Introduction

The oxygen isotopic composition of precipitation (δ18Op) recorded in natural archives, such as ice cores, speleothems, and tree rings, is one of the most widely used proxies for reconstructing past climate change. Despite its widespread application, the climatic significance of δ18Op and stalagmite δ18O (δ18Os) records, especially in the context of the East Asian summer monsoon (EASM), has been the subject of much debate1,2,3,4,5. The orbital to millennial variations of δ18Op and δ18Os in the EASM region are thought to be primarily influenced by the large-scale circulation and latitudinal migration of the EASM rain belt, ultimately showing integrated monsoonal signals from the moisture source to the cave sites1,4,6,7. Nevertheless, a consensus understanding of various hydrological processes governing variations in EASM δ18Op and δ18Os on short timescales has not been reached. Studies employing modern observations and general circulation model water isotope modeling have attributed seasonal to decadal δ18Op and δ18Os variability to multiple factors, including “upstream” processes8,9,10,11, moisture source dynamics12,13, local precipitation amount14, and precipitation seasonality15,16,17. These diverse interpretations highlight the complexities inherent in δ18Op and δ18Os, which may differ across monsoonal China, especially between the northern and southern regions.

Northern China, located at the margin of the EASM region, experiences a more continental climate. In contrast to typical EASM areas, the δ18Op in this region displays a different seasonal variation, exhibiting a pattern of depleted values in winter and relatively enriched values in summer18. However, the understanding of the seasonal δ18Op patterns in northern China remains incomplete due to a scarcity of modern observational records. Initially, the seasonal δ18Op variability in northern China was explained in terms of local climate conditions following the paradigm of the “temperature effect” and “amount effect”19,20. Furthermore, recycled moisture from local evapotranspiration was also considered as a contributing factor21,22,23. However, the relationships between δ18Op and local conditions may be masked by the complex, large-scale processes, such as moisture sources and upstream convection. Li et al.24 proposed that the “temperature effect” exists during the non-monsoon season, while summer δ18Op can be affected by upstream convection. Tang et al.25 suggested that the interplay of moisture source and upstream rainout contribute to the variation of summer δ18Op. While Zhang et al.18 and Wang et al.26 suggested a minimal influence of upstream convection on the δ18Op in northern China. They proposed that the seasonal δ18Op cycle in northern China is driven by a shift in westerly-derived continental moisture to the Pacific Ocean moisture, with the single continental air mass during the non-monsoon season being more susceptible to the temperature effect18,26. Moreover, Xie et al.27 suggested that the summer δ18Op is not affected solely by Pacific moisture but rather the remote monsoonal sources like the Indian Ocean and the Bay of Bengal moisture. Contrarily, westerly-derived continental moisture can prevail throughout the year during weak monsoon periods, underscoring the intricate moisture dynamics at the margin of the EASM area28. Overall, the influence of moisture source dynamics on δ18Op varies spatially12. It remains uncertain whether and how the varying moisture sources have a discernible effect on the δ18Op in northern China.

Here, we use the Community Atmosphere Model version 3 (CAM3), with embedded water-tagging and stable water isotopes modules, to trace the journey of water vapor from the evaporation in the source region to precipitation in the target region without the time limitation for tracing water vapor parcels29 (see Methods for details). Our primary objectives encompass identifying the sources of water vapor in northern China, quantitatively analyzing moisture source variability, and evaluating their influence on the seasonality of δ18Op. Additionally, we aim to establish connections between the moisture sources and the decadal variations observed in δ18Op.

Results

Experiments design

The CAM3 simulation has been effectively validated for various regions and is extended here to precipitation and δ18O in precipitation and water vapor (δ18Ov) in northern China. Spanning 70 years (1951–2020), the simulation successfully reproduces climatology and annual cycle in precipitation and δ18O over northern China with impressive accuracy, which is directly comparable to the observational data (see Supplementary discussion, Supplementary Figs. 1–5), indicating that CAM3 is highly effective in reproducing large-scale moisture dynamics and hydrological cycle over the northern China.

Based on previous studies24,30, eight potential moisture sources are selected to account for the precipitation falling within northern China. These regions encompass high-latitude land areas (HLA), northern China (NC), southern China (SC), low-latitude land areas (LLA), Mediterranean Sea (MS) & North Atlantic Ocean (NA), Pacific Ocean (PO), South China Sea (SCS), and North Indian Ocean (NIO) (Fig. 1). The contribution from the rest of the globe is calculated as the residual between the precipitation of the target region and the sum of the precipitation that originates from the eight moisture sources. We define the area from 32°N–42°N and 102°E–120°E as the target region.

Fig. 1: Schematic representation of moisture source regions considered in this study.
figure 1

The eight moisture sources including high-latitude land area (0–150°E, 42–70°N, HLA), northern China (102–120°E, 32–42°N, NC), southern China (102–120°E, 20–32°N, SC), low-latitude land area (30–105°E, 20–42°N, LLA), Mediterranean Sea & North Atlantic Ocean (0–30°E, 30–42°N & 60°W–0°, 30–70°N, M_N), Pacific Ocean (120°E–125°W, 20–50°N, 120°E–90°W, 10°S–20°N, PO), South China Sea (105–120°E, 10°S–20°N, SCS), North Indian Ocean (40–105°E, 10°S-20°N, NIO). The black box represents the defined areas for northern China.

Full size image

Influence of local meteorological conditions on δ18Op

In the context of the predominantly continental climate in northern China, the local meteorological factors may play a vital role in shaping the δ18Op seasonality31. Strong relationships between temperature and δ18Op have been confirmed in northern China from International Atomic Energy Agency data32. Such a relationship exists in our data for the non-monsoon seasons. A positive correlation between monthly weighted average δ18Op and temperature was observed in winter (NDJF), pre-monsoon (MAM), and post-monsoon (SO) periods in CAM3 simulation (Table 1). Notably, the temperature effect diminishes during the summer monsoon season (JJA), coinciding with a significant increase in precipitation. While previous studies have suggested that the amount effect overshadows the temperature effect during this period20, our findings propose that the influence of precipitation amount on δ18Op appears relatively weak (Table 1). These results suggest that the local meteorological factors cannot fully explain the variations of the δ18Op. It is imperative to consider other large-scale processes, including moisture sources, transport distances, and upstream convection.

Table 1 The Pearson correlation between the amount weighted δ18Op and local meteorological factors.
Full size table

Climatology of moisture contributions

In this section, we quantify the relative contributions of moisture from the eight source regions to northern China. The annual cycles of simulated moisture contributions to total precipitation (longitudinally averaged from 102°E to 120°E) from each tagged region between 10°N to 50°N are illustrated in Hovmöller diagrams (Fig. 2e). During the non-monsoon season, moisture contributions to northern China are primarily from PO (~22–34%) and LLA (~25–32%) source regions (i.e., recycled vapor originating from low latitude land surfaces (Fig. 2).

Fig. 2: Results from the CAM3 tagging simulation.
figure 2

Histograms showing relative contributions of each vapor source (%) to total precipitation averaged over northern China for winter (a), pre-monsoon (b), summer monsoon (c), and post-monsoon (d) seasons. The Pearson correlation between CAM3 simulated monthly anomalies of amount weight oxygen isotopic composition of precipitable (δ18Op) and contribution (%) for each vapor source was marked, * means significant at 95% confidence level (one-tailed, t test; after accounting for autocorrelation). Moisture sources that contribute less than 10% are considered minimal effects. e Hovmöller diagram showing the seasonal evolution of tagged vapor contribution (%) from each source region (longitudinally averaged between 102°E and 120°E) and spanning between 10°N and 55°N. Dotted black lines highlight the June to August monsoon season. The period of the analysis is from 1951 to 2020.

Full size image

Previous studies have indicated that the westerlies serve as a primary carrier of water vapor transport in northern China during the non-monsoon season, with the Atlantic Ocean, Black Sea, Caspian Sea, and recycled moisture during the pathway being the main moisture sources27,33. Despite the strength of the westerlies, some studies have pointed out that the main source of winter precipitation is oceanic moisture from the eastern transport12. Our results suggest a combined continental and oceanic moisture contribution during the non-monsoon season, with the westerlies and western Pacific subtropical high (WPSH) playing key roles26,34,35. The westward flow of western Pacific moisture, driven by the WPSH (depicted by the blue dashed line in Fig. 3), intersects with LLA moisture driven by the southern branch of tropospheric westerlies, known as the India-Burma trough (depicted by the red dashed line in Fig. 3), deflecting moisture northward to northern China.

Fig. 3: Climatological water vapor flux patterns from CAM3 (1951–2020).
figure 3

Vertical integral water vapor flux (shaded and arrows) climatology for winter (a), pre-monsoon (b), summer monsoon (c), post-monsoon (d) from CAM3. The blue dashed line indicates the Pacific Ocean moisture channel, the red dashed line indicates the westerlies moisture channel, and the white dashed line indicates the North Indian Ocean moisture channel.

Full size image

During the summer monsoon season, the cross-equatorial low-level jet intensifies, serving as the dominant pathway of moisture transport across the equatorial Indian Ocean and the Arabian Sea into northern China (white dashed line, Fig. 3), and the moisture contribution from NIO significantly increases (~13%). In contrast, the moisture contribution from the LLA region decreases (~18%) due to the weakening of westerlies (Figs. 2 and 3). The contribution from the PO remains relatively consistent at around 24%, which is attributed to the development and reinforcement of the western and equatorial Pacific moisture channel with the formation of EASM18,36. This channel converges with the NIO channel at about 110°E and turns northward, transporting the PO moisture to northern China (blue dashed line, Fig. 3). Overall, the contribution of land moisture decreases significantly during summer monsoon season, the PO and the NIO become the major oceanic water vapor contributors.

Influence of moisture contribution and upstream convection on δ18Op

Linear correlations between monthly weighted δ18Op and moisture contributions from each tagged source region are summarized in Fig. 2. During the non-monsoon season, the moisture source contributions exhibit a minor impact on the δ18Op values. With relatively stable moisture sources and low precipitation amount, the temperature-dependent isotope fractionation assumes significance, making temperature the primary factor driving δ18Op variations. However, during the summer monsoon season, the influence of moisture sources becomes apparent. Specifically, summer δ18Op exhibits a notable negative correlation with NIO moisture contribution (r ~ −0.51), and a positive correlation with contributions from PO and LLA source regions (r ~ 0.30 and 0.26, respectively) (Fig. 2c).

These correlations might be attributed to differences in the initial moisture δ18Ov values in the source region30. Contrary to the expectation, our analysis reveals no significant correlation between the initial δ18Ov of each moisture and δ18Op in northern China (Table 2). While there is a positive correlation between local δ18Ov and δ18Op, this is more likely due to local processes. Thus, we conclude that the initial moisture δ18Ov signal does not exert a discernible influence on δ18Op in northern China. Instead, modification of the initial moisture δ18Ov along the trajectories toward the sink region plays a pivotal role37.

Table 2 The Pearson correlation between the amount weighted δ18Op and each initial moisture δ18Ov.
Full size table

Transport distance from moisture sources to the sink region is proposed to influence EASM δ18Op, with extended Indian Ocean moisture leading to depleted δ18Op value, while adjacent western Pacific Ocean moisture exhibits an opposing effect13,38. However, in the context of the vast area of defined PO moisture, the transport distance may act as a secondary contributing factor. More significantly, upstream convective rainout processes during the water vapor transport pathway, affected by the seasonal movement of the Intertropical Convergence Zone, have the potential to alter the precipitation and vapor isotope compositions in the upstream regions, subsequently extending this influence to downstream sites8,9,10,37,39,40.

So far, it remains uncertain whether upstream convection affects δ18Op variation in northern China due to the extended distance from the tropical convection zone18,26. To address this, we analyzed the relationship between δ18Op and δ18Ov in northern China and outgoing longwave radiation (OLR) as an indicator for convection intensity41. During the summer monsoon season, we find significant positive correlations between δ18Op and OLR in most of the NIO source region, suggesting that lower δ18Op is associated with stronger convection in this rainout zone (Fig. 4). These high correlation areas cover especially in the India sub-continent, Bay of Bengal and Arabian Sea (Fig. 4), where the convection is the most active24,26,42. When evaporated vapor moves through the convection zone, deep convection-induced intensified re-evaporation or enhanced downdrafts, carrying lighter vapor from upper atmospheric layers, could lead to a depletion of the NIO δ18Ov10,43,44. This influence is subsequently propagated to and impacts the δ18Op in downstream areas. Therefore, the isotopically depleted NIO moisture, caused by the intensified upstream convection process during the moisture transport pathway, contributes to negative summer δ18Op values in northern China in comparison with moisture transported from PO and LLA, which typically feature enriched δ18Ov.

Fig. 4: Spatial correlation fields of CAM3 simulated δ18O in northern China and outgoing longwave radiation.
figure 4

a Spatial correlations between CAM3 simulated JJA amount weighted precipitation δ18O (δ18Op) in northern China and JJA outgoing longwave radiation (OLR) at each grid cell. b Same as a but for CAM3 simulated water vapor δ18O (δ18Ov) at lower-troposphere (1000–750 hPa) in JJA. The red rectangular is the defined sink region (102°–120°E, 32°–42°N). The correlation was performed after removing the climatological mean and trend from the data. Stippling represents data points that are statistically significant at the 95% confidence level obtained after accounting for serial correlations in data at each grid point followed by the application of FDR procedure with a 5% threshold (see “Methods”). FDR represents the anticipated fraction of null hypotheses that are mistakenly rejected in multiple-hypothesis testing.

Full size image

Moisture control on decadal variability of summer δ18Op

The δ18Op in East Asia has been proposed as a reliable index of EASM activity2,3,4,5,13, considering the minimal contribution of non-monsoon precipitation in the northern region17, we use the summer δ18Op (δ18Op_JJA) as an index of EASM intensity. The northern China δ18Op_JJA exhibits a decadal-scale weakening in the EASM intensity during the mid-1980s (Fig. 5a), which is observed across EASM region (Supplementary Fig. 6). Changes in the amplitude of decadal EASM variability over the twentieth century have been associated with significant changes in precipitation pattern over eastern China45,46,47. Previous studies have shown that the decadal-scale climate variability over the twentieth century was partly attributable to changes in regional water vapor transport30,47,48. The Pacific Decadal Oscillation (PDO)49,50,51, and the SST warming over the North Atlantic Ocean52 are considered the main influencing factors. The correlation between 7-year low-filtered δ18Op_JJA and global SST indeed resembles a PDO pattern (Supplementary Fig. 7). Therefore, we focus on the PDO phase transition on the δ18Op_JJA pattern and moisture supply in northern China.

Fig. 5: Time-series comparisons.
figure 5

Z score transformed time-series comparisons between a the modeled JJA amount weighted precipitation δ18O in northern China, b the modeled JJA precipitation in northern China, c the Pacific Decadal Oscillation index, d stalagmite δ18O in Shihua Cave for 1951–2020. The black curves indicate the 7-year low-pass Butterworth filter.

Full size image

We use composite analysis to examine decadal variations in δ18Op_JJA and the related moisture contributions from each tagged source over northern China during different PDO phases. Two periods before and after 1985 (i.e., 1967–1984 and 1985–2002) are selected as the typical cold and warm PDO phases. Figure 5 shows the synergistic decadal variation between the δ18Op_JJA and summer precipitation (r ~ −0.48, 7-year low-filtered). The warm PDO phase is characterized by positive δ18Op_JJA values and reduced precipitation amount in northern China, and vice versa for the cold PDO phase (Fig. 5a, b). During the warm PDO phase, PO moisture converges over most of the EASM domain due to a weaker and/or eastward-shifted WPSH53 (Fig. 6b). On the other hand, moisture contribution from NIO is substantially reduced over northern China as a result of weakened low-level circulation (Fig. 6a). There is also a reduction of land recycled moisture, especially the local recycled moisture (Fig. 6c). In contrast, moisture from LLA, HLA, M_N, SC, and SCS shows a minor variation (Supplementary Figs. 8 and 9). These alternations are more significant in the moisture contribution compared to the change in water vapor volume (Fig. 6d–f).

Fig. 6: Positive minus negative PDO phase JJA tagged moisture and contribution anomalies.
figure 6

a–c JJA moisture anomalies from NIO (a), PO (b), and NC (c) are superimposed by simulated vertically integrated moisture flux anomalies. d–f JJA moisture contribution anomalies from NIO (d), PO (e), and NC (f) are superimposed by simulated vertically integrated moisture flux anomalies. The black box represents the defined areas for northern China.

Full size image

We also examine how variations in moisture contributions from NIO, PO, and NC (local recycled moisture) affect the decadal variation of δ18Op_JJA, as these three moisture sources show the largest change between warm and cold PDO phases. During warm PDO phases, diminished isotopically depleted NIO moisture coupled with the increased isotopically enriched PO moisture result in enriched δ18Op_JJA over northern China (Figs. 5 and 6). Notably, local recycled moisture is substantially reduced during the warm PDO phase (Fig. 6c, f). However, the lack of correlation with δ18Op_JJA (Fig. 2c) implies that the reduction doesn’t necessarily result in enriched δ18Op_JJA, which could be attributed to the possibility that local recycled moisture comes from transpiration instead of evaporation54. Therefore, the local recycled moisture does not significantly influence the decadal variation of δ18Op_JJA. However, reduced local recycled and NIO moisture, despite increased PO moisture, contributes to decreased precipitation in northern China (Figs. 5 and 6). Both large-scale atmospheric circulation and local processes thus govern the decadal variation of precipitation in northern China. Different from precipitation, the decadal δ18Op_JJA variation reflects the relative oceanic moisture contribution related to the large-scale atmospheric circulation between the NIO and PO across different PDO phases (Fig. 6d, e). We find that such a decadal signal is recorded in the published stalagmite δ18O in Shihua Cave55 (Fig. 5d). Our CAM3 tagging simulation data shed light on potential mechanisms driving the decadal variability of the natural δ18O record from this area.

Discussion

Over the past two decades, stalagmite records have provided valuable insights into hydroclimate changes in the EASM region. However, the focus of stalagmite-based studies has mainly been on the monsoon region of southern China, with limited records from northern China, particularly modern ones, hindering accurate interpretations of stable isotopic records in this region. In this context, our analysis contributes to a better understanding of the role of various moisture sources on δ18Op in northern China on the seasonal to decadal time scale.

Our data highlight the significantly distinct controlling factors for northern China δ18Op composition in different seasons. During the non-monsoon season, the δ18Op is primarily influenced by the temperature effect, resulting in the lowest and highest values in the winter and the pre-monsoon seasons, respectively. During the monsoon season, there is actually a large shift in moisture sources, contributions from the LLA, PO, and NIO (17.8%, 24.4%, and 13.0%, respectively) are significant moisture sources impacting the summer δ18Op. Our results suggest that summer δ18Op and δ18Os variability in northern China reflect large-scale changes in moisture source and dynamic interplay of upstream convection processes.

Moreover, this study provides further evidence that the variation of oceanic moisture from the PO and the NIO plays a vital role in the decadal variation of the δ18Op_JJA, which is modulated by the atmospheric circulation anomalies influenced by PDO. While our study emphasizes the dominant role of moisture sources in influencing the decadal variation of δ18Op_JJA, we acknowledge that δ18Op_JJA in northern China also serves as an indicator of long-term precipitation changes. These changes can arise not just from the “amount effect”, but also from shifts in the relative contributions of different moisture sources. Taken together, the interpretation of decadal δ18Op_JJA and δ18Os variation in northern China can be potentially interpreted in a framework of overall changes in the moisture source dynamics and regional precipitation intensity.

Methods

CAM3 model design and applications

The CAM3 was developed at the National Center for Atmospheric Research (NCAR) with a horizontal resolution of 2.8° × 2.8°, and 26 verticals with the model top at 3.5 hPa29. We chose the finite-volume core (CAM-FV) to ensure optimal mass and energy balance, separating the dynamical core from physical parameterization through a time-split approximation29,56,57. CAM3 employs the Community Land Model as its surface scheme58 and uses a prognostic parameterization for its cloud water scheme, initially developed by Rasch and Kristjánsson59 and later updated by Zhang et al.60. Convective cloud fraction is calculated using functions of convective updraft mass flux, following the approach of Xu and Krueger61.

A moisture-tagging module is integrated into CAM362, providing the amount of water vapor from each source region that turns into precipitation at each grid for each month63,64. This module employs additional tracers for each source region to tag and track the water vapor from evaporation to precipitation62,65,66. Additionally, CAM3 has a stable water isotopes module, which allows the simulation of isotopic fractionation processes related to surface evaporation and cloud physics29,67,68.

Statistical methods for correlation analysis

We employed the least-squares linear regression method to calculate the linear Pearson correlation coefficients between the independent and dependent variables. To establish the confidence intervals for these coefficients, we employed a pairwise moving-block bootstrap method using the PearsonT3 software69. The block length of this bootstrap method was determined on average proportional to the estimated data autocorrelation. This approach effectively preserves the serial dependence of time-series data and provides corrected 95% confidence intervals that remain valid in the presence of autocorrelation. The coverage accuracy is increased by applying calibration to standard error-based bootstrap Student’s t confidence interval.

In the statistical significance of field correlations (i.e., Fig. 4), we implemented the false discovery rate (FDR) procedure using a MATLAB code to control the proportion (q = 5%) of falsely rejected null hypotheses, where q guarantees that 5% or fewer of the locations where the null hypothesis is rejected are false detections70. FDR generally provides a more robust method for correcting for multiple comparisons71,72 than procedures like Bonferroni correction that provide robust control of the familywise error rate (i.e., the probability that one or more null hypotheses are mistakenly rejected).