Increased tropical cyclone intensification time in the western North Pacific over the past 56 years

It has been projected that the influence of anthropogenic climate change on tropical cyclone (TC) intensity could be detected by the end of the century although significant increasing trends in TC intensity metrics have been found based on the currently available historic records. The human influences on TC intensity have been debated for about two decades because of a lack of quantitative assessment of the contributions of large-scale environmental factors and track shifting. As an extension of a previous study, we show that the observed rise in the percentage of intense TCs in the western North Pacific basin over the past 56 years resulted from the combined influence of the track shifting and temporary changes in environmental factors. The influence of environmental factors was primarily owing to the decrease of environmental vertical wind shear and the warming of sea surface temperature (SST). While a small part of the observed rise in the percentage of intense TCs resulted from SST warming, the track shifting increased the TC intensification time by 18.2% (11.3 h) over the past 56 years, accounting for more than one-third of the observed percentage increase in intense TCs. Since track shifting is also projected in the global warming experiments, this study suggests that anthropogenic climate change may intensify TCs by shifting TC prevailing tracks.


Introduction
It has been projected that the influence of anthropogenic climate change on tropical cyclone (TC) intensity could be detected by the end of the century although a significant increasing trend has been found based on currently available historic records (Bender et al 2010, Sobel et al 2016, Knutson et al 2019. Many studies have argued that the observed intensification of TCs resulted from uncertainty in the historic records and/or multidecadal natural variability (Chan 2006, Landsea et al 2006, Wu et al 2006, Klotzbach and Landsea 2015. For this reason, the latest Intergovernmental Panel on Climate Change (IPCC) report indicates that we have relatively low confidence in the observed influence of climate change on TC intensity (IPCC 2021). A key to reconciling the inconsistency between the projection of theory and modeling and the observation is to understand what has contributed to the observed TC intensification (Knutson et al 2019, IPCC 2021, Wu et al 2022. Current studies on the impact of anthropogenic climate change are based primarily on our understanding of how large-scale environmental factors affect TC intensity, including sea surface temperature (SST), vertical wind shear (VWS), and SST cooling induced by TC-ocean interaction (Gary 1968, Price 1981, Emanuel 1987, DeMaria and Kaplan 1994. However, relatively less attention has been paid to two outstanding issues. First, the impact of anthropogenic climate change on TC intensity is assessed on the global and basin-wide scales. In other words, we should focus on physical factors affecting the global and basin-wide TC intensity rather than a single TC. Taking TC activity as a whole, the basin-wide TC intensity can be affected by changes in TC prevailing tracks due to the non-uniform spatial distribution and change in the intensification duration of TCs Camargo 2009, Wang andWu 2015). Second, most studies detected the trends in TC intensity and qualitatively compared them with the temporary changes in large-scale environmental factors (e.g. Emanuel 2005, Webster et al 2005, Elsner et al 2008, Kossin et al 2013, Mei and Xie 2016. The lack of quantifying the contributions of individual large-scale environmental factors hinders further understanding of the impact of anthropogenic climate change on TC intensity. Wu et al (2018) quantitatively assess the contributions of individual environmental factors and TC track shifting by using a simple TC intensity model adopted from Emanuel et al (2008). In addition to SST and VWS, the influences of TC-induced SST cooling and the change of the TC outflow temperature (OFT) are also considered. They found that the percentage increase of categories 4 and 5 TCs (hereafter intense TC, or ITC) was consistent with changes in TC tracks and atmosphere/ocean environmental parameters during 1980-2015. They further suggested that the increased percentage of ITCs was mainly due to the ocean mixed layer depth (MLD) change due to the temporary change of the MLD and TC track shifting. Since the percentage increase of ITCs in Wu et al (2018) was discussed over a relatively short period, their results may be a result of multidecadal natural variability (Chan 2006).
This study is an extension of Wu et al (2018) by examining the percentage change of ITCs over the period 1965-2020. We also examined the percentage change of ITCs over the period 1958-2020 and found that the results are very similar to those over the period 1965-2020. Here we only report the results over the period with the satellite monitoring of TC tracks. The model and experimental design are generally the same as in Wu et al (2018). Together with the datasets used in this study, the model and experimental design are briefly described in section 2. In sections 3 and 4, we demonstrate the physical consistency of the observed TC intensity change with changes in the environmental factors and TC track shifting and their contributions, respectively, followed by a summary in section 5.

Data and numerical experiments
In this study, we utilized the best-track data from the Joint Typhoon Warning Center (JTWC) from the International Best Track Archive for Climate Stewardship (IBTrACS, v04r00). Following Emanuel (2005), the pre-1973 records of maximum wind speed were adjusted. Figure 1 compares the July-September TC frequency between the JTWC best track data and the adjusted data during 1958-2020. The July-September frequency was reduced before 1973 due to the adjustment. Considering that the historic track data were relatively reliable in the satellite period, our analysis focused on the period 1965-2020.
The VWS between 850 hPa and 200 hPa was computed as an ensemble mean of the three reanalysis datasets: (1) the fifth-generation European Centre for Medium-Range Weather Forecasts reanalysis (ERA5) (Hersbach et al 2020), (2) the National Centers for Environmental Prediction/National Center for Atmospheric Research Reanalysis (NCEP/NCAR) (Kalnay et al 1996), and (3) the Japanese 55 year Reanalysis (JRA55) (Kobayashi et al 2015). The TC OFT is the tropopause temperature in the NCEP/NCAR reanalysis, and the SST data are from the NOAA extended reconstructed SST (ERSST version (5) data (Huang et al 2017)). The ocean mixed layer refers to where the temperature change is 0.5 • C less than the SST (e.g. Kelly and Qiu 1995), and the ocean MLD was calculated from the ocean reanalysis dataset of the Met Office Hadley Centre (EN.4.2.1, Good et al (2013)).
The future changes in environmental steering flow are based on a selection of relatively reliable models from the Coupled Model Intercomparison Project Phase 6 (CMIP6). We compared the model performance of the 34 CMIP6 models in simulating the climatology mean state and linear trend of the steering flow in the historical runs. The pattern correlation and root mean square error (RMSE) with the observed counterparts are considered during 1979-2014 over the Western North Pacific (WNP) (10 • -30 • N, 110 • -150 • E). The top 50% models with both a larger pattern correlation and smaller RMSE were selected. The selected seven models are ACCESS-ESM1-5, BCC-CSM2-MR, CanESM5, FGOALS-g3, GFDL-ESM4, KACE-1-0-G, and KIOST-ESM. The projected wind fields under the RCP8.5 scenario were used to derive the large-scale environmental steering flow in this study. The linear trends are detected with the least square method and their significance is examined with F-test.
The numerical simulations were conducted with the highly idealized models, which are based on our understanding of the influences of large-scale environmental factors on TC intensity and movement.
It is an axisymmetric atmospheric model coupled with a simple one-dimensional ocean model, in which the effect of VWS is parameterized (Emanuel et al 2008). In Wu et al (2018), the model was modified to include the change of MLD. In the intensity model, the influences of the environmental parameters of VWS, OFT, SST and MLD can be simulated. All the simulations are initialized with the same Figure 1. Comparison of the July-September frequency of ITCs between the simulation and the JTWC best track data during 1958-2020. Control experiment: Same as E1, but the parameters averaged over 1965-1992 and 1993-2020 are used for the corresponding periods.

E3
The overall effect of environmental parameters: Same as E2, but the observed tracks during 1965-1992 are used for the two periods. E4 The overall effect of environmental parameters: Same as E2, but the observed tracks during 1993-2020 are used for the two periods. E5 The overall effect of environmental parameters: Same as E2, but the observed tracks during 1965-2020 are used for the two periods. E6 Track effect: Same as E2, but July-September mean environmental parameters over 1965-2020 are used for the two periods. SE1 Individual effects of environmental parameters: Same as E5, but one parameter is fixed during the two periods. SE2 Individual effects of environmental parameters: Same as E5, but only one parameter is allowed to vary during the two periods.
warm-core cyclonic vortex and integrated along the observed TC tracks. The simulated TCs are affected by four environmental parameters: SST, MLD, VWS, and OFT. Table 1 describes the numerical experiments for TC intensity simulations. The track model was adopted from Wang and Wu (2015). It is used to simulate the TC track shifting during the period 1965-2020 and in the future. In the track model, TC movement is controlled by climatologic mean β-drift and large-scale steering flow. The numerical experiments for track simulations are described in section 4.

Physical consistency of the observed TC intensity
The observed formation location is where a TC first reached the tropical storm intensity. Figures 2(a) and (b) compare the frequency and percentage of ITCs between the observation and the simulation in E1 (table 1). The adjusted frequency and percentage of ITCs are consistent with the simulation (figures 2(a) and (b)). As indicated by the linear trend over the past 56 years, the frequency of the observed ITCs

Temporal changes in environmental factors
By considering the presence of the linear trend in the percentage of ITCs, we divided the 56 years equally into two sub-periods: 1965-1992 and 1993-2020 and performed E2 (table 1). In E2, the monthly mean environmental parameters within the subperiods (1965-1992 and 1993-2020) are averaged over the corresponding subperiods. That is, the environmental parameters are not allowed to vary with time within the corresponding subperiods. The percentage difference between the two subperiods is 11.3% from 16.8% to 28.1% in E2 (figure 3), comparable to 9.4% in the observation. The correlations between the simulation in E2 and the observation are 0.73 for the frequency and 0.77 for the percentage of ITCs. The influences of the environmental parameters and TC track shifting were quantified based on the experiment. The overall effect of environmental parameters is first distinguished from that of track shifting. Two sensitivity experiments (E3 and E4) are conducted by using the same track data in the two subperiods. That is, the observed TC tracks in the first (second) subperiod are also used for the second (first) subperiod in E3 (E4). The overall effect of environmental parameters can be estimated with the percentage difference between the first and second subperiods, which is 6.8% in E3 and 6.5% in E4. To examine the sensitivity of the sample size, an additional sensitivity experiment (E5) was also conducted by using all the observed TC tracks (797 TCs) in the two subperiods. The overall effect of environmental parameters in E5 is 6.6%. Based on the results of E3, E4, and E5, it can be estimated that the temporary change of environmental factors accounts for a percentage increase of about 7%, accounting for more than half of the simulated increase (11.3%) in E2.  Then the influences of individual environmental factors are further examined by performing two sets of sensitivity experiments (table 1). The first set (SE1) is the same as E2, but one of the four environmental parameters in the second subperiod is the same as that in the first subperiod, meaning that the influence of the fixed factor is removed. In this case, the smaller (larger) percentage difference between the two subperiods indicates the larger (smaller) influence of the fixed parameter. In SE1, the percentage differences are 4.3% for VWS, 4.8% for SST, 5.9% for OFT, and 6.5% for MLD, respectively (table 2). Therefore, it is indicated that decreasing VWS and warming SST are important to the increasing percentage of ITCs.
The second set of experiments (SE2) is also the same as E2, but only one of the four environmental factors is allowed to vary. In SE2, the larger (smaller) percentage difference between the two subperiods indicates the larger (smaller) contribution of the varied parameter. The percentage difference is 3.8% for VWS, 3.5% for SST, 1.9% for OFT, and 1.0% for MLD. In agreement with SE1, decreasing VWS and warming SST are two important factors for the increasing percentage of ITCs.
Based on SE2, we can roughly estimate the contribution of SST warming, which is about one-third of the contribution of the temporal changes of environmental parameters and about 2.4% of the total increase of 11.3%. It is suggested that the overall contribution of the temporary changes in the environmental parameters results mainly from the changes in VWS and SST, and the SST warming only accounts for a small portion of the observed increase in the proportion of ITCs, in agreement with theory and modeling (Knutson and Tuleya 2004, Bender et al 2010, Knutson et al 2019. Figure 4 displays the climatological mean SST and VWS and their linear trends over 1965-2020, as well as the projection under the RCP8.5 emission scenario. The observed SST warmed in the entire basin ( figure 4(a)), and the SST in the high RCP8.5 emission scenario is also projected to warm in the entire basin ( figure 4(b)). Previous studies have suggested that anthropogenic climate change can warm the SST in the WNP basin (e.g. Xie et al 2010). Despite the relatively small influence, it is suggested that climate change has increased the proportion of ITCs in the WNP basin through SST warming. However, the spatial distribution of the observed VWS change is inconsistent with the projection in the RCP8.5 emission scenario ( figure 4(d)). The influence of climate change on VWS in the selected CMIP6 models may not be correctly simulated in the experiments or the observed change in VWS may result from natural variability.

Track shifting
The influence of track shifting can be estimated by comparing the percentage change of the same subperiod in E3 and E4, which is 4.8% for the first subperiod and 4.5% for the second subperiod. An additional experiment (E6) was also designed to examine the influence of track shifting. The environmental parameters in E6 were averaged over the period 1965-2020 in the two sub-periods. The influence of track shifting is 4.3% in the percentage increase. Based on the results of E3, E4, and E6, we can find that the track shifting contributed to more than one-third observed increase in the percentage of ITCs.
The track shifting can be seen in terms of the frequency of TC occurrence, which is counted with a grid box of 2.5 • latitude by 2.5 • longitude during the peak TC season. Three prevailing tracks over the WNP can be identified with the relatively large magnitude of the accumulated frequency of TC occurrence ( figure 5(a)). One is the westward prevailing track to the South China Sea, while the other two are the northwestward prevailing tracks toward East Asia and the recurving prevailing track east of 130 • E, respectively. Figure 5(a) also shows the shifting of TC prevailing tracks in terms of the change of the frequency of TC occurrence between the two subperiods. The two positive regions west of the two prevailing tracks suggest the westward shift of the prevailing tracks over the WNP, while the TC influence in the South China Sea was reduced.
To understand how track changes influence TC intensity on the basin-wide scale, the changes in  Figure 6 shows the time series of the July-September mean duration of intensification, traveling distance, and latitude of lifetime maximum intensity (LMI) of TCs. The intensification time and traveling distance of a TC refer to the duration and distance from the initial formation time and location and the time and location at which it reaches its LMI. Since almost all ITCs formed to the east of the Philippines islands, our calculation did not include TCs formed in the South China Sea. Figure 6 presents the significant linear trends in the TC intensification time (p = 0.08) and traveling distance (p = 0.06). The trend in the latitude of LMI is also in accordance with previous studies (Kossin et al 2014, Zhan and Wang 2017, Moon et al 2020 although it is insignificant (p = 0.24). Based on the linear trends, the traveling distance increased by 245.2 km from 1093.9 km to 1339.1 km from 1965 to 2020. The intensification time increased by 11.3 h from 62.1 h to 73.4 h, an increase of 18.2%. It is indicated that the track shifting over the WNP allowed TCs to have a longer time to reach their LMI, leading to an increase in the percentage of ITCs in the WNP basin. Figure 5 also displays the projected track changes under global warming over the periods 2020-2039, 2040-2069, and 2080-2099. Note that the projected track changes result only from changes in the environmental steering while the formation locations are the same as the observation during 1965-2020. The changes in the steering flow and the frequency of TC occurrence were derived from the simulations in the RCP8.5 emission scenario. Compared with the period 1965-2020, the projected track shifting over the WNP is similar to the observed shifting. It is implied that the observed track shifting is linked to climate change although further study is needed to understand factors for the observed TC track shifting. Note that similar track shifting has been found in previous observational analyses (Wu et al 2005, Figure 6. (a) The July-September mean duration between the formation and reaching lifetime maximum intensity (hour), (b) the July-September mean traveling distance (km), and the July-September mean latitude of lifetime maximum intensity ( • N). 2018) and the projection of climate change experiments (Colbert et al 2015, Wang andWu 2015).

Summary
To understand the influence of anthropogenic climate change on TC intensity in the WNP basin, we extended the study of Wu et al (2018) by covering the period 1965-2020 and quantified the individual contributions of environmental factors and track shifting. We demonstrate that the observed increases in the frequency and percentage of ITCs are physically consistent with changes in the large-scale environmental factors and track shifting over the period 1965-2020. The observed increase in the percentage of ITCs resulted from the combined influence of the temporary change of environmental factors and track shifting.
The temporary change of environmental parameters contributed to about half of the observed increase in the percentage of intense TCs, which resulted mainly from the changes in VWS and SST. The SST warming alone can account for a small portion of the observed increase in the percentage of ITCs, in agreement with previous studies (Bender et al 2010, Sobel et al 2016, Knutson et al 2019. The track shifting over the WNP contributed to more than one-third of the observed increase in the percentage of ITCs. The track shifting allowed TCs to have more time for intensification. The intensification time increased by 18.2% (11.3 h) over the past 56 years.
The results in this study are different from Wu et al (2018) in the contributions to the percentage of ITCs. Wu et al (2018) found that the increased ocean MLD played an important role in increasing the proportion of intense TCs during the period 1980-2015. Consistent with Wu et al (2018), Zhao et al (2018) also found that the increase of rapidly intensifying TCs was related to the increase of the ocean heat content after 1998. Gao et al (2020) argued that global warming amplified the ocean heat content in the WNP basin. Note that these studies have primarily concentrated on a relatively limited time frame. However, as we expand the scope of our analysis period, our study suggests that the change in MLD does not hold significant influence on the increase of the percentage of ITCs.
No new data were created or analysed in this study.