade-offs among water, ecosystems and agriculture are highly sensitive to water management strategies (Sun et al., 2018).

Closing the Surface Energy Balance

At the top of the CZ, the surface flux exchanges (i.e., the energy, water, and carbon fluxes) that occur across the interface between the land and atmosphere influence not only the weather and climate but also the chemical and thermal energy gradients. Natural and anthropogenic disturbances (e.g., climate change, erosion, contaminant accumulation, land use change, spills) can produce measurable perturbations at temporal (“hot moment”) or spatial (“hot spot”) scales, push systems out of stationarity, produce nonanalog ecosystems, and dramatically alter surface fluxes and CZ processes. Therefore, quantifying the surface fluxes is important for studying the CZ processes. Currently, the eddy-covariance method provides the most direct estimates for the surface fluxes between ecosystems and the atmosphere and is the most direct and common method used to measure heat, water vapor, and CO₂ fluxes at the top of CZs (i.e., the interface between the biosphere and the atmosphere) around the world (e.g., FLUXNET), the observation data of the EC method is widely used in the Earth's system (Baldocchi et al., 2001). However, this method suffers from the surface energy disclosure problem (Foken, 2008; Mauder et al., 2020), which may affect the reliability of studies based on EC observations. Considering that the surface energy balance is an essential cornerstone of the Earth's system, the surface energy disclosure problem has a significant influence on the calibration and validation of Earth system models and is one of the biggest challenges when employing the EC measurement technique (Mauder et al., 2020).

Based on remote sensing data and the flux observation matrix built using 22 EC systems belonging to the HiWATER project, Xu et al. (2017) found that heterogeneous land surfaces critically affect flux disclosure. In addition, the flux disclosure ratio also decreased when irrigation occurred and increased after irrigation was completed. The relatively large sensible heat fluxes measured by the LAS compared to the EC-measured values indirectly suggest that large-scale turbulent eddies, which may be caused by natural and human disturbances such as land use changes and irrigation, contribute greatly to the surface energy disclosure (Liu, Xu, et al., 2011; Mauder et al., 2020).

Therefore, inadequate sampling of large-scale turbulent eddies (i.e., the mesoscale flux transport) has been increasingly acknowledged as a leading contributor to surface energy disclosure (Foken et al., 2011). Based on large eddy simulations and multiscale data collected by the HRB observation system, we found that the ratio of the boundary-layer height to the Obukhov length (z_i/L), the integral length scale of the vertical velocity (l_w), the mean horizontal velocity (U), and the average interval (T) control the energy closure ratio (EBR) (Zhou and Li, 2019; Zhou et al., 2018). Among these four variables, l_w determines the size of the turbulent coherent structures (i.e., the large eddies; see Figure 7), whereas z_i/L affects the form of these large eddies (i.e., the flux distribution over different eddies). Moreover, U and T determine the number of these large eddies that can be sampled by EC sensors over a finite averaging period. The relationships among these four variables are sketched in Figure 7. Based on the above understanding, a diagnostic equation for I is proposed as follows: EBR = [az/z_i + b] [–K × z_i/L × l_w/UT + C], where a, b, K, and C are empirical constants and b should, in theory, be equal to one. Compared to previously established relationships and equations, the above equation is more physically based and more easily understood, thus laying a foundation for energy closure corrections in the future.

Figure 7. Sketch of the physical meaning of the proposed equation. The energy closure ratio, zi/L, lw, U, and T denote the energy closure ratio, the ratio of the boundary-layer height to the Obukhov length, the integral length scale of the vertical velocity, the mean horizontal velocity, and the average interval, respectively. The inset represents the effects of zi/L on the cospectrum.

Water Use Efficiency

Water use efficiency is an important physiological indicator that couples the carbon and water cycles (Wang, Li, et al., 2021) and is a key ecohydrological function in CZ systems. Investigating the patterns and drivers of WUE at different scales is important for understanding the evolution and function of cold and arid region CZs in endorheic basins. The HRB observation system provide an excellent opportunity to quantify spatiotemporal WUE patterns and their underlying mechanisms in endorheic basin CZs.

Studies have shown that alpine meadow ecosystems are strong CO₂ sinks, and water and carbon dioxide exchanges are energy-limited (Sun et al., 2019; Wang, Xiao, et al., 2021). The high carbon sinks observed in wet alpine meadows are mainly the result of the inhibiting effect of nearly saturated soil conditions on soil respiration rather than the effect of low temperatures, indicating that warming-induced cryosphere changes may affect carbon dynamics in alpine ecosystems (Sun et al., 2021). Divergent variations in ecosystem and canopy WUE have been observed across alpine desert-oasis ecosystems along climatic gradients in the HRB (H. Wang et al., 2019, 2021a). Stronger relationships were observed between precipitation and WUE at natural sites than at oasis sites in the HRB. This kind of water exchange between irrigation water and groundwater greatly affects the availability of water in oases. Therefore, ET is a better proxy of water availability than other indicators, as ET drives the spatial patterns of carbon fluxes in the HRB. Additionally, agricultural management techniques, such as drip irrigation with filmed mulching, significantly reduce water consumption and increase the WUE of cropland ecosystems. These findings can provide references for water resource management in endorheic basins. The contrasting WUE patterns observed between natural and oasis systems highlight the importance of human activities in influencing the redistribution of water resources in dryland CZs.

Advancing Watershed Science to Address Complexity and Dynamics in the CZ

The biological, physical, and chemical processes in CZs are very complex and are closely linked to human activities. CZ science has provided an excellent methodological framework for studying complex and interactive CZ processes. CZ science and watershed science can complement each other, and based on the practice of watershed science in the HRB, we believe that the complexity of CZs can be better addressed by enhancing and improving observation and modeling techniques and by integrating observations and simulations.

The HRB observation system has made considerable achievements in observing the CZ and ecohydrological processes using new techniques such as the Internet of Things (Zhang & Li, 2020), fluorescence remote sensing (Liu et al., 2020), isotope observation technology (Ma et al., 2017; X. Wang et al., 2016), and subsurface observation technologies (e.g., NMR and GPR). However, the HRB observation system mainly focuses on surface ecohydrological processes and does not cover the entire CZ, which spans vertically from the top of the vegetation layer to the bottom of the aquifer layer. In the future, the HRB observation system can further utilize other new techniques to enhance the observation of the watershed and the CZ (Figure 8). For instance, subsurface remote sensing is a useful tool for revealing mountain cryospheric processes (Banda et al., 2016; R. H. Chen et al., 2019; Taylor et al., 2021). Social sensing techniques such as the mobile phone location and social media (e.g., Twitter, Microblog, and others) can support the development of ABMs based on geospatial big data to achieve a better understanding of human-nature interactions. Moreover, learning from CZ science, watershed science should give more attention to the observations of subsurface processes, biogeochemical and geomorphological processes, and slow and long-term cumulative processes to achieve a deeper understanding of the complexity and dynamics of watershed-scale CZ processes. In addition, more thematic experiments are needed to address some long-standing unresolved problems (Figure 8), such as the near-surface energy closure problem, scaling problem (i.e., the great range of spatiotemporal scales of processes in the watershed and the CZ) (Cheng & Li, 2015), anomalous carbon sink problem in deserts, underestimated solid precipitation (R. Chen et al., 2018), and underrepresented human activities. The HRB observation system is a very representative basin-scale observatory in the endorheic regions of China, and therefore, it should be a part of the Critical Zone Collaborative Network and contribute to understanding CZ processes across different landscapes and climate zones.

Figure 8. Vision of the development of future observation systems for linking watershed science with critical zone science.

Numerical modeling is an important tool for understanding and managing complex CZ and watershed ecosystems. Although significant advances have been made with HRB models, they are still deficient in modeling geomorphological processes (Pan, Cai, & Geng, 2021), lateral aquatic carbon transport processes (Song & Wang, 2021), and mineral-organic interactions in the CZ. ABMs provide flexible methodologies for representing human behaviors and their interconnections and have been successfully integrated within the watershed system model to represent human activities and their interactions with the natural system (Du et al., 2020, 2022; Hung & Yang, 2021; Yuan et al., 2021). CZ science can learn from the experience of watershed science in this area to improve the understanding of the impact of human activities on CZ processes. Nevertheless, a deep integration of natural processes and human dynamics is still needed for both watershed science and CZ science to enhance the ability to model the entire water-soil-air-plant-human nexus within the watershed (Li, Zhang, et al., 2021). The integrated modeling of CZ processes that are horizontally bounded by watershed limits is not an easy task from both scientific and technical perspectives. Multidisciplinary researchers need to collaborate to design and develop such integrated models by taking advantage of developments and technological advances in different disciplines (e.g., machine learning [ML] and artificial intelligence, cloud data storage and computing capabilities, and distributed model development platforms) (Hubbard et al., 2020; Li, Cheng, Lin, et al., 2018).

Observations and models have contributed to our ability to understand, predict, and manage watersheds and CZs (Figure 9), but neither of these approaches is perfect. Data assimilation enables the harmonious integration of observations and models and can facilitate the optimal design of observational systems and models (X. Li et al., 2020). However, data assimilation has rarely been used to fuse CZ models with observations, especially those related to geochemical cycles (Sullivan et al., 2017). CZ observatories collect and provide abundant observations relevant to different disciplines. In the future, multiscale and multivariate data assimilation methods need to be further developed and improved to utilize CZ observations to the maximum extent. Moreover, observation operators need to be specifically designed for new CZ observations, such as observations of isotopes, nutrients, and carbon. Furthermore, observations of human activities, such as water diversion and pumping activities, need to be assimilated into physical models to reduce the associated uncertainties and to enhance the simulations and forecasts of CZ processes and watershed systems. In addition to observations and model simulations, a variety of other information, such as expert knowledge and public opinion, needs to be integrated in next generation DSSs to improve the management of watersheds and CZs.

Figure 9. Fusion of model simulations with multidimensional observations from critical zone (CZ) observatories through data assimilation to improve our understanding of CZ processes.

Summary

Deep integrations of knowledge, techniques and expertise from multiple fields are promoted in both CZ science and watershed science. Linking CZ science with watershed science allows for a better understanding and interpretation of all relevant natural and anthropogenic processes spanning CZs within a watershed, thus helping us better characterize, predict, and manage the Earth system. We provide a systematic introduction to a river basin-scale CZ observatory in northwest China, that is, the HRB observation system. The HRB observation system consists of three representative observation areas: the mountain cryosphere area, the agricultural oasis area, and the natural oasis area; these regions are coordinated with field experiments, long-term operational observations, and existing research stations covering the entire HRB and with airborne and satellite remote sensing to achieve multiscale observations ranging from the point scale and subbasin/irrigation district scale to the basin scale.

The HRB observation system provides a large number of multidisciplinary and multiscale data sets that are fully shared with the scientific community via data information systems. These abundant observations support the development of ecohydrological, agro-hydrological, and regional climate models, eventually leading to the development of a watershed system model representing the HRB. The watershed system model will be further developed to incorporate additional human processes. To date, these models have been embedded in DSSs to build bridges between science and decision-making and to better serve watershed management practices and sustainability assessments.

With the support of the HRB observation system and model platform, significant progress and breakthroughs have been made in CZ and watershed process research in the HRB. The major new insights and findings are described as follows: (a) the cryospheric processes including glacial processes, snow accumulation and melt, and permafrost thawing are experiencing significant changes due to the regional warm-wet transition exerting critical impacts on the hydrological processes in the upstream alpine region; (b) sparse alpine vegetation and alpine meadows are the major runoff-contributing areas in the upstream HRB; (c) farmlands, forests, and deserts exchange water through soil and groundwater transfers and through extended vegetation root systems; (d) water consumption by crops has increased significantly in the midstream HRB, mainly due to the expansion of croplands and changes in the planting structure leading to continuous decreases in groundwater storage; (e) the vegetation and terminal lakes in the downstream HRB region have recovered greatly due to increased runoff from the midstream area following the implementation of the EWDP; (f) the outflow threshold from the midstream area and the appropriate area of East Juyan Lake were determined to be 11.5 × 10⁸ m³/year and 42 km², respectively; (g) the HRB experiences frequent and dynamic surface water-groundwater interactions, which play critical roles in maintaining the health of the ecosystems in the basin; (h) strict surface water use restrictions under ecological protection programs and the significant expansion of irrigation have led to a steady decline in groundwater storage in the midstream area of the HRB; (i) desert vegetations have been shown to be able to utilize water vapor under unsaturated conditions and can respond to water stress with both open-source and throttling approaches; and (j) the water availability, rather than the temperature, dominates the spatial variabilities in the carbon and water fluxes in dryland CZs.

From a future perspective, we propose that watershed science be advanced to address the complexity and dynamics of CZs via three possible avenues. First, more comprehensive observations of surface and subsurface processes, surface-atmosphere interactions, ecohydrological and geochemical processes, and natural and socioeconomic processes can be obtained by using new observation technologies and integrated observation approaches. Second, human activities and their interactions with all physical, chemical, and biological processes spanning CZs can be explicitly simulated with improved accuracies by taking advantage of developments in different disciplines and technological advances (e.g., ABMs, data assimilation, ML, and artificial intelligence). Finally, multisource information, including observations, model simulations, expert knowledge, and public opinions, should be synthesized to advance the optimal management and sustainable development of river basins.

Acknowledgments

This work is supported by the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant XDA20100104) and the National Natural Science Foundation of China (Grant 41988101).

Data Availability Statement

This is a review paper and data were not used nor produced for this research.