Listen

Accurate simulation of the snowmelt runoff process is of great significance in understanding the evolution of water resources in high-altitude cold regions and achieving efficient utilization of water resources. This study focuses on the source regions of the Yellow River basin (SYRB) and aims to improve the snow identification and snowmelt simulation methods in the WEP-L hydrological model. The results show a significant decrease in the snowfall ratio from 2002 to 2018. The fraction of snow cover decreased at lower altitudes but increased at higher altitudes, displaying an exponential relationship with negative accumulated temperature. Snowmelt was found to be negatively correlated with snowfall and snow cover, with a stronger negative correlation at higher altitudes. The decrease in the snowfall ratio intensified with increasing elevation, while snow cover increased with elevation. However, the overall trend of snowmelt runoff was not significant. These findings highlight the dynamic relationship between snowfall, snow cover, and temperature in the SYRB. By incorporating the established response function, the accuracy of snow identification and snowmelt simulation in the WEP-L model has been enhanced. This study contributes to a better understanding of water resource evolution and the efficient utilization of water resources in high-altitude cold regions.

Listen

  • This research enhances understanding of water resource dynamics in high-altitude cold regions.

  • It highlights the complex nature of snowmelt runoff processes in high-altitude cold regions.

  • The study contributes to the development of more accurate models and strategies for water resource management in high-altitude cold regions.

elevation-dependent effects, runoff variations, snow cover, snowfall, the source regions of the Yellow River basin
Listen

The utilization of meltwater resources, particularly snow and glaciers, plays a pivotal role in various human activities (IPCC 2013). The annual and seasonal variation in the snow cover area due to its response to the climatic variables directly influences the water supplies. Climate change has received much attention from countries around the world because of its impact on snow cover and hydrological conditions (Pachauri et al. 2014). Through a water balance model study, Mccabe & Wolock (2010) pointed out that since 1970, the rise in winter temperature in the Northern Hemisphere has led to a decrease in snowfall and an increase in snowmelt, thus reducing the amount of snow cover and the area covered by snow. The Qinghai–Tibet Plateau is the highest plateau in the world, and the birthplace of many important rivers, with abundant glaciers, snow, permafrost, rivers, lakes, and groundwater, which has an important and profound impact on the global climate (Immerzeel & Bierkens 2012). Snow is one of the important freshwater resources in alpine mountains. Since 1980, the area of snow on the Tibetan Plateau has been decreasing, especially after 2000, with the snow area significantly less than the previous period of the 1990s, and the snow area decreased by more than 50% (Che et al. 2019). The climate change has a certain effect on the snow water resources on the Qinghai–Tibet Plateau, so it also puts forward higher requirements for the exploitation and utilization of water resources. The Qinghai–Tibet Plateau is one of the region's most dependent on snowmelt runoff in the world. In all the river basins that depend on snowmelt runoff, the proportion of irrigation water that depends on snowmelt has decreased due to climate warming, which has caused a water resources crisis. Up to 40% of irrigation needs must obtain alternative water sources or greatly improve the efficiency of water use; otherwise, crop yield may be affected. This will exacerbate food security problems (Mankin et al. 2015; Qin et al. 2020). In the past 30 years, snowmelt in the Yangtze River, Yellow River, Yarlung Zangbo River, and other rivers has increased to varying degrees. However, continuous warming in the future will lead to less snowmelt, which will exacerbate the risk of shortage of snow water resources. It is urgent to clarify the evolution law of snow water resources in the basin, so as to assess the impact of climate change on the development and utilization of water resources (Wu et al. 2004).

Climate warming affects the process of water circulation in cold regions, so it is of great significance to construct hydrological models in cold regions for the study of water resources management and climate change. At present, there are two types of models for simulating snowmelt runoff. One is the model based on the degree-day factor, which requires few parameters and is suitable for areas lacking data, such as the SRM model and the HBV model. The other is the physical model based on energy-balance, which has high accuracy, but requires detailed snow observation data and relatively complex parameters, mainly including the SWAT model, the VIC model, and the WEP model. Bookhagen & Burbank (2010) took the Himalayan region as an example and used the SRM (snowmelt runoff model) model to study and pointed out that in the western region with higher elevation, precipitation mainly occurred in the form of snow, accumulating a large amount of snow, and snowmelt runoff accounted for more than 60% of the total runoff, while in the eastern region, as the altitude decreased, with the increase of average temperature, the accumulation of snow gradually decreases, and the amount of snow that can be melted also decreases accordingly, the proportion of snowmelt runoff is generally less than 10%. The spatial differences of snowfall and snow cover have a significant impact on the runoff process. Yu et al. (2013) integrated the FASST (Fast All-season Soil Strength model) model into the SWAT model, taking the Heihe Mountain basin as the research area, and pointed out that the snow melting speed in spring was slow, and the confluence process and infiltration time of snowmelt runoff were long, which replenished the soil flow and groundwater. In summer, snowmelt accelerated and its contribution to the river channel increased significantly. Cuo et al. (2013) took the source regions of the Yellow River (SYRB), the source regions of the Yangtze River, the source regions of the Lancang River, the Nu River, the Brahmaputra, and the Indus River in the Qinghai–Tibet Plateau as the research area, and evaluated the application effect of the VIC model in these six river basins from 1963 to 2005. The research results showed that precipitation plays a leading role in the runoff variation of the Yellow River source regions, the Yangtze River source regions, the Lancang River source regions, the Nu River, and the Brahmaputra basin, contributing 65–78% to the annual runoff of each basin. Except for the Indus River, the runoff caused by glacier melting contributes less than 7% to all the other river basins. WEP and its improved models (e.g. WEP-L, IWHR-WEP, WEP-COR, WEP-QTP, etc.) have been applied to the small, medium, and large scales of several river basins in Japan (Tamagawa River basin, Hairao River basin, Tanida River basin, etc.), Korea (Cheonggyecheon River basin, Seoul), and China (such as Haihe River basin, Yellow River basin, Yangtze River basin, and Heihe River basin). In addition, it has been applied and verified under complex underlying surface conditions and achieved good simulation results (Liu et al. 2020). Taking the Nu River Basin as a research area, Yang et al. (2021) developed a distributed hydrological model (WEP-C) to study the variation of runoff from 1979 to 2019. The results showed that the total discharge increased at a rate of 1.36 mm/a, while the snowmelt runoff decreased at a rate of −0.53 mm/a. The WEP-L model, a development of the WEP model, is a strongly non-linear model. Based on the physical process of the hydrological cycle, the model can finely describe the changes of evapotranspiration, underground runoff, surface runoff, water-heat exchange, and other water cycle elements (Jia et al. 2006b). The WEP-L model is mainly used in the Yellow River, Haihe River, and other related study (Jia et al. 2006b 2012), applied in the Qinghai–Tibet Plateau basin is less, and the division of rain and snow is simple, that is when the temperature is less than 1 °C to snow, is greater than 1 °C for rainfall, at the same time is also lack for the change on the Qinghai–Tibet Plateau basin of snowfall and snow cover on the research of the runoff. In addition, different water cycle elements have mutual influence. Current studies mainly focus on the runoff effect of single factor change of precipitation or snow cover, and there are few studies on the comprehensive effect of snowfall, snow cover, and snowmelt. However, the models widely used at present are generally insufficient to describe the difference between temperature and snowmelt at different elevations. Therefore, this study coupled the vertical response relationship between temperature and snowmelt into the model to study.

This study focused on improving the snowmelt simulation in the WEP-L model for the SYRB. By integrating the temperature–snowmelt relationship in different elevation bands, the accuracy of the simulation was enhanced. The evaluation of the improved flow simulation technology confirmed its practical applicability and its importance for water resources management in high-altitude regions.

Listen

Study area

Listen
This study selects the SYRB located in the northeast of the Qinghai–Tibet Plateau as the research object, as shown in Figure 1. The hydrological and geographical conditions of the SYRB are characterized by high-altitude terrain, significant snowmelt contribution, and complex runoff processes. The average altitude of the SYRB is 4,079 m, and the basin area is 13.1 × 104km2. The average temperature and average precipitation of the SYRB from 2002 to 2018 are −1.01 °C and 596.3 mm, respectively. According to the elevation range of the research area, eight elevation bands are divided with 500 m as the interval, namely band1 ≤ 3,000 m, 3,000 m < band2 ≤ 3,500 m, 3,500 m < band3 ≤ 4,000 m, 4,000 m < band4 ≤ 4,500 m, 4,500 m < band5 ≤ 5,000 m, 5,000 m < band6 ≤ 5,500 m, 5,500 m < band7 ≤ 6,000 m, and band8 > 6,000 m.
Figure 1
Location of the SYRB.
View largeDownload slide

Location of the SYRB.

Figure 1
Location of the SYRB.
View largeDownload slide

Location of the SYRB.

Close modal

Methodology

Listen

WEP-L model

Listen
The WEP-L model, a distributed hydrological model, serves to simulate the transport of water and heat flux within a watershed. Its notable features encompass: (1) simultaneous consideration of hydrological processes and energy transfer; (2) integration of the Mosaic method for land use evaluation within computation units (Avissar & Pielke 1989); (3) incorporation of runoff generation theory, which accounts for the influence of terrain on runoff production (Hewlett 1982). Figure 2 illustrates the vertical and horizontal structures of the WEP-L model. The watershed is initially divided into four primary sub-basins (①, ②, ③, and ④) for regional parameter adjustment. Subsequently, each primary sub-basin is further subdivided into numerous sub-basins, each of which is subsequently divided into contour zones based on elevation. Vertically, utilizing the Mosaic method, land use within the computation unit is classified into irrigated farmland, non-irrigated farmland, vegetation-bare land area, water area, and impervious water area. The water and heat fluxes of each land type are computed individually.
Figure 2
Schematic illustration of the WEP-L model structure: (a) vertical structure within a contour band and (b) horizontal structure within a sub-watershed (Jia et al. 2006b).
View largeDownload slide

Schematic illustration of the WEP-L model structure: (a) vertical structure within a contour band and (b) horizontal structure within a sub-watershed (Jia et al. 2006b).

Figure 2
Schematic illustration of the WEP-L model structure: (a) vertical structure within a contour band and (b) horizontal structure within a sub-watershed (Jia et al. 2006b).
View largeDownload slide

Schematic illustration of the WEP-L model structure: (a) vertical structure within a contour band and (b) horizontal structure within a sub-watershed (Jia et al. 2006b).

Close modal
In accordance with the attributes of frigid regions, the model adopts the ‘degree-day factor method’ to simulate the process of snow melting, with the calculation formula as follows:
(1)
(2)
where SM and S represent the daily amount of snowmelt and the extent of snow coverage, respectively, measured in millimeters (mm). and stand for the critical temperature and average temperature of snowmelt, respectively. E denotes the sublimation of snow cover, measured in mm. represents the snow melting coefficient, also known as the degree-day factor, measured in mm/(°C d). SW stands for snow water equivalent, measured in mm. Referring to prior research experiences (Jia et al. 2006a; Zhang et al. 2006), the following degree-day factors have been determined: 1 mm/(°C d) for forests, 2 mm/(°C d) for grasslands, 2 mm/(°C d) for sloping farmlands, and 3 mm/(°C d) for bare land. The critical temperature for snowmelt is typically set at 0 °C. The classification of precipitation as rain or snow is relatively straightforward. If the temperature is below 1 °C, it is deemed as snow, whereas if the temperature exceeds 1 °C, it is classified as rain.
To evaluate the model's simulation performance, this study employed the relative error (RE) and the Nash–Sutcliffe efficiency coefficient (NSE). The formulas for these two indicators are as follows:
(3)
(4)
where and represent the observed runoff and the simulated runoff, respectively, in cubic meters per second (m³/s). A smaller magnitude of the relative error (⁠ < 5%) corresponds to NSE approaching 1, signifying better model simulation results.
Furthermore, in this study, the coefficient of variation (Cv) was utilized to characterize the annual fluctuation characteristics of runoff, with the calculation formula as follows:
(5)
(6)
where represents the monthly runoff depth in millimeters (mm). denotes the mean monthly runoff depth in mm, while σ represents the standard deviation of the monthly runoff depth. A higher value of Cv indicates a more uneven distribution of runoff.

Improvement of the WEP-L model

Listen
In this investigation, the model underwent modifications based on the correlation between temperature and snowfall intensity in various elevation zones. According to a previous study (Dai 2008), the correlation between snowfall rate (SR) and temperature (TEM) can be adequately described by a hyperbolic tangent curve, with the fitting equation as follows:
(7)

In this equation, the parameters a, b, c, and d are estimated through non-linear least squares fitting utilizing the Levenberg–Marquardt method (Press et al. 1993). Parameter a represents the conversion factor, b characterizes the slope of the curve, c corresponds to the temperature associated with the median SR, and d reflects the asymmetry of the convergence value of the SR at extreme negative and positive temperatures. The SR for each elevation band is calculated based on the relationship between temperature and snowfall intensity. If the SR equals zero, the precipitation type during that period is rainfall. Conversely, if the SR exceeds zero, the amount of snowfall is determined by multiplying precipitation and the SR.

According to the analysis of the relationship between temperature and SR in this study, the equation in different elevation zones is modified as follows:

If SR = 0, the precipitation type in this period is rainfall, if SR > 0 means the snowfall is P * SR and the rainfall is P * (1 − SR). According to the above fitting relationship, the slope of the fitted curve increases with the elevation, that is, the higher the elevation, the steeper the curve, and the faster the SR decreases with the temperature increase.

The model delineates the initial snow cover as S0 (mm), and the computation of the original snowmelt module assumes that all calculation units are blanketed in snow. If the temperature surpasses the critical threshold for rainfall, the final snow cover at the end of the period, denoted as S, is derived from the discrepancy between S0 and SM. The water resource quantity is then determined by the summation of the snowmelt and rainfall. In the event that the temperature exceeds the critical temperature for snow, the final snow cover at the end of the period is calculated as S = S0 + P * SR − SM, where P * SR represents the snowfall discerned based on the relationship between snow rate and temperature. Conversely, if the temperature falls below the critical snow temperature, the final snow cover is obtained by adding S0 to the snowfall.

This study primarily aims to enhance the calculation method for the final snow cover at the end of the period. When the temperature falls below the critical threshold for snowfall, the model separately calculates the snow accumulation within snow-covered areas and the snowfall accumulation in areas that are not typically covered by snow within the computational unit. Specifically, the calculation method for S is modified as follows:
(8)
(9)

The total snow cover at the end of the period under this scenario is then determined as S = S0 + S1 + S2 − SM. Here, S0 represents the initial snow cover before the period begins. S1 is the additional snow accumulation in areas already covered by snow, while S2 accounts for the new snowfall in areas that were not previously snow-covered. SM denotes the snowmelt during the period. To incorporate the spatial heterogeneity of snow cover within the geographic unit of study, S1 and S2 are weighted by the respective fractions of snow-covered (FSC) and non-snow-covered areas (1 − FSC). We determine the FSC using remote sensing data and ground observations, which allows us to differentiate between snow-covered and non-snow-covered areas within the study region. By adjusting the contributions of S1 and S2 based on FSC, the model can more accurately simulate the dynamics of snow cover in response to meteorological conditions.

According to the relationship between snow cover and negative accumulated temperature, add the calculation of negative accumulated temperature and snow cover:
where P is precipitation, mm. According to the method of dividing rain and snowfall mentioned above, if the precipitation type is snow, the snow cover will be generated; if the precipitation type is rain, the snow cover will not be increased. FSC is the fraction of snow cover, %; NAT is negative accumulated temperature, °C.
Figure 3 shows a flowchart that outlines the step-by-step process of our calculations.
Figure 3
Integrated snow accumulation and melting process flowchart for the WEP-L model.
View largeDownload slide

Integrated snow accumulation and melting process flowchart for the WEP-L model.

Figure 3
Integrated snow accumulation and melting process flowchart for the WEP-L model.
View largeDownload slide

Integrated snow accumulation and melting process flowchart for the WEP-L model.

Close modal

Identification of precipitation types based on the wet bulb temperature

Listen

In this study, we utilized the classification methodology proposed by Ding et al. (2014) to discern various types of precipitation (such as snow, sleet, and rain) based on the wet bulb temperature (⁠⁠). contains air temperature, humidity, and pressure information. Since precipitating droplets (including rain, sleet, and snow) have a temperature closer to than to air temperature, it is expected that is more suitable than air temperature for indicating the precipitation type. In order to explore the influence of other parameters on precipitation type, Ding et al. (2014) used the artificial neural network to analyze the dependence of precipitation type on meteorological and geographic parameters such as wind speed, relative humidity (RH), air pressure, and elevation, and found that the influence of altitude and RH on precipitation type was more important. The procedure for distinguishing precipitation patterns is as follows:

Initially, is computed on the scheduled day of precipitation using the formula:
(10)
(11)

Among them, represents the temperature in Celsius, signifies the saturated vapor pressure at in kilopascals. corresponds to relative humidity, ranging from 0 to 1, while denotes the atmospheric pressure in kilopascals. Δ signifies the slope of the saturated vapor pressure-temperature curve (in kilopascals per Celsius), calculated as ⁠.

Next, critical temperatures and are evaluated to differentiate between precipitation types:
(12)
(13)

Among them, ⁠, ⁠, and rely on the RH and altitude (Z, in kilometers): ⁠, ⁠, and ⁠.

Lastly, by comparing the wet bulb temperature value with the critical temperatures, we determine the type of precipitation occurring on a given day. The options include:
(14)

To calculate the snowfall ratio (SR), we divide the amount of snowfall by the total precipitation volume (SR = snowfall/precipitation).

Data

Listen

The ground observation data used in this study were sourced from the China Meteorological Data Service Center (http://www.cma.gov.cn/), including air temperature, precipitation, RH, wind speed, pressure, and other meteorological data. The distribution of meteorological observation stations on the Qinghai–Tibet Plateau is very uneven, mainly in the central and eastern parts of the plateau. The remote sensing meteorological data used in this study are sourced from the China Meteorological Forcing Dataset (CMFD) (1979–2018) offered by the National Qinghai–Tibet Plateau Scientific Data Center (http://data.tpdc.ac.cn/). The dataset was made through fusion of remote sensing products, reanalysis dataset, and in situ observation data at weather stations. Its record starts from January 1979 and keeps extending (currently up to December 2018) with a temporal resolution of 3 h and a spatial resolution of 0.1°. Seven near-surface meteorological elements are provided in CMFD, including 2-m air temperature, surface pressure, specific humidity, 10-m wind speed, downward shortwave radiation, downward longwave radiation, and precipitation rate.

To examine the alteration of snow cover within the research area, we employ the daily fractional snow cover dataset over High Asia provided by China Scientific Data (http://www.csdata.org/). The spatial resolution of this dataset is 500 m. The runoff data is collected from the Maqu and Tangnaihai hydrological stations' monthly runoff records from 2002 to 2016 at the SYRB.

The topographic data utilized in this investigation originate from the geospatial data cloud (http://www.gscloud.cn/), which constitutes a digital topographic elevation model (DEM). This DEM is created through data processing of radar images attained by the SRTM system, and it possesses a spatial resolution of 90 m.

Listen

The relationship between temperature and snowfall and snow cover

Listen
Between 2002 and 2018, the annual snowfall ratio at the SYRB exhibited a value of 0.30, with the annual SR ranging between 0.26 and 0.36. Figure 4 illustrates a noteworthy decrease in the annual snowfall ratio within the 5,000–6,000 m region. As elevation rises, the snowfall ratio in regions surpassing 6,000 m tends to stabilize, whereas the snowfall ratio in areas below 5,000 m shows no discernible trend. Furthermore, the variation of snowfall ratio throughout the year indicates a distinct U-shaped distribution in areas below 5,500 m. In contrast, upper regions above 5,500 m experience relatively low temperatures with insignificant fluctuations. ANOVA analysis reveals no significant difference in the snowfall ratio above 5,500 m (band7–band8) (p = 0.097 > 0.05). Conversely, the snowfall ratio varies significantly in other elevation bands primarily due to the lower temperatures found at higher altitudes, resulting in stable snowfall probabilities. As altitude decreases, temperatures gradually rise, leading to more noticeable fluctuations in snowfall.
Figure 4
Annual and intra-annual variation trend of the snowfall ratio at different elevation bands in the SYRB from 2002 to 2018.
View largeDownload slide

Annual and intra-annual variation trend of the snowfall ratio at different elevation bands in the SYRB from 2002 to 2018.

Figure 4
Annual and intra-annual variation trend of the snowfall ratio at different elevation bands in the SYRB from 2002 to 2018.
View largeDownload slide

Annual and intra-annual variation trend of the snowfall ratio at different elevation bands in the SYRB from 2002 to 2018.

Close modal
The annual mean fraction of snow cover in the SYRB is 6.57%, declining by 44.6% below 3,500 m and rising by 8.3% above 3,500 m, as illustrated in Figure 5. With global warming, snow cover reduces at lower and middle elevations while higher elevations, being typically below freezing, experience more precipitation, leading to an increase in snow cover (Stewart 2010). The general trend indicates a gradual decline in snow cover, with the snow primarily present at elevations above 6,000 m, which is in-line with prior research on snow cover changes on the Qinghai–Tibet Plateau (Pu et al. 2007). The snow cover experiences considerable reduction between July and September and a greater accumulation from December to February, exhibiting a significant ‘U’ curve in the altitude range of 5,000–5,500 m. However, changes in other areas are comparatively insignificant. ANOVA analysis demonstrates no noteworthy disparity (p = 0.09–0.12 > 0.05) in snow coverage between areas below 3,500 m (band1–band2) and above 5,500 m (band7–band8). Nevertheless, snow cover at other elevations exhibits substantial variation, attributable to higher temperature, limited snow cover, and less observable variation at low elevations, whereas the temperature at higher elevations remains below 0 °C all year round resulting in more stable changes in snow cover.
Figure 5
Annual and intra-annual variation trend of fraction of snow cover at different elevation bands in the SYRB from 2002 to 2018.
View largeDownload slide

Annual and intra-annual variation trend of fraction of snow cover at different elevation bands in the SYRB from 2002 to 2018.

Figure 5
Annual and intra-annual variation trend of fraction of snow cover at different elevation bands in the SYRB from 2002 to 2018.
View largeDownload slide

Annual and intra-annual variation trend of fraction of snow cover at different elevation bands in the SYRB from 2002 to 2018.

Close modal
As depicted in Figure 6, the runoff depth at the SYRB exhibited a progressively rising trend of 5.71 mm/a. Concerning the intra-annual distribution characteristics of runoff, the majority of discharge was concentrated within the period of July to September.
Figure 6
Annual and intra-annual variation trend of runoff in the SYRB from 2002 to 2018.
View largeDownload slide

Annual and intra-annual variation trend of runoff in the SYRB from 2002 to 2018.

Figure 6
Annual and intra-annual variation trend of runoff in the SYRB from 2002 to 2018.
View largeDownload slide

Annual and intra-annual variation trend of runoff in the SYRB from 2002 to 2018.

Close modal
Drawing insights from the scatter plot representing the correlation between snowfall ratio and temperature (Figure 7), alongside a previous study conducted by Dai (2008), it is evident that the connection between snowfall ratio and temperature can be effectively modeled by means of a hyperbolic tangent curve. When the temperature falls below 3 °C, the snowfall ratio gradually tends toward 1, and conversely, when the temperature exceeds 6 °C, the snowfall ratio gradually tends toward 0. Notably, when the temperature ranges between 3 and 6 °C, the curve demonstrates a distinct downward trend.
Figure 7
Scatter plots of the snowfall ratio and temperature at the SYRB and hyperbolic tangent curves of different elevation bands.
View largeDownload slide

Scatter plots of the snowfall ratio and temperature at the SYRB and hyperbolic tangent curves of different elevation bands.

Figure 7
Scatter plots of the snowfall ratio and temperature at the SYRB and hyperbolic tangent curves of different elevation bands.
View largeDownload slide

Scatter plots of the snowfall ratio and temperature at the SYRB and hyperbolic tangent curves of different elevation bands.

Close modal
Revealed through the scatter plot showcasing the proportion of snow cover and negative accumulated temperature (Figure 8), a prominent exponential relationship between snow cover and negative accumulated temperature within the basin becomes evident.
Figure 8
Relationship between the fraction of snow cover and negative accumulated temperature at the SYRB.
View largeDownload slide

Relationship between the fraction of snow cover and negative accumulated temperature at the SYRB.

Figure 8
Relationship between the fraction of snow cover and negative accumulated temperature at the SYRB.
View largeDownload slide

Relationship between the fraction of snow cover and negative accumulated temperature at the SYRB.

Close modal

Simulation effect of the improved WEP-L model

Listen
To conduct a runoff simulation, the study utilized the measured monthly mean discharge data from the Maqu and Tangnaihai hydrological stations located at the SYRB between 2002 and 2016. The calibration period of the simulation was from 2002 to 2010, and verification was carried out from 2011 to 2016. The results of the simulation are displayed in Figure 9 and Table 1. Upon conducting an analysis of the simulation, the NSEs of the simulated runoff was controlled to have a value above 0.7, and the REs were kept within ±4%, resulting in a significant improvement in the simulation effect.
Table 1

Comparison of simulated effects before and after model improvement

Hydrological stationsNSE
RE
OriginalImprovedOriginalImproved
Maqu 0.57 0.81 −4.7% −2.7% 
Tangnaihai 0.69 0.72 −4.2% −3.8% 
Hydrological stationsNSE
RE
OriginalImprovedOriginalImproved
Maqu 0.57 0.81 −4.7% −2.7% 
Tangnaihai 0.69 0.72 −4.2% −3.8% 
View Large
Figure 9
Comparison of results before and after improvement of the WEP-L model of the SYRB discharge process.
View largeDownload slide

Comparison of results before and after improvement of the WEP-L model of the SYRB discharge process.

Figure 9
Comparison of results before and after improvement of the WEP-L model of the SYRB discharge process.
View largeDownload slide

Comparison of results before and after improvement of the WEP-L model of the SYRB discharge process.

Close modal

The effect of changes in snow on runoff

Listen
In this investigation, the WEP-L model was employed to analyze and compute the distribution characteristics of runoff in different elevation ranges. Based on the computation outcomes (Figure 10), the flow in the basin is most substantial within the 3,000–3,500 m altitude range, gradually declining as the elevation ascends. Above 5,000 m, there is virtually no runoff due to the consistently subzero temperatures throughout the year, accounting for only 0.25% of the total area, resulting in minimal ice and snow thawing. The highest amount of snowmelt-induced runoff occurs within the altitude range of 3,000–4,000 m, diminishing as the elevation increases. Beyond 3,000 m, snowfall significantly amplifies while snowmelt runoff experiences a slight decrease. Once the altitude surpasses 4,000 m, snowmelt runoff is noticeably diminished, snow coverage substantially augmenting alongside intensified snowfall. Beyond 4,500 m, snowmelt runoff experiences a further significant decrease in conjunction with increased snow coverage. Evidently, higher altitudes observe heightened snowfall and snow coverage, concurrently with a remarkable decrease in snowmelt runoff. The snowmelt quantity is determined based on the calibration results of the hydrological model. The correlation coefficients of snowmelt, SR, and snow cover rate within different elevation ranges are presented in Table 2. Snowmelt exhibits an inverse correlation with snowfall and snow cover, with the negative correlation becoming more pronounced as the elevation increases. Among these, the Cv value is most substantial in the 4,500–5,000 m area, implying an uneven distribution of snowmelt throughout the year in this region. The Cv value diminishes with decreasing elevation, with snowmelt being greater at lower altitudes than higher altitudes. As per the evolution of runoff in the basin, climate warming leads to increased rainfall and reduced snowfall, subsequently boosting runoff and intensifying summer snowmelt and its quantity. This phenomenon is particularly notable in areas below 4,500 m.
Table 2

The correlation coefficient (CC) between snowmelt and snowfall and snow cover and variation coefficient (Cv) of snowmelt in different elevation bands

Elevation bandsBelow 3,000 m3,000–3,5003,500–4,0004,000–4,5004,500–5,000
SM-SR CC −0.85 −0.90 −0.95 −0.98 −0.99 
SM-FSC CC −0.28 −0.45 −0.51 −0.56 −0.60 
Cv 0.99 1.08 1.15 1.32 1.42 
Elevation bandsBelow 3,000 m3,000–3,5003,500–4,0004,000–4,5004,500–5,000
SM-SR CC −0.85 −0.90 −0.95 −0.98 −0.99 
SM-FSC CC −0.28 −0.45 −0.51 −0.56 −0.60 
Cv 0.99 1.08 1.15 1.32 1.42 
View Large
Figure 10
Evolution characteristics of snowmelt runoff, snowfall, and snow cover at different elevations in the SYRB from 2002 to 2016.
View largeDownload slide

Evolution characteristics of snowmelt runoff, snowfall, and snow cover at different elevations in the SYRB from 2002 to 2016.

Figure 10
Evolution characteristics of snowmelt runoff, snowfall, and snow cover at different elevations in the SYRB from 2002 to 2016.
View largeDownload slide

Evolution characteristics of snowmelt runoff, snowfall, and snow cover at different elevations in the SYRB from 2002 to 2016.

Close modal
The SR, snow cover rate, and snowmelt runoff of band1 (below 3,000 m) served as the foundation for calculating the fluctuations of each elevation band. As depicted in Figure 11, as altitude ascends, the deceleration of snowfall diminishes by 0.06–2.7%/a, while the reduction in snow cover fraction slows by 0.5–1.8%/a. The snowfall ratio remained relatively stable below 5,000 m but intensified beyond that threshold. Below 5,000 m, the fraction of snow cover exhibited rapid growth, but this growth slowed down above 5,000 m. Results from runoff simulation analyses conducted by Cuo et al. (2013) and Meng et al. (2016) at the Tangnaihai hydrological station indicated that the overall inter-annual variation trend of snowmelt runoff in the SYRB was not prominent. The key factors influencing runoff alterations align with the climate and underlying surface factors, thus echoing the pertinent findings of this study. From 2002 to 2018, the snowfall ratio diminished as elevation increased, while snow cover augmented with elevation. Nevertheless, the overall trend of snowmelt runoff did not exhibit a significant pattern.
Figure 11
Inter-annual variation trends of snowfall ratio, fraction of snow cover, and snowmelt runoff with elevation at the SYRB.
View largeDownload slide

Inter-annual variation trends of snowfall ratio, fraction of snow cover, and snowmelt runoff with elevation at the SYRB.

Figure 11
Inter-annual variation trends of snowfall ratio, fraction of snow cover, and snowmelt runoff with elevation at the SYRB.
View largeDownload slide

Inter-annual variation trends of snowfall ratio, fraction of snow cover, and snowmelt runoff with elevation at the SYRB.

Close modal

Potential limitations and implications

Listen

Based on our investigation into the impact of changes in snowfall and snow cover on variations in runoff within the source region of the Yellow River basin, we enhanced the WEP-L hydrological model by refining the calculations of snowfall and snow cover in the snowmelt module and formulating a revised equation to enhance the precision of runoff simulations.

Regarding the data utilized, although remote sensing grid meteorological observations served as the primary data source, the potential limitations in accuracy and spatial resolution of remote sensing data could pose challenges to the reliability of model outcomes. Due to the uneven distribution and small number of ground meteorological observation stations on the Qinghai–Tibet Plateau, the error between remote sensing data and ground observation data may be affected by different terrain and climate change (Gao et al. 2014; Wang et al. 2018), which may also affect the reliability of the model results to a certain extent.

During the model parameter adjustment process, we made informed corrections based on the temperature–snowfall ratio relationship across different elevation bands, adapting the snowfall and snow cover calculations in the snowmelt module to enhance the model's capability in accurately simulating runoff changes. However, there exists uncertainty in the estimation of the model parameters. The spatial continuity of the modified model equation in relation to the natural conditions was not comprehensively considered in this study. Factors such as topographic and geomorphic conditions, alterations in vegetation cover, and human activities within the study area can influence the model results, necessitating further verification and refinement of our model through additional observational data support.

The findings of this study highlight the dynamic relationship between snowfall, snow cover, and temperature in the SYRB. By incorporating the established response function, the accuracy of snow identification and snowmelt simulation in the WEP-L model has been enhanced. Overall, this study contributes to a better understanding of water resource evolution and the efficient utilization of water resources in high-altitude cold regions. The improved snow identification and snowmelt simulation methods have practical implications for managing water resources in high-altitude cold regions. Further research can build upon these findings to develop more accurate models and strategies for water resource management in similar areas.

Listen

This study aimed to enhance the understanding of the elevation-dependent effects of snowfall and snow cover changes on runoff variations in the SYRB. Through the integration of an improved response function within the WEP-L hydrological model, we have successfully demonstrated the model's enhanced capability in simulating snow identification and snowmelt processes. We draw the following principal conclusions:

During the period from 2002 to 2018, a discernible downward trend in the snowfall ratio was observed in the SYRB, particularly in the region spanning from 5,000 to 6,000 m. With ascending elevations, the snowfall ratio in the area above 6,000 m exhibited a tendency toward stability, whereas no perceptible trend was observed in the region below 5,000 m. The snowfall ratio exhibited a hyperbolic tangent relationship with air temperature, showcasing its greatest sensitivity within the temperature range of 3–6 °C.

Over the same period, the proportion of snow cover in the SYRB saw a reduction of 44.6% in the area below 3,500 m, juxtaposed with an increment of 8.3% in the area above 3,500 m. A noteworthy exponential association was observed between snow cover and negative accumulated temperature.

Through our improvements to the WEP-L model, the simulation outcomes for the SYRB demonstrated significant enhancement. The NSEs of the refined simulation results all exceeded 0.7, and the REs remained well-controlled within ±4%. Snowmelt evinces a negative correlation with both snowfall and snow cover, with the inverse relationship becoming more pronounced at higher elevations. The decline in the snowfall ratio was further exacerbated with increasing elevation, while the snow cover exhibited an opposing trend. Nonetheless, the overall patterns in snowmelt runoff did not exhibit substantive significance.

In conclusion, this research provides valuable insights into the complex interactions between snowfall, snow cover, and temperature in the SYRB. The improved WEP-L model, with its enhanced snowmelt simulation, offers a robust tool for predicting water resource evolution in high-altitude cold regions, contributing to more effective water resource management strategies.

Listen

This work was financially supported by the National Vocational Education Teaching Innovation team research project ‘Innovation and Practice of team Teacher Education and Teaching Reform in the field of Green Ecological Environment in Vocational Colleges in the New Era’ (ZH2021040101), Teaching Innovation Team Construction (Double University Construction Project), the Second Tibetan Plateau Scientific Expedition and Research Program (STEP) (Grant No. 2019QZKK0207), the Young Talent Think Tank of Science and Technology of the China Association of Science and Technology (20220615ZZ07110156), the National Natural Science Foundation of China (No. 51909275 and 51679252), the Qinghai Central Government Guided Local Science and Technology Development Fund Project (2022ZY020), the IWHR Research & Development Support Program (WR110145B0052021), and the Open Research Fund of State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water resources and Hydropower Research (Grant No. IWHR-SKL-KF202204).

Listen

Y.Y.: ideas; development of methodology; creation of models; designing computer programs; verification; application of statistical, mathematical, computational techniques to analyze study data; management activities to annotate (produce metadata), scrub data and maintain research data (including software code, where it is necessary for interpreting the data itself) for initial use and later reuse; writing the initial draft; visualization/data presentation. Y.Z.: formulation of overarching research goals and aims; provision of study materials; writing review; oversight and leadership responsibility for the research activity planning and execution; management and coordination responsibility for the research activity planning and execution. M.L.: oversight and leadership responsibility for the research activity planning and execution; management and coordination responsibility for the research activity planning and execution. W.X.: designing computer programs; visualization/ data presentation; management activities to annotate (produce metadata), scrub data and maintain research data (including software code, where it is necessary for interpreting the data itself) for initial use and later reuse. J.L.: visualization/data presentation. Y.H.: visualization/data presentation.

Listen

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

Listen

The authors declare there is no conflict.

Bookhagen
B.
 &
Burbank
D. W.
2010
Toward a complete Himalayan hydrological budget: Spatiotemporal distribution of snowmelt and rainfall and their impact on river discharge
.
Journal of Geographical Research-Earth Surface
115
,
F03019
.
Google Scholar
 
Che
T.
,
Hao
X.
,
Dai
L.
,
Li
H.
,
Huang
X.
 &
Xiao
L.
2019
Snow cover variation and its impacts over the Qinghai-Tibet Plateau
.
Bulletin of Chinese Academy of Sciences
34
,
1247
–
1253
.
Google Scholar
 
Dai
A.
2008
Temperature and pressure dependence of the rain-snow phase transition over land and ocean
.
Geophysical Research Letters
35
,
62
–
77
.
Google Scholar
 
Ding
B.
,
Yang
K.
,
Qin
J.
,
Wang
L.
,
Chen
Y.
 &
Xiaobo
H.
2014
The dependence of precipitation types on surface elevation and meteorological conditions and its parameterization
.
Journal of Hydrology
513
,
154
–
163
.
Google Scholar
Crossref
Search ADS
 
Hewlett
J. D.
1982
Principles of Forest Hydrology
.
University of Georgia Press
,
Athens, GA
.
Google Scholar
 
Immerzeel
W. W.
 &
Bierkens
M. F. P.
2012
Asia's water balance
.
Nature Geoscience
5
,
841
–
842
.
Google Scholar
Crossref
Search ADS
 
IPCC
2013
Climate Change 2013: The Physical Science Basis
.
Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change
.
Cambridge University Press
,
Cambridge and New York
.
Jia
Y.
,
Wang
H.
 &
Yan
D.
2006a
Distributed model of hydrological cycle system in Heihe River basin I. Model development and verification
.
Journal of Hydraulic Engineering
37
,
534
–
542
.
Google Scholar
 
Jia
Y.
,
Ding
X.
,
Wang
H.
,
Zhou
Z.
,
Qiu
Y.
 &
Niu
C.
2012
Attribution of water resources evolution in the highly water-stressed Hai River Basin of China
.
Water Resources Research
48
,
2513
.
Google Scholar
Crossref
Search ADS
 
Mankin
J.
,
Viviroli
D.
,
Singh
D.
,
Hoekstra
A.
 &
Diffenbaugh
N.
2015
The potential for snow to supply human water demand in the present and future
.
Environmental Research Letters
10
114016
.
Google Scholar
Crossref
Search ADS
 
Meng
F.
,
Su
F.
,
Yang
D.
,
Tong
K.
 &
Hao
Z.
2016
Impacts of recent climate change on the hydrology in the source region of the Yellow River basin
.
Journal of Hydrology: Regional Studies
6
,
66
–
81
.
Google Scholar
 
Pachauri
R.
,
Allen
M.
,
Barros
V.
,
Broome
J.
,
Cramer
W.
 &
Christ
R.
2014
Climate Change 2014: Synthesis Report
.
Contribution of Working Groups I, II and III to the fifth assessment report of the Intergovernmental Panel on Climate Change. Intergovernmental Panel on Climate Change, Copenhagen
.
Google Scholar
 
Press
W. H.
,
Teukolsky
S. A.
,
Vetterling
W. T.
 &
Flannery
B. P.
1993
Numerical Recipes in Fortran 77
, 2nd ed., Vol.
2
The Art of Scientific Computing
. pp.
678
–
680
.
Google Scholar
 
Pu
Z.
,
Xu
L.
 &
Salomonson
V. V.
2007
MODIS/terra observed seasonal variations of snow cover over the Tibetan Plateau
.
Geophysical Research Letters
34
,
137
–
161
.
Google Scholar
 
Qin
Y.
,
Abatzoglou
J. T.
 &
Siebert
S.
2020
Agricultural risks from changing snowmelt
.
Nature of Climate Change
10
,
459
–
465
.
Google Scholar
Crossref
Search ADS
 
Wang
X. J.
,
Pang
G. J.
 &
Yang
M. X.
2018
Precipitation over the Tibetan Plateau during recent decades: A review based on observations and simulations
.
International Journal of Climatology
38
,
1116
–
1131
.
Google Scholar
Crossref
Search ADS
 
Wu
T.-w.
,
Qian
Z.-a.
 &
Cai
Y.
2004
Improvement study on snow scheme in LSM scheme of NCAR CCM3 model (Ⅱ): The GCM simulation experiment analyses
.
Plateau Meteorology
23
,
569
–
579
.
Google Scholar
 
Yang
Y.
,
Weng
B.
 &
Yan
D.
2021
Partitioning the contributions of cryospheric change to the increase of streamflow on the Nu River
.
Journal of Hydrology
598
,
1
–
12
.
Google Scholar
 
Yu
W.
,
Nan
Z.
,
Zhao
Y.
 &
Li
S.
2013
Improvement of snowmelt implementation in the SWAT hydrologic model
.
Acta Ecologica Sinica
33
,
6992
–
7001
.
Google Scholar
 
Zhang
Y.
,
Liu
S.
 &
Ding
Y.
2006
Spatial variation of degree-day factors on the observed glaciers in western China
.
Acta Geographica Sinica
61
,
89
–
98
.
Google Scholar
 
© 2024 The Authors
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).