Volume 49, Issue 19 e2022GL099646
Research Letter
Free Access

Contrasting Climatic Trends of Atmospheric River Occurrences Over East Asia

Qiang Wang

Qiang Wang

School of Geography and Ocean Science, Nanjing University, Nanjing, China

Frontiers Science Center for Critical Earth Material Cycling, Nanjing University, Nanjing, China

Contribution: Conceptualization, Formal analysis, Data curation, Writing - original draft, Funding acquisition

Search for more papers by this author
Long Yang

Corresponding Author

Long Yang

School of Geography and Ocean Science, Nanjing University, Nanjing, China

Frontiers Science Center for Critical Earth Material Cycling, Nanjing University, Nanjing, China

Correspondence to:

L. Yang and X. Chen,

[email protected];

[email protected]

Contribution: Conceptualization, Methodology, Writing - review & editing, Supervision

Search for more papers by this author
Yixin Yang

Yixin Yang

School of Geography and Ocean Science, Nanjing University, Nanjing, China

Frontiers Science Center for Critical Earth Material Cycling, Nanjing University, Nanjing, China

Contribution: Methodology, Writing - review & editing, Visualization

Search for more papers by this author
Xiaodong Chen

Corresponding Author

Xiaodong Chen

Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, Richland, WA, USA

Correspondence to:

L. Yang and X. Chen,

[email protected];

[email protected]

Contribution: Conceptualization, Methodology, Writing - review & editing, Supervision

Search for more papers by this author
First published: 01 October 2022
Citations: 3

Abstract

Atmospheric rivers (ARs) are increasingly recognized as a major driver of hydrological cycles, and are projected to increase around the world under a changing climate. However, the long-term trends of ARs over East Asia (EA) remains less elucidated. Here we fill the gap by developing a longest-ever archive of EA ARs, and examine its role in determining spatiotemporal precipitation variability over EA. We find contrasting changes in AR occurrences, with more frequent ARs in low latitudes but less in high latitudes during the period 1950–2020. The “dipole” pattern of decadal changes in AR occurrences is dictated by atmospheric dynamics (i.e., winds) in the north but thermodynamics (i.e., moisture) in the south. The reduced AR occurrences explain 49% of decreased annual precipitation in northern China, while more AR-related precipitation is observed in southern China. Our results provide new insights into regional hydroclimate over EA by connecting it with large-scale weather systems.

Key Points

  • The mean annual occurrences of atmospheric rivers (ARs) show increases in low latitudes but decreases in high latitudes over East Asia

  • The “dipole” pattern of long-term AR trends is associated with contrasting dynamic and thermodynamic responses

  • Such changes in annual AR occurrences contribute to the “Southern-Flood Northern-Drought” precipitation pattern over eastern China

Plain Language Summary

Atmospheric rivers (ARs) feature long and narrow filaments with intense moisture transport in the lower troposphere. It is a vital component in the regional and global hydroclimate. Previous studies have documented the characteristics of ARs at both regional and global scales. However, studies about the mechanism of long-term trends and its contribution to precipitation changes are still lacking. Here we examine the annual variation of ARs and its impacts on regional hydroclimate over East Asia. Our analysis is based on a long-term, high-resolution reanalysis data set. We show distinct spatial patterns of the changes in annual AR occurrences, with more ARs in low latitudes but decreased frequency in high latitudes. The “dipole” pattern is mainly related to wind changes in the north but moisture changes in the south. The reduced annual AR occurrences in high latitudes explain 49% of decreased annual precipitation in northern China, while more AR-related precipitation is expected in southern China. The dipole-like pattern of AR changes provides a good elucidation of the “Southern-Flood and Northern-Drought” pattern for East China.

1 Introduction

Atmospheric rivers (ARs) are those long and narrow bands in the atmosphere that feature intense moisture transport (Newell et al., 1992; Zhu & Newell, 1998). They are a major component of the regional and global water cycle (Paltan et al., 2017; Ralph et al., 2018; X. Chen et al., 2019). ARs have also been recognized as a key driver of precipitation, runoff, winds and relevant hazards over the coastal regions worldwide (e.g., Cao et al., 2020; Ralph et al., 2006; Waliser & Guan, 2017; X. Chen et al., 2018).

Despite their fundamental roles in regional water cycles, the knowledge of ARs in several coastal regions (e.g., East Asia, Oceania) is still lacking, partially due to a lack of long-term AR archives in these regions. Compared to other hotspots, such as western US, western Europe, where ARs are the dominant drivers of precipitation, the hydroclimate of East Asia (EA) is collectively modulated by monsoon as well as the interactions between tropical systems and the westerlies (C. Chen et al., 2021a, F. Chen et al., 2021b; Dettinger et al., 2011; Guan & Waliser, 2015; Lavers & Villarini, 2015; J. J. Rutz et al., 2014). Improved understandings with respect to ARs and their spatial-temporal changes over EA contribute to advanced characterization of ARs under complex synoptic settings.

With the recent development of global AR tracking algorithms, EA ARs have been systematically identified at climatic scales (Guan & Waliser, 2015; J. Rutz et al., 2019; Pan & Lu, 20192020; Shields et al., 2018). These archives allow quantification of the interactions between ARs and the large-scale ambient environment as well as their impacts on the regional water cycle. For instance, Park et al. (2021) and Pan and Lu (2020) analyze the annual cycle of ARs, and highlight the contribution of ARs to intra-annual rainfall variability over EA. Their results show strong connections between heavy rainfall days and ARs over midlatitudes during the monsoon season. Despite existing knowledge, the inter-annual variations of ARs (in terms of their intensity and frequency) have not been elucidated so far over EA. This is important considering the prominent contribution of ARs to monsoon rainfall over EA as revealed by previous studies (Pan & Lu, 2020; Park et al., 2021). Additionally, there is a growing interest in quantifying ARs' contributions to the spatiotemporal rainfall variability (e.g., Kamae et al., 2021; Lavers & Villarini, 2015; Waliser & Guan, 2017). EA offers a perfect avenue to examine such contributions under a changing climate, due to the verified role of EA ARs in determining regional hydrological extremes.

A prominent feature of rainfall changes in EA can be phrased as the “Southern-Flood and Northern-Drought” pattern, indicating more rainfall and thus flooding in low latitudes but less rainfall and thus more droughts in high latitudes during the past decades (Day et al., 2018; Zhou et al., 20092020). Previous studies attribute the climatic trends of rainfall to variations in large-scale circulations, such as changes in East Asian Summer Monsoon (EASM, e.g., Ding et al., 2009; Meehl et al., 2008; T. Zhou et al., 2009), shifts in the Asian jet stream (e.g., Herzschuh et al., 2019; Wei et al., 2017), and other climate-change related processes (e.g., Menon et al., 2002; Z. Jiang et al., 2017). For instance, Day et al. (2018) shows that changes in frontal systems contribute overwhelmingly to the spatial pattern of rainfall changes in eastern China. B. Zhou et al. (2020) highlight the synoptic circulations that are responsible for rainfall changes in eastern China. They further attribute the changing synoptic circulations to dynamic (i.e., wind) and thermodynamic (i.e., moisture) factors. In addition to large-scale synoptic environments, there are specific weather systems/patterns that are directly responsible for spatial and temporal rainfall variability over EA, such as atmospheric moisture transport (represented by ARs, Liu et al., 2020), landfalling tropical cyclones (e.g., Yang et al., 2020). The connections between those weather systems and climatic rainfall changes offer improved characterizations of historical and future drivers for regional hydroclimate. Such connections at decadal scales, however, is still missing over EA.

Here we develop a long-term AR archive over EA during 1950–2020. It is the longest archive yet in the literature. By analyzing this AR archive and the associated climate variables, we address three questions: (a) What are the long-term trends in annual AR occurrences over EA? (b) How do the trends and their driving factors vary spatially over EA? (c) How much can ARs explain precipitation changes (both the annual mean and extremes) in eastern China? We focus on AR occurrences mostly because it captures the integral impacts of climate variations on weather systems, and thus facilitate the connections between ARs and key features of regional precipitation changes over EA, that is, the “Southern-Flood and Northern-Drought” pattern.

2 Data and Methods

The AR archive used in this study is developed based on the ERA5 Reanalysis product which is developed by the European Centre for Medium-Range Weather Forecasts (ECMWF, Hersbach et al., 2020; Hoffmann et al., 2019). The performance of the ERA5 Reanalysis in reconstructing historical climate has been verified in previous studies, and ERA5 data set has been widely used as a climate benchmark in various studies (include EA, B. Zhou et al., 2020; Guan & Waliser, 2015; Q. Jiang et al., 2021; Tarek et al., 2020). The ERA5 Reanalysis provides hourly reconstruction of global atmospheric and near-surface meteorological conditions at 0.25° × 0.25° resolution during the period 1950–2020. The following meteorological variables are retrieved at 6-hourly interval: meridional wind (u), zonal wind (v), compound wind (w), specific humidity (q), geopotential height (z), precipitation (P, including both rain and snow) and temperature (T).

ARs are usually identified with specific integrated vapor transport (IVT), integrated water vapor (IWV), duration and morphology requirements (Guan & Waliser, 2015; Pan & Lu, 2019; Zhu & Newell, 1998). We establish the AR archive during 1950–2020 based on the criteria from the existing AR tracking algorithms (Guan & Waliser, 2015; Mundhenk et al., 2016; Pan & Lu, 2019; Reid et al., 2020). Due to the relatively higher intensity of IVT and IWV for the EA ARs (e.g., Pan & Lu, 2019), we make additional constraints on intensity, duration and morphology to identify those more significant ARs. Since the algorithm shares many features with previous studies, we do not claim methodological novelty for the archive. The details of our algorithm are listed below.

IVT and IWV are derived using the data between 1,000 hPa and 300 hPa, as shown in Equations 1 and 2. Potential AR patches are delineated based on the following criteria: (a) IVT needs to exceed 250 kg/m/s and the value of local 85th percentile (whichever is larger) during 1950–2020; (b) IWV is higher than 15 mm and the value of local 85th percentile (whichever is larger). Both criteria need to be satisfied for a grid to be labeled as a potential patch of a AR. Further, if the major axis of potential AR patches is greater than 1,500 km with length-to-width ratio greater than 2, it is recorded as an AR event.
urn:x-wiley:00948276:media:grl64926:grl64926-math-0001(1)
urn:x-wiley:00948276:media:grl64926:grl64926-math-0002(2)

Inter-comparisons of different AR tracking algorithms show overall similar patterns of AR frequency of occurrences over EA (Collow et al., 2022). The results of our AR identification algorithms using the JRA55 reanalysis product (Harada et al., 2016; Kobayashi et al., 2015) remain strikingly similar with those based on ERA5 during the overlapping period (see Figure S1 in Supporting Information S1). The AR archive based on ERA5 will be analyzed due to its high spatial resolution and length of record.

Based on this AR archive, we examine long-term trends in annual AR frequency over EA. Since the long-term trends are associated with changes in mean and/or variation, we decompose the components through controlled experiments (see Table S1 in Supporting Information S1). The procedures are as follows. The experiments decompose the variable X (referring to IWV or IVT) at time t of year i into its annual mean and the instantaneous fluctuation components at the grid scale (Equation 3):
urn:x-wiley:00948276:media:grl64926:grl64926-math-0003(3)
where urn:x-wiley:00948276:media:grl64926:grl64926-math-0004 is the annual mean of X at year i, and urn:x-wiley:00948276:media:grl64926:grl64926-math-0005 is the coefficient of variation at year i. The inter-annual variability of IVT and IWV are quantified using the annual mean and standard deviation, respectively. urn:x-wiley:00948276:media:grl64926:grl64926-math-0006 denotes the constant term at time t of year i and remains consistent across different experiments. Three experiments are designed to investigate the impacts of annual mean (i.e., urn:x-wiley:00948276:media:grl64926:grl64926-math-0007), annual variability (i.e., urn:x-wiley:00948276:media:grl64926:grl64926-math-0008), and both (urn:x-wiley:00948276:media:grl64926:grl64926-math-0009) on the long-term trends of annual AR occurrences, respectively. These experiments remove the trends of IVT and IWT in mean and/or variability during the entire period 1950–2020. ARs are individually tracked based on the modified IWV and IVT fields. The linear trend of annual AR occurrences based on original data set and the modified datasets are examined through a modified Mann-Kendall test (at the significance level of 5%; Hamed & Rao, 1998).
We further attribute the long-term trends of annual AR occurrences to dynamic and thermodynamic processes based on the controlled experiments. Thermodynamic effects reflect the variation of atmospheric water-holding capacity, and dynamic changes reflect the atmospheric circulations (i.e., wind). We design three experiments, urn:x-wiley:00948276:media:grl64926:grl64926-math-0010, urn:x-wiley:00948276:media:grl64926:grl64926-math-0011, and urn:x-wiley:00948276:media:grl64926:grl64926-math-0012. These experiments are conducted similarly to urn:x-wiley:00948276:media:grl64926:grl64926-math-0013, but with the trends in specific humidity (i.e., urn:x-wiley:00948276:media:grl64926:grl64926-math-0014), wind (i.e., urn:x-wiley:00948276:media:grl64926:grl64926-math-0015) and both (i.e., urn:x-wiley:00948276:media:grl64926:grl64926-math-0016) removed. This is because analyses show that changes in annual mean contribute to the long-term trends of annual AR occurrences (see results below). Previous studies also examined the thermodynamic or dynamic processes in dictating changes in precipitation minus evaporation (e.g., Morrill et al., 2018; Seager et al., 2010), summer rainfall (e.g., Lee et al., 2017), the tropical cyclones (e.g., Tao & Zhang, 2019). Compared to those previous investigations of ARs in other geographic regions, we not only perturb the moisture and wind fields within the AR boundary, but also its regional conditions. This allows us to examine the changes of existing ARs, as well as the generation of potential ARs. By doing this, we can provide a better connection between ARs and its ambient thermodynamic and dynamic environments (similarly see Gao et al., 2015). Long-term trends of annual AR occurrences based on the original ERA5 Reanalysis fields are denoted as the CTRL scenario. We determine the individual impacts of dynamic and thermodynamic processes as well as their interactions on long-term trends of annual AR occurrences based on the factor separation analysis (e.g., Stein & Alpert, 1993):
urn:x-wiley:00948276:media:grl64926:grl64926-math-0017(4)
urn:x-wiley:00948276:media:grl64926:grl64926-math-0018(5)
urn:x-wiley:00948276:media:grl64926:grl64926-math-0019(6)

Relative contributions of different driving factors can be quantified as: urn:x-wiley:00948276:media:grl64926:grl64926-math-0020, where urn:x-wiley:00948276:media:grl64926:grl64926-math-0021 and urn:x-wiley:00948276:media:grl64926:grl64926-math-0022 represents the changing rates (estimated by the Theil-Sen slope) of annual AR occurrences in the CTRL scenario and the respective experiments at each grid (i.e., urn:x-wiley:00948276:media:grl64926:grl64926-math-0023, urn:x-wiley:00948276:media:grl64926:grl64926-math-0024, and urn:x-wiley:00948276:media:grl64926:grl64926-math-0025).

We further evaluate the predictability of ARs for extreme rainfall based on Gilbert Skill Score (GSS, Schaefer, 1990, see Text S1 in Supporting Information S1). GSS reflects how well the extreme rainfall events (hit or miss) correspond to ARs. Previous studies show that the GSS value around or larger than 0.2 indicates high predictability of extreme rainfall associated with ARs (e.g., X. Chen et al., 2018). Extreme rainfall is defined as the day with daily rainfall accumulation exceeding 90th percentile of the annual rainfall series. ARs are responsible for the extreme rainfall when an AR is identified during the rainy day over the grid.

3 Results and Discussions

Figure 1a shows the annual AR occurrences and their long-term trends over EA during 1950–2020. ARs mainly occur in eastern China, and extend toward the Korean Peninsula. Its annual frequency decreases from the southeastern coast to northwest inland. The peak frequency of annual ARs occurs in the Bay of Bengal and the Yellow Sea, that is, the front region of coastal mountains. More than 95% of ARs occur during the warm season (i.e., May to October), indicating that ARs are a key driver of the regional rainfall extremes (and thus flooding) over EA. Our results are consistent with previous studies (Pan & Lu, 2020; Park et al., 2021), highlighting the robustness of our AR identification algorithm over EA. ARs occurrences are frequently generated along with large-scale weather systems (e.g., tropical cyclones, extra-tropical cyclones, monsoon front; see Pan and Lu (2020) for more details).

Details are in the caption following the image

(a) Spatial distribution of mean annual atmospheric river (AR) occurrences (contour) and its long-term trends (shade) during 1950–2020. The red (blue) box outlines the extent of southern (northern) China. The red (blue) points in (a) indicate the grids with significant positive (negative) trends at the level of 0.05. (b and c) boxplots of long-term trends (in days decade−1) of annual AR occurrences from different experiments in northern and southern China. The box spans the 25th and 75th percentiles, and the whiskers represent 10th and 90th percentiles. The median values are shown with black lines.

During the past seven decades, annual AR occurrences exhibit an increasing trend in most regions of EA, except that northern China demonstrates a notable decreasing tendency. More specifically, the frequency of annual AR occurrences shows significant increasing trends (i.e., up to +1.5∼2.0 days/decade) in coastal regions of southeastern China, while significant decreasing trends (i.e., −0.5∼−1.0 days/decade) are observed in northern China. The contrasting spatial patterns of changes in annual AR occurrences are significant in both southern and northern China (i.e., two hotspots). In addition, annual AR occurrences slightly increase before declining from south to north in eastern China. The transition occurs at approximately 30°N, and is geographically consistent with the spatial divide of contrasting rainfall trends in eastern China (e.g., B. Zhou et al., 2020). The dipole pattern of the changes in annual AR occurrences is further highlighted by integrating the trends into different latitudinal bands over EA (Figure S2 in Supporting Information S1).

The relative importance of mean trend, natural variability as well as their interactions is quantified through control experiments (Table S1 in Supporting Information S1). Figures 1b and 1c summarize the trends of annual AR frequency in two hotspots. that is, northern China and southern China (see Figure S3 in Supporting Information S1 for the spatial patterns). When the trends of annual mean IVT and IWV are removed (i.e., urn:x-wiley:00948276:media:grl64926:grl64926-math-0026, the resulted AR trends are distinct from the CTRL scenario. In contrast, AR trends are highly consistent between the CTRL and the urn:x-wiley:00948276:media:grl64926:grl64926-math-0027 experiments (the leftmost three columns in Figures 1b and 1c). This indicates that changes in annual means in IWV and IVT dominate the long-term trends of annual AR occurrences. This impact is more prominent in the south where the AR trends in urn:x-wiley:00948276:media:grl64926:grl64926-math-0028 are almost equal to zero. The urn:x-wiley:00948276:media:grl64926:grl64926-math-0029 scenario is almost identical to the urn:x-wiley:00948276:media:grl64926:grl64926-math-0030 scenario, suggesting that the interaction of long-term trends in the annual mean and variability of IVW and IWV is negligible.

The impact of dynamic (wind), thermodynamic factors (water vapor) and their interactions on AR occurrences are shown in Figure 1 (the rightmost three columns in Figures 1b and 1c). We see that when only the q trend is removed (urn:x-wiley:00948276:media:grl64926:grl64926-math-0031), the increasing trend of AR frequency is significantly reduced in the south (see Figure S3d in Supporting Information S1 for the spatial pattern), while the trends are similar to the CTRL scenario only when the trends of wind speed are removed (urn:x-wiley:00948276:media:grl64926:grl64926-math-0032). This indicates that thermodynamic processes play a dominant role in annual AR occurrences in southern China. The removal of q and w trends simultaneously leads to similar changes with only q removed, indicating a minimal interaction between the two factors. By contrast, the removal of w trend leads to greater changes in annual AR frequency in the north, highlighting the important role of dynamic factors in controlling trends of ARs in this region.

Relative contributions of these effects are quantified in Figure 2 by scaling them with their sum (Figure S4 in Supporting Information S1). Thermodynamic and dynamic factors show dominantly positive contributions over EA. Thermodynamic effect dominates the trends in annual AR occurrences in the south, with relative contributions ranging from 50% to 80%. The regions dominated by dynamic effects coincide with decreasing AR occurrences, indicating the dominant role of wind anomalies in reducing AR occurrences in northern China (i.e., north of 30°N). For instance, the dynamic contribution shows small variations, with their values ranging from 40% to 60%. As shown in Figure 2c, the dynamic-thermodynamic interactions exhibit slightly mixed effects in the south, while mostly inhibition effects are observed in northern China. These results collectively highlight contrasting roles of thermodynamic and dynamic processes in determining the long-term trends of EA ARs.

Details are in the caption following the image

Relative contributions of (a) thermodynamic effect, (b) dynamic effect, and (c) their interactions to the long-term trends of annual atmospheric river occurrences, (d) relative contribution averaged over latitudinal bands (i.e., 5° for each band) in eastern China (i.e., between the longitudes 105° ∼ 125°E).

To better reveal the underlying mechanisms, we analyze the trends of warm-season averaged 850 hPa air temperature, geopotential height, wind speed and specific humidity over EA (Figure 3). As expected, mean 850 hPa temperature shows significant increasing trends, consistent with the general warming trend (Figure 3a). As a results of thermal expansion of warming atmosphere, the increasing trends of mean 850 hPa geopotential height mainly occur in eastern China, highlighting a tendency of westward extension of the Western Pacific Subtropical High (WPSH, Figure 3b). The increased geopotential height is accompanied by a distinct downward trend of low-level wind speeds, mostly notable along the rim of WPSH and northern China (Figure 3c). The changes in water vapor mixing ratio at 850 hPa indicate significant wetting (drying) tendencies in southern (northern) China (Figure 3d). The synoptic contrasts between the first and second 30 years of the study period show similar patterns (Figure S5 in Supporting Information S1).

Details are in the caption following the image

The spatial distribution of trends in warm-season (i.e., May to October) mean (a) temperature (in K decade−1), (b) geopotential height (in gpm decade−1), (c) zonal and meridional winds (in m s−1 · decade−1), and (d) specific humidity (in g · kg−1 · decade−1), at the level of 850 hPa during the period 1950–2020. The black vector represents the average wind field (in m s−1). The red (blue) points indicate the grids with significant positive (negative) trends at the level of 0.5.

In southern China, water vapor is mainly provided by Asian summer monsoons from the North Pacific Ocean and the Indian Ocean (Figure 3c and Figure S5c in Supporting Information S1). The EASM has been weakened during past decades, resulting increased water vapor (Yang et al., 2013) and then increased AR occurrences in low latitudes. The increased atmospheric moisture content is partially related to the increased water-holding capacity under a warming climate (e.g., Allen & Ingram, 2002), as more water vapor would be evaporated from the land surface and absorbed in the atmosphere. The transport of water vapor in northern China is mainly through the Asian westerly jets and EASM (as shown for the average wind field in Figure 3 and Figure S5 in Supporting Information S1, see also e.g., C. Chen et al., 2021a2021b). Instead of increases, water vapor is decreased due to the significant changes of zonal and meridional winds in northern China, contributing to reduced annual AR occurrences. The weakened EASM prevents water vapor from reaching far to the north, further inhibiting the occurrences of ARs.

Long-term trends of annual AR occurrences are responsible for the spatial and temporal variability of precipitation (both the annual mean and extremes) over EA. We define AR-related precipitation when AR is identified during the day of precipitation over a given grid. Our analyses show that ARs are responsible for more than 42% of annual precipitation in central and eastern China (Figure S6 in Supporting Information S1). The most notable contribution is observed in eastern China, where ARs contribute up to 70% of annual precipitation trends (Figure 4a). The contribution of AR trends to annual precipitation trends fluctuates from south to north over EA, with the peaks attained around 30∼40°N (Figure 4b). Relatively lower contribution to the precipitation trends along 25°N compared to its neighboring bands is mainly due to the diverse precipitation changes across the longitudes and the contrasting roles of ARs in dictating them (see Figure 1a). For example, dynamic rather than thermodynamic processes dominate the long-term changes of ARs occurrences and precipitation for the inland region of the 25°N band (i.e., west of 110°E), while the opposite is true along the coasts (i.e., east of 110°E).

Details are in the caption following the image

Contribution of changes in annual atmospheric river (AR) occurrences to the changes in annual precipitation. (a) Spatial pattern, (b) contribution summarized by latitudinal bands in eastern China (within the longitudes 105° ∼ 125°E). The contribution of AR changes to annual precipitation changes is defined as the mean linear slope between the anomalies of annual AR precipitation and total precipitation (see Figure S8 in Supporting Information S1 for illustration). The box spans the 25th and 75th percentiles, and the whiskers represent the 10th and 90th percentiles. The median values are plotted using the black lines in the box.

The spatial patterns of annual precipitation and the AR-related precipitation are comparable in eastern China (Figure S7 in Supporting Information S1). This indicates the important role of AR variability in shaping the contrasting patterns of annual precipitation changes in eastern China. Most areas in northern China exhibit significant decreasing trends in both annual precipitation and AR-related precipitation. The long-term trends for annual AR frequency, annual precipitation, and annual AR-related precipitation are highly consistent in northern China. The long-term trends in annual precipitation and annual AR-related precipitation are −30.63 mm/decade and −13.04 mm/decade, respectively (Figure S8a in Supporting Information S1). Therefore, changes in AR-related precipitation contribute approximately 49% to the annual precipitation changes in northern China (Figure S8b in Supporting Information S1). Such contributions of similar magnitudes extend from northern China to the Korean Peninsula.

ARs shows a high predictability of extreme rainfall over EA (Figure S9 in Supporting Information S1). The region with GSS score exceeding 0.2 extends from North China to the Korean Peninsula. However, the GSS values for extreme rainfall have been increased (decreased) significantly in southern (northern) China during the past decades. This imply that the predictability of ARs for extreme rainfall over EA evolves under a warming climate. Since accumulated extreme rainfall accounts for a dominant portion of annual total precipitation over eastern China (i.e., the ratio exceeds 40% for most regions, Figure S10 in Supporting Information S1), the high predictability of ARs for extreme rainfall and its temporal evolution applies for annual total precipitation as well. Our results thus highlight the important role of ARs in modulating annual precipitation (both the mean and extremes) over EA during the past seven decades. These changes partially contribute to the “Southern-Flood-Northern-Drought” pattern in eastern China.

4 Concluding Remarks

In this study, we examine the long-term trends of annual AR occurrences over EA, and shed light on the driving factors for regional hydroclimate based on a 70-year AR archive. We highlight a distinct spatial pattern of trends in mean annual AR occurrences, with decreases in high latitudes but increases in low latitudes. This indicates that increases and decreases in the annual AR occurrences co-exist at regional scale due to complex synoptic conditions, even though the decreasing tendency is often overlooked in global-scale analyses (Payne et al., 2020). Our results highlight the interactions of different climate variables in dictating the long-term trend of ARs over EA. The dipole pattern is related to atmospheric dynamic factors (i.e., winds) in the north and thermodynamic factors (i.e., moisture) in the south. The thermodynamic impact on the changes in annual AR frequency is reduced gradually from low to high latitudes over EA, while the opposite is true for the dynamic impact. AR-related precipitation accounts for up to 42% of annual precipitation in eastern China, with its long-term trends of annual AR occurrences alone explaining about 49% of annual precipitation variability in northern China. A limitation of the study is that the AR dynamics and the partitioning of dynamic and thermodynamic impacts are based on data-driven analyses. Well-designed regional climate simulations can provide further validation of our results (e.g., Hagos et al., 2016; Martin et al., 2018; Rhoades et al., 2020). This would be an endeavor to pursue in future studies. We establish the connections between large-scale weather systems and long-term changes in regional precipitation over EA. Such knowledge can better prepare policymakers and stakeholders for flood and drought risk management under a changing climate (Payne et al., 2020).

Acknowledgments

This study is supported by the National Natural Science Foundation of China (Grants 42101020, and U2240203), China Postdoctoral Science Foundation (Grant 2021M691529), Jiangsu Postdoctoral Research Funding Program (Grant 2021K121B), the Fundamental Research Funds for the Central Universities (Grants 0209–14380103 and 0209–14380104) and the Research Funds for the Frontiers Science Center for Critical Earth Material Cycling, Nanjing University. X.C. is supported by the US Department of Energy Office of Science Biological and Environmental Research as part of the Regional and Global Model Analysis program area. PNNL is operated Battelle Memorial Institute for the Department of Energy under contract DE-AC05-76RL01830.

    Data Availability Statement

    All data necessary, including ERA5 and JRA55 datasets, to reproduce the results of this work are available online. The ERA5 datasets used in this study are available through https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5. The JRA55 datasets are available through https://jra.kishou.go.jp/JRA-55/index_en.html. The code for ARs detection is available on Zenodo (https://zenodo.org/record/6912433#.YuEAUnZBwuU, DOI: 10.5281/zenodo.6912433).