1. Introduction
Detecting future changes in tropical cyclone (TC) activity primarily involves two aspects: one is changes in the relevant environmental factors that affect TC activity, such as sea surface temperature (SST), vertical wind shear, and humidity variables. The other is the simulation of TC activity in the projected environmental conditions that are derived from climate models in future scenarios (Emanuel 2013; Wu et al. 2014; Zhang and Wang 2017; Wang and Wu 2018a,b; Knutson et al. 2020; Emanuel 2021). Numerous studies apply downscaling methods with regional models and statistical–dynamical frameworks using the projected large-scale environments or directly analyzing statistical characteristics of TC activity in high-resolution global climate models to understand TC activity change in a warming climate (Tory et al. 2013; Knutson et al. 2019; Vecchi et al. 2019; Zhang et al. 2020; Roberts et al. 2020). The predicted, more favorable vertical wind shear and SST are conducive to TCs and thus increase the TC genesis potential in the latest projection (Murakami and Wang 2022; Lockwood et al. 2022). The largest increases in TC track density are predicted far from the land, while TC intensification increases rapidly, and landfalling TCs are supposed to accumulate more power (Emanuel 2021; Camargo and Wing 2021). Whatever approaches are used, the prediction of TC activity is substantially controlled by changes in TC-relevant, large-scale circulation in each basin (Wu et al. 2022). However, there is a lack of research focusing on the future change of TC-relevant, large-scale circulation systems (e.g., the monsoon trough) and the evaluation of models’ performance in simulating these circulations remains elusive.
Since the confidence of TC activity projection considerably relies on models’ capacity in reproducing large-scale circulation relevant to TC activity in historical simulations and future projections, it is necessary to evaluate models’ performance on the related circulation systems in the state-of-the-art phase 6 of the Coupled Model Intercomparison Project (CMIP6). As the initial step in simulating TCs, the TC formation location is the basic aspect in predicting TC activity, and it is closely related to the monsoon trough (MT) over the western North Pacific (WNP). A majority of TCs spend most of their lifetime in the MT because of its accompanied favorable thermodynamic and dynamic large-scale conditions for TC formation and intensification (Molinari and Vollaro 2013), such as strong cyclonic vorticity, high midlevel relative humidity, and weak vertical wind shear (Chia and Ropelewski 2002; Chen et al. 2006; Zong and Wu 2015). The observations and high-resolution model simulations confirmed that the eastward (westward) shift of the MT is closely related to more (less) TC formation over the WNP (Wu et al. 2012, 2014). Moreover, many studies suggested that TC track and intensity are also associated with MT variations over the WNP on multiple time scales (Yumoto and Matsuura 2001; Wang and Chan 2002; Matsuura et al. 2003; Huangfu et al. 2017). In addition, the orientation of the MT further determines the formation location of the TCs and the position where the TC reaches its maximum intensity. Thus, the simulation of MT mean location and its changes is crucial for projecting TC activity, since the future TC formation location, track, and intensification are closely related to this large-scale circulation system. In phase 5 of the Coupled Model Intercomparison Project (CMIP5) models, it is found that climate models have difficulty reproducing TC activity with the diversity in simulating TC-relevant, large-scale circulation (Camargo 2013; Wang and Wu 2018a). How is the simulation and prediction of the MT in the CMIP6 models? This can help us better understand its impact on predicting TC activity, especially TC formation.
In this study, we first evaluated models’ capability in simulating TC-relevant, large-scale circulations in CMIP6 historical runs. Focusing on the spatiotemporal characteristics of the simulated MT, climate models are filtered to examine future MT shifts and their possible impacts on TC formation location over the WNP. Through this evaluation, we expect to gain an in-depth physical understanding of the model biases in simulating the MT and its major implications for guiding future efforts aimed at improving TC activity prediction.
2. Data and methods
a. Reanalysis and CMIP6
We used monthly mean winds data to examine large-scale circulation variations over the WNP from the ERA5 reanalysis (Hersbach et al. 2020) for the period 1979–2018. The monthly SST was obtained from the National Oceanic and Atmospheric Administration (NOAA) Extended Reconstructed SST, version 5 (ERSSTv5) (Huang et al. 2017). Our analysis mainly focuses on the TC peak season [July–September (JAS)]. We used 34 historical simulations for the period 1979–2014 provided by CMIP6 (Eyring et al. 2016). Detailed information on the 34 climate models is listed in Table 1.
List of the 34 climate models in CMIP6. The historical simulations from the first ensemble member run of variant label r1i1p1f1 in each model are used, where r1 is realization index, i1 is the initialization index, p1 is the physics index, and f1 is the forcing index. The boldface text denotes models with both historical and future scenario experiments. The expansions of most models can be found at https://www.ametsoc.org/PubsAcronymList. Additional expansions are Community Integrated Earth System Model for CIESM; BCC-CSM2 medium resolution for BCC-CSM2-MR; Korea Meteorological Administration Advanced Community Earth-System model for K-ACE-1-0-G; Manabe Climate Model, University of Arizona, for MCM-UA-1-0; Nanjing University of Information Science and Technology Earth System Model, version 3, for NESM3; Norwegian Climate Prediction Model for NorCPM1; SNU Atmosphere Model, version 0, with Unified Convection Scheme, for SAM0-UNICON; and Taiwan Earth System Model, version 1, for TaiESM1.
CMIP5 features representative concentration pathways (RCPs) to detect different possible, future greenhouse gas emissions. The CMIP6 has developed a new set of emission scenarios driven by different socioeconomic assumptions; this set is named the shared socioeconomic pathways (SSPs). A number of these SSP scenarios, including greenhouse gases, aerosols, and other climate and socioeconomic forcing, have been selected to force climate models to project what might happen in the future. In this study, we examine the SSP5-8.5 case, which has substantially higher CO2 emissions than RCP8.5 in CMIP5 and more optimistic outcomes (SSP2-4.5) to investigate how the world might be if it fails to enact any climate policies. The SSP2-4.5 is the updated RCP4.5, which is designed to explore scenarios that are generally warming to around 3°C by 2100, featuring a higher starting point and slightly slower decline in the middle of the twenty-first century. The cases of SSP5-8.5 and SSP2-4.5 will help us to assess the impact of global warming on climate systems at lower and higher degrees of CO2 emissions.
b. The definition of the MT location
Since the orientation and length of the simulated MT vary widely, the full length of the MT is considered in this study. Following previous studies (Simpson et al. 1968; Holland 1995), the MT location indices are defined as the mean location of the trough, represented by the mean latitude and longitude averaged from the starting point and the eastern extremity of the MT. The starting point of the MT is defined as the location where the strongest positive relative vorticity occurs in the region 10°–30°N, 100°–120°E at 850 hPa. The eastern extremity of the MT is determined by the eastern boundary of 850-hPa easterlies and westerlies (the wind line of zero zonal winds) with positive relative vorticity in the region 5°–20°N, 110°–170°E. The locations of TC formation are closely related to the MT location indices in the observations, especially in longitude (Fig. S1 in the online supplemental material).
c. Self-organizing map clustering method
To objectively assess the performance of the multimodels in terms of their successes in replicating the observed spatial features of the MT, we used the self-organizing map (SOM) method to perform cluster operation on spatial patterns of 850-hPa wind fields derived from each model. This method is a type of artificial neural network algorithm that is trained using unsupervised learning to cluster the most common and major spatial features into the same group across all input patterns. The SOM can recognize patterns, relying on nearby locations that are attributed to similar characteristics (Kohonen 1990). This method has been extensively used to analyze temporal dynamics of model errors in hydrological modeling (e.g., Herbst et al. 2009; Reusser et al. 2009), but applications for evaluating climate model performance remain limited.
Here, we used this method to cluster a set of 850-hPa wind trends derived from CMIP6 models into several well-distinguished groups. We also used two other assessment metrics, pattern correlation and root-mean-square error (RMSE), to evaluate spatial patterns of 850-hPa winds in each model against the observed counterparts. Thus, these three methods (i.e., pattern correlation, RMSE, and SOM) collectively provide a comprehensive examination that aids us in better characterizing and analyzing multimodel performance in replicating the observed spatial characteristics of 850-hPa atmospheric circulation systems.
3. The simulated TC-relevant, large-scale circulation in CMIP6
The prerequisite for predicting TC changes primarily relies on reasonable simulations of TC-relevant, large-scale circulation systems. There are four major circulation systems affecting TC activity over the WNP in boral summer: the South Asian high (SAH) and the tropical upper-tropospheric trough (TUTT) in the upper levels, in conjunction with the MT and the western North Pacific subtropical high (WNPSH) in the lower troposphere (Wang and Wu 2018a,b; Feng and Wu 2022; Wu et al. 2022). The SAH is characterized by a strong and stable anticyclone centered over the Tibetan Plateau at 200 hPa in the climatological mean wind field (Fig. 1a). The easterly winds at the eastern boundary of the SAH along 15°N are juxtaposed with the bottom of the TUTT. At 850 hPa, the convergence of westerlies and easterlies along 5°–20°N represents the mean location of the MT, usually along with relatively strong, positive relative vorticities (Fig. 1c). To the east of the MT, the WNPSH is the dominant anticyclonic circulation. These observed structures can be reasonably simulated in the multimodel ensemble means of CMIP6 historical runs, although considerable model biases remain in reproducing the exact location of the boundary between the easterly and westerly winds (Figs. 1b,d). At the lower levels, the multimodel ensemble means of simulated MT locate more northward than in the observations. In particular, the mean latitude of the simulated MT varies between 10° and 20°N in each model, exhibiting noticeable differences in its orientation (Fig. 2), even some models failing to reproduce a clear transformation of the westerlies over the WNP. These intermodel biases in simulating the climatological mean location of the MT may further lead to uncertainty in predicting TC formation location and its further development.
The 200-hPa mean winds (vectors; m s−1) in (a) ERA5 during 1979–2018 and (b) the multimodel ensemble mean in CMIP6 historical simulations during 1979–2014 in the TC peak season (JAS). (c),(d) As in (a) and (b), but for 850-hPa winds (vectors; m s−1) and relative vorticity (shading; ×105 s−1). Red lines indicate that the zonal wind speed is equal to zero. In (b) and(d), stippling indicates grids where a majority (>70%) of the 34 models exhibit the same sign as that of the corresponding wind pattern in observations in (a) and (c).
Citation: Journal of Climate 36, 15; 10.1175/JCLI-D-22-0395.1
The mean location of the monsoon trough (MT) in the TC peak season (JAS) in ERA5 (thick black curve) and the CMIP6 historical simulation for each model. The thick red curve indicates the multimodel ensemble mean location of the MT. The definitions of the starting point and the eastern extremity of the MT are described in detail in section 2b.
Citation: Journal of Climate 36, 15; 10.1175/JCLI-D-22-0395.1
The linear trend of these systems is another important item to inspect model performance in historical simulations, as the physical processes of a model are designed to reflect the atmospheric response to both internal and external forcing. Thus, the linear trends of 200-hPa and 850-hPa winds are examined during the period 1979–2014 in CMIP6 historical simulations. When comparing with observed easterly trends along 30°N over Eurasia and cyclonic trends in the central North Pacific around 160°E at 200 hPa, models can basically capture these features in the multimodel ensemble means (Figs. 3a,b). However, changes in low-level winds do not match the observations in general. The 850-hPa winds exhibit a cyclonic trend along the MT over the WNP in the observation, while it appears to be an anticyclonic trend in the multimodel ensemble means (Figs. 3c,d). The strong increase in the observed relative vorticity along the MT is hard to find in the multimodel ensemble means. This indicates that models have larger uncertainty in reproducing low-level circulation variations than that in the upper levels. Since the MT is an important system influencing TC activity, it is necessary to evaluate the performance of each model in simulating the MT.
Linear trends of 200-hPa winds (vectors; m s−1 decade−1) in (a) ERA5 during 1979–2018 and (b) the multimodel ensemble mean trend in CMIP6 historical runs during 1979–2014 in the TC peak season (JAS). (c),(d) As in (a) and (b), but for 850-hPa winds (vectors; m s−1 decade−1) and relative vorticity (shading; ×105 s−1 decade−1). Red lines indicate that the zonal wind speed is equal to zero. Grid points with trends in zonal winds [in (a) and (b)] and relative vorticity [in (c) and (d)] that are statistically significant at the 95% confidence level are denoted by purple stippling.
Citation: Journal of Climate 36, 15; 10.1175/JCLI-D-22-0395.1
4. Model performance in simulating the MT
a. Evaluating the simulated low-level circulation over the WNP
To further evaluate model performance with regard to spatial characteristics of the simulated MT, we first applied two widely used metrics—pattern correlation and RMSE—for the climatological mean 850-hPa winds and the trend field in the MT active region (5°–30°N, 100°–160°E), separately. Pattern correlation coefficients of the climatological mean winds calculated in each model against ERA5 generally exhibit significant positive values (ru850 = 0.87 and rv850 = 0.54, on average) with the RMSE varying from 1 to 13 (RMSEu850 = 6.82 and RMSEv850 = 2.58, on average), indicative of a relatively good simulation of the 850-hPa wind structure with certain spread in magnitude (Figs. 4a,b). In contrast, pattern correlation coefficients of zonal and meridional wind trends in each model against ERA5 spread from −0.77 to 0.73 for zonal winds and from −0.62 to 0.68 for meridional winds (ru850 = −0.08 and rv850 = −0.09, on average), with the RMSE ranging from 0.05 to 0.25 for zonal winds and from 0.03 to 0.9 for meridional winds (RMSEu850 = 0.11 and RMSEv850 = 0.05 on average; Figs. 4c,d). Consistent with our comparison of the observed and simulated 850-hPa winds using the multimodel ensemble means in the previous section, these results suggest that the models have a common limitation in simulating the observed spatial patterns of low-level circulation mean states and trends in the MT active region, especially the linear trend.
(a) Pattern correlations and (b) RMSE of 850-hPa mean zonal (horizontal axis) and meridional (vertical axis) winds between CMIP6 models and ERA-5 in 5°–30°N, 100°–160°E (blue-outlined box in Figs. 3c,d). (c),(d) As in (a) and (b), but for linear trends of the 850-hPa winds. Red dashed lines represent the multimodel ensemble mean of the pattern correlations and RMSE in zonal and meridional directions.
Citation: Journal of Climate 36, 15; 10.1175/JCLI-D-22-0395.1
Considering that these two methods only provide an overall assessment of model performance without detailed information on how each model performs in simulating specific regional features of 850-hPa wind trends, we additionally use the SOM clustering method to classify linear trends of 850-hPa winds derived from the 34 models. This also provides us an opportunity to visualize how close models’ simulations are to the observed counterparts in a spatial sense. Since the explained covariance increases with the increasing number of clusters, considering a reasonable number of clusters, which contributes to the largest increasing rate of the squared covariance of total variations (Kohonen 1990), 34 wind-trend fields are classified into nine groups that explain 58% of the total variations. The nine groups yield diversified spatial structures, and the mean position of the MT averaged from each group largely was spread out (Fig. 5). The mean latitude of simulated MT locates more northward than in the observations, and the eastern extremity of simulated MT extends from 140°E to east of 160°E in the nine groups. Approximately two-thirds of the models exhibit westerly trends to the western side of the MT, indicating an eastward extension of the MT (nodes 1, 4, 5, 7, 8, and 9, including 22 models), while the rest present easterly trends to the western side of the MT, indicating a westward retreat of the MT (nodes 2, 3, and 6, including 12 models; Fig. 7). Their pattern correlation coefficients of zonal and meridional winds against observed counterparts vary from −0.78 to 0.85. Only two groups (node 1 including five models and node 4 including three models) bear some resemblance to the observations (pattern correlation: ru850 = 0.85 and rv850 = 0.80 for node 1; ru850 = 0.80 and rv850 = 0.74 for node 4; Figs. 5a,d), containing eight models (i.e., ACCESS-ESM1-5, CESM2-FV2, CIESM, CanESM5, and GISS-E2-1-G in node 1 and BCC-CSM2-MR, GFDL-ESM4, and SAM0-UNICON in node 4). Consistent with the other two assessment methods (Fig. 4), the SOM method reinforces the conclusion that most CMIP6 models have a noticeable limitation in reproducing the mean position of the MT and the variations of the low-level circulation related to MT activity.
Classifying patterns of 850-hPa wind trends using the SOM method. Linear trends of 850-hPa winds (vectors; m s−1 decade−1) from 34 models are clustered into nine groups, considering spatial patterns of meridional and zonal winds (shading; m s−1 decade−1) as two components. The number of models classified in each group is indicated in the title of each panel. The models in each cluster are presented in Fig. 7, below. Red curves indicate the mean location of the MT averaged from the models in each group. The definitions of the starting point and the eastern extremity of the MT are described in detail in section 2b.
Citation: Journal of Climate 36, 15; 10.1175/JCLI-D-22-0395.1
b. Limitations of diversified tropical SST in simulations
Why are the models limited in their ability to reproduce the low-level circulation associated with the MT? We speculate that this limitation is possibly linked to a poor simulation of the SST mean state in the tropical Pacific. The warm SST anomalies in the eastern tropical Pacific serve as a positive heating that generates a pair of symmetric cyclones along the equator (Gill 1980). The cyclonic circulation over the North Pacific extends the MT eastward (Wu et al. 2012; Wang and Wu 2018b). This suggests that changes in the east–west SST contrast in the North Pacific can shift the basinwide mean location of the MT by modulating tropospheric large-scale circulations (Feng and Wu 2022). Similar to these previous studies, shifts in the simulated MT and relative cold SST share a consistent movement (Fig. 6a). The cold (warm) SST extension favors a westward (eastward) shift of the MT. To represent the zonal migration of the cold anomalous SST extension, we use the zero line of SST anomalies that is relative to the basinwide Pacific SST by subtracting the zonal mean SSTs averaged over the Pacific at each latitude. The western boundary of the zero line varies from ∼140°E to ∼170°W in each model, which indicates a large intermodel bias in simulating the mean state of the east–west SST contrast in the tropical Pacific (Fig. 6b). The boundary of the zero line in the multimodal ensemble means is located more northwestward than that in observations along 10°N. This bias in simulating SST mean states in the tropical Pacific might further limit a model’s skill in reproducing large-scale circulations including the MT. The climatological mean circulation in boreal summer is commonly harder to simulate than in the other seasons, which is possibly due to an imperfect simulation of the strength of the Hadley cell or eddy–mean flow feedbacks (Webster 1961; Hartmann 2007; Wu et al. 2012). More efforts are needed to improve simulations of air–sea interactions, such as tropical convection in future studies.
(a) The longitude of the mean location of the MT and the western point of the zero SST line of anomalous SST over the tropical Pacific in each model and the observations (the black circle). (b) The zero SST line of anomalous SST over the North Pacific in ERSST5 (thick black curve) and 34 historical simulations with the multimodel ensemble mean (thick red curve). (a) The asterisk indicates that the fitting curves are statistically significant at the 95% confidence level.
Citation: Journal of Climate 36, 15; 10.1175/JCLI-D-22-0395.1
5. MT changes in the warming scenarios
Although models have biases in reproducing TC-relevant, large-scale circulation, some that capture the observed characteristics of atmospheric circulations in historical runs may deliver valuable messages of the atmosphere response to anthropogenic forcing in their future simulations. To select relatively reliable models to predict future changes of the MT, we compared model performance in simulating the MT in historical runs by considering the three assessment methods (pattern correlation, RMSE, and SOM). Climate models that meet the two requirements—having a larger pattern correlation and smaller RMSE among 50% of the total models, and being grouped in nodes 1 and 4 by the SOM—are selected to investigate future changes of the MT in the SSP2-4.5 and SSP5-8.5 scenarios (Fig. 7). Five models passed the criteria: ACCESS-ESM1-5, BCC-CSM2-MR, CIESM, CanESM5, and GFDL-ESM4. It is noticed that TaiESM1 (model number 34) is clustered in node 7, which does not primarily present the features of the spatial distribution as in the observations (i.e., the cyclonic trend located along the MT), although the pattern correlations of the trends with the observations are relatively higher than in other models (ru850 = 0.46 and rv850 = 0.34). This result suggests a weakness in using pattern correlation to compare the similarity of two spatial fields, since the correlation coefficient as a number roughly estimates the overall correlation, but it ignores the characteristics of the spatial distribution. In addition to the pattern correlation, the SOM approach is helpful in detecting how relevant the two spatial fields are in model evaluation.
Classification of CMIP6 models using the SOM method and pattern correlations in evaluating model capabilities in simulating 850-hPa wind fields over the WNP (5°–30°N, 100°–160°E; blue-outlined box in Figs. 3c,d). Models in each node are ordered according to the correlation of trends from high to low.
Citation: Journal of Climate 36, 15; 10.1175/JCLI-D-22-0395.1
We focus on the long-term trend, which provides an approximation of the response to anthropogenic forcing in the climate system (Wang and Wu 2018a). The projections are expressed as changes in three periods: 2020–39, 2040–59, and 2080–99 relative to the reference period of 1979–2014. The ensemble mean changes of the selected models in 850-hPa winds and relative vorticity in the SSP2-4.5 scenario feature easterly trends east of 120°E in the tropical WNP (Figs. 8a,c,e). These easterly trends and reduced relative vorticity weaken the MT with a slight northwestward withdrawal of the MT in the periods 2020–39 and 2040–59. The MT has no significant shift in 2080–99 since the easterly trends tend to be weakened in the end of the trough east of 140°E. With the rise in CO2 emissions in the SSP5-8.5 scenario, the easterly trends are mainly intensified west of 120°E and gradually weakened east of 120°E, even transforming into westerly trends east of 130°E along 10°N. This wind adjustment contributes to the eastward extension of the MT in 2040–59 and 2080–99, while it is not obviously seen in 2020–39 (Figs. 8b,d,f). These results suggest that atmospheric circulation variations in response to the strength of CO2 emissions may not comply with a relatively linear relationship. Responses to changes of atmospheric circulation characterize regional features in quasi-linear increases in emission scenarios rather than simply a uniform response. We also examined the MT changes in the multimodel ensemble means of all models and found similar patterns with weaker signals in the concerned region south of 20°N (Fig. 9).
Ensemble mean changes of 850-hPa winds (vectors; m s−1) and relative vorticity (shading; ×106 s−1) over the WNP in the future scenario of SSP2-4.5 in (a) 2020–39, (c) 2040–59, and (e) 2080–99 from the five selected models (ACCESS-ESM1-5, BCC-CSM2-MR, CIESM, CanESM5, and GFDL-ESM4). (b),(d),(f) As in (a), (c), and(e), but for the future scenario of SSP5-8.5. Black lines indicate that the zonal wind speed is equal to zero in the five-model ensemble means in the historical simulation, and red lines indicate the same for the future scenario in the specific periods. Stippling indicates grids where the five selected models agree on the sign of the change.
Citation: Journal of Climate 36, 15; 10.1175/JCLI-D-22-0395.1
As in Fig. 8, but for the multimodel ensemble means of all models. Stippling indicates grids where a majority (>70%) of the models agree on the sign of the change.
Citation: Journal of Climate 36, 15; 10.1175/JCLI-D-22-0395.1
To clarify temporal distributions of the MT, we use the MT location indices (see section 2b) to present migrations of the projected MT in the two SSP scenarios. The projected MT exhibits a slight northward migration in both the SSP2-4.5 and SSP5-8.5 scenarios, which indicates that global warming might induce a poleward shift of the MT. However, this shift is insignificant, with a weak northward trend of 0.06° decade−1 in the SSP5-8.5 and 0.08° decade−1 in the SSP2-4.5 while time series of the MT latitude oscillate on decadal to multidecadal scales (Fig. 10a). The longitude of the projected MT has a westward shift in the next 80 years in the SSP2-4.5 scenario, with a linear trend of –0.19° decade−1 (Fig. 10b). This westward contraction of the MT is intensified in the SSP5-8.5 scenario (–0.28° decade−1 in 2015–55) until the middle of the twenty-first century, while it converts to an eastward extension at the end of this century (0.34° decade−1). This result implies that the migration of the MT location might not vary linearly with increased emission scenarios, which is consistent with the 850-hPa wind changes over the WNP in the two scenarios.
Time series of the mean (a) latitude and (b) longitude of the MT in the future scenario SSP2-4.5 (black curves) and SSP5-8.5 (red curves) during 2015–2100 in the selected models. The solid curves correspond to the three-point running mean. Black dashed lines indicate the long-term linear trend of the mean location of the MT in the SSP2-4.5, and red dashed lines indicate the same for the SSP5-8.5. In (b), blue dashed lines indicate the shorter-term linear trend of the mean longitude of the MT in the SSP5-8.5 during 2015–55.
Citation: Journal of Climate 36, 15; 10.1175/JCLI-D-22-0395.1
6. Summary and discussion
CMIP6 models can commonly simulate the mean structure of the four large-scale circulation systems relevant to TC activity over the WNP, although with a slight intermodel bias near the boundary of the SAH and the TUTT at the upper level and the MT and the WNPSH at the lower level. The models exhibit more noticeable limitations in reproducing variations in the low-level circulation than in the upper levels. The mean latitude of simulated MT spreads between 10° and 20°N in each model, which is associated with the diversified simulated mean state of the east–west SST contrast in the tropical Pacific. Previous studies found that SST variability has a profound impact on tropical circulations (Nitta and Yamada 1989; Rasmusson and Carpenter 1982; Trenberth et al. 1998). The east–west Pacific SST contrast modifies the trade wind and, thus, the Walker circulation (Collins et al. 2010; Kociuba and Power 2015; Tokinaga et al. 2012; Vecchi and Soden 2007), which is associated with the east–west migration of the MT and further TC formation locations (Wang and Wu 2018b; Feng and Wu 2022). Current climate models still struggle to fully capture tropical SST variability (Deser et al. 2012a,b; Fasullo et al. 2020; Feng et al. 2021), which obstructs models’ performance in replicating atmospheric circulation over the WNP.
The performance of models’ historical simulations is generally assessed using a variety of metrics. Here, we propose an effective clustering method (the SOM) for assessing climate model performance. Other measures of model performance may be justified in specific applications. Considering the large intermodel biases in simulating low-level circulation related to the MT, we filter models to obtain relatively reliable models by using two widely used assessment methods (pattern correlation and RMSE) and an unsupervised machine-learning method (the SOM). The filtered models show that the MT has no significant zonal and poleward shifts in the SSP2-4.5 and SSP5-8.5 scenarios. The prediction in the selected models is consistent with the multimodel ensemble means, which exhibits an insignificant weak signal of MT shift in warming scenarios in this study. It is worth noticing whether the predicted future change is more reliable if the selected models have better performance in historical simulations, which needs to be further investigated in future studies.
In addition, this study primarily analyzes climate trends to investigate the effects of anthropogenic forcing on the TC-relevant, large-scale circulation, while the contribution of internal climate variability is also important in climate change projections. Deser et al. (2012a) demonstrated that the intrinsic atmospheric variability is a major source of uncertainty in simulating atmospheric circulation in the extratropics, and the coupled ocean–atmosphere variability plays a dominant role in the tropics. The internal variability, together with climate model simulation and the climate response to external forcing, are proposed as three key sources of uncertainty in future climate change projections (Tebaldi and Knutti 2007; Hawkins and Sutton 2009). Detecting and attributing observed climate change and multimodel assessments is a common way to characterize the uncertainty in climate change projections. In particular, the quantification of the uncertainty in predicting future changes in atmospheric and oceanic systems remains a fundamental issue and needs to be further explored.
Furthermore, the results of future changes in the WNP low-level circulation under relatively low and high CO2 emissions show that atmospheric circulation responses to the strength of CO2 emissions may characterize distinct spatial features on the regional scale rather than on the global scale. This might result from the response of atmospheric circulation to anthropogenic forcing overlaying internal variations in a certain region of space. More studies are required to investigate regional feedback and local effects on the influence of human activity.
Acknowledgments.
This research was supported by the National Natural Science Foundation of China (41730961, 42192551, and 42150710531). Author Feng was jointly supported by Shanghai Post-Doctoral Excellence Program (2021021) and Shanghai Typhoon Research Foundation (TFJJ202203).
Data availability statement.
The ERA5 reanalysis (https://www.ecmwf.int/) and Global SST data from the NOAA ERSSTv5 (https://www.ncei.noaa.gov/products/extended-reconstructed-sst) are available online. The CMIP6 simulations employed in this study are available from the Earth System Grid Federation (https://esgf-node.llnl.gov/search/cmip6/). The SOM package used in this study is developed by MATLAB.
REFERENCES
Camargo, S. J., 2013: Global and regional aspects of tropical cyclone activity in the CMIP5 models. J. Climate, 26, 9880–9902, https://doi.org/10.1175/JCLI-D-12-00549.1.
Camargo, S. J., and A. A. Wing, 2021: Increased tropical cyclone risk to coasts. Science, 371, 458–459, https://doi.org/10.1126/science.abg3651.
Chen, T.-C., S.-Y. Wang, and M.-C. Yen, 2006: Interannual variation of the tropical cyclone activity over the western North Pacific. J. Climate, 19, 5709–5720, https://doi.org/10.1175/JCLI3934.1.
Chia, H. H., and C. F. Ropelewski, 2002: The interannual variability in the genesis location of tropical cyclones in the northwest Pacific. J. Climate, 15, 2934–2944, https://doi.org/10.1175/1520-0442(2002)015<2934:TIVITG>2.0.CO;2.
Collins, M., and Coauthors, 2010: The impact of global warming on the tropical Pacific Ocean and El Niño. Nat. Geosci., 3, 391–397, https://doi.org/10.1038/ngeo868.
Deser, C., A. Phillips, V. Bourdette, and H. Teng, 2012a: Uncertainty in climate change projections: The role of internal variability. Climate Dyn., 38, 527–546, https://doi.org/10.1007/s00382-010-0977-x.
Deser, C., and Coauthors, 2012b: ENSO and Pacific decadal variability in the Community Climate System Model version 4. J. Climate, 25, 2622–2651, https://doi.org/10.1175/JCLI-D-11-00301.1.
Emanuel, K. A., 2013: Downscaling CMIP5 climate models shows increased tropical cyclone activity over the 21st century. Proc. Natl. Acad. Sci. USA, 110, 12 219–12 224, https://doi.org/10.1073/pnas.1301293110.
Emanuel, K. A., 2021: Response of global tropical cyclone activity to increasing CO2: Results from downscaling CMIP6 models. J. Climate, 34, 57–70, https://doi.org/10.1175/JCLI-D-20-0367.1.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Fasullo, J. T., A. S. Phillips, and C. Deser, 2020: Evaluation of leading modes of climate variability in the CMIP archives. J. Climate, 33, 5527–5545, https://doi.org/10.1175/JCLI-D-19-1024.1.
Feng, X., and L. Wu, 2022: Roles of interdecadal variability of the western North Pacific monsoon trough in shifting tropical cyclone formation. Climate Dyn., 58, 87–95, https://doi.org/10.1007/s00382-021-05891-w.
Feng, X., and Coauthors, 2021: A multidecadal-scale tropically driven global teleconnection over the past millennium and its recent strengthening. J. Climate, 34, 2549–2565, https://doi.org/10.1175/JCLI-D-20-0216.1.
Gill, A., 1980: Some simple solutions for heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc., 106, 447–462, https://doi.org/10.1002/qj.49710644905.
Hartmann, D. L., 2007: The atmospheric general circulation and its variability. J. Meteor. Soc. Japan, 85B, 123–143, https://doi.org/10.2151/jmsj.85B.123.
Hawkins, E., and R. Sutton, 2009: The potential to narrow uncertainty in regional climate predictions. Bull. Amer. Meteor. Soc., 90, 1095–1108, https://doi.org/10.1175/2009BAMS2607.1.
Herbst, M., H. V. Gupta, and M. C. Casper, 2009: Mapping model behaviour using self-organizing maps. Hydrol. Earth Syst. Sci., 13, 395–409, https://doi.org/10.5194/hess-13-395-2009.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Holland, G. J., 1995: Scale interaction in the western Pacific monsoon. Meteor. Atmos. Phys., 56, 57–79, https://doi.org/10.1007/BF01022521.
Huang, B., and Coauthors, 2017: Extended Reconstructed Sea Surface Temperature, version 5 (ERSSTv5): Upgrades, validations, and intercomparisons. J. Climate, 30, 8179–8205, https://doi.org/10.1175/JCLI-D-16-0836.1.
Huangfu, J., R. Huang, W. Chen, T. Feng, and L. Wu, 2017: Interdecadal variation of tropical cyclone genesis and its relationship to the monsoon trough over the western North Pacific. Int. J. Climatol., 37, 3587–3596, https://doi.org/10.1002/joc.4939.
Knutson, T., and Coauthors, 2019: Tropical cyclones and climate change assessment: Part I: Detection and attribution. Bull. Amer. Meteor. Soc., 100, 1987–2007, https://doi.org/10.1175/BAMS-D-18-0189.1.
Knutson, T., and Coauthors, 2020: Tropical cyclones and climate change assessment: Part II: Projected response to anthropogenic warming. Bull. Amer. Meteor. Soc., 101, E303–E322, https://doi.org/10.1175/BAMS-D-18-0194.1.
Kociuba, G., and S. B. Power, 2015: Inability of CMIP5 models to simulate recent strengthening of the Walker circulation: Implications for projections. J. Climate, 28, 20–35, https://doi.org/10.1175/JCLI-D-13-00752.1.
Kohonen, T., 1990: The self-organizing map. Proc. IEEE, 78, 1464–1480, https://doi.org/10.1109/5.58325.
Lockwood, J. W., M. Oppenheimer, N. Lin, R. E. Kopp, G. A. Vecchi, and A. Gori, 2022: Correlation between sea-level rise and aspects of future tropical cyclone activity in CMIP6 models. Earth’s Future, 10, e2021EF002462, https://doi.org/10.1029/2021EF002462.
Matsuura, T., M. Yumoto, and S. Iizuka, 2003: A mechanism of interdecadal variability of tropical cyclone activity over the western North Pacific. Climate Dyn., 21, 105–117, https://doi.org/10.1007/s00382-003-0327-3.
Molinari, J., and D. Vollaro, 2013: What percentage of western North Pacific tropical cyclones form within the monsoon trough? Mon. Wea. Rev., 141, 499–505, https://doi.org/10.1175/MWR-D-12-00165.1.
Murakami, H., and B. Wang, 2022: Patterns and frequency of projected future tropical cyclone genesis are governed by dynamic effects. Commun. Earth Environ., 3, 77, https://doi.org/10.1038/s43247-022-00410-z.
Nitta, T., and S. Yamada, 1989: Recent warming of tropical sea surface temperature and its relationship to the Northern Hemisphere circulation. J. Meteor. Soc. Japan, 67, 375–383, https://doi.org/10.2151/jmsj1965.67.3_375.
Rasmusson, E. M., and T. H. Carpenter, 1982: Variations in tropical sea surface temperature and surface wind fields associated with the Southern Oscillation/El Niño. Mon. Wea. Rev., 110, 354–384, https://doi.org/10.1175/1520-0493(1982)110<0354:VITSST>2.0.CO;2.
Reusser, D. E., T. Blume, B. Schaefli, and E. Zehe, 2009: Analysing the temporal dynamics of model performance for hydrological models. Hydrol. Earth Syst. Sci., 13, 999–1018, https://doi.org/10.5194/hess-13-999-2009.
Roberts, M., and Coauthors, 2020: Projected future changes in tropical cyclones using the CMIP6 HighResMIP multimodel ensemble. Geophys. Res. Lett., 47, e2020GL088662, https://doi.org/10.1029/2020GL088662.
Simpson, R. H., N. Frank, D. Shideler, and H. M. Johnson, 1968: Atlantic tropical disturbances, 1967. Mon. Wea. Rev., 96, 251–259, https://doi.org/10.1175/1520-0493(1968)096<0251:ATD>2.0.CO;2.
Tebaldi, C., and R. Knutti, 2007: The use of the multimodel ensemble in probabilistic climate projections. Philos. Trans. Roy. Soc., A365, 2053–2075, https://doi.org/10.1098/rsta.2007.2076.
Tokinaga, H., S.-P. Xie, C. Deser, Y. Kosaka, and Y. M. Okumura, 2012: Slowdown of the Walker circulation driven by tropical Indo-Pacific warming. Nature, 491, 439–443, https://doi.org/10.1038/nature11576.
Tory, K. J., S. S. Chand, J. L. McBride, H. Ye, and R. A. Dare, 2013: Projected changes in late-twenty-first century tropical cyclone frequency in 13 coupled climate models from phase 5 of the Coupled Model Intercomparison Project. J. Climate, 26, 9946–9959, https://doi.org/10.1175/JCLI-D-13-00010.1.
Trenberth, K. E., G. W. Branstator, D. Karoly, A. Kumar, N.-C. Lau, and C. Ropelewski, 1998: Progress during TOGA in understanding and modeling global teleconnections associated with tropical sea surface temperatures. J. Geophys. Res., 103, 14 291–14 324, https://doi.org/10.1029/97JC01444.
Vecchi, G. A., and B. J. Soden, 2007: Effect of remote sea surface temperature change on tropical cyclone potential intensity. Nature, 450, 1066–1070, https://doi.org/10.1038/nature06423.
Vecchi, G. A., and Coauthors, 2019: Tropical cyclone sensitivities to CO2 doubling: Roles of atmospheric resolution, synoptic variability and background climate changes. Climate Dyn., 53, 5999–6033, https://doi.org/10.1007/s00382-019-04913-y.
Wang, B., and J. C. L. Chan, 2002: How strong ENSO events affect tropical storm activity over the western North Pacific. J. Climate, 15, 1643–1658, https://doi.org/10.1175/1520-0442(2002)015<1643:HSEEAT>2.0.CO;2.
Wang, C., and L. Wu, 2018a: Projection of North Pacific tropical upper-tropospheric trough in CMIP5 models: Implications for changes in tropical cyclone formation locations. J. Climate, 31, 761–774, https://doi.org/10.1175/JCLI-D-17-0292.1.
Wang, C., and L. Wu, 2018b: Future changes of the monsoon trough: Sensitivity to sea surface temperature gradient and implications for tropical cyclone activity. Earth’s Future, 6, 919–936, https://doi.org/10.1029/2018EF000858.
Webster, F., 1961: The effect of meanders on the kinetic energy balance of the Gulf Stream. Tellus, 13A, 392–401, https://doi.org/10.3402/tellusa.v13i3.9515.
Wu, L., Z. Wen, R. Huang, and R. Wu, 2012: Possible linkage between the monsoon trough variability and the tropical cyclone activity over the western North Pacific. Mon. Wea. Rev., 140, 140–150, https://doi.org/10.1175/MWR-D-11-00078.1.
Wu, L., and Coauthors, 2014: Simulations of the present and late-twenty-first-century western North Pacific tropical cyclone activity using a regional model. J. Climate, 27, 3405–3424, https://doi.org/10.1175/JCLI-D-12-00830.1.
Wu, L., H. Zhao, C. Wang, J. Cao, and J. Liang, 2022: Understanding of the effect of climate change on tropical cyclone intensity: A review. Adv. Atmos. Sci., 39, 205–221, https://doi.org/10.1007/s00376-021-1026-x.
Yumoto, M., and T. Matsuura, 2001: Interdecadal variability of tropical cyclone activity in the western North Pacific. J. Meteor. Soc. Japan, 79, 23–35, https://doi.org/10.2151/jmsj.79.23.
Zhang, C., and Y. Wang, 2017: Projected future changes of tropical cyclone activity over the western North and South Pacific in a 20-km-mesh regional climate model. J. Climate, 30, 5923–5941, https://doi.org/10.1175/JCLI-D-16-0597.1.
Zhang, G., H. Murakami, T. R. Knutson, R. Mizuta, and K. Yoshida, 2020: Tropical cyclone motion in a changing climate. Sci. Adv., 6, eaaz7610, https://doi.org/10.1126/sciadv.aaz7610.
Zong, H., and L. Wu, 2015: Synoptic-scale influences on tropical cyclone formation within the western North Pacific monsoon trough. Mon. Wea. Rev., 143, 3421–3433, https://doi.org/10.1175/MWR-D-14-00321.1.
