Projected near term changes in the East Asian summer monsoon and its uncertainty

This work uses the models MetUM-GOML1 (hereafter: GOML1) and MetUM-GOML2 (hereafter: GOML2) (Hirons et al 2015): both are near-globally-coupled models that comprise the atmospheric components of two state-of-the-art climate models coupled to a multi-level mixed-layer ocean.

The atmospheric component for GOML1 is the Met Office Unified Model (MetUM) at the fixed scientific configuration Global Atmosphere 3.0 (GA3.0; Walters et al 2011). The atmospheric component of GOML2 is MetUM GA6.0 (Walters et al 2017). Comparing with GA3.0 used in GOML1, the largest change in GA6.0 is that the 'New Dynamics' dynamical core is replaced with 'ENDGame' (Wood et al 2014). The inclusion of the ENDGame dynamical core is an important upgrade to the Global Atmosphere configuration of the MetUM. ENDGame maintains the benefits of 'New Dynamics', whilst improving on its accuracy, stability and scalability. The improved accuracy significantly reduces the model's implicit damping, leading to a beneficial improvement in various modes of variability, such as the depth of extra-tropical cyclones and the structure of frontal systems. The resolution for both atmospheric models is 1.875° longitude and 1.25° latitude (N96) with 85 vertical layers. The models include an interactive tropospheric aerosol scheme which is able to simulate the direct, indirect, and semi-direct effects of aerosols (Jones et al 2011, Walters et al 2011).

The oceanic component of GOML1 and GOML2 is a Multi-Column K Profile Parameterization (MC-KPP) mixed-layer ocean model (Large et al 1994). The vertical resolution is 1.2 m at the surface, and less than 2 m over the first 41.5 m. The oceanic and atmospheric components are coupled once every 3 h via the Ocean Atmosphere Sea Ice Soil (OASIS) coupler (Valcke et al 2003). The atmospheric and ocean components are not coupled in regions of sea ice defined using the AMIP sea ice data set (Taylor et al 2012). Outside the coupled region, daily SST and sea ice climatologies are prescribed. More details on the oceanic component and the coupling are documented in Hirons et al (2015).

The experiments performed in this study are summarised in table 1. We first performed a relaxation experiment (R0) for 12 years for each model, in which the temperature and salinity of the ocean (MC-KPP) are relaxed to a present day observed climatology (Smith and Murphy 2007). The relaxation experiment uses 1994–2011 average anthropogenic GHG concentrations and anthropogenic aerosol (AA) emissions (Lamarque et al 2010, 2011).

Based on the relaxation experiment, seasonally varying 3D flux corrections for ocean temperature and salinity are diagnosed for each model. Then two sets of experiments are performed with the same prescribed 3D ocean temperature and salinity corrections. Two numerical experiments are carried out for each model: present-day (PD, 1994–2011) and future period (FP, 2045–2055). PD uses present day anthropogenic GHG concentrations and AA emissions averaged for 1994–2011. FP is driven with 2045–2055 average GHG concentrations and AA emissions from the RCP4.5. The changes in both GHG concentrations and AA emissions can change air-sea heat flux, and therefore lead to changes in SST and ocean heat content. However, different changes of air-sea heat flux in the two models could lead to different resulting SST patterns. The PD and FP simulations in each model started from the same initial conditions from the end of the corresponding relaxation run. The same set of simulations have been used to investigate projected changes of heat waves in China (Su and Dong 2019).

Relative to PD, GHG concentrations for FP under RCP4.5 scenario are higher, with CO₂ increased by 30%, CH₄ by 4.5%, and N₂O by 10.4% respectively. AA emissions decrease with, in particular, sulphur dioxide concentrations decreasing by more than 50% over large areas of the North America and Eurasian continents (figure 1 of Su and Dong 2019). All experiments are run for 50 years and use the climatological PD sea ice extent from HadISST (Rayner et al 2003).

Zoom In Zoom Out Reset image size

The last 45 years of each experiment are used for analysis (see table 1 for more details). We define the future changes as the difference between FP and PD experiment (FP–PD). Statistical significance of the mean changes in model experiments is assessed using a two tailed Student's t-test.

The biases of simulated present day SST distributions in both GOML1 (Dong et al 2017) and GOML2 (Luo et al 2018) are much smaller (typically between −0.5 °C and 0.5 °C) than in the CMIP5 ensemble (e.g. Wang et al 2014). To evaluate the performance of the models in simulating the EASM, the circulation and precipitation climatology over East Asia in GOML1 and GOML2 are compared with observations. Here we used the Climate Research Unit Precip 3.2 dataset (CRU) on a 0.5° × 0.5° grid (Mitchell and Jones 2005). The observed wind fields are from the NCEP/NCAR reanalysis version 1.0 on a 2.5° × 2.5° grid (Kalnay et al 1996). Observed sea level pressure (SLP) data are obtained from the Hadley Centre's monthly historical mean SLP data set (HadSLP2) at a horizontal resolution of 5° × 5° (Allan and Ansell 2006).

In observations, the subtropical anticyclone over the Western North Pacific is associated with a south-westerly wind along the southern and eastern coast of East Asia. The south-westerly wind converges with easterly winds from the tropical western Pacific in the monsoon trough and becomes a strong southerly (figure 1(a)). This southerly wind transports moisture to East Asia enhancing precipitation. Summer precipitation in China decreases with latitude with mean precipitation being more than 6 mm d⁻¹ south of 30°N, but less than 2 mm d⁻¹ north of 35°N (figure 1(b)).

The general features of the summer circulation field in observations are reproduced reasonably well by GOML1 (figure 1(c)) and GOML2 (figure 1(e)). The models also accurately capture the precipitation centres over the lower reaches of the Yangtze River (figures 1(d) and (f)). However, in GOML1, the precipitation is overestimated over northern India and eastern China, which may be related to the bias in simulating the monsoon trough. As shown in figure 1(c), GOML1 overestimates the westerly wind over the Indian Ocean and westerlies extend too far to the east into the Pacific, making the simulated monsoon trough deeper than in observations. Corresponding to the biases in circulations, the simulated precipitation over south-western China and Myanmar is stronger than observations. In contrast to the wet bias in GOML1, in GOML2 the simulated monsoon precipitation is weaker over East China, especially northeastern China. As shown in figure 1(e), the simulated Western North Pacific subtropical high (WNPSH) is weaker in GOML2 compared with observations, thus, the monsoon trough and related southerly wind from the South China Sea (SCS) are located farther east from the coast.

Despite some deficiencies, the differences between the atmospheric circulation of the EASM in GOML1 and GOML2 are at least representative of differences between CMIP5 models. The simulated precipitation and lower tropospheric circulation over East Asia are comparable with observations, which suggests that both GOML1 and GOML2 are appropriate tools to study the future changes in the EASM.

GOML1 projects that precipitation will increase over northern China (30°N–50°N, 100°E–120°E) (figure 2(a)). The regional average increase is 0.41 mm d⁻¹ with the maximum increase larger than 0.5 mm d⁻¹, which is more than 10% of the PD mean precipitation and is statistically significant at the 10% level.

Zoom In Zoom Out Reset image size

The increased precipitation over northern China is consistent with the increased land-sea thermal contrast driving enhanced circulation (figure 3(a)): the surface warming over the Asian continent is larger than over the surrounding ocean (Chou 2003). The increased zonal temperature gradient between the Asian continent and the Western Pacific Ocean enhances the zonal pressure gradient and leads to stronger southerly winds over eastern China around 110°E and easterly wind around 120°E, transporting more water vapour from the adjacent ocean to northern China (figures 3(b) and (c)). On the other hand, over south-eastern China (20°N−30°N, 110°E−120°E), the regional averaged precipitation decreases significantly by about 0.35 mm d⁻¹ with the maximum above 1.0 mm d⁻¹ (figure 2(a)). The decreased precipitation is related to the circulation anomalies: there is an anomalous low SLP centre at northern China (dashed box in figure 3(b)), thus, the surface pressure over south-eastern China (solid box in figure 3(b)) is relatively high, consistent with anti-cyclonic circulation anomaly of surface wind.

Zoom In Zoom Out Reset image size

GOML2 also projects increased precipitation over northern China for about 0.43 mm d⁻¹ (in figure 2(b)). However, in contrast to GOML1, over south-eastern China the regional averaged precipitation significantly increases by about 1.0 mm d⁻¹. This difference is consistent with the circulation differences: in GOML2, the simulated SLP change over the SCS is higher than that in GOML1 (compare figures 3(b) and (e)), with an anti-cyclonic circulation anomaly over the SCS. The anomalous south-westerly flow on the northwest of the anti-cyclonic circulation anomaly transports more water vapour to southern China (figures 3(e) and (f)), leading to increased precipitation in GOML2.

Overall, this difference in the response of precipitation over south-eastern China to global warming reflects the large uncertainty in future precipitation change over the East Asian monsoon area, in particular related to the uncertainty in the projected circulation change. This is consistent with Zhou et al (2017) who found with 18 CMIP5 models that circulation changes contributed most of the uncertainty in projected future EASM rainfall.

In order to explore the different processes that contribute to the future changes in precipitation, we analysed the changes of vertical integrated total moisture transport convergence $\left(\delta {\rm{T}}{\rm{M}}{\rm{C}}\right)$ which is based on 6-hourly data. As shown in equation (1), $\delta {\rm{T}}{\rm{M}}{\rm{C}}$ can be decomposed into transport convergence based on monthly mean data $\left(\delta {\rm{M}}{\rm{C}}\right)$ and transport convergence in transient eddies (TE). To investigate the changes due to anomalous circulation and the part due to anomalous humidity, the monthly moisture transport convergence can be further separated into changes in dynamic component $\left(\delta {\rm{D}}{\rm{Y}}\right),$ thermodynamic component $\left(\delta {\rm{T}}{\rm{H}}\right),$ and the quadratic term of covariance between changes in humidity and winds $\left(\delta {\rm{Q}}{\rm{T}}\right)$ (Li et al 2015, Li and Ting 2017):

$$\begin{eqnarray}&&\begin{array}{l}\delta {\rm{T}}{\rm{M}}{\rm{C}}=\delta {\rm{M}}{\rm{C}}+\delta {\rm{T}}{\rm{E}}=\delta {\rm{D}}{\rm{Y}}+\delta {\rm{T}}{\rm{H}}\\ \,\,\,\,+\,\delta {\rm{Q}}{\rm{T}}+\delta {\rm{T}}{\rm{E}}\,.\end{array}\end{eqnarray} \tag{ 1 }$$

Here the difference between FP and PD is represented by $\delta ,$ the dynamic components represented by $\delta {\rm{D}}{\rm{Y}}$ involves only changes in circulation. The thermodynamic contribution represented by $\delta {\rm{T}}{\rm{H}}$ involves only changes in specific humidity.

The regional averaged changes in precipitation over south-eastern China and northern China are compared with changes in moisture transport convergence in figure 4. In GOML1 (figure 4(a)), $\delta {\rm{D}}{\rm{Y}}$ explains around 74% of the decreased precipitation over southern China and explains 60% of increased precipitation over northern China. In GOML2 (figure 4(b)), $\delta {\rm{M}}{\rm{C}}$ explains 95% of increased precipitation over south-eastern China with both $\delta {\rm{D}}{\rm{Y}}\,\,$and $\delta {\rm{T}}{\rm{H}}$ being positive.

Zoom In Zoom Out Reset image size

This indicates that changes in moisture transport convergence by monthly mean flow are predominantly responsible for the changes in precipitation over south-eastern China. However, over northern China, the precipitation changes show large differences from the moisture transport convergence based on monthly mean data. By analysing the different contributions from evaporation and transient eddies, we find that the changes in evaporation are very small (not shown), the differences between precipitation and moisture transport convergence based on monthly mean data are mainly from the impact of transient eddies (figure 4).

Over northern China, consistent with the changes in precipitation (figures 2(a) and (b)), both models show an increase in moisture transport convergence (figures 5(a) and (d)). This is mainly driven by the dynamical component ($\delta {\rm{D}}{\rm{Y}}$) related to the enhanced monsoon circulation due to amplified land-sea thermal contrast (figures 5(b) and (e)). In GOML1 (figure 4(a)), the regional averaged $\delta {\rm{D}}{\rm{Y}}$ explain 88% of the increase in $\delta {\rm{M}}{\rm{C}}.$ In GOML2 (figures 4(b)), the increase in $\delta {\rm{D}}{\rm{Y}}$ is 83% of the $\delta {\rm{M}}{\rm{C}}.$ Therefore, in both GOML1 and GOML2 the increased moisture transport convergence over northern China is dominated by the changes of the dynamical component $\left(\delta {\rm{D}}{\rm{Y}}\right)$ related to the enhanced monsoon circulation.

Zoom In Zoom Out Reset image size

Over south-eastern China, the decrease of regional averaged $\delta {\rm{M}}{\rm{C}}$ in GOML1 (figure 4(a)) is mainly driven by DY. In GOML2 (figure 4(b)), the area averaged $\delta {\rm{M}}{\rm{C}}$ is contributed from both $\delta {\rm{D}}{\rm{Y}},$ about 60% which causes convergent anomalies over south-eastern China, and the $\delta {\rm{T}}{\rm{H}}$ by about 33%, which is caused by the increased moisture. This indicates that the inconsistency in projected precipitation over south-eastern China between GOML1 and GOML2 is largely due to the difference of the changes in circulation.

In summary, over northern China, the increased precipitation is driven by the enhanced monsoon circulation. This is consistent in both GOML1 and GOML2. Over south-eastern China, the thermodynamic component increases precipitation in both models. However, the changes of dynamic component are opposite, showing a decrease in GOML1 and an increase in GOML2, leading to the decreased precipitation in GOML1 and increased precipitation in GOML2. Therefore, the uncertainty of projected precipitation changes arises largely from differences in the projected circulation changes.

As shown above, uncertainty in projected precipitation change is mainly due to uncertainty in how circulation changes. In GOML1, decreased precipitation over south-eastern China is located at the south of a low SLP anomaly over northern China (in dashed red box of figure 3(b)). Over southern China, the SLP is less negative (in solid red box of figure 3(b)). This is consistent with an anticyclonic circulation anomaly and related to less precipitation over south-eastern China. In GOML2, the increased precipitation over south-eastern China is related to a low SLP anomaly over south-eastern China and a related cyclonic circulation anomaly (in solid box of figure 3(e)). Therefore, to understand why the projected precipitation over south-eastern China is different in GOML1 and GOML2, our next question is why the SLP response in GOML1 over southern China is higher than that in GOML2?

In GOML1 the regional averaged precipitation is increased over the SCS (figure 6(a)). This is associated with the warming of SST due to global warming which induces more convection over the SCS. Although the mean SST is also increased in GOML2, the change of meridional SST gradient between the SCS and Western North Pacific (WNP) is much larger in GOML2 than in GOML1 (figures 6(c) and (d)).

Zoom In Zoom Out Reset image size

In GOML1, the warming of SST in the tropical ocean is related to anomalous ascent in the tropical atmosphere and anomalous descent around 20°N (figure 7(a)), which is consistent with higher SLP and less precipitation over south-eastern China. In GOML2, however, the weakened meridional SST gradient between the SCS and WNP (figure 6(d)) is associated with an anomalous local Hadley circulation, characterized by anomalous ascent at 20°N–30°N and anomalous descent at 5°N–15°N (figure 7(b)). This anomalous local Hadley circulation is consistent with the lower SLP and enhanced precipitation over south-eastern China resulting from the anomalous ascent in GOML2.

Zoom In Zoom Out Reset image size

In summary, the differences in the projected circulation fields over south-eastern China in GOML1 and GOML2 are mainly due to different tropical SST anomalies particularly the changes in the meridional SST gradient. In GOML1, the warming of SST over the SCS enhances the local convection and weakens the local Hadley cell. The corresponding descending anomaly over south-eastern China is consistent with higher SLP and reduced precipitation.

In GOML2, however, the SST warming in the western tropical and North Pacific is non-uniform, which decreases the meridional SST gradient between the SCS and Western North Pacific. This large decrease in meridional SST gradient is associated with an anomalous local Hadley cell with anomalous ascent and increased precipitation over south-eastern China.