Increasing Rainfall and Runoff in Alpine Mountainous Areas Can't Completely Eliminate Drought Driven by Temperature Rise

Drought research under climate change is of great scientific significance. For Land Use and Land Cover Change (LUCC), temperature and rainfall in climate change, which factor has a greater impact on runoff change in alpine mountainous areas? Can the increase of rainfall in the alpine mountainous area completely eliminate the drought driven by temperature rise? This study takes the upper reaches of Heihe River basin (URHRB) as an example, the URHRB's Soil and Water Assessment Tool (SWAT) model is constructed. Based on 58 scenarios and The Budyko Framework, here we show that a)climate change has a greater contribution to runoff than LUCC, effect of increased rainfall greater than temperature rising on runoff in alpine mountainous area; b)the drought of 57.14% of UHRRB’s sub-basins have eased, 42.86% of the sub-basins is more serious, the increase in rainfall can't completely eliminate the drought driven by temperature rise. This study coupling SWAT simulation with Budyko Framework and other methods solves the problem of lack of data in alpine mountainous areas, and more accurately quantifies the impact of climate change, LUCC on runoff changes, realizing theoretical and method innovation. The results of this study provide a scientific paradigm for solving scientific problems in similar regions in China and other countries, and have important promotion value.


Introduction
The alpine areas are the sources of many rivers, and they have powerful functions at water conservation and regulation, water ecological environment, and water security 1,2 . There are many river runoff formed at the upper alpine areas in the world. The upstream river channel determines the amount of water resources sent to the middle and lower reaches, thereby restricting the sustainable development of social economy in the middle and lower reaches. Therefore, exploring the characteristics, paths and formation processes of runoff changes in alpine areas has important practical significance for the prevention and control of water pollution, improvement of water environment, protection of water safety, scientific management, and sustainable development of social economy 3 .
The formation process of runoff in alpine regions is more complicated than that in warm regions. Many scientific questions still need to be answered. The most prominent ones are as follows: What is the main reason for the change of runoff in the alpine areas? What mechanism is it regulated? For temperature and rainfall in climate change, which factor has a greater impact on runoff change in alpine mountainous areas? Can the increase of rainfall in the alpine mountainous area completely eliminate the drought driven by temperature rise?
Drought research under climate change is of great scientific significance. In recent years, many domestic and foreign scholars have conducted analysis and research on the attribution analysis of watershed runoff changes 4,5,6 . Commonly used methods include simple empirical curve method (such as rainfall-runoff double accumulation curve, etc.), hydrological model analysis method. Budyko Framework and various empirical formulas derived from it comprehensively consider the interaction between various factors in the watershed 7,8,9 , and the introduction of underlying surface parameters makes the method have a certain physical mechanism 10,11,12 , with the calculation being relatively simple and convenient 13,14,15 . The Budyko hydrothermal coupling equation uses a large number of parameters in parameter calculations 16,17,18 . At present, domestic and foreign researches are very dependent on monitoring data, and are unable to do anything about sub-catchments with missing data; they often replace sub-catchments with the entire watershed; the statistics are too small 19 . How to realize the transformation from the whole basin to sub-basin and grid is very important, and SWAT model considers the comprehensive influence of various important factors. Therefore, carrying out relevant research based on the results of SWAT simulation is of great significance to solve the problem of missing data in some watersheds 20,21,22 .
The temporal and spatial evolution of surface runoff is to a large extent caused by land use and climate change, and is caused by the interaction of elements in the basin system 23,24 . Using the SWAT model to predict future hydrological responses is conducive to proposing reasonable watershed management measures for different future scenarios. This study coupling SWAT simulation with Budyko Framework and other methods solves the problem of lack of data in alpine mountainous areas, and more accurately quantifies the impact of climate change, LUCC on runoff changes. The major objectives and motivation of this study are to use these methods and models to a) identify the trend and sudden year with temperature, rainfall, potential evaporation (ET0), relative moisture index (M), etc.; b) simulate the UHRRB runoff, explore the impact of LUCC and climate (rainfall, temperature) changes on runoff, and find the main influencing factors of runoff changes; c) respond to people's following concerns: Compared with LUCC, which factor has a greater impact on runoff changes in alpine mountainous areas? Compared with rainfall, which factor has a greater impact on runoff changes in alpine mountainous areas? Can the increase in rainfall in the alpine mountainous area completely eliminate the drought driven by temperature rise?
The disposable income of URHRB residents is low. The National Development and Reform Commission had approved the Qilian Mountain Ecological Protection and Comprehensive Management Plan. The main construction content includes wetland ecological protection, water and soil conservation, water resources protection, glacier environmental protection, and scientific and technological support projects. During the "Fourth Five-Year Plan" period, the Qilian Mountain Heihe River Basin Ecological Protection Project will be listed as a provincial major project. Therefore, this study selects UHRRB as the study area. Based on Budyko Framework, SWAT model, etc. to carry out related research, provide theoretical support for the ecological protection of the basin, and provide new ideas for economic growth and ecological development.
This has important practical significance and practical value for regional human survival, production, life, ecological environment and the formulation of related policies.

Annual runoff and summer, autumn and winter runoff having a better correlation with rainfall;
fall and winter runoff having a better correlation with temperature than spring and summer.
In order to explore the relationship between rainfall, temperature and runoff, the distribution of runoff during the year is divided into four seasons. Spring is from March to May, Summer is from June to August, Autumn is from September to November, and Winter is from December to February of the following year. Fig. 1 is an analysis of the correlation between UHRRB temperature (a), runoff(b) and precipitation at seasonal scales. It can be seen from Fig. 1 that annual runoff and summer, autumn and winter runoff have a better correlation with rainfall; fall and winter runoff have a better correlation with temperature than spring and summer.

Climate change has a greater impact on runoff than LUCC
In order to compare the impact of LUCC and climate changes on runoff, the SWAT model is built to calculate the runoff, based on S1-S4 scenarios such as LUCC and climate changes. The results are shown in Table 1. It can be seen from Table 1 that the runoff of scenarios 1, 2, 3, and 4 are 58.16m 3 /s, 52.88m 3 /s, 58.94m 3 /s, 53.66m 3 /s, respectively.
(1) Taking S4 as the base period, compared with S2, we can get: The climatic conditions remain unchanged from 1981 to 2003, considering the change of single factor of land use relative to 1980 in 2015, runoff was only reduced by 0.78m 3 /s, and the change rate was -1.45%. According to the analysis results of the land use transfer matrix 32,33 , the change rates of AGRL, BARR, FRST, PAST, URLD and WATR in this area from 2004 to 2014 were 0.00%, -7.30%, 0.82%, 5.91%, 0.00%, 0.57%, respectively. 71.58％ of the reduced bare land area is converted into grassland, 23.21％ is converted into forest land, and 5.13％ is converted into water bodies and wetlands. Therefore, from 2004 to 2014, from the perspective of LUCC, the conversion of bare land to grassland and woodland caused a decrease in runoff in the study area.
(2) Taking S4 as the base period, by comparing S3 and S4, we can get: Under the condition of  (4) As the results above, we can find that from 2004 to 2014, the decrease in annual runoff caused by UHRRB's LUCC (-1.45%) was not only smaller than the increase in annual runoff caused by climate change (+9.84%), but also less than the combined effect of LUCC and climate change (+8.39%). Compared with land use, climate change played a leading role in UHRRB runoff changes.
(5) The contribution rate of climate change and land use LUCC are 85.23% and 14.77%, respectively, indicating that the impact of climate change on runoff is far greater than that of LUCC. Under the combined effects of climate change and LUCC, runoff is still keep the trend of increasing, which is conducive to the conservation and regulation of water resources in the basin, the improvement of water ecological environment and water security.

Increasing rainfall has a greater impact on runoff than temperature rising under climate change scenarios
In order to compare the impact of rainfall and temperature changes on runoff and ET0, the SWAT model is built to calculate the runoff and ET0, based on S1 and S5-S58 scenarios such as rainfall and temperature changes(see Supplementary Table 1). On account of the influence of rainfall on ET0 is extremely insignificant, it is ignored in this study. The results are shown in Fig. 2.
It can be seen from Fig. 2 that when the temperature decreases, surface runoff increases; when the temperature rises, the changes in surface runoff present uncertainty, due to the conflicting effects caused by evaporation and snowmelt runoff; when precipitation increases and temperature decreases, runoff increases the most; when temperature increases and precipitation decreases, runoff decreases most significantly. The main results were as follows: It can be seen from Fig. 2 that when the temperature keeps invariant, and the precipitation increases or decreases by 10% and 20%, the average annual runoff increases by 7.45%, 14.69% and decreases by 20.85% and 30.61%, respectively. The temperature drops or rises by 0.5℃, 1℃, 1.5℃, 2℃ and 3℃, the average annual runoff increases by 21.61%, 17.40%, 13.14%, 9.10%, 4.35%, 9.20%, 6.28%, 5.36%, 4.43% and 3.95%. The absolute average of the former is 4 times the absolute average of the latter, suggesting that the runoff of alpine mountain rivers is most affected by precipitation. Although temperature has a certain effect on runoff, it is relatively small.
When the temperature decreases and the precipitation decreases (S19-S28), the decrease in temperature will lead to an increase in runoff, and the decrease in precipitation will lead to a decrease in runoff. The response rate of average temperature rise to average runoff is -3.29m 3 /(s·℃), and the percentage response rate of average temperature rise to average runoff is -4.16%/72.88%, that is, -1.94%/10%; The response rate of average rainfall to average runoff is +0.1179m 3 /(s·mm), and the percentage response rate of average rainfall to average runoff is +8.34%/10%, indicating that the response rate of average annual rainfall to average annual runoff is much greater than the response rate of average annual temperature rise to average annual runoff.
On account of the influence of rainfall on ET0 is extremely insignificant, when the temperature decreases and the precipitation consist (S19-S23),the response rate of the average temperature rise to the average ET0 is 19.98mm/℃. The percentage response rate of the average temperature rise to the average ET0 is 8.77%/72.88%, which is 1.20%/10%.The response rate of the average temperature rise to the average H is -0.0245/℃, and the percentage response rate of average temperature rise to average drought intensity(H) is -8.22%/72.88%, that is, -1.13%/10%. The response rate of average rainfall to average H is +0.0002/mm, and the percentage response rate of average rainfall to average runoff is +0.067%/10%.
When the temperature decreases and the precipitation increases (S29-S38), the decrease in temperature will lead to an increase in runoff, and the increase in precipitation will lead to an increase in runoff. And when the temperature drops by 3℃ and the rainfall increases by 20%, the runoff has a maximum value; The response rate of the average temperature rise to the average runoff is -5.15m 3 /(s·℃), and the percentage response rate of the average temperature rise to the average runoff is -22.15%/72.88%, which is -3.44%/10%; The response rate of average rainfall to average runoff is +0.1780m 3 /(s·mm), and the percentage response rate of average rainfall to average runoff is +12.59%/10%, indicating that the response rate of average annual rainfall to average annual runoff is much greater than that of average annual temperature rise to average annual runoff. The response rate of the average temperature rise to the average ET0 is 19.98mm/℃.
The percentage response rate of the average temperature rise to the average ET0 is 8.77%/72.88%, which is 1.20%/10%.The response rate of the average temperature rise to the average H is -0.033/℃, and the percentage response rate of average temperature rise to average H is -11.07%/72.88%, that is, -1.52%/10%. The response rate of average rainfall to average H is +0.0002/mm, and the percentage response rate of average rainfall to average runoff is When the temperature rises and the precipitation decreases (S39-S48), the increase in temperature will lead to a decrease in runoff, and the decrease in precipitation will lead to a decrease in runoff. And when the temperature rises by 3°C and the precipitation decreases by 20%, the runoff is at a minimum. The response rate of average temperature rise to average runoff is -1.820m 3 /(s·℃), and the percentage response rate of average temperature rise to average runoff is -7.75%/72.88%, that is, 1.06%/10%; The response rate of average rainfall to average runoff is +0.1572m 3 /(s·mm), and the percentage response rate of average rainfall to average runoff is +11.12%/10%, indicating that the response rate of average annual rainfall to average annual runoff is much greater than the response rate of average annual temperature rise to average annual runoff.
When the temperature decreases and the precipitation consist (S39-S43), the response rate of average temperature rise to average ET0 is 21.96mm/℃. The percentage response rate of the average temperature rise to the average ET0 is 9.64%/72.88%, which is 1.32%/10%. The response rate of the average temperature rise to the average H is -0.021/℃, and the percentage response rate of average temperature rise to average H is -7.04%/72.88%, that is, -0.97%/10%.The response rate of average rainfall to average H is +0.00017/mm, and the percentage response rate of average rainfall to average runoff is +0.057%/10%.
When the temperature rises and the rainfall increases (S49-S58), the increase in temperature will lead to an increase in runoff, and an increase in rainfall will lead to an increase in runoff. The reason of increased runoff caused by rising temperatures may be a conflict effect between evaporation and snowmelt. The response rate of average temperature rise to average runoff is +2.36m 3 /(s·℃), and the percentage response rate of average temperature rise to average runoff is +10.16%/72.88%, that is, +1.39%/10%.The response rate of average temperature rise to average ET0 is 21.96mm/℃. The percentage response rate of the average temperature rise to the average ET0 is 9.64%/72.88%, which is 1.32%/10%.The response rate of average rainfall to average runoff is +0.169m 3 /(s·mm), and the percentage response rate of average rainfall to average runoff is +11.99%/10%, indicating that the response rate of average annual rainfall to average annual runoff is much greater than that of average annual temperature rise to average annual runoff. The response rate of the average temperature rise to the average H is -0.028/℃, and the percentage response rate of average temperature rise to average H is -9.39%/72.88%, that is, 1.29%/10%.The response rate of average rainfall to average H is +0.00017/mm, and the percentage response rate of average rainfall to average runoff is +0.057%/10%.
Through the scenario settings of S19-S58, the variation values and rates of runoff, ET0 and H caused by meteorological factors in this study area are shown in the Supplementary Fig.2. and is mainly related to the change of temperature, which is consistent with the research conclusions of Yao 34 . ET0 increases with the increase of temperature, for every 1℃ increase in temperature, ET0 increases by 20.92mm.  Fig. 3(a)-(b). It can be seen from Fig. 3(a)-(b) that the UHRRB runoff has an overall increasing trend. The runoff of the No. 14-21 sub-basin in the Dongcha (Babao River) of the Heihe River in the southeast has slightly decreased, with the highest decrease of 5.40%; The runoff of Xicha (Yeniugou River) and UHRRB main stream of Heihe River in the northwest increases significantly, with the highest increase of 27.62%.

Increasing rainfall and runoff can't completely eliminate the drought driven by
The response results of the 21 sub-basins of URHRB to each influencing factor are shown in Fig. 3(c), and the contribution rate of each influencing factor to the change of runoff in each sub-basin is shown in Fig. 3(d).
It can be seen from Fig. 3(c) that the contribution rate of rainfall to runoff in 100% of the sub-basins is positive; the contribution rate of ET0 to runoff in 100% of the sub-basins is negative; The contribution rate of the underlying surface change to runoff in 33. Global climate change, land use patterns, and human activities make the hydrological process in alpine regions very complex [37][38][39] . Factors such as air temperature, rainfall, evapotranspiration, runoff, glacier, snowfall, frozen soil, vegetation, groundwater 40 , land use patterns, and human activities interact with each other and cause each other, and the relationship is very complex 41,42 .
Clarifying the relationship between them is currently a hot topic and scientific issue in domestic and foreign research. At present, domestic and foreign scholars have carried out a lot of research on the situation in different regions, but due to the large uncertainty of climate change, the lack of monitoring data in alpine mountainous areas, and the differences in research data and methods, there are still big differences in relevant research results.

3.Discussion
In previous studies, changes in runoff (m 3 /s) (dimension is m 3 /s/℃ and m 3 /s/ mm) are often compared when rainfall and temperature change is 1℃ and 1mm respectively, and relevant conclusions are drawn. The main problem of this comparison method is the inconsistency of dimensions. This study believes that when the temperature decreases and the rainfall decreases, the percentage response rate of the average temperature rise to the average runoff is -1.94%/10%,and the percentage response rate of the average rainfall to the average runoff is+8.34%/10%; When the temperature decreases and the rainfall increases, the response rate of the average temperature rise to the percentage change of the average runoff is -3.44%/10%, and the response rate of the average rainfall to the percentage change of the average runoff is +12.59%/10%; When the temperature rises and the rainfall decreases, the response rate of the average temperature rise to the percentage change of the average runoff is -1.06%/10%, and the response rate of the average rainfall to the percentage change of the average runoff is +11.12%/10%; When the temperature rises and rainfall increases, the response rate of the average temperature rise to the percentage change of the average runoff is +1.39%/10%; the response rate of the average rainfall to the percentage change of the average runoff is +11.99%/10%. It shows that the response of rainfall to runoff is much greater than the response to temperature; the response rate of the average temperature rise to the percentage change of the average ET0 is +1.32%/10%. Obviously, this study compares the percentage of change in runoff with the percentage of change in rainfall and temperature rise under each scenario, and the dimensions are unified, making the results comparable. Therefore, this method overcomes the main problems and shortcomings in previous studies, and the conclusion is more reliable.
In previous studies, relevant data based on hydrological sites were often used as the basis for the study of the research area. There was no detailed study on the sub-basins of the research area above hydrological sites, which could not fully reflect the heterogeneity of the research area, and the research conclusions often cover up many facts. This study couples SWAT simulation with Budyko Framework and other methods, and conducts the research from the scale of the sub-basin, solving the problem of missing data in alpine mountainous areas. It more accurately quantified the impact of climate change and human activities on runoff changes, and came to the important conclusion that the increase in rainfall can't completely eliminate the drought driven by temperature rise. It is more scientific and more realistic, which is conducive to the scientific planning of the green hills, clear water, cultivated land, woodland, grassland and the upper and lower reaches and the left and right banks of the Heihe River Basin.
It should be pointed out that although the research results under different scenarios in this study have the same trend, due to the monitoring errors of hydrology, meteorology, soil, land use and other data, as well as the limitations of the SWAT model and its calibration, and the representativeness of the verification parameters, the calculation results inevitably have certain errors, indicating that the results have a certain degree of uncertainty, which is precisely the inexhaustible motivation and source of our in-depth and continuous research.

Methods
The Heihe River is the second largest inland river in Northwest China. It originates from the Qilian Mountains, flows through the Hexi Corridor, and ends at Ejinaqi in northern Inner Mongolia; With Yingluoxia Hydrological Station and Zhengyixia Hydrological Station as the boundary, it is divided into upper, middle and lower reaches.

URHRB includes most of Qilian County in Qinghai Province and part of Sunan County in
Gansu Province, covering an area of 9,920 km 2 ( Supplementary Fig. 3). The main rivers include Xicha Yeniugou, Dongcha Babao River, the main stream of the central Heihe Rive. Dongcha is about 75km and is between 97°28′-101°16′E and 37°30′-39°41′N. It is the main runoff producing area of the Heihe River Basin, which is composed of surface runoff, glacier snowmelt water and groundwater 43 . The annual runoff at the outlet of Yingluoxia is 1.58 billion m 3 . Due to the heavy rainfall and the supply of snowmelt water from glaciers, the underlying surface is a rocky mountainous area with good vegetation. It is a runoff-forming area in the Heihe River Basin. The runoff above the mountain mouth accounts for 88.0% of the river's natural water volume. The snowline height in the study area gradually rises from east to west, and the glacier coverage rate is only 0.5%. The annual average glacier meltwater accounts for only 3.5% of the runoff (the proportion in dry years is close to 5%, and the proportion in dry months is as high as 16%). The remaining 96.5% of the runoff is made up by rainfall. Glacier retreat has little effect on runoff 44 , but glaciers have an important role in stabilizing runoff and regulating abundance and dryness.

Data and sources
See Supplementary Table. 2.

Calculate M and H
M is suitable for drought monitoring and assessment, the calculation formula is shown in formula (1), and the classification of M is shown in Supplementary Table 3.
Where, P is precipitation and ET0 is potential evapotranspiration.
H is used to evaluate the severity of drought, which is reflected by the average relative humidity index. The calculation formula is shown in formula (2).
Where, n is the number of years of drought, and Mi is the relative humidity index value when drought occurs.

Identification of trends and mutation points
First, diagnose the correlation between runoff, temperature, rainfall, ET0 and M on the annual scale and the spring, summer, autumn, and winter scales to assess the correlation between hydrological elements in the inter-annual and intra-annual seasons; Secondly, conduct trend and mutation point analysis and evaluation of runoff, temperature, rainfall, ET0 and M to determine the base period and change period of the age change. There are many ways to analyze trends and mutation points 45 . This study uses the hydrological time series trend and mutation analysis system developed by Wang 46 to comprehensively analyze the trend and change point characteristics of the runoff series. The trend test method adopts the linear trend correlation analysis method; the mutation point test adopts the MK and CDC methods.

SWAT simulation analysis
Using the SUFI-2 algorithm in SWAT-CUP software 47 Supplementary Table 4 and Supplementary Fig. 4) are used to evaluate the fit and applicability of the simulated and measured values until they meet requirements, and then perform parameter verification and return. The calculation formulas are as follows: Where: Q m is the simulated runoff; Q s is the observed runoff.

Runoff simulation scenario setting and contribution rate calculation under different land use and climate change based on SWAT model
After diagnosing the applicability of the SWAT model in UHRRB, set and compare the values of 4 (see Supplementary Where: Qi represents the runoff simulated by scenario i.

Attribution analysis of runoff changes
Only by accurately quantifying the impact of precipitation, ET0, underlying surface and other factors on runoff can we plan and use water resources more effectively 48 .
This study uses Budyko Framework for attribution analysis. The main formula of Budyko analysis is shown in formula (15)- (19): Where: E is the actual evapotranspiration (mm), P is the precipitation (mm), E0 is the amount of ET0 (mm), n is the underlying surface parameter [49] , ε is the elastic coefficient, R is the runoff (mm), and ΔR is the runoff change, εp, εE0, and εn represent the precipitation elasticity coefficient, ET0 elasticity coefficient and the underlying surface elasticity coefficient respectively, η is the relative change rate of runoff, R1 and R2 represent the average annual runoff in the base period and the influence period, respectively, and △Rx is the influence of the corresponding factor on the change of runoff; ԑx is the elastic coefficient of each factor to the change of runoff; x is a certain influencing factor of runoff change, including P, E0 and n; △x is the amount of change of a certain impact factor in the influence period compared with that in the base period.

Competing interests
The authors declare no competing interests.

Figures
………………………………………………………………………………………………………………………… Fig. 1 Correlation analysis of temperature(a),runoff(b) and rainfall. The slope of the four lines in Fig.1(a) represents the influence degree of temperature on runoff. The higher the slope is, the greater the 22 influence degree of rainfall on runoff in that season is. The shadowed part in Fig.1(a) shows the distribution of temperature in that season. The slope of the four lines in Fig.1(b) represents the influence degree of rainfall on runoff. The higher the slope is, the greater the influence degree of rainfall on runoff in that season is. The shadowed part in Fig1. (b) shows the distribution of rainfall in that season. It should be noted that because the units of temperature and rainfall are different, the slopes between a and b cannot be compared.