Introduction

The Tibetan Plateau (TP) stands as a vast elevated landmass in the mid-latitude region of the Northern Hemisphere, encompassing more than a quarter of the land area in China. With an average altitude exceeding 4000 m, it is often referred to as the Earth’s third pole1. The unique topography and geomorphology of the TP exert a substantial impact on energy exchange and atmospheric circulations, driven by thermal and mechanical effects2,3,4,5,6,7. Moreover, the TP also plays an essential role in modulating atmospheric moisture transport and the Asian water cycle3,8,9,10,11,12. During the boreal spring and summer, a notable moisture convergence zone develops over the TP13. The monsoon and westerlies are primary sources of the summer moisture over the TP14,15. External moisture, regulated by Indian monsoon and Westerlies, converges over the TP16. The TP is often called the “Asian Water Tower” because of its crucial role in accumulating atmospheric moisture, promoting cloud-precipitation and supporting hydrological processes17.

Although water vapor constitutes a small proportion of the atmospheric composition, it is essential for the global water cycle18,19. Simultaneously, moisture is a prerequisite for the cloud-precipitation processes, and exert substantial influences on the clouds and precipitation the TP and its downstream regions9. The dynamic and thermal effects of the TP profoundly impact on the development of the Asian summer monsoon and contribute to adverse weather conditions in downstream areas of Asian monsoon regions20. Therefore, a comprehensive understanding of moisture sources and transport mechanisms is imperative for regulating the regional climate of the TP and mitigating the risk of droughts and floods in downstream of the TP.

The complex geographical effects of the TP pose significant challenges for climate studies the southern slope of the TP (SSTP) serves as the primary channel for moisture transport to the TP21,22,23. The intricate terrain and steep slopes of the SSTP alter the moisture transport from the tropical ocean to Asian monsoon region and the TP. Currently, numerical models provide the most effective approach for analyzing precipitation processes and future climate changes over the TP24,25,26. The numerical models are widely employed to investigate the moisture transport and precipitation changes over the TP27,28,29,30,31,32. However, model frequently overestimates precipitation due to uncertainties in parameterization of convective processes and orographic effects, especially the SSTP effects on clouds and precipitation33,34,35.

Previous studies have underscored the significance of model resolution influencing the simulation of moisture transport and precipitation over the TP36,37. Increasing the model resolutions can mitigate the simulated wet bias38,39,40,41. The model resolutions can reduce the wet bias in precipitation simulations on the TP by retarding the northward water vapor transport, which is attributed to a reduction in the positive bias of the wind speed29,37,42. Weisman et al.43 found that the lifetime of the convective systems tends to be long and the intensity becomes stronger as the model grid spacing increases. Subsequently, Petch et al.44 examined the sensitivity to horizontal resolution in a viewpoint of precipitation diurnal cycle, and found that the diurnal maximum occurs earlier as the horizontal resolution becomes higher. With higher resolution, model afford more realistic wind field and atmospheric vertical heating simulations, leading to more precise characterizations of the moisture transport processes45. The high-resolution simulations also ameliorate the spatial distribution of summer precipitation and effectively diminish the systematic underestimations of average precipitation in the model46. However, even in high-resolution simulations, non-negligible precipitation anomalies induced by topography effects47,48,49, highlighting the challenge of simulating clouds and precipitation over the TP. Previous studies mostly focused on refining the model resolution at coarse horizontal resolutions within relatively small regions27,28,29,30,37,50.

Terrain complexity refers to the characterization of the Earth’s surface complexity with the amplitude of elevation changes51. In model, the elevations between adjacent grids undergo interpolation and smoothing to maintain computational stability. Disparities in resolution between the modeling horizontal grids and original elevation data can lead to the underestimation of terrain complexity, introducing inaccuracies in terrain analysis. Consequently, for the complex topography on the TP, current elevation data resolution proves insufficient for accurately identifying the terrain features. The complex topography of the TP cannot be accurately described in the model, resulting in discrepancies in the modeling precipitation variations over the TP9,52. Therefore, it is critical to evaluate the ability to resolve the topography of the TP in the model. More refined complex terrain in the model rely on the higher resolutions. As for the TP region, finer terrain allows more small-scale valleys to be identified, which may serve as conduits for moisture transport across the SSTP. Thus, more accurately describing the topography of the TP in the model facilitate improving the usability of simulated moisture transport, precipitation formation, and convection initiation. Moreover, the spatial distributions of precipitation over the TP are simulated more reasonably realistically by applying higher resolutions in the model53. The increased terrain complexity resulting from higher simulation resolution caused an increase in the moisture budget of TP by ~11%, and also significantly optimized the spatial distribution of simulated precipitation on the SSTP42.

In this study, we explored the impact of terrain complexity on precipitation over the TP by using multiple sets of elevation data with varying horizontal resolutions. Employing the Weather Research and Forecasting Model (WRFv4.2) and multiple-source model terrain datasets with different resolutions, we conducted a series of sensitivity experiments with varying terrain complexities. The simulation results were compared to investigate the impacts of terrain complexity on cloud and precipitation changes over the SSTP and TP Platform. This study aims to quantitatively evaluate the response of moisture transport, precipitation patterns, and convection initiation to terrain complexity over the TP. These results are expected to guide the modulation of the TP on East Asian and global climate changes and improve cloud and precipitation simulations over the regions with fine topography complexity.

Results

Input data from different complex terrain

To evaluate the effectiveness of interpolation, smoothing, and other processing in restoring natural topography, Fig. 1 illustrates the spatial distribution differences between the terrain used in experiments and the neighboring grids derived from the SRTM. Clearly, input data exhibits a markedly better topographic smoothing effect at Coarser Horizontal Resolution (CHR) (Fig. 1a–e) than Higher Horizontal Resolution (HHR) (Fig. 1f–j). Complex terrain (e.g., including the ridges and valleys) is more accurately represented by HHR, whereas CHR tends to overlook or underestimate it. Moreover, the limited ability of CHR grid in accurately representing actual terrain variations, interpolation smoothing lead to a notable decrease in slope between neighboring grids, especially in the Himalayan54. Elevation discrepancies between various experiments and SRTM primarily concentrate on the SSTP, followed by TP platforms and the Transverse Mountain Area. At the foot of the SSTP, elevation discrepancies among different terrain datasets are minimal, all under 20 m. However, in complex terrains above 1500 m, primarily valleys, elevation resolution errors are substantial (Supplementary Fig. S4). To maintain model stability, interpolation and smoothing operations between neighboring grids significantly underestimate elevations in complex terrains. Elevation is underestimated over the SSTP particularly with CHR. Additionally, experiment elevations are compared with those at observation stations (Fig. 1 Top right corner), using the Correlation Coefficient (COR) and Root Mean Square Error (RMSE) to assess interpolation effectiveness. Higher resolutions in horizontal and terrain dimensions improve the COR and reduce the RMSE, enhancing the accurate reproduction of terrain features in the WRF.

Fig. 1: Spatial distribution of topographic elevation differences.
figure 1

a–j Elevation differences (units: m) between WRF preprocessing input data and SRTM for Cases 1–10. The COR and RMSE values for the topographic elevations are shown at the top of the figures. The black lines mark the region above 3000 m.

Full size image

Despite its high elevation, the terrain of the TP platform is relatively flat. In contrast, the SSTP features exceptionally complex terrain with numerous valleys. The SSTP, one of the wettest regions on the TP, serves as a crucial channel for moisture transport4,6,13,55,56. However, the model lack the necessary precision to accurately represent the TP’s complex terrain. Topography limits regional-scale processes, and the water vapor transport in the TP could produce wet biases. In addition to errors in microphysics and cumulus parameterization, inappropriate surface precision also enlarge precipitation inaccuracies by generating unrealistic local moisture cycles57. Such differences could significantly affect the meteorological conditions of these two regions. Subsequently, incorrect response in mechanical and thermodynamic mechanisms may alter precipitation patterns on the SSTP and TP platform. Therefore, the terrain complexity index (TCI) is calculated, as shown in Table 1, to quantify the TP’s terrain complexity. When elevation datasets are interpolated to the coarse grids in CHR, the insufficient resolution cannot capture a more significant valleys and ridges. As a result, CHR exhibits a smoother terrain profile and a lower TCI. Coarse grids minimize the impact on adjacent grids, leading to lower sensitivity to terrain datasets and more minor discrepancies of the TCI among the CHR. Nevertheless, as the horizontal resolution increases, the terrain complexity is more sensitive to terrain data resolution, resulting in notable discrepancies in TCI among the HHR. A higher TCI indicates that the Model can resolve more intricate valleys and ridges, potentially identifying specialized terrain features that serve as key channels for water vapor transport. In the WRF, TCI disparities result in varying precision in terrain representation. Consequently, substantial variations in drag resistance in mountainous areas significantly influence the wind field, moisture transport, and precipitation-cloud processes associated with TCI, meriting further investigation.

Table 1 The TCI for each experiment
Full size table

Modeling validation

Changes in TCI can directly retard the near-surface wind at the 10 m level. Figure 2 shows the mean daily cycle of both simulated and observed wind components at all stations and precipitation on the TP platform (above 2500 m altitude). The zonal wind component, predominantly influenced by the westerly belt, contributes more to the absolute wind speed than the meridional wind component31. Since the SSTP is crucial for the SASM to transport moisture from South Asia to the TP platform23. The complex terrain of the SSTP primarily affects the meridional wind component.

Fig. 2: Diurnal variations of near-surface wind and precipitation.
figure 2

a Differences in the diurnal variations of near-surface zonal wind (u10; unit: m s−1) compared to observations at all stations on the TP platform (above 2500m altitude). b Similar to a but for near-surface meridional wind (v10; unit: m s−1). c Diurnal cycles of precipitation (mm hour−1) with observations over the TP Platform. Observed estimations based on the TPHiPr are shown as a black solid line.

Full size image

The difference between the simulated near-surface wind and the observation under the different TCIs are shown in (Fig. 2a, b). Currently, the model tends to overestimate near-surface wind speeds over the TP31. However, adjustments are made to counteract the overestimation of near-surface wind as TCI increases. In the CHR, terrain dataset resolution presented a weak impact on near-surface wind speeds. Conversely, higher terrain resolution produces a more significant correction in HHR. With increasing horizontal resolution, the TP’s complex terrain is more accurately represented, reducing the bias in simulated near-surface winds, especially in the zonal wind anomalies (Fig. 2a). Increased TCI enhances the sensitivity to surface resistance. As terrain resolution increases, the meridional wind anomalies significantly decrease across the entire TP (Fig. 2b). At the same horizontal resolution, an increase in TCI can increase the meridional component driving across the TP platform to a certain extent. Higher terrain resolution leads to a slight increase in moisture transport from SASM42. This may be attributed to significant variations in SSTP topography at identical horizontal resolutions, leading to alterations in meridional wind patterns. The diurnal cycle of surface winds exhibits marked variations in turbulent diffusion intensity, with the maximum occurring later as horizontal resolution increases. Additionally, differences in TCI significantly influence nocturnal near-surface winds.

As TCI increases, precipitation visibly declines and the wet bias decreases over the TP platform (Fig. 2c). Precipitation increases at 9 LT (local time) and this pattern reverses at dusk. The rainfall amount is largest in CHR and smallest in HHR, suggesting that the nighttime precipitation is also sensitive to the TCI. The daily precipitation peak occurs around 14–16 LT. The TCI significantly affects both the timing and intensity of precipitation. Typically, precipitation in CHR starts later but is stronger and more abrupt when it begins. Petch et al.44 and Sato et al.58 also discovered similar diurnal rainfall variations over the TP platform. Daytime precipitation peaks are reduced, and nocturnal precipitation overestimations approach TPHiPr. Zhou et al.59 and Jiang et al.60 also suggested that the TPHiPr has a reduced wet bias and more realistically simulates found precipitation over the TP platform. Similar to the zonal component behavior (Fig. 2b), precipitation disparities primarily occur during daytime. Therefore, the evolution of TCI significantly influences the precipitation characteristics over the Tibetan Plateau, through interactions with complex terrain, modulated by the large-scale circulation patterns associated with the ISM.

Enhancements in TCI decrease the intensity of the zonal component and increase that of the meridional component in the same horizontal resolution, largely due to the interpolation of varied terrain datasets on the same grids. The SSTP’s complex terrain affects the transmission of meridional winds, which causes a delay in the peak of near-surface wind and an advance in the peak of precipitation. The changes in the zonal and meridional components caused by TCI result from the changes that reflect the influence of friction or terrain on the near-surface wind field. The terrain complexity amplifies these effects, creating spatial variability in precipitation distribution, cloud cover, and moisture transport dynamics61,62. In summary, HHR enhances terrain resistance and related processes over complex terrain, resulting in more distinct corrections in near-surface winds and precipitation than those caused by varying weights of terrain datasets in the same model grids.

Terrain complexity impacts precipitation patterns

Previous studies have investigated the influence of the SSTP’s complex terrain on precipitation and moisture transport, focusing on the critical “barrier” effect at different horizontal resolutions29,37. However, interpolating terrain datasets of different resolutions to the same horizontal resolution can also generate anomalous valleys or ridges as channels for water vapor transport. Thus, quantifying the impact of TCI is essential to understand how topographic differences influence precipitation pattern balance.

Figure 3 shows the spatial distributions of the average daily precipitation from various experiments. It is noteworthy that the main precipitation differences are concentrated in the SSTP, particularly along the slope and its foot (Fig. 3a, b). However, topography’s impact on precipitation simulation within the TP platform is minimal (Fig. 3c–f). As TCI increases, terrain-driven precipitation intensifies at the SSTP and its base, showing a “decrease-increase-decrease” pattern along the meridional direction at the southern boundary (1500 m elevation, 84°-93°E). HHR enables identification of small-scale terrain features not captured by CHR. Consequently, in HHR, precipitation becomes more sensitive to and more strongly dependent on terrain (Fig. 3g–i). The complex terrain of the SSTP results in more terrain-driven precipitation. Precipitation decreases at the foot of the SSTP and increases on the slopes. Furthermore, as TCI increases, precipitation differences between experiments are amplified. The impact of topographical differences on precipitation indirectly controls through their influence on other mechanisms. As a result, the steep and complex topography of the TP generates a distinctive precipitation pattern.

Fig. 3: Spatial patterns of daily precipitation and their variations.
figure 3

a Average daily precipitation (unit: mm day−1) for Case 1. b Similar to a but for Case 10. c–l Differences in the spatial distribution of average daily precipitation between simulation experiments: c Case2-Case1, d Case5-Case3, e Case8-Case6, f Case10-Case9, g Case3-Case1, h Case6-Case3, i Case8-Case1, j Case10-Case8, k Case10-Case5, l Case9-Case2. The top-right color map applies to panels a–b, the middle-right color map applies to panels c–f, and the bottom-right color map applies to panels g–l. Dashed lines mark the 3000 m elevation for the TP (unit: m).

Full size image

To explore the precipitation pattern over the SSTP, the latitudinal distribution of the average and maximum precipitation is depicted in Fig. 4. A large amount of precipitation over South India along the SSTP, decreasing towards the TP platform (Fig. 4a). With increased TCI, HHR simulations show more precipitation over North India and the SSTP (peaks at 25°N and 27°N), and less along the Himalayan ridge and the TP platform (troughs at 26°N and 28.5°N). Precipitation predominantly concentrates in the frontal region of the SSTP, exhibiting an upward trend. After crossing the SSTP, precipitation exhibits a decreasing trend. In contrast, precipitation sharply diminishes uphill, minimally impacting the TP platform. As TCI increases, the average precipitation shows an overall upward trend, likely due to the enhanced orographic drag associated with complex terrain, which directly impacts atmospheric moisture transport63,64,65. The complex terrain of the Himalayas strongly controls water vapor transport. The zonal near-surface wind distribution also shows a distinctive pattern over the SSTP (Fig. 4). However, the precipitation significantly exceeded OBS in HHR and CHR. Previous studies have found that current numerical models tend to overestimate precipitation on the SSTP9,35,66. Computational errors in the model’s pressure gradient force, potentially due to complex terrain, are identified as a factor67,68. These errors arise from coupling abnormal topographic details due to incorrect representation of terrain-following coordinates and the advection-condensation processes that could lead to overestimation in topographic precipitation. However, with gentler terrain changes, precipitation variations on the TP platform are relatively minor.

Fig. 4: Meridional average and maximum cumulative precipitation.
figure 4

a Meridional average of cumulative precipitation (unit: mm day−1) in the region 84°-93°E. b Similar to a but for meridional maximum cumulative precipitation. The insert figures in upper right corner show the average and maximum precipitation over the TP Platform area. Shaded areas show the corresponding terrain (unit: m).

Full size image

An examination of the meridional maximum precipitation (Fig. 4b), reveals a trend similar to that of average precipitation (Fig. 4a). Compared to CHR, precipitation intensity in HHR demonstrates greater sensitivity to terrain. The differences in the average precipitation are mainly in the response of precipitation intensity to TCI. After categorizing the regions, further analysis is conducted on the precipitation characteristics at the foothills of northern India, the SSTP, and the TP platform. Figure 5 illustrates variations in the probability density distribution of daily accumulated precipitation across experiments, compared to the reference (Case1, Fig. 5d). We compared the impact of TCI on precipitation magnitude across all grids in these regions. With increasing TCI, the proportion of grids experiencing heavy precipitation at the foot increases, while those with weak precipitation decrease. This leads to more pronounced average precipitation at the foot (Fig. 5a). The precipitation probability density distribution on the TP platform resembles that of the near-surface meridional wind component (Fig. 2c). The proportion of heavy precipitation on TP platforms decreases and more weak precipitation is concentrated on the platform (Fig. 5b). Increased grids of weak precipitation can regulate atmospheric energy and inhibit deeper convection66. Furthermore, deep convection is identified as the primary driver of diurnal variations in warm-season precipitation69. Consequently, this results in a weakened peak in daily average precipitation (Fig. 2c) and a reduction in wet bias on the TP platform. The frequency of slopes receiving daily precipitation exceeding 50 mm is higher over the SSTP (Fig. 5c). Higher frequency of deep convection systems over the SSTP compared to the TP platform, resulting in greater heavy precipitation concentration due to terrain-driven factors70. TCI indirectly modifies moisture transport and the energy cycle by influencing complex terrain, significantly increasing precipitation intensity over the SSTP.

Fig. 5: Differences in the probability density distribution of daily precipitation.
figure 5

a The variation in the probability density distribution of daily precipitation (%) is compared between Case1 and other experiments for all grid points over the SSTP foot. b Similar to a but over the TP platform. c Similar to a but over the SSTP. The dashed line represents a value of 0. d The probability density distribution of daily precipitation over the Foot, TP Platform and SSTP in Case1.

Full size image

Therefore, overlooking the impact of mesoscale and small-scale terrain contributes significantly to discrepancies in precipitation simulations in complex topographic areas. TCI plays a crucial role in precipitation simulations, influencing moisture transport and cloud microphysics responses to complex terrain. We believe that during heavy precipitation events, processes like orographic lifting are critical for transporting moisture-rich air masses to the TP platform and altering cloud precipitation mechanisms.

Changes in the moisture transport and microphysical processes

In all experiments, the SSTP exhibits the most pronounced terrain contrast (Fig. 2). Moreover, the SSTP is critical to the moisture transport between the TP and the Southern Ocean. The complex terrain of the SSTP significantly increases the moisture supply to the SASM71. Figure 6 illustrates the estimated mean horizontal distribution of moisture for intense precipitation events along the southeastern areas of the SSTP. At the same resolution, increased terrain complexity enhances the blocking effect on IVT (Fig. 6a–d). Additionally, as grid intervals decrease, the complex terrain increasingly blocks IVT (Fig. 6e–h). An anticyclonic anomaly consistently occurs in the southeastern areas in northern India. The magnitude of divergence in each experiment closely relates to the TCI. With increasing TCI, intensified terrain blocking gradually enhances the strength of the anticyclonic anomaly, deflecting IVT at the southeastern areas. However, anomalous divergence results from the interplay between upper and lower atmospheric systems. Low-level cyclonic conditions enhance surface moisture convergence in this region. Typically occurring around the Himalayas’ southeastern areas, this cyclonic activity facilitates the lifting of moist air masses, critical for vertical moisture transport72. Upper-level anticyclonic conditions augment this lifting by stabilizing the upper atmosphere, thereby facilitating sustained upward moisture movement73,74. This combination of upper-level anticyclonic and lower-level cyclonic activity creates a conducive environment for moisture ascent, influencing precipitation patterns. Therefore, despite the Himalayas separating Northern India and the southeastern areas, the combined effects of low-level cyclonic and upper-level anticyclonic conditions promote vertical moisture transport over the southeastern areas.

Fig. 6: Spatial distribution of IVT and circulation differences.
figure 6

a The spatial distribution of IVT (shaded and stream flow, unit: kg m−1 s−1) for Case2-Case1. b Similar to a but for Case5-Case3. c Similar to a but for Case8-Case6. d Similar to a but for Case10-Case9. e Similar to a but for Case6-Case3. f Similar to a but for Case10-Case5. g Similar to a but for Case3-Case1. h Similar to a but for Case6-Case1. The color map on the right applies to panels a–d, and the one at the bottom applies to panels e–h. Dashed lines denote the 3000 m elevation contour (unit: m).

Full size image

Dong et al.75 also found that this vertical structure is assisted primarily by “up-and-over” moisture transport, in which moisture is lifted by convective systems over northern India and the Himalayan foothills, dominantly driven by upslope flow over the Himalayas. Therefore, increased TCI facilitates moisture transport from lower to higher levels toward the TP, enhancing anomalous moisture transport at the same horizontal resolution (Fig. 7). Simultaneously, the more complex terrain reduces the IVT supply at the southern and southeastern areas of the SSTP (not shown). Additionally, stronger vertical velocities, induced by complex terrain in all areas, enhance vertical moisture transport to varying degrees over the SSTP (Figs. S1, S2). Jiang et al.76 indicated that more complex terrain significantly impacts wind fields, thereby affecting the vertical distribution of moisture. Complex terrain impedes low-level moisture transport by diminishing horizontal wind speeds. Specifically, the SSTP’s intricate terrain fosters strong convection by enhancing ascending motions. This dynamic is instrumental in the rapid development of mesoscale vortices and the intensification of precipitation events throughout the region. The SSTP’s steep and varied topography amplifies atmospheric vertical motions, triggering convective processes as moist air is lifted and cooled. Although these processes are often localized, they exert widespread effects through the development of mesoscale systems that transport moisture and energy over larger areas. Furthermore, the topography of the SSTP can induce the lifting of air masses that enhance condensation and precipitation, effectively redistributing water vapor across a broader area. Therefore, the complex terrain of the SSTP is not merely a passive barrier but an active and dynamic system that plays a central role, influencing both local precipitation patterns and convection initiation.

Fig. 7: Spatial variations in vertical moisture flux density.
figure 7

a The vertical distribution (unit: hPa) of vertical moisture flux density (unit: kg m−2 s−1) in the southwest areas for Case2-Case1. b Similar to a but for Case5-Case3. c Similar to a but for Case8-Case6. d Similar to a but for Case10-Case9. e Similar to a but for Case6-Case3. f Similar to a but for Case10-Case5. g Similar to a but for Case3-Case1. h Similar to a but for Case6-Case1. The average areas same as the left black box in Fig. 3. The color map on the right applies to panels a–d, and the one at the bottom applies to panels e–h. Shaded areas show the corresponding terrain (unit: m).

Full size image

Besides studying moisture transport over the SSTP, investigating it across the TP platform is crucial. The moisture column is divided into lower (TP surface-400 hPa) and upper (400–100 hPa) levels to investigate the vertical distribution of moisture transport (Fig. 8). In the CHR (Fig. 8a–e) and HHR (Fig. 8f–j), an increase in moisture entering the lower level of the TP platform from the SSTP is found, enhancing moisture convergence at this level as TCI increases (Red label). Concurrently, moisture divergence at the upper levels intensifies (Green label). Convergence at lower levels and divergence at upper levels create conditions conducive to convection system formation, enhancing upward moisture transport. However, the impact of terrain differences on dynamic mechanisms primarily affects the lower levels directly, while their effect on the upper levels remains relatively minor. Consequently, the limited divergence, compared to significant convergence, results in an increased moisture budget on the TP platform. Upper-level divergence and lower-level convergence favor the formation of medium and high-level stratiform clouds and deep convection systems, which influence precipitation over the TP platform77.

Fig. 8: Moisture budget at TP platform boundaries.
figure 8

a Moisture budget at each boundary (30°-38°N,85°-102°E) on the TP platform (unit: 1011 g s–1) for Case1, b–j Similar to a but for Case2-Case10.The western boundary (indicated by a large black arrow on the left) and the southern boundary (large blue arrow in the center) serve as input boundaries. The northern boundary (large black arrow at the top) and the eastern boundary (large red arrows on the right) serve as output boundaries. The total moisture budget is presented in the top-left corner of the figure (Purple label). The upper rectangle details the moisture budget between 100-400 hPa, while the lower rectangle shows moisture at the 400 hPa-surface. Red labels denote the moisture budget at lower levels, and green labels represent the upper-level moisture budget. The 3D spatial distribution illustrates the Cases topography of the TP (unit: m).

Full size image

At the same terrain resolution, moisture across all boundaries decreases as TCI increases (Fig. 8a-c-f, Fig. 8b-d-g-i, Fig. 8e-h-j). Compared to CHR, HHR creates a drier environment, reducing moisture content at lower levels along the SSTP (Supplementary Fig. S3). More moisture remains in the SSTP, taking the form of terrain-driven precipitation. Compared to changes in terrain resolution, the impact of horizontal resolution on the moisture budget is more pronounced due to the terrain smoothing effect. Precipitation concentrates on the SSTP, where complex terrain acts as a special moisture channel, indirectly promoting moisture transport to the TP platform. Reducing horizontal resolution decreases terrain resistance, allowing more moisture to be transported to the TP rather than being trapped in the SSTP. Furthermore, complex terrain creates a stationary moisture front between moist and dry air masses, leading to significant diurnal variability in wind speeds and moisture transport. This diurnal pattern, especially pronounced during the monsoon season, is a critical factor in the TP’s unique hydrological and climatological phenomena78.

Representation of topography may significantly influence the performance of convective systems, as mountains and valleys can profoundly affect the developments and evolutions of rainfall and clouds79,80. Diurnal variations in cloud characteristics between the SSTP and TP platform are compared using sensitivity experiments (Fig. 9). Generally, cloud cover on both the SSTP and TP platform exhibits consistent diurnal variations. Low cloud cover initially increases then decreases (Fig. 9a–c), whereas medium and high cloud cover shows the opposite trend (Fig. 9b–d). Many small cloud cells form in the TP between 9 and 11 LT. Daytime developments of convective systems are marked by a pronounced increase in cloud cover. Favorable environmental conditions for convective system development occur as surface temperatures rise during the daytime81. Cloud systems developing in the afternoon generate more medium and high clouds while reducing low cloud cover between 12 and 18 LT, consistent with the results of Sato et al.82. With increasing TCI, the peak of diurnal cloud cover variations and convection initiation occur earlier. Thus, topographic effects are key in initiating clouds and causing phase differences in the diurnal cycle. Yang83 also discovered that the proliferation of convective activities is influenced by varying topographical conditions within plateau valleys. With increasing TCI, medium and high clouds become more prevalent (Fig. 9b), while low clouds decrease (Fig. 9a) in the SSTP. This fact implies that enhanced vertical moisture transport on the SSTP (Fig. 7) promotes deep convection development, which can be benefit to the development from the low clouds to medium and high clouds in the afternoon. Near mountains, deep convection development involves interacting processes between mountain–valley circulations and rain evaporative cooling83. Moreover, the substantial latent heat released by condensation in the precipitation center (the peak of diurnal precipitation variation) enhances the formation and intensification of medium and high clouds (Fig. 9b, d), resulting in the second peak of these clouds at 15 LT (Fig. S3e, f). However, the reduction in moisture entering the TP platform is an important reason for the reduction of cloud cover over the TP Platform. (Fig. 8). On the TP platform, complex terrain influences the vertical distribution of moisture and thermal conditions, resulting in a drier and colder near-surface environment. This environment complicates cloud initialization and reduces low cloud formation (Fig. 9c). Simultaneously, the compressed troposphere causes the medium and high cloud growth to be restricted at lower levels (Fig. 9d). It is noteworthy that a smaller TCI delays cloud initiation and induces more downward shortwave radiation. It also increases sensible heat on the TP platform, intensifies the water cycle in low TCI experiments, and increases and lags precipitation. Therefore, insufficient terrain resolution may be a key factor for the wet bias over the TP platform in current model.

Fig. 9: Diurnal cycle of cloud fraction over SSTP and TP platform.
figure 9

a Diurnal cycle low cloud cover cloud fraction (%) over the SSTP. b Similar to a but for medium and high cloud. c Diurnal cycle low cloud cover cloud fraction (%) over the TP platform. d Similar to c but for medium and high cloud.

Full size image

The variations in cloud hydrometeor content in the SSTP and TP platform are displayed in Fig. 10 to investigate the TCI effects on the cloud and precipitation processes. Diurnal variations in hydrometeor content mirror cloud cover, with peaks at night and troughs between 9 and 12 LT. Additionally, noticeable differences exist in the diurnal variations among experiments. Similar to precipitation, increased TCI results in earlier and more consistent cloud cover build-up. Correspondingly, the consumption of hydrometeors and the increase in precipitation occur in the morning. This fact implies that precipitation persistence is significantly influenced by the consumption of locally accumulated moisture and hydrometeors before dawn84.

Fig. 10: Diurnal cycle composite of hydrometeor content.
figure 10

a Diurnal cycle composite of total column hydrometeor content (Qrain, unit: 10−5 kg kg−1) on the SSTP. c Similar to a but for Qcloud. e Similar to a but for Qice, g Similar to a but for Qgraupel+Qsnow. b Diurnal cycle composite of total column hydrometeor content (Qrain, unit: 10−5 kg kg−1) on the TP platform. d Similar to b but for Qcloud. f Similar to b but for Qice, h Similar to b but for Qgraupel+Qsnow. a1–h1 Vertical distribution (unit: hPa) of hydrometeoric content during the daytime on the SSTP and TP platform. a2–h2 Similar to a1–h1 but during the nighttime.

Full size image

Along with the TCI increase, rain particles in the SSTP similarly increase to precipitation (Fig. 10a), while the TP platform shows the opposite trend (Fig. 10b). Rain drop is concentrated in the SSTP at 850–500 hPa, while they are concentrated on the surface of TP platform. This is mainly because rain droplets are the only precipitating hydrometeors in the lowest atmosphere over the SSTP. The maximum rain mixing ratio occurs just above the surface, indicating that most hydrometeors remain suspended in the air and do not fall out to form precipitation (Fig. 10a1–b2). Both regions exhibit a decreasing trend in cloud droplets and ice particles (Fig. 10c–f). Notably, some cloud water is distributed at higher levels (Higher 0 °C isotherm in SSTP), resulting in abundant supercooled liquid water in the cold regions. This is a distinctive feature of the TP, as the melting layer remains low even in summer (Figs. 10c1–d2). Simultaneously, the abundance of cloud water increases the likelihood of reaching the rain droplet conversion threshold in triggering models76. In addition, the amount of ice crystals is an order of magnitude smaller than that of cloud droplets (Fig. 10g, h). However, solid precipitation particles (e.g., graupel, snow) are increasing in both areas (Fig. 10e, f). With an increase in TCI, both the vertical speeds of SSTP and TP Platform exhibit an upward trend (Figs. 7, 8). Increased vertical velocities facilitate the upward transport of lower-layer water vapor and liquid cloud water, creating favorable conditions for ice crystal growth. As ice crystals enlarge, they form snow and graupel particles, enhancing convective cloud development and prompting the medium and high cloud cover over the SSTP. Consequently, with increased TCI, more ice particles develop into snow and graupel (Fig. 10e1–h2). The primary microphysical process of convective precipitation on the TP involves ice crystals developing into heavier snow particles, which fall and accumulate supercooled water to transform into graupel, ultimately melting above the ground to form rain85. Vertically, the transformation of graupel into rain is evidenced by the decrease in graupel particles and the simultaneous increase in rainwater. This suggests that numerous graupel particles descend into warmer regions and melt before reaching the surface. Therefore, as TCI increases, the transformation of graupel particles into raindrops in SSTP results in a corresponding increase in raindrops similar to the increase in graupel particles. Simultaneously, the relatively small variation in raindrop levels at different heights indicates that ice-phase particles continuously replenish the consumption of raindrop particles. Furthermore, during intense precipitation events, ice microphysical processes generate precipitation more efficiently than warm-rain processes over the TP70. This explains why precipitation in SSTP increases with the rise of TCI. However, on the TP platform, although the number of snow and graupel particles increased, the number of raindrops decreased. In the near-surface atmosphere, raindrops are the only precipitation hydrometeor (with the graupel mixing ratio dropping to zero). The maximum rainwater mixing ratio occurs in the air, indicating that most hydrometeors remain suspended and do not fall to form precipitation, leading to an increased proportion of liquid phase precipitation. This explains why precipitation on the TP platform decreases and weakens with the increase of TCI.

In summary, TCI primarily changes moisture transport across the TP by facilitating various mechanisms, such as moisture transport, atmospheric convergence, vertical transport, and cloud microphysical processes, thereby affecting cloud and precipitation processes. Enhanced topographic descriptions could modulate cloud precipitation processes and reduce moisture transport via mechanisms like terrain resistance, thus establishing deeper and earlier convection initiation over the TP. Complex topography directly influences circulation and moisture at various altitudes transported to the TP. It also modulates the vertical distribution of water vapor transport through microphysical mechanisms, leading to distinct precipitation patterns and a new equilibrium in response.

Discussions

In this study, simulation experiments on complex terrain utilized varying horizontal and topographic resolutions to explore how the terrain complexity affects cloud and precipitation processes on the TP. These studies also assessed the impact of unrealistic climatic responses on precipitation patterns and convection initiation. As the horizontal resolution increases, the terrain complexity is more sensitive to terrain data resolution, resulting in notable discrepancies in terrain complexity at the HHR. This setting enhances the representation of complex terrain on the TP. The model preprocessing often significantly underestimates elevation in these areas to ensure stability. The WRF shows a markedly greater topographic smoothing effect at CHR compared to HHR, with elevation differences mainly concentrated on the SSTP. Terrain complexity significantly affects the timing and intensity of precipitation peaks by delaying near-surface wind speeds. In the HHR, precipitation becomes more sensitive and increasingly dependent on complex terrain, resulting in more pronounced precipitation on SSTP. The proportion of heavy precipitation at the SSTP increases, while those with weak precipitation decrease. Heavy precipitation on TP platforms decreases, while weak precipitation increases. Precipitation noticeably increases over the SSTP, reducing the wet bias over the TP platform.

The complex terrain of the SSTP acts not merely as a passive barrier but as an active and dynamic system, playing a central role in influencing local precipitation patterns and convection initiation. As terrain complexity increases, moisture transport to the SSTP encounters increased resistance from the complex terrain, impeding lower-level airflow and moisture. Stronger vertical velocities, induced by the complex terrain, enhance vertical moisture transport to varying degrees. SSTP topography can induce air mass lifting, enhancing condensation and precipitation, thereby redistributing moisture more broadly. As terrain complexity intensifies, there is a notable peak of diurnal cloud cover variations occurring earlier, thereby leading to earlier convection initiation. There is an increase in the prevalence of middle and high clouds, accompanied by a corresponding decrease in low clouds over the SSTP. The proportion of solid precipitation increases. Compared to lower-level convergence, the limited upper-level divergence contributes to an increased moisture budget on the TP platform. The further development of low clouds is curtailed, and the growth of middle and high clouds is constrained at lower levels on the TP platform. On the TP platform, the colder conditions and weaker precipitation lead to the retention of more ice-phase particles in clouds, with precipitation occurring predominantly in liquid form.

Although finer terrain complexity can better reproduce natural topography, a wet bias still occurs in the SSTP. This issue likely arises from significant mismatches between the actual elevation and the interpolated elevation in the model’s neighboring grid cells. These elevation discrepancies can lead to a dynamic barrier effect due to sub-grid topography variations that the model fails to accurately capture, thereby affecting airflow and precipitation distribution. However, we acknowledge that this overprediction of precipitation might also be influenced by numerical factors, and a definitive explanation remains uncertain. In our study, we ranged from large-scale simulations to convection-permitting simulations. To ensure consistency and comparability across various research findings, we ultimately selected the same cumulus parameterization scheme, facilitating a more direct assessment of terrain-induced changes. However, considerable uncertainty remains regarding the impact on precipitation within the grid spacing of the gray area. The scale-aware cumulus parameterization scheme may have an advantage in simulating extreme precipitation events over the TP in the gray zone86,87. We will evaluate the performance of these newer schemes in future work.

Furthermore, advancements in computational power now allow for the expansion of simulation domains and the extension of integration times, enabling long-term, large-scale climate simulations at convection-resolving resolutions. This will enhance our understanding of the dependence of TP cloud and precipitation processes on high-resolution topography at convection-permitting scales.

Methods

The National Surface Station Hourly Mean Observation Dataset provided by the China Meteorological Administration Information Center (http://data.cma.cn) was employed in this study. The representativeness of the station observations can be limited due to the observation continuity and the spatial distribution of the stations. However, ground-based station observation data are currently the most reliable source of observation data at present. Validating the model results with ground observation data will be representative. The TP has few ground observation stations, due to its challenging topography and harsh environmental conditions. Compared with the western TP, stations in the eastern TP are densely and relatively evenly distributed. In order to match the research period of the model, a total of 371 stations were filtered out via data quality control (e.g., excluding stations with too many missing data) as model validation points, which are conducted from July 1 to July 30, 2020 with the hourly data. The considerable benefits gained from resolving detailed cloud microphysical processes with increased grid spacing are significant. The precipitation dataset for the Third Polar region (TPHiPr) datasets (~0.033°), which has higher accuracy than the current mainstream precipitation products in the Third Pole region, is used to assess the rainfall diurnal variations of the model59,60. Therefore, validating high-resolution simulations with higher-resolution rainfall dataset is representative.

The Weather Research and Forecasting Model (WRFv4.2) was used in this study88. The model employs a horizontal Arakawa-C grid and third-order Runge-Kutta time integration scheme with a terrain-following coordinate. The WRF provides an elevation field dataset by default for static initialization. Users can use the WRF Preprocessing System (WPS) to generate appropriate static fields for integration into WRF. Currently, the low-resolution elevation data was provided by WRF, which was replaced by high-resolution data from USGS GMTED 2010 (30 s) after WRF v3.889. Nevertheless, with the development of higher-resolution simulation, the complex terrain becomes more precise between adjacent grids. The low-resolution datasets still pose challenges in accurately representing the ridges and valleys of the TP.

The SRTM (Shuttle Radar Topography Mission) dataset originated from the space shuttle radar topography mapping mission90. It was measured through a joint effort between the USA, Germany, and Italy (http://srtm.csi.cgiar.org). The data covers a measured range of 60°S to 60°N and provides coverage of over 80% of the earth’s surface, encompassing a total area of over 119 million square kilometers. Generate WPS-compliant binary file datasets (3s) based on SRTM files. The HGT_M section in the WPS directory and select the same interpolation method (Weighted average of 16 surrounding grid point values). Figure 11 shows the comparison of precision between SRTM_3s and GMTED_30s elevation on the SSTP. High resolution of complex terrain on the SSTP is necessary to reproduce the climatic effects of complex valleys and ridges in the WRF.

Fig. 11: Preprocessing Terrain on SSTP.
figure 11

a Different terrain preprocessing dataset (unit: m) used for the SSTP, based on SRTM_3 arc second. b Similar to a but for 10 arc minute.

Full size image

Meteorological initial and boundary conditions for the experiment originated from the European Centre for Medium-Range Weather Forecast (ECMWF) reanalysis dataset, featuring a spatial resolution of 0.25° × 0.25° and temporal intervals of 6 h91. The center grid cell is located at (89.5°E, 32°N). The model consists of 51 vertical layers and the highest level of 50 hPa at the top layer. Simulations are conducted from May 26 to July 1, 2020 with the hourly output. The first 5 days serve as the spin-up time for the simulation, and the simulation results of the subsequent 30 days in June 2020 are utilized for the cloud and precipitation analyses in this study. The WRF domain, with the primary focus region highlighted by a white box (d02), is illustrated in Fig. 12. The configuration details of the WRF experiments are summarized in Table 2. Sensitivity experiments, employing topography from three distinct terrain dataset resolutions, were conducted at horizontal resolutions of 0.625°, 0.25°, 0.1°, and 0.04°. The configurations of these experiments are detailed in Table 3. These experiments aim to quantify the impact of various complex topographies on cloud and precipitation processes within the WRF grids.

Fig. 12: WRF Domain.
figure 12

Topography heights (Above Sea Level; unit: m) of the WRF modeling domain, with meteorological observation stations marked as red points.

Full size image
Table 2 WRF modeling setup
Full size table
Table 3 Sensitivity experiment design
Full size table

To examine the spatial variations in both dynamic and thermodynamic aspects resulting from disparities in surface morphology, the information entropy of terrain curvature is employed to reflect the complexity of the terrain through TCI. The terrain curvature is applied to describe the abundance and variations of the slope surface in a region, indicating how convex or concave the surface is in different directions. Simultaneously, entropy can capture these nonlinear characteristics, revealing the inherent complexity of the system. Therefore, the TCI is defined for capturing non-linear, multi-scale, and diverse variations, which makes it more effective quantification of terrain complexity.

$${S}_{{ix}}=\frac{{\partial }^{2}h}{{\partial d}_{x}^{2}}=\frac{\partial }{{\partial d}_{x}}\left(\frac{\partial h}{{\partial d}_{x}}\right)$$
(1)
$${S}_{{iy}}=\frac{{\partial }^{2}h}{{\partial d}_{x}^{2}}=\frac{\partial }{{\partial d}_{y}}\left(\frac{\partial h}{{\partial d}_{y}}\right)$$
(2)
$$\Delta {d}_{x}=\left(\Delta {lat}\right)* a$$
(3)
$$\Delta {d}_{y}=\left(\Delta l{on}\right)* a* \cos {\varphi }_{i}$$
(4)
$${P}_{i}=\frac{\left|{S}_{i}\right|}{\mathop{\sum }\limits_{i=1}^{n}\left|{S}_{i}\right|}$$
(5)
$${TCI}=-\mathop{\sum }\limits_{i=1}^{n}{P}_{i}\log {P}_{i}$$
(6)

Where \({S}_{{ix}}\) and \({S}_{{iy}}\) are the longitudinal and latitudinal terrain curvature, and \(\Delta {\rm{d}}\) (m) is the horizontal distance between adjacent grids. a is the average radius of the earth, a = 6.37 × 106 m, and the \({\varphi }_{i}\) indicates the latitude of the grid point. \({P}_{i}\) is the probability that the random variable takes the value i; n is the number of possible values of the random variable in latitude and longitude directions; TCI represents the entropy of the random variable.

The hourly vertical integrated water vapor transport (IVT) from 1000 to 100 hPa is used to characterize the moisture transport92,93.

$${{\rm{IVT}}}_{u}=-\frac{1}{g}{\int }_{\!\!1000}^{100}{qu}{\rm{d}}p$$
(7)
$${{\rm{IVT}}}_{v}=-\frac{1}{g}{\int }_{\!\!1000}^{100}{qv}{\rm{d}}p$$
(8)
$${\rm{IVT}}=\sqrt{{{\rm{IVT}}}_{u}^{2}+{{\rm{IVT}}}_{v}^{2}}$$
(9)

Where \({{\rm{IVT}}}_{u}\) and \({{\rm{IVT}}}_{v}\) (\({kg}\cdot {m}^{-1}\cdot {s}^{-1}\)) are the zonal and meridional components of the \({\rm{IVT}}\), respectively; u and v (\(m\cdot {s}^{-1}\)) are the zonal and meridional wind speeds, respectively; g (\(m\cdot {s}^{-2}\)) is the acceleration of gravity; q (\({kg}\cdot {{kg}}^{-1}\)) is the specific humidity; and \({\rm{d}}p\) is the pressure difference between two adjacent pressure levels. Eastward (northward) \({{\rm{IVT}}}_{u}\) (\({{\rm{IVT}}}_{v}\)) is defined as positive.

To improve our quantifying of the moisture budget on the TP, various rectangular areas were chosen as the water vapor transfer balance regions, and the IVT at the four boundaries was calculated.

Moisture budget at the western boundary:

$${{Flux}}_{W}={\int }_{{\!\!\varphi }_{S}}^{{\varphi }_{N}}{Q}_{{\lambda }_{W}}{\rm{ad}}\varphi$$
(10)

Moisture budget at the eastern boundary:

$${{Flux}}_{E}={\int }_{{\!\!\varphi }_{S}}^{{\varphi }_{N}}{Q}_{{\lambda }_{E}}{\rm{ad}}\varphi$$
(11)

Moisture budget at the southern boundary:

$${{Flux}}_{S}={\int }_{{\!\!\lambda }_{W}}^{{\lambda }_{E}}{Q}_{{\varphi }_{S}}{\rm{acos}}\,{\varphi }_{S}d\lambda$$
(12)

Moisture budget at the northern boundary:

$${{Flux}}_{N}={\int }_{{\lambda }_{W}}^{{\lambda }_{E}}{Q}_{{\!\!\varphi }_{N}}{\rm{acos}}\,{\varphi }_{N}d\lambda$$
(13)

Where \({\lambda }_{W}\) is the longitude of the western boundary, \({\lambda }_{E}\) the longitude of the eastern boundary, \({\varphi }_{S}\) the latitude of the southern boundary, and \({\varphi }_{N}\) the latitude of the northern boundary. \({Q}_{{\lambda }_{W}}\) and \({Q}_{{\lambda }_{E}}\) represent the IVT passing through the western and eastern boundaries, while \({Q}_{{\varphi }_{S}}\) and \({Q}_{{\varphi }_{N}}\) represent the IVT passing through the southern and northern boundaries. During the summer, moisture transportation across the TP involves the western and southern boundaries as the input boundaries for water vapor, while the northern and eastern boundaries are treated as output boundaries.

Net budget of the moisture \({{Flux}}_{{All}}\) is calculated as follow:

$${{Flux}}_{{All}}={{Flux}}_{S}-{{Flux}}_{N}+{{Flux}}_{W}-{{Flux}}_{E}$$
(14)