Moisture Source and Atmospheric Circulation Differences for Summer Rainfall in Different Intensity Classes over Mu Us Sandy Land, China

Abstract

Although heavy rainfall occurs infrequently during summer (June–August, JJA) in the Mu Us Sandy Land (MUSL), it has almost the same contribution to summer precipitation as light rainfall. However, it remains unclear on forcing mechanism of heavy rain events and their differences with moderate and light rainfall events from the perspective of moisture sources. In this paper, based on the Dynamical Recycling Model (DRM), we analyze moisture source and atmospheric circulation differences for summer rainfall in different intensity classes over MUSL. The results show that the moisture of summer precipitation in MUSL comes primarily from external terrestrial moisture supplies from the west and southwest directions. As the precipitation intensity increases, moisture contributions from the southwest direction increase significantly, especially for the northeastern part of the Tibet Plateau (defined as Key Region), which accounts for about 39.3% of all moisture sources for heavy rainfall events. Further analysis reveals that anomalous atmospheric circulations, such as the cyclonic circulation anomaly at lower troposphere and anomaly wave train at middle level, also favor the occurrences of different precipitation intensities. Based on these findings, our paper possibly contributes to the conservation of this fragile ecosystem and the prevention of damage caused by precipitation extremes.

1. Introduction

The Mu Us Sandy Land (MUSL), located in the semiarid region to the northwest of the Loess Plateau (Figure 1a,b), is one of the regions with the most abundant rainfall (with an annual precipitation of 250–400 mm) among deserts/sandy lands in China or even the whole world [1]. Due to the lack of water input from large rivers, the local surface water resources depend almost entirely on precipitation; therefore, studying the variations in precipitation is of great significance for both local water resource management and ecosystem security assessment. Because MUSL lies in the marginal zone of the East Asian summer monsoon, its summer precipitation accounts for the majority (>60%) of its annual precipitation, with large interannual variability as a result of fluctuations in the summer monsoon system [2]. In summer, although most of MUSL rainfall occurs as light rain grade (0.1–10 mm d⁻¹) that accounts for approximately 80.4% of all precipitation events (>0.1 mm d⁻¹), the cumulative precipitation contribution of all light rain events to total summer precipitation amount (31.6%) is comparable to that (33.0%) of heavy rain grade (>25 mm d⁻¹, Figure 1c). Although the number of heavy rain events accounts for only 5.4% of all rainfall days, it can cause secondary disasters, such as flooding and soil erosion, which severely threaten the local fragile ecosystem [3]. As a result, directly averaging the climate conditions of all summer rainfall events with equal weights will seriously underestimate the important impact of heavy rainfall events [4,5,6]. In addition, most current research focuses on the study of the mechanisms underlying extreme precipitation events [7,8], while few focus on discrepancies of rainfall in different intensities as well as their moisture sources and driving mechanisms. Therefore, it is imperative to analyse the moisture source for summer precipitation events at different intensity grades in MUSL and explore their forcing circulation mechanisms, thereby posing an important basis for protecting the local ecological environment.

Two main methods have been widely used to classify the precipitation events, including the absolute and relative threshold methods [9,10]. The absolute threshold method for precipitation levels in China generally adopts the China Meteorological Administration (CMA) rainfall intensity classification criteria (heavy rainfall: 25.0–49.9 mm d⁻¹; moderate rainfall: 10.0–24.9 mm d⁻¹; light rainfall: 0.1–9.9 mm d⁻¹) [11]. CMA is designed for operational and impact-based classification and serves for capturing physical intensity of precipitation [12,13]. The relative threshold methods, which are intended to identify climatological extremes relative to local norms and aim to capture statistical extremity [14,15], generally follow the percentile threshold as suggested by ETCCDI (Expert Team on Climate Change Detection and Indices, http://etccdi.pacificclimate.org/list_27_indices.shtml (accessed on 4 January 2026)), such as R90p, R95p, and R99p, and by selecting those events that exceed the mean ± k times standard deviation (e.g., k = 1, 1.5, or 2). Since the purpose of this study is to compare the differences in moisture sources and circulation mechanisms among precipitation events of different grades and the percentiles corresponding to different precipitation intensity grades of 11 meteorological stations in MUSL during summer obtained using the absolute threshold method are relatively consistent (Table S1), the absolute threshold method can better help us accurately distinguish and assist in further analysis (see Section 2.1 for details).

The water vapor for a precipitation event originates from either the local evapotranspiration or moisture input from upwind regions [16,17], and unraveling the relative contribution from different moisture source regions to the precipitation helps to understand the causes of extreme drought and flood events and is of vital significance for anomalous precipitation prediction and regional water resources management. Currently, various methods have been used for identifying moisture sources, including isotope composition analysis [18,19] and moisture source tracing models [20,21]. Although the isotope composition analysis method can quantitatively determine moisture source at a single point, it is difficult to obtain the spatial and temporal variation distribution of moisture source over a large range and for a long time because this method depends heavily on in situ measurements [22]. Moisture source tracing models, most of which assume the atmosphere is “well mixed” in that the proportion from surface evaporation is relative to external moisture input, is consistent with those during the precipitation process [23,24], and this hypothesis is fit for most regions (including the study area), except for where strong vertical wind shear is present, can be roughly divided into Eulerian and Lagrangian methods [25,26]. Although the moisture tracing model based on Eulerian formula has high computational efficiency [27], it is difficult to get a detailed tracking path of moisture transport. By comparison, Lagrangian model is capable of featuring the moisture source tracks [28], but it needs substantial computation costs, particularly for long-term and high-resolution studies [29]. Therefore, we can use a simpler but efficient model to identify the moisture source, such as DRM. The DRM, proposed by Dominguez et al. [30], originates from the atmospheric water balance equation (appearing to be Eulerian-based), while its coordinate system moves with the wind (with moisture content weighting, i.e., the ratio of vertically integrated moisture flux to precipitable water), focusing on air parcels moving with the wind, making it as a Lagrangian model widely used for solving the precipitation recycling in many regions (e.g., middle and lower Yangtze River) [31] and has been refined by Martinez and Dominguez [32] in order to calculate the moisture contribution of any area to the sink and the moisture transport trajectory. Moreover, most models assume a negligible changes in water vapor content with time when compared with other terms of atmospheric cycle, which is applicable on monthly time scale or longer but does not hold for daily time scale or shorter, while DRM fully accounts for atmospheric moisture variations over time and is suitable for investigating moisture sources of precipitation events in different intensities on daily scale (e.g., light rain, moderate rain, and heavy rain).

Although DRM has been widely used for many regions in existing research [33], most of which only discuss the attribution of precipitation from the perspective of moisture source. However, we not only investigate the moisture sources in MUSL during summer based on DRM but also compare the source differences among precipitation events of varying intensities. Furthermore, by integrating atmospheric circulation mechanisms, we also explore the forcing mechanisms behind rainfall events in different intensities, thereby offering to contribute to the conservation of this fragile ecosystem and the prevention of damage caused by precipitation extremes.

2. Data and Methodology

2.1. Station Data and the Classification Method of Precipitation Events

The daily precipitation data from a total of 11 meteorological stations in MUSL are utilized in this paper (Figure 1), which was collected and preprocessed by CMA. According to grade of precipitation “GB/T28592-2012” [34], the summer precipitation events of MUSL are categorized into light rain, moderate rain, and heavy rain by the following steps: (1) When the daily precipitation at one station is between 0.1–10 mm d⁻¹, it is defined as a light rain event at this station; between 10–25 mm d⁻¹ as a moderate rain event; and exceeds 25 mm d⁻¹ as a heavy rain event. (2) When precipitation events occur at no less than 25% of all stations in MUSL on the same day, that day is defined as a regional precipitation event [35]. From 1980 to 2020, based on step 1, a total of 2112 light summer rain events, 775 station moderate rain events, and 343 station heavy rain events were selected for all stations in MUSL, from which we obtained 1404 regional light rain events, 248 regional moderate rain events, and 94 regional heavy rain events in MUSL based on step 2.

2.2. ERA-5 Reanalysis Data

To analyze the moisture transport characteristics and atmospheric circulation of summer precipitation events in different classes, we use the ERA-5 from the European Centre for Medium-Range Weather Forecasts (ECMWF, with a spatial resolution of 0.25° × 0.25° and a temporal resolution of hourly) [36,37], including the daily precipitation, evaporation, precipitable water, vertically integrated moisture flux (VIMF, the calculation formula is $\stackrel{⃑}{\mathrm{Q}}={\displaystyle \frac{1}{\mathrm{g}}}{\int }_{{\mathrm{p}}_{\mathrm{s}}}^{{\mathrm{p}}_{\mathrm{t}}}\mathrm{q}\stackrel{⃑}{\mathrm{v}}\mathrm{d}\mathrm{p}$, where q is the specific humidity (kg kg⁻¹), $\stackrel{⃑}{v}$ is the wind vector (m s⁻¹), p_s is the surface pressure (hPa), p_t is the pressure of the top atmosphere (hPa), and g is the gravitational acceleration (m s⁻²), as well as geopotential height and wind fields at different isobaric layers from 1980 to 2020. Here, we employ the composite analysis method to examine atmospheric circulation conditions and moisture source contribution patterns for precipitation events in different intensities. Specifically, both the geopotential height, wind field, and moisture flux from ERA-5 and moisture contribution of various source regions from DRM of each regional rainfall event at different intensities were calculated, the differences in relation to their corresponding multi-year averages (after a 10-day filtering). These anomalies were then averaged to make a composite analysis, and the significance of their differences was tested using a t-test. Additionally, since the daily station data are collected and observed based on (BJT), from 20:00 Beijing Time (BJT) of the previous day to 20:00 BJT of the current day, all the ERA-5 data are shifted from Coordinated Universal Time (UTC) to BJT.

2.3. Dynamical Recycling Model

The DRM, initially proposed by Dominguez et al. [30] and later refined by Martinez and Dominguez [32], which has been widely used to quantify the relative contribution percentage from surface evaporation to precipitation, is utilized in this paper to analyze the moisture supply from various source regions during precipitation events in MUSL. Since fully accounting for temporal variations in atmospheric moisture content, this model is able to track moisture of precipitation at daily scale, which is suitable for this study to analyze the moisture attribution for individual precipitation events. In essence, the DRM calculates, when moisture is transported horizontally by an effective 2D wind field derived from the vertical integral of the horizontal moisture flux and the total column moisture, how much is collected from surface evaporation by an atmospheric column and how much this air column loses via precipitation. All calculations can be directly implemented through this model [33]. For a given grid point (i) in the target area (i.e., MUSL), by calculating the integral of its backward moisture transport trajectory, its contribution percentage ρ_i(s) of surface evaporation to precipitation over the grid i at time t is

$${\rho }_i\left(s\right)=1-exp\left[-{\int }_{s_0}^{s_t}E/w ds\right]$$

(1)

where E is evaporation, w is precipitable water, s₀ and s_t indicate the start and end position of the target grid point (i), and s stands for the moisture transport trajectory. The backtracking trajectory along the path (s) can be divided into different source grids, thereby enabling the calculation of the total precipitation contribution ratio a_k from all grid points k along the path to the target grid point (i) as

$$a_k\left(x,y,t\right)=\sum _{n\in A_k}\left\{\left[\prod _{i=1}^{n-1}\left(1-{\rho }_i\left(s\right)\right)\right]{\rho }_n\left(s\right)\right\}$$

(2)

where the sum is done over all parts of the trajectory that fall into the region A_k. Finally, absolute contribution P_m of the k-th region to the target region A_i and the relative contribution rate (ρ_r,i) are obtained as follows:

$$P_m\left(A_k,A_i,t\right)=\sum _{\left(x,y\right)\in {\mathrm{A}}_{\mathrm{i}}}{a}_k\left(x,y,t\right)P\left(x,y,t\right)\delta A\left(x,y\right)$$

(3)

$${\rho }_{r,i}={\displaystyle \frac{P_m\left(A_k,A_i,t\right)}{P\left(x,y,t\right)}}={\displaystyle \frac{{\sum }_{\left(x,y\right)\in {\mathrm{A}}_{\mathrm{i}}}{a}_k\left(x,y,t\right)P\left(x,y,t\right)\delta A\left(x,y\right)}{{\sum }_{\left(x,y\right)\in {\mathrm{A}}_{\mathrm{i}}}P\left(x,y,t\right)\delta A\left(x,y\right)}}$$

(4)

where P is precipitation, P_m refers to total contribution from evaporation in the k-th region, and $\delta A\left(x,y\right)$ represents the area of the grids.

Moreover, to better understand the moisture contribution from different source subregions during summer rainfall in MUSL, we divide all the source grids of precipitation in MUSL into 24 major subregions (Figure S1): Key Region in the northeastern TP (Key Region, southwestern source region), the rest part of Tibet Plateau (RP. TP, southwestern source region), Central Asia (C. Asia, western source region), Northwest China (NW. China, western source region), South Asia (S. Asia, southwestern source region), Western Russia (W. Russia, northern source region), North China (N. China, eastern source region), West Asia (W. Asia, southwestern source region), Mongolia (northern source region), MUSL, Eastern Russia (E. Russia, northern source region), Indian Ocean (southwestern source region), Europe (western source region), Atlantic (western source region), Pacific (eastern source region), the Black Sea and Caspian Sea (BC. Sea, western source region), South China (S. China, southern source region), Mediterranean (western source region), Africa (southwestern source region), Southeast Asia (SE. Asia, southern source region), Arctic (northern source region), America (western source region), Oceania (southern source region), and Korea and Japan (eastern source region). Here, the northeastern TP is considered Key Region because of the largest amount of moisture contribution to summer rainfall in MUSL from the moisture contribution pattern of source regions.

3. Results

3.1. Temporal and Spatial Distribution in Summer Precipitation in MUSL

Climatologically, both ERA-5 and station data show increasing trends in summer precipitation from northwest to southeast in MUSL. However, the ERA-5 data appear to overestimate the summer precipitation when compared with station data, especially in the southeastern region of MUSL, and their regional-averaged summer daily precipitation is 2.54 mm d⁻¹ and 2.26 mm d⁻¹, respectively (Figure 2a). Although ERA-5 precipitation data show overestimation to some degree, the regional mean time series of summer precipitation for both data sources are closely correlated with the correlation coefficient as high as 0.85 (p < 0.001), demonstrating their high consistency in interannual variability (Figure 2b). Moreover, the scatterplot of daily precipitation from two sources (Figure 2c) also shows a good agreement at the daily scale (R² = 0.7243). From 1980 to 2020, the overwhelming majority of summer precipitation events in MUSL occurred at light rain grade (80.4%), while only 14.2% and 5.4% of the precipitation events were found at moderate and heavy grades, respectively. However, their contributions to the total summer precipitation are relatively balanced, at 31.57%, 35.37%, and 33.05%, respectively (Figure 1c). Although heavy rain events have low frequency, they do contribute a rather considerable proportion to the total summer precipitation, indicating that heavy rain plays a significant role in summer precipitation over MUSL.

3.2. Characteristics of Moisture Contribution for Different Rainfall Intensities

The results of DRM show evident differences in the absolute contribution for different precipitation levels: as the level increases, the extent of grids that have moisture contribution of over 10³ m³ expand evidently, and the moisture contribution per grid in the major source regions also increases substantially (illustrations in Figure 3). For instance, the moisture of light rain in MUSL mainly comes from its adjacent southwestern region with the moisture contribution per grid exceeding 20 × 10³ m³ in Key Region (purple box in Figure 3a). As the class increases, the regions where the moisture contribution of unit grid exceeding 10³ m³ expand toward distal region (Figure 3b,c), with the region of higher value extending toward the southwest. When moderate rain occurs, the moisture contribution of each grid in Key Region continues to increase, exceeding 100 × 10³ m³ (Figure 3b), and rises to 400 × 10³ m³ when heavy rain happens (Figure 3c). It can be observed that as the level increases, although the moisture source expands, the area with substantially high amount of contribution gradually concentrates in southwestern regions to MUSL, particularly in Key Region. However, because the absolute moisture contribution is proportional to the precipitation in sink region, it is inevitable that the contributions from source regions increase substantially as the precipitation levels rise. To better compare the relative importance of different source regions for precipitation events of different intensities, we mainly focus on the relative contribution percentage from different source regions.

From the vertically integrated water vapor flux, summer rainfall in MUSL mainly stems from western and southwestern directions (Figure 4), while from the distribution of relative contribution, Key Region in the northwestern TP contributes the most (32.3%) of moisture to all the summer rainfall events. Other major source regions from the western and southwestern direction, including the RP. TP, C. Asia, NW. China, and S. Asia, also supply about 10% of the moisture for summer precipitation (Figure 4d). By comparison, a total of no more than 20% of moisture comes from the east and north directions, such as Mongolia, N. China, E. Russia, Pacific, etc. (

Table 1. The moisture contribution percentage (%) of source regions from different directions to precipitation events in different classes.

Regions	Light Rain	Moderate Rain	Heavy Rain
Southwestern Source Regions	41.3%	54.5%	61.2%
Western Source Regions	23.0%	19.9%	14.2%
Northern/Eastern Source Regions	17.0%	9.8%	11.4%
Mu Us Sandy Land	3.1%	2.4%	1.8%

). Moreover, because of the small area and the limited surface evapotranspiration, the local MUSL contributes only 2–3% to all precipitation events (i.e., the precipitation recycling ratio is only 2–3%). Therefore, the vast majority of moisture in summer precipitation comes from external water vapor input, of which the moisture supplies from the southwestern and western regions play a significant role in occurrence of summer rainfall events in MUSL.

As the rainfall grade grows, the moisture contribution percentage from southwestern directions exhibits increasing trend, while that from other directions generally shows decreasing trends (Figure 4d, Table 1). For instance, moisture contribution percentage from source regions located to the southwest of MUSL increases as the precipitation level increases, including Key Region, RP. TP, S. Asia, and W. Asia (Figure 4d), which collectively occupy about 61% moisture of heavy rain in MUSL (Table 1). In particular, Key Region, as the largest moisture supply region for summer precipitation, shows an evident increase in contribution rate from 27.3% for light rain to 39.1% for heavy rain (Figure 5). In contrast, although the source regions from the western direction (including NW. China and C. Asia) contribute 23.0% moisture to the light rainfall, their relative contribution percentage decreases dramatically to 14.2% for heavy rainfall. The source regions from the east and north directions (including N. China, Mongolia, and E. Russia) have relatively low contribution percentages and generally decrease as precipitation intensity increases. In addition, the recycling ratio of MUSL is relatively low and decreases with increasing precipitation levels. Compared to terrestrial regions, the moisture contribution rate from marine is quite low (only 3–4%), which also shows a slight decrease as precipitation grade grows. Generally, light rain events tend to have relatively dispersed moisture sources (Figure 5) with relatively more comparable contributions from different directions, whereas heavy rain events tend to have more concentrated moisture sources from southwestern direction (such as the entire TP and S. Asia), highlighting the importance of the southwestern source regions in supplying moisture for extreme summer precipitation events in MUSL.

3.3. Atmospheric Circulation Responsible for Different Rainfall Intensities

Results of moisture tracing model display the important role of Key Region to the southwestern of MUSL in supplying ample moisture to summer precipitation events, particularly for heavy rain, which implies the potential forcing mechanism from southwest monsoon. Therefore, we also analyse the atmospheric circulation pattern responsible for different precipitation level events. When summer rainfalls occur, the lower troposphere (850 hPa) is characterized by two significant high-pressure anomaly systems over the west and east sides of MUSL, which are divided by an anomalous low pressure that extends from southwest to northeast across western China on southern side of MUSL (Figure 6). With the increase in precipitation levels, the high-pressure anomaly on the east side intensifies evidently, accompanied by a corresponding enhancement in anticyclonic circulation anomaly. Simultaneously, the low-pressure anomaly on the southwestern China also markedly strengthens and extends northward as precipitation upgrades, reaching the southeastern boundary of MUSL during moderate rain (Figure 6b) and fully covering the entire MUSL during heavy rain (Figure 6c), which is accompanied by a strong cyclonic wind shear with intensified convergence and ascending motion at low level of troposphere, thereby providing favorable conditions for precipitation occurrences in MUSL.

At mid-level of troposphere (500 hPa), when MUSL experiences precipitation in summer, there is a significant anticyclonic anomalous circulation centered over the Liaodong Peninsula to the east of MUSL, which keeps intensifying as the precipitation level increases (Figure 7), indicating that the Western Pacific Subtropical High (WPSH) is enhancing with its position moving northwards and westwards compared to its climatological condition. Under its control, the intensified southwesterly anomalous winds on the western side of this anomalous anticyclonic system facilitate the transport of abundant warm and wet moisture into MUSL, providing substantial moisture supply for precipitation (Figure 7). Meanwhile, a significant low-pressure anomaly exists to the western side of MUSL, with a cyclonic anomaly circulation. The enhanced southwesterly wind on the eastern side of this circulation overlaps with that on the western side of the high-pressure anomaly, further increasing the moisture transport of the southwesterly anomalous winds to MUSL. With the intensification of precipitation level, high-pressure anomaly on eastern side further strengthens, which helps to enhance the warm moisture transport into MUSL. Additionally, there is a high-pressure anomaly in the West Siberian, forming a “high-low-high” anomaly wave train along a northwest-southeast direction over the mid-high latitudes of East Asia (Figure 7b,c), which becomes stronger as precipitation level increases. This wave train guides cold air entering the study area from Siberia, which converges with warm and wet moisture over the study area, favoring precipitation there.

At upper level of troposphere (200 hPa), when precipitation events occur in the MUSL, the South Asia High (SAH, with gpm-200 > 12,500 gpm) is positioned further eastwards compared to its climatological condition (Figure 8), which is accompanied by a large-scale anticyclonic circulation anomaly over northern China. Under the influence of this anticyclonic circulation anomaly, the axis of the westerly jet stream (U-200 > 28 m s⁻¹) shifts eastward. MUSL is located at the southwestern corner of this axis anomaly; as a result, the marked eastward in westerly enhances the divergence over upper troposphere, which favors ascending motion with intensified moisture condensation, thereby promoting the occurrence of precipitation (Figure 8). With the increasing intensity, the SAH moves further northeastward; the position of the westerly jet axis shifts more eastward and extends eastward to approximately 130° E during heavy rain (Figure 8c). Additionally, the eastward displacements of the westerly jet and SAH maintain the westward and northward movement of the WPSH, further promoting the occurrence of rainfall in MUSL.

4. Discussion

4.1. Impact of Threshold Selections for Different Precipitation Intensities

Based on the absolute threshold method (mentioned in Section 2.1), precipitation levels have been divided into light rain, moderate rain, and heavy rain. However, this artificial standard might have caused the results to be affected by the selection of threshold values. Therefore, we further refine the precipitation level classification by using daily precipitation amounts of 1, 5, 10, 25, and 50 mm d⁻¹ as absolute threshold values for division, from which the results (Figures S2 and S3) are consistent with those mentioned above (Figure 4d and Figure 5), illustrating that the precipitation classification method selected in this paper has a negligible impact on the results. Additionally, we also calculate the regression coefficients of the relative contribution percentage from each source grid against the regional mean daily precipitation of all summer precipitation events in MUSL, showing that as the amount of precipitation increases, moisture contribution ratio from the southwestern source regions increases with the highest increase observed in northeastern TP (Figure 9) that passes the significance test (p < 0.05), while the source regions from other directions generally exhibit decreasing trends, particularly in the local MUSL and its adjacent northeastern region, as well as NW. China, where the reduction is most pronounced. These results are consistent with those obtained in Figure 5 and Table 1, highlighting the crucial positive influence of moisture from the TP and S. Asia for summer precipitation in MUSL, while also demonstrating the negative impact of moisture from eastern and northern regions for the increase of precipitation.

4.2. Limitations

In this study, although there is a high correlation coefficient in daily precipitation between ERA-5 and stations (Figure 2), their root mean square error (RMSE) is 0.51 mm d⁻¹, and relative bias (RBias) is 5.9%, indicating overestimation in ERA-5 precipitation data. This is possibly because the ERA-5 reanalysis data assimilate multiple sources, including surface observation, satellite images, radar measurements, profile data, and other data sources [38,39], while relatively lower quality of the reanalysis data in regions with sparse distribution of stations may not fully represent the detailed precipitation pattern in MUSL [40,41]. Essentially, in the DRM, precipitation is required when calculating the absolute moisture contribution; the overestimation might have caused overestimation for the absolute contribution pattern. According to a previous study, Li et al. [42] used different data sources as the input data for moisture track models; they found that, compared to MERRA-2, which shows a minor overestimation of precipitation, those driven by ERA-5 tend to have a slight overestimation of the absolute moisture contribution of moisture, mainly for the remote source regions. Nevertheless, this overestimation does not affect the relative contribution pattern because precipitation is not included as a parameter in its calculation. Although overestimation of precipitation has impacts on the absolute contribution pattern, our results about moisture source distributions of MUSL are still consistent with those of previous studies, based on other moisture source tracing models [43,44,45], all of which supports that moisture of semiarid regions of China is transported by the Asian summer monsoon from the Indian Ocean and the westerlies from northwestern China–eastern central Asia. Besides, South China and the North China Plain have been identified as the critical moisture source regions for the summer extreme precipitation events over the semiarid regions in both Hu et al. [44] and Zhang et al. [17]. In addition to these important moisture sources, in this study, we find a key moisture source region in northeastern part of TP that supplies substantial moisture to summer rainfall over MUSL, particularly for the heavy precipitation events, which stresses necessity of analyzing the moisture source of precipitation in different intensity.

On the other hand, as performed by other commonly used moisture tracking models, the DRM also assumes the atmosphere is “well mixed.” Although this assumption holds true in most cases, compared with more complex models, the hypothesis of complete vertical mixing may simplify the real atmospheric condition so that multilayer models are required to capture moisture transport in the vertical direction, as locally evaporated water is more concentrated near the surface [46,47]. However, for the research area (MUSL) in this paper, which is located in the semiarid region of China, moisture transport is usually weak, and most of the water vapor is concentrated in the middle and lower layers of the troposphere [48]. Therefore, the one-layer DRM model adopted in this paper can effectively identify the moisture transport characteristics of different intensity precipitation events in MUSL during summer and can also save a large amount of time and computational cost. Although the DRM might have some limitations [49,50], the results of which are still similar to those of existing research that focus on similar study areas, they all indicate that land evaporation moisture from the west and southwest is the main source of summer precipitation in northern China and semiarid regions of China [17,43,44,45]. In the future, it is necessary to utilize multiple moisture tracing methods that incorporate numerical models to get more precise moisture sourcing analysis.

4.3. Forcing Mechanism Behind Rainfall Events in Different Intensities

The results of atmospheric circulation indicate that southerly winds prevail over MUSL during summer rainfall (Figure 7), but the moisture contribution from the southern regions is relatively small (Table 1), possibly because, although the southerly wind transports abundant moisture northward, most of which precipitates before reaching MUSL. By the time it arrives, the amount of warm and wet moisture is significantly reduced, making heavy precipitation unlikely. In addition to the impacts of the moisture transport, the influences of anomalous atmospheric circulation on heavy rainfall have also been confirmed by existing studies [51,52]. When precipitation events occur, cyclonic wind shear anomaly exists at lower troposphere to the southwest of MUSL, conducive to the convergence, and uplift of atmospheric moisture further contributes to form precipitation (Figure 6). As this wind shear anomaly continues to intensify and extend northward, eventually fully controlling the entire MUSL, the precipitation process will gradually strengthen. Meanwhile, the WPSH at mid-level of troposphere enhances gradually with its position moving westwards compared to its climatological condition (Figure 7b,c), which intensifies southwesterly wind and facilitates the transport of abundant warm and wet moisture into MUSL, providing substantial moisture supply for heavy rainfall and making Key Region dominate in heavy precipitation. At the same time, there is a “high-low-high” anomaly wave train along a northwest-southeast direction over the mid-high latitudes of East Asia, guiding cold air entering the study area from Siberia, which converges with warm and wet moisture over the study area, favoring precipitation there. At upper level of troposphere, the SAH and the westerly jet have been verified as the large-scale systems with major impact on summer precipitation over northern China [53]. When the SAH increases and moves eastward (Figure 8), together with an enhanced WPSH, the intensified southwesterly wind would transport more moisture supplies from the southwest directions with increases in the contribution from Key Region. Besides, as the westerly jet stream intensifies and shifts eastward, MUSL is located at the southwestern corner of an anticyclonic circulation anomaly. The intensified westerly jet enhances divergence over upper troposphere and then favors ascending motion with intensified moisture condensation, further enhancing precipitation in MUSL.

5. Conclusions

In MUSL, although the number of heavy rain events makes up a much smaller percentage of all summer precipitation events than that of light rain, they have a roughly equal share of the precipitation amount, while the heavy rainfall may cause more catastrophic damage over such an ecologically fragile region. However, research on the discrepancies at different classes from the perspective of moisture sources remains limited. Therefore, based on the DRM, this paper analyses the variations of moisture sources in different intensity classes during summer in MUSL from 1980 to 2020 and investigates the differences of atmospheric circulation. The results show that MUSL is majorly supplied by moisture mainly from four regions during summer, including Key Region (32%), RP. TP (11%), C. Asia (10%), and NW. China (9%), while the local precipitation recycling ratio is much smaller (only 2.6%). However, the moisture contribution rates from four regions vary with the intensity of rainfall. Key Region and RP. TP from the SW. source regions exhibit more significant contributions during heavy rain, while C. Asia and NW. China contribute more to light rain. According to the analysis of atmospheric circulation patterns, compared to light rain, the forcing mechanisms behind heavy rain are more intense, such as the strong cyclonic wind shear with intensified convergence and ascending motion at low level of troposphere; the WPSH at mid-level of troposphere significantly strengthens with the strengthening southwesterly wind, as well as the SAH and the westerly jet at upper level of troposphere extending eastward and northward with the enhanced divergence and ascending motion. This paper analyzes the main moisture source regions for MUSL summer rainfall from the perspective of moisture tracing and the forcing mechanisms behind rainfall events in different intensities. Not only does our research supplement current studies on the mechanisms that influence the occurrence of precipitation in different levels in MUSL, but also, in conjunction with previous research, it contributes to the protection of the ecological environment and provides a strong scientific basis for improving the accuracy of precipitation predictions.