Influence of sub-cloud secondary evaporation and moisture 1 sources on stable isotopes of precipitation in Shiyang River 2 Basin ， Northwest China

11 The fractionation of stable isotopes in precipitation runs through all links of the hydrological cycle. 12 Studying its composition will help to understand the hydrological cycle process and the interaction 13 between land and atmosphere. Based on the data of measured precipitation isotopes and related 14 meteorological elements from 11 sampling points from January 2018 to September 2019, the existence 15 of sub-cloud secondary evaporation is verified. Used the water vapor flux and the improved Lagrangian 16 model, the moisture source affecting precipitation is tracked. On the basis of them, the influence of 17 sub-cloud secondary evaporation and moisture sources on stable isotopes of precipitation is analyzed. 18 The results show that the sub-cloud secondary evaporation exists in the Shiyang River Basin, and it is 19 stronger in spring and summer than that in autumn and winter, which makes the stable isotopes of 20 precipitation higher in summer and lower in winter. Besides, the sub-cloud secondary evaporation is 21 stronger in the midstream and downstream, which results in the heavy isotopes of precipitation are 22 generally more enriched. In the vertical direction, the secondary evaporation between 850 hPa and 700 23 hPa is the strongest, which makes heavy isotopes enrich and d -excess decreases in this layer. The 24 moisture source of precipitation in the Shiyang River Basin is dominated by westerly air masses, that 25 mid-high latitude continental sources have a large contribution to precipitation but the supply of sea 26 sources is very limited, which makes the d -excess of precipitation is higher and does not show regional 27 consistency and seasonality well.


Introduction
As a natural source tracer, stable isotopes of precipitation can not only record water information but 32 also reflect changes in climate and the environment 1 . Observation and research on stable isotopes of

116
Sample collection and data sources 117

Collection and processing of precipitation sample 118
As shown in Fig. 1, this study selected 11 collection points of precipitation sample in the Shiyang River 119 Basin, and the sampling period is from January 2018 to September 2019. All precipitation samples 120 were collected by the definition of precipitation events stipulated by meteorological observation, sealed 121 in polyethylene sampling bottles, and stored in cold storage. Using the automatic weather observation 122 instruments, the data of temperature, dew point, relative humidity, and atmospheric pressure were also 123 collected. A total of 670 precipitation samples was transported to the isotope laboratory of the College 124 of Geography and Environmental Sciences of Northwest Normal University for testing and analysis. 125 The testing instrument is the DLT-100 liquid water isotope analyzer developed by Los Gatos Research 126 in the United States, and the measurement result is expressed by the difference between the test sample 127 isotope ratio and the standard sample ratio (‰): 128 Where R s is the ratio of 18 O/ 16 O in the precipitation sample, and R v-SMOW is the ratio of 18 O/ 16 Where n is the number of samples, x is the δ 18 O, and y is the δD. 149 The theoretical slope (S T ) of the LMWL is calculated based on the condensation temperature 31  Where 2 α and 18 α are the equilibrium fractionation coefficients of hydrogen and oxygen isotopes 152 between water and gas, respectively 32 . 153

Raindrop evaporation model of improved STEWART 154
Due to the sub-cloud secondary evaporation, there are some differences in the isotopes between the 155 raindrops at the cloud bottom and the raindrops falling on the ground (∆d Where 2 γ, 18 γ, 2 β, and 18 β are defined by Stewart 11 ; 2 α and 18 α are the same as formula (4); f is the 160 proportion of the remaining mass of raindrops after evaporation during falling 14 Where T mean is the average temperature between the lifting condensation level (LCL) and the 169 ground (℃). P and P LCL are the pressure (hPa) at the ground and the LCL, respectively. 170 According to Kinzer et al 34 , the r ev can be calculated as: 171 Where Q 1 is a function (cm) of temperature (T) and raindrop diameter (D); Q 2 is a function 173 Where e is the natural constant; I is the precipitation intensity (mm·h -1 ); F is the cumulative 179 volume percentage of raindrops with a diameter less than D in the atmosphere. Where D c is the diameter of raindrops at cloud base; m end and m ev are the same as formula (6); ρ is 185 the density of water. 186

187
Local meteoric water line Similar to the GMWL, the relationship between δD and δ 18 O in precipitation in a certain area is called 189 the LMWL. The slope of the meteoric water line is composed of the ratio of the fractionation rates of 190 the δD and δ 18 O, which can reflect the fractionation effects between them, and the intercept of its can 191 reflect the deviation degree of δD from the equilibrium state 36 . Relevant studies have shown that 192 precipitation affected by the sub-cloud secondary evaporation will deplete the lighter isotope but enrich 193 the heavy isotope 37 , which makes the slope and intercept of the meteoric water line decrease. So the 194 changes of the slope and intercept of the LMWL under different meteorological conditions are 195 considered to evaluate the intensity of sub-cloud secondary evaporation 38,39 . 196  Table 1  Xiyingwugou is lower than other sampling points of the upstream, the mean of δD and δ 18 O is the 203 lowest. This may be due to the fact that the sampling point is close to the reservoir and is affected by 204 the combined influence of moisture recirculation intensity, relative humidity, and temperature, which 205 masks the elevation effect of the stable isotope. 206 According to the results in Table 1, it can be seen that the slopes and intercepts of the LMWL at 207 sampling plots in the Shiyang River Basin deviate to a certain extent from the GMWL. The slope of 208 LMWL less than 8 indicates that the precipitation process occurs under non-Rayleigh conditions 42 . In 209 this study, the slopes of the LMWL except for the sampling point of Dengjiazhuang are below 8, 210 indicating that the Shiyang River Basin has a dry climate and raindrops are affected by sub-cloud 211 secondary evaporation. Compared with the slope, the intercept change of the LMWL is more 212 complicated, which is related to the degree of fractionation. The factors that affect fractionation include 213 humidity, temperature, and wind speed. Generally, high temperature and low relative humidity make 214 the isotopes of precipitation more deviate from the equilibrium state, which results in the intercept of 215 LMWL decrease 43 . Except for Lenglongling, Hulinzhan, Xiyingwugou, and Dengjiazhuang, the 216 intercepts of the LMWL at other sampling points are much lower than the GMWL, which reflects that 217 there is sub-cloud secondary evaporation in the Shiyang River Basin 44 . At the same time, the S T of each 218 sampling point is greater than 8 and greater than the slope of the LMWL, which also shows that there 219 exists sub-cloud secondary evaporation in the Shiyang River Basin 45 . Besides, the slopes of most 220 sampling points in the midstream and downstream are lower than those in the upstream, indicating that 221 the sub-cloud secondary evaporation is stronger in the midstream and downstream than in the 222

upstream. 223
Sub-cloud secondary evaporation 224 Based on the measured meteorological elements of 11 sampling points, assumed the atmosphere of 225 raindrops falling from the cloud base to the ground as a homogeneous form, and used the measured 226 meteorological data of ground sampling points as input parameters, the sub-cloud secondary 227 evaporation was calculated at different sampling points (Table 2). Relevant studies have shown that 228 temperature, relative humidity, raindrop landing height, and raindrop diameter have a certain impact on 229 the strength of the sub-cloud secondary evaporation 11,46,47 . Therefore, this study focuses on the 230 environmental factors to clarify their impact on the sub-cloud secondary evaporation. 231 From the results in Table 2, it can be seen that the changing trend of f and ∆d is basically the same, 232 and the f of most sampling points in the midstream and downstream is smaller than that in the upstream 233 (except for Hongqigu, Dengjiazhuang, Jiuduntan). This indicates that the sub-cloud secondary 234 evaporation in the midstream and downstream is stronger than that in the upstream, which is consistent 235 with the results of the LMWL. Comparing the environmental elements in table 2, it can be shown that 236 the temperature in the midstream and downstream is higher than that in the upstream. This is mainly 237 because the underlying surface causes temperature differences, that the upstream is mountainous but 238 the downstream is desert. At the same time, the downstream belongs to an extremely arid area, where 239 the precipitation is concentrate in summer and scarce or even no precipitation in winter, while the 240 precipitation in the upstream is distributed in every season. Although the average raindrop diameter of 241 most sampling point in the midstream and downstream is larger than that in the upstream, the lower 242 altitudes, long time for raindrops to fall, temperature and precipitation conditions are favorable for 243 evaporation, so the sub-cloud secondary evaporation of the former is stronger than that in the latter. The 244 f of Hongqigu is relatively larger in midstream and downstream. This is because of the influence of the 245 Hongyashan Reservoir nearby, so the larger diameter of raindrops is not conducive to evaporation. The 246 f of Dengjiazhuang and Jiuduntan is also relatively larger. The reason is that the precipitation samples 247 are mostly concentrated in summer with abundant precipitation, and the larger raindrop diameters, the 248 higher relative humidity, the lower height of raindrops, and the short time for raindrops make the 249 sub-cloud secondary evaporation weaker than other sampling plots in the midstream and downstream. 250 Under the influence of the sub-cloud secondary evaporation, the light isotopes are depleted and 251 the heavy isotopes are enriched. In the Shiyang River Basin, the sub-cloud secondary evaporation in 252 the upstream is weaker than that in the midstream and downstream, so the δ 18 O is more enriched in the 253 former than that in the latter. This conclusion is consistent with previous research results 48,49 . 254 Temporal and spatial changes of d-excess provides a basis for inferring the source of precipitation moisture. Therefore, this study divides the 263 d-excess by time and region and further analyzes its change characteristics (Fig. 2). 264  According to the calculation results of the stratification assumption in Table 3, it can be seen that 309 the f values and the d-excess of the first group (Lenglongling) are relatively low, which indicates the 310 sub-cloud secondary evaporation is strong. This may be related to the raindrop diameter that is the 311 smallest among all sampling points, which is conducive to evaporation. For the sampling points of the 312 secondary group, the f values at 700 hPa-500 hPa (except for the Hulinzhan) reach the maximum, 313 which indicates the sub-cloud secondary evaporation is weaker. The f value at ground-700 hPa is lower 314 than that at 700 hPa-500 hPa at Huajianxiang, Xiyingwugou, and Xiyingzhen, indicating that heavy 315 isotope enriches and light isotope deplete, and they further make the d-excess decrease, but the 316 opposite for Hulianzhan. Comparing the environmental factors, the temperature, relative humidity and 317 raindrop diameter of the Hulinzhan is highly consistent with other sampling points in the same group, 318 while they have the opposite sub-cloud secondary evaporation, which may be related to the landing 319 height of raindrops. The average raindrop landing height at ground-700 hPa is only 303.37 m at 320 Hulinzhan, which causes the raindrop landing time to be shorter than other sampling points, so the 321 sub-cloud secondary evaporation is weaker. The changing trends of the f of each sampling point are the 322 same in the third group, that all of them are ground-850 hPa>700 hPa-500 hPa>850 hPa-700 hPa. 323 Comparing environmental factors, it can be seen that this phenomenon is related closely to temperature, 324 raindrop landing height, and raindrop diameter. At ground-850 hPa, the temperature is the highest 325 among the three levels, and the raindrop diameter is the smallest, this creates good conditions for 326 evaporation. However, the landing height of raindrops is the lowest, so the raindrop landing time is the shortest and the sub-cloud secondary evaporation effect is weaker. At 700 hPa-500 hPa, the landing 328 height of raindrops is the highest among the three levels and the raindrop landing time is the longest, 329 but the temperature is the lowest and the raindrop diameter is the largest, which is not conducive to 330 evaporation and the sub-cloud secondary evaporation is weaker relatively. At 850 hPa-700 hPa, not 331 only temperature is higher, but also the landing height of raindrops is higher, and the raindrops 332 diameter has not reached the maximum, so the sub-cloud secondary evaporation effect is the strongest. 333 Because the meteorological elements at different levels and the height of the cloud base are quite 334 different, which makes the sub-cloud secondary evaporation is different, further makes the stable 335 isotopes of precipitation change in the vertical direction. The third group of sampling points is stratified 336 in detail, so the difference of sub-cloud secondary evaporation is more obvious. At the 850 hPa-700 337 hPa, the sub-cloud secondary evaporation is the strongest, indicating that the heavy isotopes of 338 precipitation are most concentrated in this layer. At the 700 hPa-500 hPa, the sub-cloud secondary 339 evaporation is larger, but the evaporation is restricted by the lower temperature, so the stable isotope of 340 precipitation is gradually depleted as the altitude rises and the temperature drops. At the ground-850 341 hPa, the f reaches more than 90%, indicating that the sub-cloud secondary evaporation near the ground 342 is weaker. At the same time, the ∆d of each sampling point at this layer is greater than -1‰, indicating 343 that the isotope composition of precipitation is less affected by environmental conditions and the 344 change of it is not obvious. 345 In summary, the natural environment of the sampling points is quite different, and there are 346 differences in the sub-cloud secondary evaporation under the various heights of precipitation, but the 347 overall regular pattern is basically the same. In the Shiyang River Basin, the sub-cloud secondary 348 evaporation is the strongest at 850 hPa-700 hPa, making the heavy isotope of δ 18 O is enriched and the 349 d-excess is low. When the air pressure is lower than 700 hPa, the sub-cloud secondary evaporation is 350 weak, and the stable isotopes of precipitation continue to deplete as the altitude rises. 351 Analysis of the moisture source based on wind field and water vapor flux 352 This study used the reanalysis data to analyze the geopotential height and wind field at 500 hPa and 353 700 hPa (Fig.3, Fig.4), the vapor flux fields, and the atmospheric precipitation from the ground to 500 354 hPa (Fig. 5)

362
From Fig. 3 and Fig. 4, it can be seen that the water vapor from westward transport plays a leading 363 role in the Shiyang River Basin, which is consistent with previous studies 49 . At 500 hPa (Fig.3), the 364 geological height contours and wind direction are parallel basically with latitude (except for the 365 summer), and the wind speed has obvious seasonal changes. The wind speed is the largest in winter (40 366 m/s), and that is the smallest in summer (10 m/s). Although the higher wind speed is conducive to 367 sub-cloud secondary evaporation, the study area is deep inland, and the winter is dominated by dry and 368 cold air from the northwest. Therefore, except for the upper mountainous, the precipitation of the 369 winter half-year is sparse and mainly solid precipitation, which causes that the sub-cloud secondary 370 evaporation is generally weaker. At 700 hPa, the wind fields are similar relatively in spring, autumn, 371 and winter, and all of which are dominated by westerly winds. The wind speed is highest in winter 372 (20m/s), and it is weaker in spring, summer, and autumn (10m/s). Compared with 500 hPa, the 373 geopotential height and wind field at 700 hPa are more complicated, the geopotential height is no 374 longer parallel to the latitude, and even a large closed circulation of low-pressure is formed near the 375 study area in summer, which changes the wind direction. This is mainly affected by the influence of the 376 topography of the Qinghai-Tibet Plateau. The water vapor from the Indian Ocean in summer is blocked 377 and enters the study area along the edge of the Qinghai-Tibet Plateau, which increases the water vapor 378 of relatively humid from the south. 379

382
The seasonal characteristics of heavy isotopes in precipitation enriched in summer and depleted in 383 winter are considered widely to reflect the dominance of westerly water vapor 17 . This understanding is 384 also supported by reanalysis data. It can be seen from Fig. 5 that the Shiyang River Basin is affected by 385 westerly water vapor throughout the year. The atmospheric precipitation and water vapor flux from the 386 ground to 500 hPa have similar seasonal changes. Contrary to winter, the vapor transport is stronger in 387 summer and the atmospheric precipitation is the largest. Considering the relative humidity of water 388 vapor that forms precipitation, the lower d-excess in the Shiyang River Basin corresponds to the season 389 when water vapor is replenished in large quantities. That is d-excess is lower in summer and higher in 390 winter. However, the increase of recirculated water vapor of inland from local evaporation will lead to