Introduction

The Tibetan Plateau (TP), recognized as one of the most climate-sensitive regions under global climate change, plays a critical role in the Earth’s climate system. As the largest repository of glaciers outside the polar regions, the TP is often referred to as the “Water Tower of Asia”, sustaining the hydrological systems of several major rivers across the continent1,2,3. The TP hosts approximately 1400 lakes, each with an area exceeding 1 km2 4, and is often termed the “Roof of the World” or the “Third Pole” due to its average elevation exceeding 4000 m above sea level5. Under global warming, the TP has experienced amplified warming and increased precipitation, with these changes being more pronounced compared to other regions6. In recent decades, extreme precipitation events have been witnessed on the TP, resulting in devastating consequences such as floods, debris flow, urban waterlogging, and soil erosion. These extreme precipitation events not only impact the TP but also exert significant influences on the weather and climate patterns of surrounding regions7,8.

Recent studies have demonstrated a significant increasing trend in summer precipitation over the western TP under global warming9. The Inner TP (ITP; also called the Qiangtang Plateau) encompasses much of the western TP and parts of the central TP. It contains approximately 820 lakes larger than 1 km2 4, accounting for 62% and 70% of the total number and area of TP lakes, respectively4. Notably, these lakes have undergone substantial expansion since the mid-1990s10,11, a phenomenon closely linked to the observed increase in precipitation over the past two decades, as evidenced by various datasets12,13. However, due to the harsh natural conditions of the ITP, meteorological stations are sparsely distributed, with most concentrated in the central and eastern TP. This scarcity of observational data, combined with the region’s unique topography, has hindered a comprehensive understanding of the spatiotemporal variability of summer precipitation and its underlying mechanisms9.

Summer precipitation over the ITP is influenced by a complex interplay of latitudinal atmospheric systems14. For instance, the El Niño-Southern Oscillation has been shown to significantly modulate interannual variability in summer precipitation over the southwestern TP15. Additionally, changes in sea surface temperature in the Atlantic, Pacific, and Indian Oceans have enhanced moisture influx into the ITP, contributing to increased precipitation and subsequent lake expansion4. Furthermore, the positive phase of the North Atlantic Multidecadal Oscillation has likely played a role in the observed increase in ITP precipitation since the mid-1990s, through the weakening or northward shift of the westerly jet stream9. Previous studies have also identified distinct precursor wave trains associated with circulation anomalies during extreme precipitation events over both the northwestern and southeastern TP16. Given that extreme precipitation events are often accompanied by anomalous synoptic conditions, investigating these events provides valuable insights into the drivers of precipitation variability over the ITP. Climate projections indicate that ITP will experience increased wetness due to future global warming17, with a more pronounced rise in extreme precipitation18. Despite these projections, the characterization of summer extreme precipitation over the ITP and its underlying causes remains poorly understood. Most prior research has focused on the southeastern TP19, with limited attention given to the western TP or the ITP. Therefore, it is imperative to explore whether the multimodal characteristics of extreme precipitation over the ITP are related to different Rossby wave train sources and whether distinct moisture transport mechanisms are involved.

This study aims to elucidate the spatiotemporal characteristics of summer extreme precipitation over the ITP and to identify two typical dominant circulation patterns governing extreme precipitation events and their potential impacts by employing a clustering method. Furthermore, we examine the precursor sources and propagation dynamics of large-scale remotely correlated wave trains. These findings are expected to provide scientific insights for improving the prediction of summer extreme precipitation events over the ITP and deepen our understanding of the mechanisms driving lake expansion in the region.

Results

Regional characteristics and patterns of summer extreme precipitation over the ITP

Summer extreme precipitation over the ITP is generally limited, with the primary precipitation zone situated to the south and the maximum extreme precipitation occurring in the southeastern region (Fig. 1b). However, the northern part of the ITP exhibits a higher frequency of extreme precipitation (Fig. 1c). The R90P index, representing the cumulative amount of precipitation when daily rainfall exceeds the 90th percentile during summer, contributes 29.60% to the average summer precipitation over the ITP (Fig. 1d). Among the two categories of moderate-intensity precipitation, those falling within the 50th to 75th percentile range account for 25.78% of total precipitation, while the range from the 75th to 90th percentile contributed 25.20% (Fig. 1d). In comparison, light precipitation accounts for a smaller proportion, at 19.42%. The annual averages for R90P, R90D (number of extreme precipitation days), and R90I (intensity of extreme precipitation) over the ITP are 70.88 mm, 8.37 d, and 8.47 mm d−1, respectively. Before 2010, R90P and R90D exhibited a modest increasing trend, with rates of 0.33 mm y−1 and 0.05 d y−1, respectively. However, post-2010, these trends intensified significantly, with R90P and R90D increasing at rates of 4.85 mm y−¹ and 0.44 d y−¹, respectively. Multiple datasets consistently demonstrate this increasing trend (Fig. S1). However, the southeastern TP exhibits a significant decreasing trend in extreme precipitation events19, which stands in marked contrast to the increasing trend observed in the western region18 (Fig. S2). This observed spatial heterogeneity provides robust evidence for the divergent trends of extreme precipitation between the eastern and western sectors of the TP against the backdrop of global climate change20. R90I exhibited a relatively stable climatological trend since R90P and R90D displayed consistent variability, indicating that the overall increase in R90P over the ITP is primarily driven by the rising frequency of extreme precipitation events rather than changes in their intensity (Fig. 1e).

Fig. 1: Climatology of summer extreme precipitation over the ITP.
figure 1

a Geographic location of the ITP, b composite annual JJA-mean extreme precipitation (shading, unit: mm), and c its frequency (shading, unit: d). d Contribution from the five classifications of daily precipitation intensity (<25th, 25–50th, 50–75th, 75–90th, and >90th percentile) to the JJA-mean precipitation over the ITP. The precipitation and the percentage values are marked on the bars. e Interannual variation of R90P (unit: mm), R90D (unit: d), and R90I (unit: mm d−1) over the ITP.

Full size image

Based on the definition of regional extreme precipitation events (REPEs), a total of 220 REPEs were identified during the summer periods from 1979 to 2020. Cluster analysis identifies two (cluster numbers are referenced by objective scores: silhouette coefficient (SC)21 and Caliński–Harabasz (CH) score22) synoptic patterns responsible for the summer REPEs over the ITP (Fig. 2). The first pattern, REPE1, occurred 111 times (50.45%) during 1979–2020 and is characterized by a precipitation center located in the southeastern part of the ITP (Fig. 2a). REPE1 is associated with southwesterly winds to the south of the ITP and the presence of high-pressure systems at middle and high latitudes to the west of the TP (Fig. 2a). Upper-level circulation during REPE1 events is dominated by anomalous westerlies across the ITP (Fig. 2c). The second pattern, REPE2, occurred 109 times (49.55%) and features a precipitation center situated in the southwestern ITP. REPE2 is marked by the development of a low-pressure trough to the west of the TP and a cyclonic vortex system over the Indian Peninsula, which drives southeasterly winds to the south of the ITP (Fig. 2b). In contrast to REPE1, REPE2 events are characterized by pronounced upper-level easterly anomalies across the ITP (Fig. 2d). These two synoptic patterns highlight the contrasting atmospheric circulation mechanisms that govern summer extreme precipitation events over the ITP. Notably, the circulation systems governing two types of extreme precipitation events over the ITP exhibit both similarities and differences compared to those driving southeastern TP events identified by Cao et al.19. Both events are influenced by the Middle Eastern high-pressure system and the Indian Peninsula low-pressure system (Fig. 2), indicating a common synoptic-scale forcing mechanism. While the ITP’s REPE1 event shares similar characteristics with the Type 1 event of southeastern TP, its systems shift westward (Fig. 2a). This similarity may be associated with the concentration of REPE1 precipitation in the southeastern TP (Fig. 2a). Mid-latitude circulation contrasts are more pronounced: the trough of SETP’s Type 1 event is more in the west than REPE2’s counterpart (Fig. 2b), and SETP’s Type 2 trough exhibits greater intensity than REPE1 (Fig. 2a)19. These results demonstrate that extreme precipitation over different TP sectors arises from both interconnected circulations and region-specific mid-latitude adjustments, refining our understanding of summer precipitation extremes across the TP.

Fig. 2: Two typical summer REPEs over the ITP and their climatology.
figure 2

a Composite precipitation (shading, unit: mm d−1), geopotential height (contours; unit: gpm), and horizontal wind (vectors; unit: m s−1) at 500 hPa on the onset day (D0) of REPE1 and number of occurrences of REPE1 in June, July, and August during 1979–2020 in inset. c The climatology (green contours, unit: m s−1) and anomalous (shading, units: m s−1) of U-wind at 200 hPa, areas significant at the 0.05 level are denoted by dots. b, d are the same as (a, c), respectively, but for REPE2. e The interannual variation of REPE1 and REPE2.

Full size image

REPE1 occurs 30 and 34 times in June and August, respectively, with the most occurring in July at 47 times (Fig. 2a). In contrast, REPE2 is largely concentrated in July and August, with 52 and 53 times, respectively, but is notably rare in June, with only 4 times (Fig. 2b). This seasonal disparity in REPE2 frequency may be attributed to the northward migration of the westerly jet stream during the summer months23,24, which is observationally supported by the distinct contrast in westerly jet stream positions between REPE1 and REPE2 (Fig. 2c, d). Consequently, nearly all extreme precipitation events over the ITP in June are driven by the REPE1 pattern. Correlation analysis of interannual variability in REPE1 and REPE2 event frequencies reveals a negligible correlation (r = −0.06) with no statistical significance, further confirming the independence of these two event types. Before 2010, the frequencies of both REPE1 and REPE2 remained relatively stable. However, post-2010, REPE1 exhibited a modest increase at a rate of 0.05 times y−1, while REPE2 showed a more pronounced upward trend of 0.16 times y−1 (Fig. 2e). This marked rise in REPE2 frequency post-2010 is a key driver of the significant increases observed in R90P and R90D during the same period (Fig. 1e).

Circulation characteristics of two typical summer extreme precipitation events over the ITP and their precursor signals

The analysis above reveals two distinct types of REPEs over the ITP, each characterized by unique atmospheric circulation features. The synoptic conditions during REPE1 and REPE2 exhibit marked differences in their dynamical structures. For REPE1, a negative-positive-negative geopotential height anomaly pattern is observed at middle and high latitudes. On the onset day (D0), an anomalous cyclonic circulation is evident over the northwestern TP at 200 hPa (Fig. 3a), while at 500 hPa, the anomalous cyclone extends over the ITP (Fig. 3c). Significant positive vorticity anomalies are present below 400 hPa over the ITP, driving strong upward vertical motion (Fig. 3e). Concurrently, anomalous southwesterly winds facilitate substantial moisture transport from the Indian Ocean to the southeastern ITP (Fig. 3c, g)25, creating favorable conditions for extreme precipitation in this region. In contrast, REPE2 is associated with a positive–negative–positive geopotential height anomaly pattern at middle and high latitudes. On D0, an anomalous cyclonic circulation is located to the northwest of the TP, while an anomalous anticyclone dominates the northeastern TP (Fig. 3b, d). This configuration induces southwesterly winds to the west of the ITP. Positive vorticity anomalies below 300 hPa over the ITP enhance upward motion (Fig. 3f). Additionally, southeasterly winds driven by a low-pressure system over the Indian Peninsula converge with southerly winds in the west of the ITP, resulting in moisture convergence over the southwestern ITP (Fig. 3h), which promotes the occurrence of REPE2. The occurrence of REPE1 and REPE2 is attributed to different circulation anomalies. REPE1 is primarily driven by westerly wind anomalies (Fig. 3e), whereas REPE2 is influenced by easterly wind anomalies (Fig. 3f). Furthermore, the moisture supply for REPE1 is more pronounced compared to that for REPE2. These fundamental differences in circulation patterns and moisture transport mechanisms lead to significant variations in the spatial and temporal characteristics of the two REPE types (Fig. 2).

Fig. 3: Composite atmospheric circulation and moisture flux anomalies on the onset day (D0) of REPE1 and REPE2 occurrence.
figure 3

Composite geopotential height anomalies (shading; unit: gpm) and horizontal wind anomalies (vectors; unit: m s−1) at 200 hPa (a, b) and 500 hPa (c, d) on the onset day (D0), vertical cross-sections of composite vorticity anomalies (shading; unit: 10−6 s−1) and wind anomalies (vectors; horizontal airflow unit: m s−1, vertical airflow unit: 10−2 Pa s−1) averaged at 30°–35°N on D0 (e, f), vertically integrated moisture flux anomalies (vectors; unit: kg m−1 s−1) and moisture divergence anomalies (shading; unit: 10−5 kg m−2 s−1) on D0 (g, h) for REPE1 (first column) and REPE2 (second column) over the ITP. Black dotted areas indicate that the geopotential height, vorticity, and moisture flux divergence anomalies are statistically significant at a confidence level of 95%. Bolded green vectors indicate that the wind field is statistically significant at a confidence level of 95% in (a–f). Red vectors in (g, h) indicate that the moisture flux divergence is also statistically significant at a confidence level of 95%. The region enclosed by the black polygon represents the TP and ITP.

Full size image

To elucidate the evolution of atmospheric circulation anomalies associated with REPE1 and REPE2 and to identify the large-scale precursor signals leading to these extreme precipitation events, we analyzed the circulation field anomalies before the onset day (D0): the eighth (D-8), sixth (D-6), fourth (D-4), and second (D-2) days before the event, as well as the second day after the onset (D2). The atmospheric patterns preceding and following REPE1 and REPE2 exhibit distinct evolutionary characteristics.

For REPE1, a Rossby wave source originating from the western North Atlantic on D-8 initiates an eastward-propagating Rossby wave train, which reaches western Europe by D-6 (Figs. S3a and 4c). By D-4, the wave train progresses further eastward (Fig. 4e), coinciding with the development of convective anomalies in the southeast of the ITP (Fig. S5e). A pronounced negative-positive-negative geopotential height anomaly pattern emerges by D-2 (Fig. 4g), accompanied by intensified convective activity (Fig. S5g) and a significant increase in moisture over the Indian Peninsula (Fig. S4g). On D0, an anomalous cyclonic circulation intensifies over the northwestern TP, serving as the primary synoptic system driving REPE1 (Figs. 4i, S4i, and S5i). By D2, the energy of the Rossby wave train weakens as it propagates eastward and moves away from the ITP (Fig. 4k).

Fig. 4: Evolution of composite atmospheric circulation anomalies before and after the occurrence of REPE1 and REPE2 during 1979–2020.
figure 4

Evolution of the composite anomalies in geopotential height at 200 hPa (shading, bottom color bar; unit: gpm) and 500 hPa (contours, right color bar; unit: gpm) and horizontal wave activity fluxes at 200 hPa (vectors; unit: m2 s−2) from D-8 to D2 with an interval of 2 days for REPE1 (a, c, e, g, i, k) and REPE2 (b, d, f, h, j, l) over the ITP. Regions marked by black dots and solid contours indicate statistically significant geopotential height anomalies at the 95% confidence level at 200 and 500 hPa, respectively. The horizontal wave activity flux values less than 0.5 m2 s−2 are omitted. The region enclosed by gray and black polygons represents the TP and ITP, respectively.

Full size image

In contrast, for REPE2, a Rossby wave source over the northeast Atlantic excites a downstream-propagating Rossby wave train (Fig. S3b). A positive-negative-positive-negative-positive wave train forms across Eurasia by D-6 and D-4 (Fig. 4d, f). By D-2, the eastward propagation of wave energy strengthens the positive-negative-positive geopotential height anomaly upstream of the ITP (Fig. 4h), which further intensifies by D0, becoming a critical synoptic system for REPE2 (Fig. 4j). The Rossby wave train continues to transmit energy downstream, influencing weather patterns in regions east of the ITP (Fig. 4l)26. Moisture over the ITP begins to increase by D-6 and persists throughout the event (Fig. S4), while convective activity initiates and strengthens along the Indian Peninsula (Fig. S5). By D-2 and D0, southerly winds over the western ITP significantly enhance moisture levels in the ITP (Fig. S4), and convective activity over the ITP peaks (Fig. S5), creating optimal conditions for the occurrence of REPE2. Notably, the governing Rossby wave activity for REPE2 exhibits a near-antiphase relationship with that of REPE1, constituting a critical distinction between the two event types. These findings highlight the distinct precursor signals and dynamical processes associated with REPE1 and REPE2, underscoring the role of Rossby wave trains and moisture transport in shaping the evolution of extreme precipitation events over the ITP.

Discussion

The ITP, characterized by its high elevation and extensive lake systems, plays a critical role in regional hydrology and ecology. The expansion of ITP’s lakes, driven by climatic changes, has profound societal and ecological implications27. In the context of global warming, summer extreme precipitation over the ITP has intensified, significantly influencing lake dynamics. This study investigates the characteristics and formation mechanisms of two dominant types of summer extreme precipitation events (REPE1 and REPE2) over the ITP. Notably, REPE2 has exhibited a more pronounced increase in recent years compared to REPE1, likely due to changes in the synergistic effects of westerly winds and monsoons24, as well as alterations in moisture transport pathways under global warming28. REPE1 is primarily driven by an eastward-propagating Rossby wave train originating from the northwest Atlantic. This wave train induces a cyclonic anomaly over the northwestern TP, which transports substantial unstable energy to the southeastern ITP. Coupled with enhanced moisture influx from the Indian Ocean and strong upward motion over the ITP, these conditions facilitate the occurrence of REPE1 (Fig. 5a). In contrast, REPE2 is associated with a positive-negative-positive geopotential height anomaly pattern, sustained by a Rossby wave train propagating from the northeast Atlantic. The convergence of wave energy maintains the intensification of this pattern. Southeasterly winds south of the TP, combined with southerly winds west of the ITP, channel moisture to the southwestern ITP. This moisture convergence, alongside robust upward motion, drives the formation of REPE2 (Fig. 5b). The increasing frequencies of REPE1 and REPE2 have contributed to a significant rise in R90P and R90D over the ITP in recent years, although R90I has not shown a comparable increase.

Fig. 5: Characterization of atmospheric circulation and precursors during the occurrence of REPEs over the ITP.
figure 5

a REPE1, b REPE2. C: cyclonic circulation anomaly. A: anticyclonic circulation anomaly. The dark green arrow indicates moisture transport and its direction, and the light green arrow indicates enhanced upward motion. The purple curve stands for the Rossby wave train.

Full size image

Previous studies have shown that extreme precipitation over northwestern TP is associated with highly anomalous positive-negative-positive geopotential in the upper and middle troposphere on D018, and our study finds that this presented feature is jointly contributed by the occurrence of two different extreme precipitation events, which refines the understanding of extreme precipitation in the western part of TP, the results of the study (e.g., the two typical antiphase wave train precursor) also enhances the potential for improved climate and precipitation prediction in this region. Notably, summer extreme precipitation events over the ITP exhibit analogous low-latitude synoptic drivers to those influencing the southeastern TP, albeit with distinct spatial configurations, while demonstrating pronounced divergence in mid-high-latitude circulation systems—a disparity likely associated with differential Rossby wave activity upstream. Specifically, REPE1 events display a structural analogy to Type 1 extreme precipitation events over the southeastern TP19. This advances a unified framework for understanding regional heterogeneity in extreme precipitation over the TP. However, this study focuses on synoptic-scale characteristics and precursor signals of extreme precipitation, the factors influencing extreme precipitation over the ITP on longer timescales, such as decadal variability and large-scale climate modes, remain to be explored29,30. Additionally, the ITP’s extreme altitude poses challenges for data accuracy, necessitating future efforts to improve observational and modeling capabilities in this region31. As extreme precipitation events over the ITP become more frequent, their impact on lake expansion and regional hydrology warrants further investigation32. Future research should aim to refine the mechanisms by which extreme precipitation influences lake dynamics. Furthermore, the use of CMIP6 models will be essential to assess projected trends in REPE1 and REPE2 under future climate scenarios33, providing critical insights for adaptation and mitigation strategies in this climatically sensitive region34.

Methods

Data

This study focuses on the summer extreme precipitation over the ITP during the period 1979–2020. The datasets utilized in this research are as follows:

The precipitation data employed in this study comprise a long-term, high-resolution dataset spanning 1979–2020, with a spatial resolution of 1/30° on a daily scale, specifically tailored for the Third Pole region. This dataset was generated through advanced atmospheric modeling techniques, integrated with extensive observational data sourced from the National Tibetan Plateau/Third Pole Environment Data Center in China35. The high resolution and accuracy of this dataset facilitate the robust detection and analysis of extreme precipitation events36.

To analyze atmospheric circulation patterns, meteorological parameters were extracted from the ERA5 reanalysis dataset, provided by the European Centre for Medium-Range Weather Forecasts. The ERA5 dataset includes hourly data on single levels and pressure levels37, with a horizontal resolution of 0.25° × 0.25°. This dataset has demonstrated high reliability and accuracy over the TP region38,39. Key meteorological variables, such as geopotential height, horizontal wind components, air temperature, and other relevant parameters, were computed as daily averages, covering the period 1979–2020. This study clarifies the relationship between atmospheric circulation fields and extreme precipitation events by temporally aligning their occurrence. These data establish a comprehensive foundation for diagnosing synoptic-scale atmospheric processes linked to extreme precipitation over the ITP.

Study region and definition of the REPEs

The ITP is situated in the west-central region of the TP, as shown in Fig. 1a. In this study, REPEs are defined as follows based on previous studies18,40,41: Specifically, the 90th percentile of all historical daily precipitation records with a daily intensity ≥0.1 mm d−1 during the summer months (June–August) from 1979 to 2020 is adopted as the extreme precipitation threshold for each grid point. An extreme precipitation event at a given grid point is identified when the daily precipitation exceeds this 90th percentile threshold. A regional extreme precipitation day is defined when the number of grid points experiencing extreme precipitation on that day exceeds the 90th percentile of all summer days.

A REPE is characterized as a period encompassing one or more consecutive regional extreme precipitation days. For single-day REPEs, the onset day is designated as D0, with Dn (D-n) representing n days after (before) D0. For multi−day REPEs, the variables on D0 are computed as the average across all days within the event, while Dn (D-n) refers to n days after (before) the last (first) day of the event. The REPEs identified through this methodology are presented in Fig. S6. We also implemented a comparative analysis employing 90th and 95th percentile thresholds to identify REPEs, and the results demonstrated substantial consistency between these methodological frameworks, thereby validating the robustness of our findings. This framework ensures a consistent and systematic approach to analyzing the temporal evolution and spatial characteristics of REPEs over the ITP.

Clustering method

In this study, we employed a spectral clustering method to classify extreme precipitation events over the ITP, considering the influence of atmospheric circulation patterns. The spectral clustering method was selected due to its superior adaptability to data distribution compared to traditional methods such as K−means40 and hierarchical clustering19, offering enhanced clustering effectiveness and reduced computational demands42. Moreover, its effectiveness has been validated in several studies41. Specifically, we applied spectral clustering to analyze wind fields at three pressure levels (500, 300, and 200 hPa) within the geographical domain of 20°–45°N latitude and 70°–100°E longitude, during REPEs over the ITP. To ensure the robustness of our clustering results, we conducted a sensitivity analysis to evaluate the impact of varying the size of the cluster analyzed wind fields. The results demonstrated that the classification of atmospheric circulation patterns remained consistent and was largely unaffected by changes in the spatial extent of the region defined for atmospheric circulation cluster investigation. (Fig. S7). This finding underscores the reliability of our clustering method and its applicability to the study of extreme precipitation events over the ITP.

Diagnosis of Rossby waves

To diagnose the propagation of Rossby waves, this study employs the Takaya and Nakamura (T-N) wave activity flux, a robust diagnostic framework specifically designed for detecting and analyzing large-scale quasi-stationary Rossby waves in a propagating context43. The wave activity flux is mathematically expressed as follows:

$$W=\frac{P}{2{|U|}}\left(\begin{array}{c}\overline{u}\left({\psi }_{x}^{{\prime} 2}-{\psi }^{{\prime} }{\psi }_{xx}^{{\prime} }\right)+\overline{v}\left({\psi }_{x}^{{\prime} }{\psi }_{y}^{{\prime} }-{\psi }^{{\prime} }{\psi }_{xy}^{{\prime} }\right)\\ \overline{u}\left({\psi }_{x}^{{\prime} }{\psi }_{y}^{{\prime} }-{\psi }^{{\prime} }{\psi }_{xy}^{{\prime} }\right)+\overline{v}\left({\psi }_{y}^{{\prime} 2}-{\psi }^{{\prime} }{\psi }_{yy}^{{\prime} }\right)\end{array}\right)$$
(1)

Let U = (u, v) represent the horizontal wind vector, where u and v denote the zonal and meridional wind components, respectively, and ψ represents the stream function. In Eq. (1), the horizontal and vertical lines denote the climatological mean and perturbation components, respectively, providing a clear distinction between the background state and wave-induced anomalies.