Borneo Vortices (BVs) are cyclonic weather systems that frequently occur in boreal winter and are often associated with extreme precipitation and flood events over Southeast Asia. They contribute significantly to extreme winter precipitation over northern Borneo (up to 60–70% of the total extreme precipitation above the 95th percentile daily precipitation rate) and the east coast of Peninsular Malaysia (15–25%)1. Furthermore, the equatorward propagation of BVs can lead to torrential rainfall and flooding over North Sumatra and inundate the most populated cities of the region, which include Jakarta2 and Padang3. Extreme BVs can also develop into tropical cyclones (e.g. Tropical Storm Greg in 1996 and Typhoon Vamei in 2001) and have further devastating effects on life and property4,5,6. This has been highlighted by the recent Tropical Depression 29W which developed from a BV that formed to the northwest of Borneo on 13th December 2021. In a week, this storm made landfall in the east coast of Peninsular Malaysia and triggered extreme flooding, leaving 67,629 people displaced and severe traffic congestion in the major cities of the region7.
Although the socio-economic impacts of BVs can be substantial, there remains relatively little research seeking to understand their response to global warming. There are two reasons for this knowledge gap. First, the relationship between BVs and precipitation extremes stems from the vigorous rainfall-favoring environments near the vortex center, including the intense low-tropospheric convergence of cyclonic winds and the development of cumulus convection at a fine spatial scale of 20–200 km8. Simulating these fine-scale features using climate models requires a sufficiently fine horizontal resolution of at least ~ 65-km, as the frequency and wind structures of BVs are poorly resolved when coarser model resolutions are used1. Thus, current general circulation models (GCMs), such as many of the GCM members (at resolutions corresponding to ~ 100–300 km)9 of the Coupled Model Intercomparison Project Phase 6 (CMIP6) and used by the latest IPCC assessment reports, cannot adequately resolve these synoptic processes. Second, the dynamical characteristics of BVs are usually distinctly observed when an equatorward cold surge exists within the northeasterly winter monsoon flow from the South China Sea (SCS)10. The development of such a winter circulation relates to winter high-pressure disturbances over the upstream mid-latitudes via a teleconnection associated with meridional dispersion of the zonal Rossby wave group11. Precise detection and analysis of these multi-scale characteristics are essential for understanding the changes in BVs and their dynamical drivers, but are technically challenging.
Recently, state-of-the art climate model outputs from the High Resolution Model Intercomparison Project (HighResMIP12), an experimental protocol aiming to deliver a multi-GCM ensemble at horizontal resolutions finer than the current CMIP6 experiments, have demonstrated an ability to simulate the historical climatology13,14 and natural variability15 of regional precipitation over Southeast Asia. HighResMIP also plays an important role in understanding near-future projected changes in regional precipitation and their uncertainties due to changes in climate model resolution12. This implies a necessity, but also an opportunity, to fill the knowledge gap on possible future changes in BVs using the state-of-the-art high-resolution GCMs.
Here, we undertake the first study to project the near-future response of BV characteristics to a high-level emission scenario of greenhouse gases as used for the current IPCC AR6 assessment report. This is achieved by using six high-resolution GCMs (referred to as HR-GCMs and summarized in Supplementary Section 1) of HighResMIP and an advanced objective identification algorithm1,16,17 (see Methods) to detect the full-lifetime trajectories of BVs in these models. To demonstrate the importance of the use of high-resolution climate models for BV projection, a validation of the models and their low-resolution counterparts, compared to one of the latest high-resolution climate reanalysis datasets (ECMWF Reanalysis v5, ERA518), is discussed in detail in Supplementary Section 1. This study investigates changes in BV characteristics (including frequency, intensity, precipitation and associated large-scale environmental controls) for the near future climate (2030–2050) under the Shared Socioeconomic Pathway 5–85 (SSP5-85) scenario19 relative to the historical baseline period (1979–1999).
Changes in vortex features
Figure 1 shows the anomalies of BV frequency relative to the mean frequency for the historical baseline period (1979–1999). The historical time series for the ERA5 reanalysis dataset exhibits a multi-decadal variability of BV frequency, as the period 1986–1999 shows relatively high frequency and a lower frequency is shown for 2000–2014 for all-regions (Fig. 1a). The cause of such variability remains unclear, with a possible association with decadal climate variability in the tropical Pacific20. With such variability, historical climate reanalysis data with a limited time period does not allow for robust generalization of long-term trends in BV frequency forced by anthropogenic global warming. The ensemble mean of the HighResMIP GCMs shows some ability to reproduce the historical variation of BV frequency for North Borneo (Fig. 1c), but the variation for the southwestward propagating BVs is poorly simulated (Fig. 1b). Analyses of the BV frequency based on the ensemble mean from 1979 to 2050, with the period 2015–2050 projected under SSP5-85, indicates significant decreasing trends (p-value < 0.05 in Mann-Kendall test) in BV frequency by -0.46 and − 0.42 per decade for Peninsular Malaysia-Sumatra and North Borneo respectively. Such decreasing trends are captured by all the selected GCMs except HadGEM3-GC3.1 (Supplementary Table 2), indicating a limited but noticeable model uncertainty.
The spatial distribution of BVs in the historical baseline period (Fig. 1d-f) and projected future changes (Fig. 2g-i) are analyzed using the multi-GCM ensemble mean. Figure 1g indicates a projected decrease (p-value < 0.1 according to the Wilcoxon rank-sum test) in the genesis density of BV (up to 1.8 per 3.5° spherical area per season) to the north of Borneo which is the primary BV formation region (Fig. 1d), a change consistent with the decreasing trend in BV frequency (Fig. 1c). For the mean track density of BVs, a significant reduction (p-value < 0.05) by up to 2.5 cyclones per month per 3.5° spherical area is projected over the south of Peninsular Malaysia and the east of Sumatra. Some decrease (p-value < 0.1) is seen over the most active region near the western coast of North Borneo. For the lysis density, Fig. 1i shows a significant decrease by up to 2.1 cyclones per month per 3.5° spherical area per season near the Karimata Strait (105°E, 0°N). This reveals a decrease in the southwestward propagating BVs; hence, BVs are projected to be confined within the offshore region of North Borneo and become more stationary.
Changes in BV-associated precipitation
Figure 2 shows the spatial distribution of total precipitation during ONDJFM and the contribution from BVs. There are significant increases (greater than 90 mm per season, p-value < 0.05) in BV-associated precipitation (defined in Methods) over most of the southern part of the SCS (Fig. 2e). Although most of the statistically significant changes are over ocean, increases in BV-associated precipitation are found in most of Borneo, with the most significant increases to the north and east of the region. Also, the fractional contribution of BVs to the total precipitation during ONDJFM increases significantly over these regions (Fig. 2f). The largest increase (up to 13.5%) is seen near the offshore region of North Borneo (110°E, 5°N), associated with both increased BV-associated precipitation and some slight decreases in the seasonal total precipitation (Fig. 2d). To the southwest of Peninsular Malaysia and east of Sumatra, there is some decrease in BV precipitation and the fractional contribution to the seasonal precipitation associated with the decrease in southwestward propagating BVs over these regions, though these changes are small compared to the general increases for most of the study region. As the occurrence of BVs decreases in most of the focused region, significant increases in the total amount of BV-associated precipitation are primarily driven by the significant enhancement of the mean precipitation intensity of BVs (Supplementary Figs. 6g), i.e. the mean daily precipitation rate within a geodesic distance of 3.5° of the BV centers. BV-associated precipitation has been observed in other studies to be enhanced by the occurrence of northeasterly cold surges6,10. However, the multi-GCM ensemble mean shows no apparent changes in the frequency, intensity (figure not shown) and precipitation enhancement by cold surges relative to the seasonal mean BV precipitation intensity (Supplementary Figs. 6i). This suggests a limited effect of the changes in cold surges on the changes in BV-associated precipitation.
A fraction of BVs can trigger extreme precipitation over Borneo1 during the winter monsoon, downstream over the Peninsular Malaysia–Sumatra–Java region21. Hence, we analyze projected changes in composite precipitation averaged over 50 of the most intense BVs (ranked by their central relative vorticity at 850-hPa) in each multi-GCM ensemble member. The ensemble mean projects a significant increase in precipitation, by up to 12–16 mm per day to the north (about 1° of geodesic distance) of the BV center (Fig. 4a). Such an intensified near-center precipitation is projected by all the selected GCMs except EC-Earth3P. This can be explained by the projected intensification of extreme BV dynamics, in terms of significant increases in the low-level wind speed (Supplementary Fig. 7g) driven by the sharper horizontal gradient of mean-sea-level pressure (MSLP, Supplementary Fig. 7h) and low-level convergence (Supplementary Fig. 7i). These structural changes also partly explain the increased mean precipitation intensity of BVs (Supplementary Figs. 6g, h). Analysis of changes in the probability density function for BV intensities indicates an increase of the mean BV intensity (i.e. increases in the proportion in the relatively high intensity range) when BV intensities are measured by the central relative vorticity at 850-hPa (Supplementary Fig. 8), which is consistent with the above-mentioned intensification of extreme BVs.
To quantify extreme precipitation from BVs, we compute the number of Heavy Rain Days (HRDs) occurring within a 3.5° geodesic radius from the identified BV center. A HRD (also referred to as R50mm22) is defined as a day with daily precipitation > 50 mm, a threshold suggested by the World Meteorological Organization23. We found a significant increase (by up to 0.8 days per season) in BV-associated extreme precipitation over most of the southern SCS during ONDJFM. An increase (up to above 12%) in the fractional contribution of BVs to the total HRDs (all days with precipitation > 50mm per day) during ONDJFM is also seen near the offshore region of North Borneo. This suggests that BVs will be more closely linked to the occurrence of extreme precipitation due to near-future climate warming.
Environmental Drivers of BV Responses
The projected reduction in frequency and increased precipitation intensity of BVs can be explained by two environmental mechanisms: (1) changes in the Northeast Monsoon flow; (2) changes in convective instability. As BVs are observed in the lower troposphere and are mainly dependent on the low-level environmental fields6,8, we analyze the historical distribution and the projected future changes of environmental variables at 925-hPa, including the wind field, relative vorticity, moist static instability and vertical velocity averaged during ONDJFM (Fig. 5). In Fig. 5e, a weakening of the Northeast Monsoon flow is seen across 2.5–10°N, with the most significant weakening occurring upstream of the monsoon flow near the Philippines (p-value < 0.05). Significant intensification of the monsoon flow is seen to the north of 10°N. This indicates a northward shift in the Northeast Monsoon, which creates a less favorable environment for BVs over their most active region near the western coast of northern Borneo. A similar result is seen for the changes in the averaged synoptic fields 24-hours prior to the genesis of the 50 most intense BVs (Supplementary Fig. 9c). Note that the analysis of the synoptic fields for the most extreme BVs indicates a significant intensification of the disturbances expanding from the Indochina Peninsula to southern China. This indicates a more active cold air mass southward extending from the Siberian High downstream in a warmer climate (Supplementary Fig. 9d), consistent with the increase in the magnitude of winter temperature variability in southern China, as projected by the CMIP5 GCMs24. These changes lead to a sharper pressure gradient that drives a significant intensification of the low-level winds across 10–20°N. This makes the northeasterly winds and the associated cold surges shift northward, making it more difficult for these events to propagate southward and trigger BVs through orographic effects8,10, an observation which is consistent with the decreased occurrence of BVs. This is further confirmed by the projected northward shift in cold-surge-related BV tracks during the days when cold surges occur according to the mean track density and occurrence fraction (Supplementary Fig. 10c, d).
The analyses of the inertial (in terms of relative vorticity, Fig. 5f) and moist static instability (moist-adiabatic lapse rate, Fig. 5g) show significant increases in the lower-tropospheric instability (p-value < 0.05). These changes favor an increase in BV-associated precipitation and extreme precipitation events. An analysis of vertical profiles shows that increased relative vorticity mainly dominates 10°N throughout the much of the troposphere and is accompanied by increased vertical ascent (Supplementary Fig. 11e), a result of the northward shift in the Intertropical Convergence Zone during winter25. The increase in moist static instability extends throughout the lower troposphere from 700 to 925-hPa (Supplementary Fig. 11f), which is driven by the larger magnitude of the increase in specific humidity (Supplementary Fig. 11g) while the decrease in the tropospheric lapse rate, associated with a greater magnitude of the projected warming in the upper troposphere, has an opposite effect (Supplementary Fig. 11h). Further analysis of the changes in the 925-hPa barotropic and baroclinic stabilities, in terms of the barotropic (MPV1) and baroclinic moist potential vorticity (MPV2, both defined in Methods), suggests a significant increase in lower-tropospheric barotropic instability (Supplementary Fig. 12c). In contrast, baroclinic instability decreases significantly to the north of 15°N (Supplementary Fig. 12d). Although this region is controlled by an intensification of the monsoon flow and associated cold surges (which act to increase the baroclinic instability through intensified cold advection), a pronounced decrease in the meridional temperature gradient (Supplementary Fig. 11h) is projected, which tends to baroclinically stabilize the low troposphere. Furthermore, changes in large-scale ascent in terms of changes in the 925-hPa vertical velocity (Fig. 4h) show a significant increase to the north of 10°N and decrease in the vicinity of the Peninsular Malaysia–Sumatra–Java region, which agrees well with the changes in the BV-associated precipitation (Figs. 2h, 3b).