The Relative Contribution of Large-Scale Circulation and Land Surface to Summer Precipitation Over Asian Mid-Low Latitudes

Understanding the contributions of large-scale atmospheric circulation and local land surface processes to precipitation is essentially important for the climate prediction. This study adopts a dynamic adjustment (DA) approach based on constructed circulation analogs to quantitatively isolate the contribution of atmospheric circulation to summer land precipitation (Pr) over Asian mid-low latitudes during 1980-2019. The atmospheric circulation factor is represented by the 500 hPa geopotential height (Z500) from the fth generation ECMWF reanalysis (ERA5), and the land surface factors, including soil moisture (SM) and net radiation and heat uxes are from the products of the Global Land Data Assimilation System (GLDAS). The residual component after DA is regarded as the contribution from land surface processes via evaporation mainly resulting from SM. The results indicate that the key SM-Pr feedback areas are mainly located in northeast China and the northern Indian Peninsula. The key inuencing area of Z500 on the land Pr anomaly shows a “-+-” tripole pattern in the mid-latitude region. Atmospheric circulation determines the magnitude of summer land Pr, while the residual components reect the land-atmosphere coupling effect and dominate Pr trend. This conclusion is helpful for better understanding the evolution mechanism of summer climate over Asia mid-low latitudes and may also have application value for climate prediction.


Introduction
The Asian mid-low latitudes contain tropical and subtropical monsoon regions and "hot spots" for landatmosphere coupling, where accurately predicting precipitation (Pr) is always a great challenge ; Koster et al. 2006). As one of the typical monsoon climate areas in the world, the Pr in summer (June-July-August, JJA) over East Asia, accounts for 52% of its annual total (Zhang 2015). Many studies have shown that large-scale atmospheric advection plays a leading role in the JJA Pr over Asian monsoon region, and the 500 hPa circulation system has a close relationship with the East Asian summer monsoon (EASM) and Indian summer monsoon Pr variability (Aizen et al. 2001; Ding and Chan 2005;Huang et al. 2012;Meehl and Arblaster 2001). Therefore, the geopotential height (Z500) and other relevant indices at the 500 hPa pressure level are often used as predictors of the JJA Pr over this region in operational climate prediction systems. The land surface processes also in uence the formation and development of the land climate through the energy and water cycles between the land and atmosphere (Shukla and Mintz 1982). The results from numerical experiments indicate that soil moisture (SM) signi cantly in uences Pr in boreal transition zones between wet and arid climates (Koster et al. 2004; Koster et al. 2006). Understanding the respective in uences of the slowly varying land surface state and atmospheric components on Pr is very helpful for its prediction.
As an important land surface hydrological component, SM determines the partition of net radiation (R net ) into sensible heat ux (SHF) and latent heat ux (LHF) (Koster et al. 2004). The SM anomalies state and its persistence are of great signi cance to the prediction of summer climate in the Northern Hemisphere  (Eltahir and Bras 1996) or indirectly affect it through altering boundary layer dynamics and mesoscale circulation (Ek and Holtslag 2004;Taylor et al. 2011). These studies either analyze the in uencing mechanism of SM on Pr or compare the impacts of SM and SST on Pr through initial condition sensitivity modeling experiments. However, few of these studies have addressed the relative contributions of atmospheric circulation and land surface components to land Pr over Asian midlow latitudes. It has been revealed that SM anomalies may last for several weeks to months. Therefore, it could in uence the successive Pr through SM-Pr feedback mechanism (Jones and Brunsell 2009). Besides, there are many other sources of predictability for Asian summer Pr, such as Arctic sea ice, SST and Eurasian snow cover. Among them, the impacts of sea surface temperature, sea ice, and remote snow cover on inland Pr are mainly achieved by adjusting the atmospheric circulation (Barnett et al. 1989; Goddard et al. 2001;Trenberth et al. 1998; Wu et al. 2016), while the local snow melt in spring may also in uence the SM through its memory to impact subsequent soil heat and water properties in summer (Yang and Wang 2019). Therefore, it is reasonable and ideal to use Z500 and SM as the respective indicators to measure the sources of land Pr in Asian mid-low latitudes.
The dynamic adjustment (DA) approach has been developed to isolate the climate change caused by atmospheric circulation (Deser et al. 2016 a global climate model ensemble and indicated that the internal atmospheric variability had a signi cant impact on the winter Pr trend in part of Europe. The above studies have proven that DA is a useful and effective approach for separating the contributions of atmospheric circulation patterns to the climate. However, DA has not been used to explore the in uence of large-scale circulation patterns on the land Pr in Asian Mid-Low Latitudes. The main purpose of this study is to quantitatively separate the contributions of atmospheric circulation and local evapotranspiration (ET) to summer Pr over Asian mid-low latitudes. Z500 is used to represent the component of atmospheric circulation, and SM is used as a comprehensive indicator of land surface processes. To execute DA, the constructed circulation analog (CCA) approach is applied to identify the in uence of atmospheric circulation on summer Pr. The residual Pr after DA is de ned as the contribution of the land surface processes. Finally, the physical mechanism of land-atmosphere coupling on Pr is intensively explored.
The paper is organized as follows. The reanalysis datasets are described in Section 2. Section 3 describes the DA-CCA approach and other analysis methods used in this study. Section 4 analyzes the relationships between summer land Pr in Asian mid-low latitudes and Z500/SM and the characteristics of the circulation and residual components of Pr after DA. The summary and discussion are provided in Section 5.   Table 1). Correspondingly, the SM in the latter version was also signi cantly higher than that in GLDAS-2.0 during 2007-2011, and soil in both products display strong wetting trend (Table1). Research has shown that Pr in GLDAS-2.1 is generally closer to observations than the GLDAS-2.0 on average, although both datasets underestimate Pr in majority river basins in China (Qi et al. 2020). The large differences in radiation and turbulent uxes between the two versions may be due to the bias-corrected of AGRMET radiation elds and SRB in the GLDAS-2.1 dataset (https://hydro1.gesdisc.eosdis.nasa.gov/data/GLDAS/GLDAS_NOAH10_M.2.1/doc/README_GLDAS2.pdf). Therefore, it can conclude that GLDAS-2.1 is overall more reliable than GLDAS-2.0 over land area in Asia. transpiration affected by SM should also be discussed within this depth. Therefore, the SM in the 0-40 cm layer (i.e., the summation of the upper two soil layers) was used in this study.

ERA5 products
ERA5 is the fth generation ECMWF reanalysis product for the global climate and weather and has been available since 1979 (Hersbach et al. 2020). The atmospheric model used to derive the ERA5 products has 137 hybrid sigma/pressure vertical levels with the top level at 0.01 hPa. The advanced 4D-Var data assimilation system was incorporated into the ECMWF's Integrated Forecast System (IFS). Compared to ERA-Interim, ERA5 bene ts from many improvements, including more enriched observation datasets and advanced assimilation systems (Hersbach et al. 2020). Because the ERA5 product has a relatively high resolution (0.25°×0.25°) and a state-of-the-art global atmospheric reanalysis, it has been applied for investigating synoptic-scale extratropical circulation, allowing for detailed study of atmospheric motion features (Rohrer et al. 2020). In this study, we used the monthly 500 hPa geopotential height (Z500), vertical integral of water vapor uxes from the land surface to the top of the atmosphere and the planetary boundary layer height (PBLH) from the ERA5 products for the period of 1980-2019. The ECMWF convention for vertical uxes is positively downward. All datasets were interpolated to 1°×1° to match the horizontal resolution of the merged GLDAS dataset. From the map distribution of climatology mean of JJA Pr and vertical integral of horizonal water vapor ux (Fig. 1b). It can be seen that regions with large water vapor convergences are well matched with the high Pr. This indicates that the water vapor transport through the large-scale atmospheric circulation is the dominant water sources of Pr in the Asian mid-low latitude land areas.

Dynamic adjustment using the constructed circulation analog approach
The CAA approach, rst proposed by Lorenz (1969), is a statistical method for weather and climate prediction. In recent decades, the CAA has been developed and widely applied to infer the dynamic contribution of atmospheric circulation to climate trends (Deser et al. 2016;Hurrell 1995). Because the origins of moisture for land Pr come from remote areas (e.g., ocean surfaces) through the advection of large-scale circulation and/or local ET, we employed the DA-CAA to isolate the contributions of the above two components to the JJA Pr in the study region.
In the following, we brie y describe the DA-CCA procedure used in the current study, and a detailed description of the method can be found in Deser et al. (2016). For each summer month during the period of 1980-2019, the domain of 0°-80°N, 20E°-180°E was used to compose Z500 analogs (Fig. 1a). Taking June 1980 as an illustrated example, we randomly subsampled 30 from 39 June (after excluding the target month, i.e., June 1980) Z500 elds, and then they were used to composite the optimally linear tting of the target month, which was then denoted as the rst circulation analog of June 1980 (C1). Having applied the above optimally linear tting procedure to the corresponding 30 June Pr elds, we obtained the rst optimal Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js tting Pr component eld (P1) for the target month that was caused by atmospheric circulation. This random sampling procedure was then repeated 100 times to reduce the sampling errors in previous study (Merri eld et al. 2017). To test the impact of resampling times on the results, four groups of tests including 100, 200, 500 and 1000 times of resampling are conducted and the results among those tests show very slightly different (the correlation coe cients between the calculated results are greater than 0.9 at the 5% signi cance level). In order to reduce the sample error to large extent, the DA-CCA results obtained after repeated 1000 times of resampling are used in our study. The average of 1000 Z500 analogs (C1, C2..., C1000) and their associated Pr components (P1, P2..., P1000) were used to quantitatively estimate the effect of atmospheric circulation on Pr in the target month. After repeating the above steps for all 120 summer months, we obtained a complete reconstruction of the dynamically induced Pr component in each month, denoted as Pr-Circ. Finally, the Pr-Circ component was subtracted from the raw Pr eld of each month, and the residuals (Pr-Res) were interpreted as being primarily driven by land surface processes

Other analysis methods
To facilitate analyze the relationships between different variables, each variable was rstly normalized by subtracting its long-term monthly mean and then dividing by its temporal standard deviation (STD). The normalization results in an approximately normal distribution of sample data with a mean of 0 and a variance of 1. Considering the lagging feedback of SM to Pr, as well as reducing the simultaneous in uence of Pr on SM in the statistical analysis of SM-Pr feedback, we chose the SM of one month ahead of Pr for the following analysis.
The singular value decomposition (SVD) method is commonly used to investigate the collocation patterns of two meteorological quantities (Prohaska 1976). Using the empirical orthogonal function (EOF) analysis technique, the SVD method aims to identify the high-relationship area of two elds by decomposing the cross-correlation coe cients between two meteorological elds. In SVD, the correlation coe cient between one variable eld (left eld) and the time coe cient of the other variable (right eld) in the paired element elds is called the heterogeneous correlation coe cient, and its spatial distribution represents the relationship pattern of two variables. Where the heterogeneous correlation passes signi cance test is the key area of the interaction between the two variable elds. The principle and applicability of SVD are described in detail in matrix theory (Wallace et al. 1992).
SVD analysis usually requires a sample size greater than the number of spatial grids in both target variable elds, which is often di cult to achieve in climate research. Therefore, when only limited climate data are available, the Monte Carlo technique is used to test the signi cance of SVD modes and then to determine whether the result is signal or noise. In this study, the random number generator was used to rst produce two data matrices with a Gaussian distribution 100 times, and then they were used to reperform SVD calculations. If the variance contribution of the original SVD modes was greater than the 95% quantile among the above 100 results, the pair of SVD modes was considered to be signi cant at the 95% signi cance level (Wallace et al. 1992 In this study, Pr was regarded as the right eld, and Z500 and SM were successively used as the left eld. The monthly Pr, Z500 and SM were rstly converted to the anomaly by subtracting the corresponding monthly climatology for 1980-2017, and then their JJA mean were used to carry out SVD analysis. Before performing SVD analysis, both latitudinal weighing and normalization were applied to all variable elds, and then 40 dimensionless JJA mean variable elds were obtained. The rst pair of signi cant SVD modes and the corresponding distribution pattern of the heterogeneous correlation coe cient were the key regions of the interaction between the two elds, which will be analyzed in detail below. Figure 3 shows the climatology of the JJA land surface hydrological variables from the merged GLDAS dataset for 1980-2019 over Asian mid-low latitudes. In order to avoid the overlapping between periods of MJJ SM and JJA Pr, the correlation of previous SM and Pr was rstly calculated month by month. For example, the correlation analysis of SM in May and Pr in June was conducted, and then Fig. 3b is obtained by averaging the correlation coe cients of three months. There is a clear spatial gradient in the distributions of Pr (Fig. 3a) and SM (Fig. 3b), showing gradual changes from dry in the northwest inland area to wet in the southeast coastal area. The wettest region appears in the South Asia land area, while the driest regions are around 40°N. Regional land-atmosphere coupling intensity is dependent on the water and energy at the land surface. The spatial gradient variables in Fig. 3 indicate that climate and land surface heat uxes over Asian mid-low latitudes are substantially variable from inland to coastal regions. The upward LW net and downward SW net over Asian mid-low latitudes are both strongest in the western part and weakest in the southeastern coastal area (Figs. 3c, d). The R net computed as the difference between SW net and LW net is then decomposed into the SHF, LHF and GHF. Both SHF and LHF show similarly spatial distribution as that of SM, but SHF displays an opposite gradient (Figs. 3e, f). Except for the southern region, GHF is positive, that is, downward, throughout the study area (Fig. 3g). The R net is mainly consist of SHF in the western region, while it is dominant by LHF in the other regions, especially in the southeast, and the GHF is relatively small in the whole region.  Fig. 4b can be regarded as the feedback strength of Pr to antecedent SM forcing. The wet (dry) shallow layer soil (0-40 cm) during May-June-July will induce more (less) Pr in the succedent month over most of the study areas (Fig. 4b). Pr in the southeast of the study area is affected by the EASM, where the correlation coe cient between Pr and Z500 is opposite to that in the western and northern inland regions (Fig. 4c).

Relationship between JJA Pr and Z500 and SM over Asian mid-low latitudes
The regions with signi cant Pr-SM correlation and Pr-Z500 correlation are not independent. Thus, it is necessary to perform further SVD analysis to extract the main coupling signals between the Z500/SM and Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js Pr elds. Using the MJJ SM and JJA Z500 as the left elds respectively and the JJA Pr as the right eld, two sets of SVD analyses are conducted (Fig. 5). The squared covariance fraction (SCF) of the rst SVD mode (SVD1) of the JJA Pr and Z500 (SM) is 30.75% (33.31%). The Monte Carlo technique test shows that the rst pair of SVD modes in both groups are signi cant at the 95% con dence level. There is an obvious "+-+" meridional wave train in Z500, and the negative centers on the west side of Lake Baikal and low latitude zone are the key regions affecting Pr (Fig. 5a). The heterogeneous correlation coe cients between Pr and Z500 (Fig. 5c) are signi cantly negative in the west of the study area (Fig. 5c). When the negative Z500 anomaly exhibits the abnormal distribution shown in Fig. 5a, the location of WPSH is to the north as compared with the normal situation, which is conducive to the transportation of water vapor to the north. While the Kazakh hills to the Tibetan Plateau are controlled by anticyclone, which is unfavorable for Pr Similarly, the heterogeneous correlation coe cient between the SVD1s of Pr and SM is shown in Figs. 5b, d.
The correlation coe cient of left eld (SM) over both Mongolian Plateau and Tibetan Plateau are signi cant negative, while it is signi cantly positive over the rest of the study area (Fig. 5b). The corresponding heterogeneous correlation coe cient of SM is much similar to that of Pr (Fig. 5d), re ecting the local coupling of antecedent SM and Pr. According to the above analysis, the key areas where SM and Z500 affects Pr obtained by the SVD method partly overlapped. Above analysis is only the interaction from a statistical point of view but cannot reveal the physical mechanisms driving the effects of Z500 and SM on Pr.

Effects of Z500 and SM on the JJA land Pr over Asian mid-low latitudes
To separate the contributions of large-scale circulation and local ET to Pr, the DA approach is used to decompose the contributions into two parts, Pr-Circ and Pr-Res (Section 2.3). The physical meaning of Pr-Circ is noticeable and indicates that water vapor for Pr formation is transported by large-scale atmospheric circulation factors (e.g., westerly wind) from faraway oceans. To verify the speculation of the physical meaning of Pr-Res, we compute and analyze the correlation coe cients between the JJA Pr-Res anomaly and the thermodynamic/dynamic quantities at the land surface after removing their corresponding seasonal cycles (Fig. 6).
The total net surface radiation consists of GHF, SHF and LHF, representing the major heat uxes of land surface processes. Considering that the magnitude of GHF is relatively small and nearly time invariant, only the relationships between the Pr-Res and SHF, LHF and PBLH are investigated. The correlation between Pr-Res and LHF shows strongly and signi cantly positive in the arid and semiarid regions, but it is signi cantly negative in southeast China, Korean Peninsula and Japan (Fig. 6a). Pr-Res and SHF are negatively correlated across almost all of the study region (Fig. 6b) anomalies may add perturbations in atmospheric humidity and temperature, they will affect the development and entrainment of the boundary layer (Ek and Holtslag 2004). Although the moisture for Pr might be transported from faraway ocean evaporation, local SM conditions in uence the boundary layer stability and growth, which determines when and where it is raining. Figure 6c shows that the PBLH is negatively correlated with Pr-Res in most part of study region except for Kazakh hills and central China.
When the soil is relatively wet, the surface net radiation is mainly distributed as LHF, that is, less SHF and more ET, so that SAT in the boundary layer decreases and the relative humidity increases, which then suppress the development of Pr. Because a high PBLH is favorable for the formation of Pr, the relationship between Pr-Res and PBLH indicates that land surface processes may indirectly impact boundary layer stability and growth. Thus, it can be concluded that Pr-Res can represent the portion of Pr induced by changes in land surface processes.
To quantify the independent effects of two factors on Pr, the SVD method is again used to analyze the relationships between Z500 and Pr-Circ and between SM and Pr-Res (Fig. 7). The results are then compared with SVD based on the original Pr and two factors (Section 3.1 and Fig. 5). The spatial distribution of the SVD1 heterogeneous correlation coe cient between Z500 and Pr-Circ shows a double-blocking anticyclone locating in the mid-latitudes. When Z500 presents a "-+-" tripole pattern, the WPSH moves easterly and southerly, under which the central of study area is controlled by the high-pressure system (Fig. 7a). Thus, Pr-Circ in most of study area except for Taklimakan Desert decreases signi cantly (Fig. 7c). When the distribution pattern of Z500 is similar to that in Fig. 7a and the northeasterly wind anomaly occurs at 850 hPa in North China, the water vapor transport from the low latitudes is reduced, which then suppresses the formation of Pr (Wu and Zhang 2011). Compared with the SVD results before DA-CCA, it is found that the patterns of the key areas that affect the total Pr and Pr-Circ are signi cantly different ( Fig. 5a and Fig. 7a), and the SCF of SVD1 increases from 30.75-53.01%, implying that the DA-CCA method can effectively separate the in uence of the atmospheric circulation on Pr.
The distribution of the SVD1 heterogeneous correlation coe cient between SM and Pr-Res reveals the key region of SM-Pr feedback. When the soil is wet in the Mongolian Plateau and the Huanghuai area of China and dry in the Tibetan Plateau, the southeast and northeast China, the Pr-Res increases in most of the study region. Compared with the SVD results before DA, it is found that although the correlation and signi cance of the key areas of SM affecting total Pr and Pr-Res are weakened, the spatial pattern is consistent (Fig. 5b and Fig. 7b). There are signi cant differences in the key areas affected by SM (Fig. 5d and Fig. 7d), indicating that it is more effective to analyze the effect of SM on Pr_Res than that on Pr directly. Figure 8 shows the mean, linear trend, and STD of Pr and its two components, Pr-Circ and Pr-Res for the period of 1980-2019 over Asian mid-low latitudes. The mean spatial pattern of Pr-Circ is similar to that of the original Pr (Fig. 8a, accounting for 97.5% of the total mean Pr), while Z500 is similar to its circulation analog ( gure not shown). The pattern of mean Pr-Res is also similar to that of total Pr, but the magnitude is much smaller (accounting for 2.5% of the total mean Pr). From the ratio of Pr-Circ to Pr, it can be seen that Pr-Cric dominates the magnitude and the spatial pattern of the JJA Pr and the circulation analog obtained through DA well re ects the average distribution of Z500 over Asian mid-low latitudes.
Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js The total Pr shows a weak increasing trend with the center in south of 30°N, but it displays slightly decreasing trend in the Indian Peninsula and the middle and lower reaches of the Yangtze River basin (Fig. 1a). Pr-Res shows an increasing trend in most of the study region, while Pr-Circ mainly weakens (Figs. 8b, e). The strong increasing trend of total Pr in Indochina Peninsula mainly depends on Pr-Res, while the weakening trend in the Indian Peninsula and the middle and lower reaches of the Yangtze River basin is jointly contributed by Pr-Circ and Pr-Res (Fig. 1a, Figs. 8b and 8e). The 500hPa constructed circulation analog shows a weakening trend over the Northwest Paci c, consistent with the WPSH and an increasing trend in the mid-high latitudes and the Iranian Plateau, which is unfavorable for Pr-Circ in northern China and the Indian Peninsula (Fig. 8b). This result indicates that the northward movement of the WPSH is the main circulation driving factor, leading to Pr-Circ increasing in northeastern Asia and decreasing on the southeastern coast. The coe cient of determination (CD), de ned as the square of the percentage of squared pattern correlation between component and total Pr (CD = 100%*R 2 ), is adopted here to quantify the contribution of the two components to the total Pr trend (Wang et al. 2018). The CD of Pr-Res (88.15%) is much higher than that of Pr-Circ (6.79%), indicating that the Pr-Res dominates the trend of total Pr. The above analysis shows that Pr-Res can better explain the trend of total Pr after removing the effect of atmospheric circulation. That is, the Pr trend over Asian mid-low latitudes is mainly controlled by land surface factors, even though its magnitude is relatively small.
The STD represents the interannual variability of the JJA Pr in the past 40 years, and the land Pr components in south of the study region both present a wetter trend (Figs. 8c, f), similar to the total Pr (CDs > 98%). The signi cant increase in the STD of Pr-Res south of 30°N may be related to the larger variability in SM in the humid area than in the arid and semiarid areas in the northwestern area.

Physical mechanism of Pr-Res trend
The DA results clearly show the different impacts of both atmospheric circulation and land surface processes on Pr over Asian mid-low latitudes. SM-Pr feedback is achieved through ET, which modi es available water vapor within the planet boundary layer. Water vapor evaporating from soil is associated with the phase change of water (from liquid to gas), which alters the partition of R net between the LHF and SHF (GHF is omitted due to its small magnitude). That is, ET affects the distribution of land surface available energy (R net -GHF) between SHF and LHF, and this effect can be quanti ed by the evaporation fraction (EF) (Seneviratne et al. 2010), For different EF ranges, the dominant controlling factors on ET are also different so that the mechanisms of water vapor transport from land to the atmosphere are also different. When EF is in the range of 0~0.5 (the red area in Fig. 9a), the available energy is mainly SHF. The corresponding area is called the "water-limited area", where the land is relatively dry and the R net is su cient. When EF is in the range of 0.5~1 (the blue area in Fig. 9a), most of the available energy is used for ET. This area is called the "energy-limited area", where ET is mainly determined by the available energy and the SM is su cient.
Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js The spatial pattern of EF is quite similar to that of Pr-Res except for in the south of Tibetan Plateau, where high and complex topography exist (Figs. 8d and 9a). The increase in EF in the Tibetan Plateau, Indian Peninsula and southeast China indicates that more land surface available energy is contributed by LHF (Fig. 9b). EF is mainly decreasing in the north of 40°N, where more available energy is consistent of SHF. The water vapor for Pr is mainly provided by the external transport of atmospheric advection and local land surface ET. Changes in the above two conditions reveal the physical mechanisms of land-atmosphere coupling on Pr from the perspective of water vapor sources. In addition, the spatial pattern of EF trend is similar to that of Pr-Res (Fig. 8h, r=0.29, α = 0.01), and the following analysis is conducted from the aspects of water and energy supply to EF. Here, the 0-40 cm SM is used to characterize the soil water condition in the water-limited area, and the available energy trend is used in the energy-limited area.
In the water-limited area, a signi cant soil wetting trend appears in the belt-shaped region from the Tianshan Mountains to the Greater Hinggan Mountains and the north side of Tibetan Plateau (Fig. 9c) However, in the periphery of the Tarim Basin where the soil shows wetting tendency, the variation trend of Pr-Res is relatively weak. This indicates that in the water-limited area, the sensitivity of Pr-Res to the SM reduction is stronger than that of SM enhancement. Central China and the Korean Peninsula are located in humid areas, where ET is limited by the available energy. In the energy-limited region, the Pr-Res decreases in eastern China but increase the Korean Peninsula and the Indian Peninsula, the pattern of which are consistent with the that of available energy (Figs. 8e and 9d). It is noted that the Mongolian Plateau is located in the transition area of EF=0.4~0.6, where the effects of enhancement in SM and reduction in available energy cancel each other, Pr-Res has a slightly strengthening trend. This means that the in uence of land surface processes on Pr in the dry-wet transition region is very complex and deserved further research in future work. Besides, we have distinguished the contribution of large-scale transport and local ET from the perspective of water vapor sources in this study. However, the indirect in uence of SM on Pr through regulating circulation is not discussed (Conil et al. 2007; Zeng and Yuan 2021), which can be addressed through the numerical experiments by using climate models in the future. The de nition of EF depends on SM, while the land surface energy balance is also affected by other factors. Study found that the intense melting of snow in west Russia and Siberia in April and May had a strong impact on the SHF, PBLH and Pr in the next month (Ye and Lau 2019).

Summary
In this study, DA-CCA approach was used to quantitatively separate the contributions of atmospheric circulation and local ET to summer Pr over Asian mid-low latitudes. First, the DA-CCA approach was applied to identify the contribution of atmospheric circulation to summer Pr (i.e., Pr-Circ), and then Pr-Res was de ned as the contribution of the land-atmosphere coupling. Z500 was used to describe the effect of atmospheric circulation, and SM was used to represent the land surface condition. Then, the in uence and Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js physical mechanisms of land-atmosphere coupling on Pr were also analyzed. Our major conclusions are as follows: 1. After removing the effect of atmospheric circulation, Pr-Res represents the part of the JJA Pr over Asian mid-low latitudes inducted by local ET resulted from land surface processes. 2. The key Z500 pattern affecting summer Pr over Asian mid-low latitudes is the zonal tripole in the midlatitude after separating the effects of atmospheric circulation through DA-CAA.
3. Although Pr-Res is small in magnitude, it dominates the trend pattern of the total JJA Pr over Asian midlow latitudes in the past 40 years. This understanding may provide information for improving regional summer climate prediction.
4. The soil wetting trend in the land water-limited areas in study regions explain the enhancement of Pr-Res in the corresponding region to a certain extent and it play dominant role on total Pr trend.      The heterogeneous correlation coe cients of the rst SVD modes a,c between the JJA mean Z500 and Pr and b,d between the SM in the antecedent month and Pr over Asian mid-low latitudes from 1980-2019. The dotted areas are statistically signi cant at the 5% level Figure 6 Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js Correlations (color shading) between the JJA Pr-Res obtained using DA and a LHF, b SHF, and c PBLH over Asian mid-low latitudes from 1980 to 2019. The dotted areas are statistically signi cant at the 5% level