1. Introduction
Tropical cyclone genesis (TCG) is a critical scientific issue that warrants further clarification. The issue is particularly complicated in the western North Pacific (WNP), where, among all basins, tropical cyclones (TCs) are the most active and are influenced by the interplay between the monsoon trough and subtropical anticyclone in the lower troposphere, which affects the moisture convergence and content, vertical wind shear (VWS), convection, and even sea surface temperature (SST). The large-scale environmental background provides the conditions necessary for TCG. In their pioneering study, Gray (1968) proposed the following necessary conditions for TCG: sufficiently high SST, conditional instability, low-level convergence and vorticity, high midlevel relative humidity (RH), high-level divergence, and weak VWS. Since then, many criteria based on this idea have been proposed, and the most well-known is that of the SST threshold being >26.5°C. Researches have suggested that the threshold SST is not fixed but rather varies among geographical and climatic background conditions (Emanuel et al. 2008; Zhao et al. 2009; Knutson et al. 2010; Merlis et al. 2013, 2016). Nevertheless, by using advanced satellite observations, researchers have confirmed that an SST of 26.5°C is a condition necessary for TCG (Dare and McBride 2011; Tory and Dare 2015). The genesis potential index (GPI) and its revised version proposed by Emanuel and collaborators (e.g., Emanuel and Nolan 2004), where SST was replaced with potential intensity, has been an indicator widely used for TCG estimation. Environmental parameters, such as low-level vorticity, midlevel RH, and deep-layer VWS, have continued to be the critical criteria in related studies (McBride and Zehr 1981; Goldenberg and Shapiro 1996; Ritchie and Holland 1999; DeMaria et al. 2001; Cheung and Elsberry 2002; Camargo et al. 2009; Lee et al. 2008; Huang et al. 2011; Jiang et al. 2012a; Kerns and Chen 2013; Zhao et al. 2015a,b; Teng et al. 2020). Fluctuations in large-scale circulations [e.g., the monsoon trough and gyre, easterly wave, El Niño–Southern Oscillation (ENSO), and boreal summer intraseasonal oscillation (BSISO)] can modulate the aforementioned environmental conditions for TCG even though the relative degree and time scale of their influence may vary. For instance, TCG is more likely to occur in an active monsoon trough environment than in an easterly environment (Ooyama 1982; Chen et al. 2008; Lin and Lee 2011; Hsieh et al. 2017; Teng et al. 2019, 2020). ENSO and BSISO phases modulate the interannual and intraseasonal variations in TCG locations and TC activities in the WNP (Wang and Chan 2002; Sobel and Camargo 2005; Camargo et al. 2007; Hsu et al. 2008; Hsu 2012; Li et al. 2012; Nakano et al. 2015; Zhao et al. 2015a,b, 2018; Weng and Hsu 2017; Chen et al. 2018; Kikuchi 2021).
Although the accuracy of numerical forecasting of TC behavior has increased considerably, numerical forecasting of TCG remains challenging. To increase the current knowledge regarding numerical forecasting of TCG, high-resolution global numerical model (e.g., less than 25 km) output is a crucial reference source (Cossuth et al. 2013; Halperin et al. 2017, 2020; Yamaguchi and Koide 2017; Tsai and Elsberry 2019). Several studies have demonstrated the benefits of higher model spatial resolution and improved physical parameterization for TC simulation and revealed the remaining challenges, particularly the long lead forecast time (Briegel and Frank 1997; Cheung and Elsberry 2002; Pratt and Evans 2009; Tsai et al. 2011; Nakano et al. 2017; Halperin et al. 2016, 2020).
Studies have also demonstrated the adeptness of the Finite Volume Cubed-Sphere Dynamical Core Global Forecast System (fvGFS) in predicting and simulating the structural characteristics, intensity, tracking, and genesis of TCs (Chen et al. 2019; Hazelton et al. 2018a,b, 2020; Huang et al. 2020, 2022; Gao et al. 2021; Zhang et al. 2021). The existing TCG analysis mostly focused on comparing the pros and cons of the models, whereas the impact of environmental conditions on the TCG prediction ability of fvGFS has not been deeply discussed. This study aimed to fill the aforementioned knowledge gap by conducting a series of hindcast experiments to evaluate the TCG forecasting ability of the fvGFS in the WNP during 2018–19 and to identify the large-scale environmental circulations most favorable for TCG. In this paper, section 2 describes the fvGFS, experimental design, diagnostic methodology, and data used in this study; section 3 presents the results of the fvGFS forecasting and environmental field analyses; and sections 4 and 5 present the discussion and conclusion of this study, respectively.
2. Data and methodology
The fvGFS is a Next-Generation Nonhydrostatic Global Prediction System (NGGPS) that replaced the global forecast system previously used for the routine weather forecast operation of the National Centers for Environmental Prediction (NCEP). The dynamic core developed at the NOAA/Geophysical Fluid Dynamics Laboratory (GFDL) has been adopted and combined with the existing GFS physical parameterization schemes. The model version used in this study was the 201806b version of the fvGFS (i.e., the 2018 version of the SHiELD model listed in Table 1 of Harris et al. 2020), which was obtained through collaboration with GFDL. Its configuration was C768 yielding a horizontal resolution of approximately 13 km, as well as 91 vertical layers with the model top at 0.64 hPa.
The initial data used in the hindcast experiments were retrieved from the initialized analysis produced by the NCEP fvGFS daily forecast, with a horizontal resolution of 0.25°. The observed TC information, including genesis time, central location, and maximum intensity [i.e., minimum sea level pressure (mSLP)], were obtained from the Japan Meteorological Agency Regional Specialized Meteorological Center (JMA-RSMC). To analyze the observed background circulation conditions of TCG, we used the real-time analysis data of the NCEP Climate Forecast System Version 2 (CFSv2; Saha et al. 2014). The CFSv2 is a fully coupled model representing the interaction between Earth’s atmosphere, oceans, land, and sea ice. The model became operational in March 2011, replacing the CFSv1 used in creating the Climate Forecast System Reanalysis (CFSR; Saha et al. 2010). For details of CFSv2, we refer readers to Saha et al. (2014). Although CFSv2 provides both analysis data and forecast products, only the analysis data (including meridional and zonal winds, RH, and SST with a horizontal resolution of 0.5°) used to validate the hindcasted large-scale background circulation by the fvGFS. Both CFSv2 data and fvGFS hindcasted results were regridded to 1° × 1° resolution for comparison.
Hindcast experiments were performed for all 42 WNP TCG events during 2018–19. The time of TCG detected by JMA-RSMC was assigned as the reference time (t = 0 h). The control (CTL) experiments were initiated at 5 days (Lead-5d) before TCG (t = −120 h). Sensitivity experiments initiated at 4 days (Lead-4d, t = −96 h) and 3 days (Lead-3d, t = −72 h) before TCG were also performed to evaluate the hindcast sensitivity to forecast lead time. All experiments were terminated 2 days after TCG. The fvGFS was coupled to a slab ocean model between 30°N and 30°S. Uncoupled experiments were also performed to assess whether the air-sea coupling could affect the TCC development before TCG since the coupling has been suggested to affect TC strength. The results showed nonsignificant contribution of the coupling (data not shown). Here, we present only the results of coupled experiments.
TCG detection in hindcast experiments was conducted within a ±48-h time window and a 10° × 10° area surrounding the observed TCG. The following conditions were adopted to detect TCG in the hindcast experiments (Vitart et al. 1997; Vitart and Stockdale 2001; Walsh et al. 2007):
-
The maximum relative vorticity at 850 hPa [ξmax(850)] is higher than 8 × 10−5 s−1 in the detection windows.
-
The mSLP within 2° of ξmax(850) is designated as the center of convective disturbance, and the increased mSLP is >5 hPa within the latitude and longitude grid 5° outward from the center.
-
The distance between the center of maximum thickness and mSLP is <2°, and the thickness should be reduced by at least 50 m within 5° of latitude and longitude outward from the center.
Tracking of TC was conducted according to these three criteria:
-
The maximum wind speed (Vmax) is >17.5 m s−1 in the detection windows.
-
The location of mSLP < 1000 hPa in the detection windows is defined by the center of convective disturbance (or the pressure rise reaches 8 hPa within 8°).
-
The distance between mSLP and Vmax is <2°.
A TCG prediction was considered successful when conditions 1–3 were fulfilled for >6 h within a 12-h period.
3. Results
a. Hindcast hit rate evaluation
To evaluate whether the fvGFS could effectively forecast TCG, we selected 42 WNP TCG events during 2018–19. Furthermore, 7-day hindcasts were initiated 5, 4, and 3 days before TCG (referred to as Lead-5d, Lead-4d, and Lead-3d, respectively). The hit probability from these sensitivity experiments was calculated to assess the ability of the fvGFS in TCG simulation. As shown in Fig. 1 and Table S1 in the online supplemental material, Lead-5d hindcasts had a 50% hit rate with 21 successes (“Hit” cases) and 21 failures (“Fail” cases). Among the Hit cases, 12 were simulated on average 19 h ahead and 530 km distant from the observations, whereas 7 were simulated on average 18 h behind and 498 km distant from the observations. In two other Hit cases, the genesis time was consistent with that in the observations. The hindcast performance was particularly poor during August–September 2018 and after October 2019; in other words, most of the Fail cases occurred in these periods. This clustering phenomenon was due to the dominance of a specific background circulation type, which will be discussed later in sections 3d and 4.
(top) Time and (bottom) spatial errors of the fvGFS hindcasts of 42 WNP tropical cyclones in 2018–19. Green, blue, and yellow bars represent the initialization at lead times of 5, 4, and 3 days before TCG (Lead-5d, Lead-4d, and Lead-3d), respectively. “Hit” cases mark the successfully hindcasted TCG events in colors, whereas those were not successfully hindcasted (“Fail” cases) were left blank.
Citation: Weather and Forecasting 38, 11; 10.1175/WAF-D-23-0056.1
When the forecast lead time was shortened to 4 days, the hindcast improved significantly, resulting in 32 Hit and 10 Fail cases (hit rate = 76%). However, no further improvement was noted in Lead-3d experiments (Fig. 1 and Table S1). These lead-time sensitivity test results suggested that the fvGFS hindcasts significantly degraded the lead time from 4 to 5 days. The reason underlying this rapid degradation, however, remains unknown. Nevertheless, the result may be case and model dependent. The existence of lead-time dependence warrants further investigation. Although a shorter lead time tended to yield a higher hit rate, whether an improvement in genesis time and location errors results in a shorter lead time remains unclear. The fine-scale, stochastic processes leading to TCG [e.g., the tropical cloud cluster (TCC) or the mesoscale convective systems (MCSs) serving as the so-called TC seeds (e.g., Ikehata and Satoh 2021)] could not be well-resolved in the fvGFS even with a 13-km resolution model (Hsieh et al. 2022; Yang et al. 2021; Yamada et al. 2021; Emanuel 2022).
Figure 2 shows the geographic location of the observed and hindcasted TCG events for three lead times. The locations of the observed TCG events could be roughly divided into two regions relative to 135°E: one was in the far WNP close to the Luzon Island, where TCG events occurred in a relatively small area, whereas the other, a much larger one was between 150°–165°E and 5°–20°N, where the TCG events were scattered widely. The hindcast results of Lead-5d demonstrated that most of the Fail cases were located in the eastern region (orange digits in Fig. 2a). The numbers of Fail cases in the eastern and western regions were 14 (67%) and 7 (33%), respectively. This difference may be attributable to the characteristics of the large-scale background flow (e.g., more active monsoon trough in the west than east of 135°N). This separation was not as clear in Lead-4d and Lead-3d hindcasts. The results suggested that better-structured, large-scale circulation can facilitate forecasting with a longer lead time; however, this advantage may be lost with a shorter lead time when the model can better capture the details required for effective simulation. This speculation appears to be consistent with the characteristics of background flows, which are discussed in later sections.
Spatial distribution of tropical cyclone genesis locations between the hindcasts and observations for the (a) Lead-5d, (b) Lead-4d, and (c) Lead-3d experiments. Blue and black numbers denote Hit case locations in hindcasts and observations, respectively, whereas brown numbers indicate Fail case locations in the hindcasts. The numbers in parentheses indicate the numbers of the Hit and Fail cases.
Citation: Weather and Forecasting 38, 11; 10.1175/WAF-D-23-0056.1
b. Forecast error analysis
Figure 3 shows the position of Hit cases relative to observed TCG location (i.e., the centers of the plots), with time biases (h) indicated in colored numbers (0–12 h, red; 12–24 h, green; 24–36 h, blue; 36–48 h, gray). The results indicate that the temporal biases of nearly half of the cases (11/21) in Lead-5d were within ±12 h (Fig. 3a); however, those of the remaining four, five, and one case were within ±24, ±36, and ±48 h, respectively. The average temporal bias in Lead-5d was −4.7 h (RMSE = 21.1 h). More than half (14/21) of Hit cases had a spatial bias within 5°, whereas the remaining seven cases had a spatial bias outside 5° (Fig. 1). The average spatial bias was 514.5 km (RMSE = 570.0 km). The temporal and spatial biases did not appear to be correlated. In other words, smaller and larger spatial biases were not necessarily accompanied by smaller and larger temporal biases, respectively.
Spatial distribution of Hit cases in the hindcasts relative to the observed tropical cyclone centers for the (a) Lead-5d, (b) Lead-4d, and (c) Lead-3d experiments. Red, green, light blue, and gray numbers denote the time differences in tropical cyclone genesis events between the Hit cases and those observed within ±12, ±24, ±36, and ±48 h. Temporal and spatial biases and root-mean-square errors are shown at the bottom.
Citation: Weather and Forecasting 38, 11; 10.1175/WAF-D-23-0056.1
As shown in Fig. 3b, the spatial bias in Lead-4d was much lower than that in Lead-5d, with an average of 359.7 km (RMSE = 436.7 km). As for the TC centers, 26 and 6 Hit cases had their genesis locations inside and outside 5°, respectively. However, no such significant reduction was noted for time biases, with the average temporal bias being −5.9 h (RMSE = 18.0 h). In particular, 15 and 13 Hit cases had a temporal bias within ±12 and ±24 h, respectively, and 2 and 1 Hit cases had a temporal bias of ±36 and ±48 h. Notably, the Lead-3d experiment results demonstrated increases in both average spatial and temporal biases as well as their RMSE (Fig. 3c). A shorter lead time did not lead to improvements in the biases, suggesting that fvGFS may not be able to accurately simulate certain key processes (e.g., the development of MCSs leading to TCG).
In summary, the Lead-4d experiments yielded a higher Hit rate with a smaller spatial bias than did the Lead-5d experiments; however, no significant differences were noted in the temporal bias between the Lead-4d and Lead-5d experiments. Although the Lead-3d forecast also demonstrated a high Hit rate (79%), the bias and RMSE were not improved. The causes for this inconsistency remain unknown and warrant more similar hindcast experiments.
For TCG forecasting, a relatively long lead time is ideal. Therefore, our subsequent analyses focused on identifying the environmental factors that affected the Lead-5d hindcast results. We conducted composite analyses of various environmental circulations and contrasted them between the Hit and Fail cases by comparing them with those derived from CFSv2 analysis. We also noted that similar diagnostics conducted for the Lead-4d and Lead-3d experiments yielded identical conclusions.
c. Environmental condition analysis
To investigate the differences in the background circulations between the Hit and Fail cases in the Lead-5d experiments, we conducted composite analyses of the 850-hPa streamfunction (SF850), VWS, 200-hPa divergence (D200), midlevel (700-hPa) RH (RH700), SST, and GPI associated with the TCG events separately for the Hit and Fail cases. Taking the location of each TCG case as the center, the above variables within a 40° × 40° region were averaged to reveal the characteristics of large-scale background circulations. Figures 4a and 4b illustrate the average SF850 5 days before TCG in the Lead-5d hindcasts. The averaged values of SF850 within 5° of the TC center (Rin5) were calculated and displayed in the lower-right corners of Figs. 4a and 4b. The results demonstrate that the Hit cases developed in the south and southwest of a subtropical high slightly weaker than its counterpart in the Fail cases. Meanwhile, both the Hit and Fail cases appeared in a cyclonic circulation, reflecting a monsoon trough structure extending from the west side to the east side. However, the cyclonic circulation in the Fail cases was weaker and less spatially extensive. In general, the Hit cases occurred in the weaker subtropical high and stronger monsoon trough background (Fig. 4c). The contrast between the Hit and Fail cases increased as the TCG day approached as seen in the averaged SF850 2 days before TCG (Figs. 4d–f). The Hit cases developed in a stronger cyclonic circulation with a further eastward-retreated subtropical high in the northeast. The enhanced contrast between the Hit and Fail cases 2 days before TCG can be seen more clearly in the difference map (Fig. 4f). The CFSv2 composites (Figs. 4g–l) also demonstrate characteristics similar to those in the hindcasts. In summary, the Hit cases tended to occur in a more pronounced monsoon trough and weaker subtropical high background (i.e., an environment favorable for TCG), and thus, were relatively easy to simulate. By contrast, the simulation was more challenging for the fvGFS in the Fail cases because the environmental flow was not as favorable as in the Hit cases.
Composites of the 850-hPa streamfunction for tropical cyclone genesis (TCG) cases in (a)–(f) hindcasts and (g)–(l) CFSv2, 5 and 2 days before TCG (L5d and L2d in titles, respectively). (left) The Hit case composite, (center) the Fail case composite, and (right) the difference between the two. Rin5 in the bottom-right corner of each panel represents the averaged value of 850-hPa streamfunction within 5° from the center. The origin in each panel represents the location of observed TCG events.
Citation: Weather and Forecasting 38, 11; 10.1175/WAF-D-23-0056.1
VWS is a factor that is less certain in its effect on TCG. Whereas some studies confirmed low VWS would favor TCG (Gray 1998; DeMaria et al. 2001; Davis and Bosart 2003), others found otherwise (Fu et al. 2012; Teng et al. 2019). Despite the uncertainty, we conducted an analysis of VWS and presented the VWS composites 2 days before TCG in Fig. 5. For the Hit cases, TCG occurred in a region of small VWS surrounded by two strong VWS regions: one associated with the midlatitude westerlies in the north of the TC center and another associated with the low-level southwesterly wind in the southwest (Fig. 5a). A similar VWS structure also appeared in the Fail cases but with slightly smaller amplitudes (Fig. 5b). The VWS differences between the Hit and Fail cases were small and did not demonstrate a more favorable condition for the Hit cases (Fig. 5c). VWS structure observed in CFSv2 were more pronounced but with mixed positive and negative values yielding no sign for a more favorable condition for the Hit cases (Figs. 5d–f). These findings suggest that the VWS in both simulation and observation did not function as a favorable environmental factor for the TCG cases examined in this study.
Composites of vertical wind shear between 200 and 850 hPa for tropical cyclone genesis (TCG) cases in (a)–(c) hindcasts and (d)–(f) CFSv2, 2 days before TCG (L2d in titles). (left) The Hit case composite, (center) the Fail case composite, and (right) the difference between the two. Rin5 in the bottom-right corner of each panel represents the averaged value of vertical wind shear within 5° from the center. The origin in each panel represents the location of observed TCG events.
Citation: Weather and Forecasting 38, 11; 10.1175/WAF-D-23-0056.1
Upper-level divergence (i.e., D200), another key TCG indicator, was also examined. At 2 days before TCG in the hindcasts (Figs. 6a–c), D200 in the Hit cases exhibited a stronger divergence near and south of the center than that in the Fail cases, leading to significant differences in Rin5 between the Hit and Fail cases. Stronger D200 in the Hit cases was also noted in CFSv2 (Figs. 6d–f), consistent with the hindcasts. The results suggested the supporting role of D200 for TCG.
Composites of upper-level divergence at 200 hPa for tropical cyclone genesis (TCG) cases in (a)–(c) hindcasts and (d)–(f) CFSv2, 2 days before TCG (L2d in titles). (left) The Hit case composite, (center) the Fail case composite, and (right) the difference between the two. Rin5 in the bottom-right corner of each panel represents the averaged value of upper-level divergence at 200 hPa within 5° from the center. The origin in each panel represents the location of observed TCG events.
Citation: Weather and Forecasting 38, 11; 10.1175/WAF-D-23-0056.1
We examined RH700, another suggested favorable indicator for TCG (Gray 1968, 1998; Cheung and Elsberry 2002; Emanuel and Nolan 2004; Kerns and Chen 2013; Zhao et al. 2018; Teng et al. 2019, 2021). The composites 2 days before TCG (Fig. 7) indicated more abundant RH700 in the Hit cases for both the hindcasts and CFSv2, suggesting that a moister environment leads to improved predictive skill for TCG in the fvGFS (Fig. 7c). Regarding the SST, another proposed favorable factor for TCG, our results revealed that the differences between the Hit and Fail cases in both the hindcasts and CFSv2 were nonsignificant (Fig. 8). The SSTs within Rin5 in both the Hit and Fail cases were as high as 29°C, which is a conducive condition for TCG. This might be the reason that slight SST fluctuations did not have significant effects on TCG.
Composites of relative humidity at 700 hPa for tropical cyclone genesis (TCG) cases in (a)–(c) hindcasts and (d)–(f) CFSv2, 2 days before TCG (L2d in titles). (left) The Hit case composite, (center) the Fail case composite, and (right) the difference between the two. Rin5 in the top-right corner of each panel represents the averaged value of relative humidity at 700 hPa within 5° from the center. The origin in each panel represents the location of observed TCG events.
Citation: Weather and Forecasting 38, 11; 10.1175/WAF-D-23-0056.1
Composites of sea surface temperature for tropical cyclone genesis (TCG) cases in (a)–(c) hindcasts and (d)–(f) CFSv2, 2 days before TCG (L2d in titles). (left) The Hit case composite, (center) the Fail case composite, and the (right) difference between the two. Rin5 in the top-right corner of each panel represents the averaged value of sea surface temperature within 5° from the center. The origin in each panel represents the location of observed TCG events.
Citation: Weather and Forecasting 38, 11; 10.1175/WAF-D-23-0056.1
Finally, we calculated the GPI to analyze the differences between the Hit and Fail cases. The overall GPI patterns in the Hit and Fail cases 5 days before TCG were weak and dispersed (Figs. 9a,b), and therefore, the differences between the Hit and Fail cases within Rin5 did not show consistent signs in both the hindcasts (Figs. 9a–c) and CFSv2 (Figs. 9g–i). At 2 days before TCG, positive GPI became much more evident in the Hit cases than in the Fail cases (Figs. 9d–f) and demonstrated a more favorable condition for TCG. In contrast to the hindcast GPI, the CFSv2 GPI did not yield a consistent increase or decrease within Rin5 between the Hit and Fail cases (Figs. 9j–l). This occurred because of the inconsistent differences in the analyzed factors shown above between the Hit and Fail cases.
Composites of genesis potential index for tropical cyclone genesis (TCG) cases in (a)–(f) hindcasts and (g)–(l) CFSv2, 2 days before TCG (L2d in titles). (left) The Hit case composite, (center) the Fail case composite, and (right) the difference between the two. Rin5 in the bottom-right corner of each panel represents the averaged value of genesis potential index within 5° from the center. The origin in each panel represents the location of observed TCG events.
Citation: Weather and Forecasting 38, 11; 10.1175/WAF-D-23-0056.1
A compilation of the results, including SF850, VWS, divergences at 850 hPa (D850) and D200, RH700, SST, and GPI, averaged within Rin5 is presented in Table 1. A t test was applied to compare two means at the 90% confidence level based on the null hypothesis of no difference. The P values for the hindcasts indicated statistically significant differences between the Hit and Fail cases in four parameters: SF850, D850, D200, and GPI. The differences in VWS and SST between the Hit and Fail cases were small, as mentioned above; the P value also confirmed that these two parameters did not significantly influence the model hindcast results. The significant differences between the Hit and Fail cases derived from CFSv2 were found only in SF850, but not in D850, D200, and GPI. This result indicates that the fvGFS hindcasts tended to amplify the differences between Hit and Fail cases in certain environmental factors. In summary, dynamic factors (i.e., low-level rotational and upper-level divergent flow) were more influential for TCG than thermodynamic factors (such as SST, RH700) in both observation and hindcasts, and the fvGFS demonstrated its capability to capture these key parameters.
Statistical analyses of seven environmental factors derived from the hindcasts and CFSv2 within a 5° range from the TC center (Rin5), 2 days before tropical cyclone genesis (TCG; L2d). The environmental factors include the streamfunction at 850 hPa (SF850), divergences at 850 and 200 hPa (D850 and D200, respectively), vertical wind shear between 200 and 850 hPa (VWS), midlevel relative humidity at 700 hPa (RH700), sea surface temperature (SST), and genesis potential index (GPI). The P value marked by an asterisk (*) indicates a statistically significant difference between the Hit and Fail cases at the 90% confidence level.
d. Large-scale environmental influences on the Hit rate
In a pioneering study, Ritchie and Holland (1999) identified five types of background flow (viz., monsoon shear line, monsoon confluence, monsoon gyre, easterly waves, and Rossby wave dispersion) leading to TCG, based on 8-yr TC cases, providing informative guidance for identifying the background flows favorable for TCG. Classifications were slightly modified in later studies (e.g., Lee et al. 2008; Yoshida and Ishikawa 2013). Our study focused on the environmental condition several days before TCG. More recent studies by Teng et al. (2019, 2020) exploring the environment favoring the transition from TCC to TC seem to be more suitable references for our study. Teng et al. (2019) studied 2248 TCCs, which were often a precursor to TCG, in the WNP and found that about 23.6% of them developed into TCs. The author examined 29-yr satellite observation and reanalysis data and applied cluster analysis to objectively classify background flows of TCCs into eight categories based on the characteristics of wind fields at 850 hPa. They further classified these eight categories on the basis of TC occurrence into three types in descending order of occurrence frequency: monsoon-related, subtropical high edge (e.g., TCCs located in the west and southwest of a subtropical high), and easterly flow. Using a similar approach, we examined the simulated 850-hPa wind field 2 days before TCG and subjectively classified the environmental flows into four types: Monsoon (monsoon trough or monsoon gyre), Sub-High (subtropical high), Easterly, and Others. The environmental flows that could not be classified into the first three types were placed as the Others type. The corresponding composites of environmental circulation are presented in Fig. 10. In the Monsoon type, TCG was located within a monsoon trough that covered most of the analyzed region, with westerlies and easterlies to the south and north of the TCG center, respectively (Figs. 10a,b). In the Sub-High type, TCG was found between a subtropical high in the northeast and a cyclonic circulation in the southwest (Figs. 10c,d). In the Easterly type, TCG was found within a prevailing easterly flow between a strong subtropical high in the north and a weaker cyclonic circulation in the south (Figs. 10e,f). In the 42 experiments, only 2 cases could not be classified into the above three circulation types and were classified as the Others type (Fig. 10g), occurring within a small cyclonic circulation in the northern flank of a strong southwesterly flow.
Composite of 850-hPa winds in the Lead-5d hindcasts, 2 days before TCG (L2d in titles) for the four environmental types: (a),(b) Monsoon; (c),(d) Sub-High; (e),(f) Easterly, and (g) Others. (left) Hit and (right) Fail cases. The case number of each classification is shown at the top of each panel. The origin cross symbol in each panel denotes the location of tropical cyclone genesis events. Shading displays wind speed (m s−1).
Citation: Weather and Forecasting 38, 11; 10.1175/WAF-D-23-0056.1
The case numbers of TCG occurrences in different circulation types are also shown in each panel of Fig. 10. In total, 14 of the 21 Hit cases were found associated with the Monsoon type (Fig. 10a), whereas the remaining 5 and 2 cases occurred in the Sub-High and Easterly type, respectively (Figs. 10d,g). By contrast, 10 of the 21 Fail cases occurred in the Easterly type (Fig. 10h), whereas the remaining 4 and 6 cases occurred in the Monsoon and Sub-High types, respectively (Figs. 10b,e). Two additional Fail cases were classified as the Others type (Fig. 10k). The differences between the distribution of Hit and Fail cases in the different circulation types were considerable. This result suggested that the fvGFS was more capable of simulating the TCG events that occur in a monsoon trough (78% hit rate); however, it lacked the ability (approximately 50% hit rate) to predict the Sub-High type and had the lowest ability (about 17% hit rate) for the TCG occurring in a strong prevailing easterly flow. The aforementioned finding suggested that the fvGFS has the best TCG forecasting ability in a cyclonic environment, which exhibits strong dynamical forcing.
4. Effects of background flow
The aforementioned results of low-level streamfunction and environmental wind field analyses revealed that the fvGFS often fails to forecast TCG occurrences in an easterly flow background, which is also the main characteristic of the prevailing mean flow east of 135°E where two-thirds of the Fail cases occurred. This result is consistent with those revealed in Figs. 4 and 10. The TCG forecasting ability of fvGFS was higher in a monsoon trough environment than in an easterly environment, because the monsoon trough that provides strong forcing for TCG is the most active in the western WNP. By contrast, more Fail cases occurred further to the east in the region, where easterly flows prevailed. This result was also observed in the circulation analysis of the Fail cases: approximately 48% (10/21) of the cases occurred in an easterly background flow. However, the ratio was only approximately 10% (2/21) for the Hit cases. This result is consistent with those of previous studies on the influence of large-scale environments on TCG (Ritchie and Holland 1999; Chen et al. 2008; Yoshida and Ishikawa 2013; Wu and Duan 2015; Teng et al. 2020).
Table S1 indicates that the hindcasts tended to fail from mid-August to September 2018 and after October 2019. There were 11 Fail cases (numbered 13–17, 34–36, 38, 40, and 42) in the two periods. Four Fail cases occurred in the easterly environment for each of the two periods and the other three were classified as the Sub-High type. The reason for the less hindcast skill in the easterly and Sub-High-type environment, which often appeared in the eastern WNP, is likely the weaker dynamical forcing for TCG. TCs originating from an easterly wave often developed from the relatively smaller-scale convective systems that were likely more difficult for the 13-km fvGFS to accurately resolve. A similar situation would occur in the Sub-High-type environment that was predominantly anticyclonic circulation and not favorable for the development of small-scale TC seeds or MCSs, which might not be easily resolved by the 13-km fvGFS.
The clustering of the Fail cases from mid-August to September 2018 and after October 2019 could also be linked to the less favorable large-scale environments in an intraseasonal time scale (e.g., BSISO). Figure S1 presents the Hovmöller diagram of 30–60-day filtered 850-hPa zonal wind and outgoing longwave radiation. One can see clearly that the TCGs in the two periods tended to occur in a large-scale background flow when and where the BSISO was weak in the subtropical WNP. Similar results can also be seen in the reconstructed intraseasonal perturbation data on the APEC Climate Center BSISO Monitoring website (https://www.apcc21.org/ser/moni.do?lang=en). In this weak background condition, the Monsoon type environment for TCG was presumably less likely to occur; instead, the Easterly and Sub-High types were more dominant as observed. The fvGFS was therefore less skillful in hindcasting the TCG.
By comparing the fvGFS hindcast results with CFSv2 analysis, we found that certain environmental characteristics were more conducive to the TCG in the Hit cases than in the real environment. For instance, the negative low-level streamfunction was identified in both the hindcasts and CFSv2 5 days before TCG (Figs. 4a,g). However, at 2 days before TCG, the simulated cyclonic circulation was notably stronger in the hindcasts than in CFSv2 (Figs. 4d,j). Similar differences were noted in other environmental factors such as D200 (Figs. 6a,d) and GPI (Figs. 10a,d,g,j). By contrast, for the Fail cases, the differences between the hindcasts and CFSv2 were nonsignificant. These results suggested that when the model efficiently comprehended TCG-associated circulation characteristics, it tended to enhance the environmental flow and make it more favorable for TCG.
The average VWS within Rin5 2 days before TCG in both the hindcasts and CFSv2 reanalysis [which was approximately 12 m s−1, a number similar to that reported by Teng et al. (2019) and Yoshida and Fudeyasu (2020)] demonstrated nonsignificant differences between the Hit and Fail cases. The differences between the Hit and Fail cases, in both the hindcasts and CFSv2 reanalysis, exhibited both positive and negative values within Rin5 (Fig. 5). This result indicated that VWS did not function as a key favorable environmental factor for the 2018 and 2019 TCG cases.
For the studied cases, SST did not appear to be a key background factor for either the Hit or Fail cases in both the hindcasts and CFSv2 analysis (Fig. 8). TCG was insensitive to the relatively small SST variation in the WNP—where SST in the typhoon season was much higher than the threshold value (i.e., 26.5°C), a sufficient condition for TCG, and was able to supply sufficient energy for TCG. Because of the mixed contributions of various environmental factors to TCG, the GPI calculated based on CFSv2 analysis did not exhibit a tendency toward either the Hit or the Fail cases. Notably, the fvGFS simulated a considerable difference in GPI between the Hit and Fail cases 2 days before TCG. Understanding whether this fvGFS-simulated enhancement is a robust feature or an unintentional occurrence warrants further investigation.
5. Discussion and conclusions
In this study, we evaluated the ability of fvGFS to simulate TCG by conducting hindcast experiments for 42 TCG events in the WNP over 2018–19. These experiments were initiated 5, 4, and 3 days before TCG so as to examine model sensitivity to forecast lead time. The results demonstrated TCG simulation success rates of 50%, 76%, and 79% in the Lead-5d, Lead-4d, and Lead-3d experiments, respectively, suggesting a marked improvement between the lead time of 5 and 4 days. Regarding the temporal and spatial biases and their RMSEs of TCG, a further shortened lead time from 4 to 3 days did not lead to consistent improvement. The mechanisms underlying these results, which may be case dependent, model dependent, or both, remain unknown and thus warrant additional studies. One potential area would be the model ability in properly simulated finer scale disturbances such as TC seeds and MCSs that could help trigger TCG.
Successful simulation tended to occur in certain large-scale background environmental characteristics. We identified these characteristics associated with all TCG cases by comparing the composites of thermodynamic and dynamic factors between the Hit and Fail cases. In the hindcasts, the four environmental factors—namely, SF850, D850, D200, and GPI—provided favorable conditions for TCG in the Hit cases. By contrast, the differences in the other suggested favorable factors (e.g., VWS, 700-hPa moisture, and SST) between the Hit and Fail cases were nonsignificant.
In CFSv2, we observed a significant difference only in SF850. The differences between the hindcasts and CFSv2 suggested that the fvGFS is relatively more sensitive to the favorable environmental factors for simulating a TCG. Our results based on hindcasts and CFSv2 analysis suggest that low-level rotational circulation and upper-level divergence provide more favorable environmental conditions for TCG development than other factors suggested previously. The fvGFS, which tended to simulate larger environmental differences between the Hit and Fail cases than those in CFSv2, demonstrated a better ability to simulate TCG in the aforementioned environmental conditions. This finding may facilitate the in-advance assessment of the TCG forecasting ability of the fvGFS.
To examine how the environmental conditions would affect the fvGFS performance, we divided the large-scale circulation 2 days before TCG of all the Hit and Fail cases further into four environmental types (viz., Monsoon, Sub-High, Easterly, and Others) according to the methodology of Teng et al. (2019). As indicated by observation, of the 42 TCG cases over 2018/19, 17, 11, 12, and 2 cases occurred in the Monsoon, Sub-High, Easterly, and Others types, respectively. Whereas more cases occurred in the Monsoon type, the number of occurrences (17, 11, 12) did not indicate an overwhelming preference for one particular type. This result is consistent with previous findings (Simpson et al. 1997; Lee et al. 2008, 2010; Yoshida and Ishikawa 2013; Zhao et al. 2015b; Fudeyasu and Yoshida 2018; Teng et al. 2019, 2021; Yoshida and Fudeyasu 2020). Interestingly, the fvGFS demonstrated the highest successful hindcast rate for the Monsoon type, with an 82% hit rate (14/17) and a moderate rate for the Sub-High type (45% hit rate; 5/11), did poorly for the Easterly type (17% hit rate; 2/12), and failed in two the others type cases. These results demonstrated the moderate ability (21/42) of the fvGFS to forecast all TCG events. The model demonstrated a much better ability for the Monsoon-type environmental circulation. This is likely due to the stronger environmental forcing that provided a more favorable condition for TCG. By contrast, the relatively poor model hindcast skill for the Sub-High- and Easterly-type environmental circulations resulted in the clustering of Fail cases in the periods from mid-August to September 2018 and after October 2019 when and where the BSISO was weak in the subtropical WNP. During the periods, TCG could still occur but presumably was more challenging for the fvGFS to predict properly because of weak dynamic forcing. This result is useful for in-advance assessment of the fvGFS’s forecasting ability. It also indicates that numerical weather forecasting systems, even with the most advanced high-resolution weather forecast models, such as the fvGFS, warrant further improvement. The evaluation approach adopted in this study can be easily applied to assess the TCG predictive skill of the existing weather forecast models in the context of the DIMOSIC (different models, same initial conditions) intercomparison (Magnusson et al. 2022), which conducted experiments using different models with the same initiation conditions to assess the skill of medium-range forecast models.
A good contrasting example from our hindcast result is Typhoon Faxai that was one of the Fail cases in the current study, but its genesis and track were successfully hindcasted in another 14-km mesh global nonhydrostatic atmospheric model (Yamada et al. 2023). Whereas the long lead time for a successful track forecast was impressive as presented in Yamada et al. (2023), their study focused on track simulation, and it is not clear by reading the paper the successful rate in TCG simulation. It would be difficult to have a direct comparison because of several reasons, e.g., the experimental design (1600 members in their experiment versus one single model run in the current study), the initial conditions (well assimilated data used in Yamada et al. versus CFSv2 analysis in our study), and different model configurations. Comparing a single case may not offer a comprehensive perspective. A well-organized intercomparison research as discussed above might be needed to identify the reason.
The marked improved Hit rate from 5- to 4-day lead time and the nonsignificant improvement from 4- to 3-day lead time is an interesting contrast but also an unresolved issue in current study. It could be model and/or time dependent (e.g., different typhoon seasons or periods). Another important aspect is the ability of the current high-resolution (e.g., 10–20 km) model in simulating smaller-scale convection such as MCSs that serve as seeds for TCG. How subtle are these issues for a successful TCG simulation and how to improve model ability remain challenging? An intercomparison project can be useful for deeper understanding.
Due to computational resource constraints, only 42 cases were hindcasted. Additional case simulations will be added in future studies to evaluate the model forecasting ability further. Other questions, such as whether a forecast barrier exists in terms of lead time as seen in this study and whether the tendency for certain environmental factors remains in other typhoon seasons, also warrant additional studies.
Acknowledgments.
This work was supported by Academia Sinica Thematic Research Program AS-TP-109-M11. This manuscript was edited by Wallace Academic Editing. The TC information was obtained from the JMA-RSMC, the environmental background data were obtained from NCEP-CFSv2, and the initial data for fvGFS hindcasts were produced by the NCEP Central Operations.
Data availability statement.
The initial data for fvGFS hindcasts were produced by the NCEP Central Operations: NCEP Products Inventory-Global Products (https://www.nco.ncep.noaa.gov/pmb/products/gfs/). The environmental background data were from NCEP-CFSv2 Operational Forecasts (https://www.ncei.noaa.gov/access/metadata/landing-page/bin/iso?id=gov.noaa.ncdc:C00877). The TC information was obtained from the JMA-RSMC: Best Track Data (https://www.jma.go.jp/jma/jma-eng/jma-center/rsmc-hp-pub-eg/besttrack.html). Data produced in this study will be provided through a request to the corresponding author.
REFERENCES
Bister, M., and K. A. Emanuel, 2002a: Low frequency variability of tropical cyclone potential intensity. 1. Interannual to interdecadal variability. J. Geophys. Res., 107, 4801, https://doi.org/10.1029/2001JD000776.
Bister, M., and K. A. Emanuel, 2002b: Low frequency variability of tropical cyclone potential intensity. 2. Climatology for 1982–1995. J. Geophys. Res., 107, 4621, https://doi.org/10.1029/2001JD000780.
Briegel, L. M., and W. M. Frank, 1997: Large-scale influences on tropical cyclogenesis in the western North Pacific. Mon. Wea. Rev., 125, 1397–1413, https://doi.org/10.1175/1520-0493(1997)125<1397:LSIOTC>2.0.CO;2.
Camargo, S. J., A. W. Robertson, S. J. Gaffney, P. Smyth, and M. Ghil, 2007: Cluster analysis of typhoon tracks. Part II: Large-scale circulation and ENSO. J. Climate, 20, 3654–3676, https://doi.org/10.1175/JCLI4203.1.
Camargo, S. J., M. C. Wheeler, and A. H. Sobel, 2009: Diagnosis of the MJO modulation of tropical cyclogenesis using an empirical index. J. Atmos. Sci., 66, 3061–3074, https://doi.org/10.1175/2009JAS3101.1.
Chen, J.-H., S.-J. Lin, L. Zhou, X. Chen, S. Rees, M. Bender, and M. Morin, 2019: Evaluation of tropical cyclone forecasts in the Next Generation Global Prediction System. Mon. Wea. Rev., 147, 3409–3428, https://doi.org/10.1175/MWR-D-18-0227.1.
Chen, J.-M., C.-H. Wu, P.-H. Chung, and C.-H. Sui, 2018: Influence of intraseasonal–interannual oscillations on tropical cyclone genesis in the western North Pacific. J. Climate, 31, 4949–4961, https://doi.org/10.1175/JCLI-D-17-0601.1.
Chen, T.-C., S.-Y. Wang, M.-C. Yen, and A. J. Clark, 2008: Are tropical cyclones less effectively formed by easterly waves in the western North Pacific than in the North Atlantic? Mon. Wea. Rev., 136, 4527–4540, https://doi.org/10.1175/2008MWR2149.1.
Cheung, K. K. W., and R. L. Elsberry, 2002: Tropical cyclone formations over the western North Pacific in the Navy Operational Global Atmospheric Prediction System forecasts. Wea. Forecasting, 17, 800–820, https://doi.org/10.1175/1520-0434(2002)017<0800:TCFOTW>2.0.CO;2.
Cossuth, J. H., R. D. Knabb, D. P. Brown, and R. E. Hart, 2013: Tropical cyclone formation guidance using pregenesis Dvorak climatology. Part I: Operational forecasting and predictive potential. Wea. Forecasting, 28, 100–118, https://doi.org/10.1175/WAF-D-12-00073.1.
Dare, R. A., and J. L. McBride, 2011: The threshold sea surface temperature condition for tropical cyclogenesis. J. Climate, 24, 4570–4576, https://doi.org/10.1175/JCLI-D-10-05006.1.
Davis, C. A., and L. F. Bosart, 2003: Baroclinically induced tropical cyclogenesis. Mon. Wea. Rev., 131, 2730–2747, https://doi.org/10.1175/1520-0493(2003)131<2730:BITC>2.0.CO;2.
DeMaria, M., J. A. Knaff, and B. H. Conell, 2001: A tropical cyclone genesis parameter for the tropical Atlantic. Wea. Forecasting, 16, 219–233, https://doi.org/10.1175/1520-0434(2001)016<0219:ATCGPF>2.0.CO;2.
Emanuel, K. A., 1995: Sensitivity of tropical cyclones to surface exchange coefficients and a revised steady-state model incorporating eye dynamics. J. Atmos. Sci., 52, 3969–3976, https://doi.org/10.1175/1520-0469(1995)052<3969:SOTCTS>2.0.CO;2.
Emanuel, K. A., 2022: Tropical cyclone seeds, transition probabilities, and genesis. J. Climate, 35, 3557–3566, https://doi.org/10.1175/JCLI-D-21-0922.1.
Emanuel, K. A., and D. S. Nolan, 2004: Tropical cyclone activity and the global climate system. 26th Conf. on Hurricanes and Tropical Meteorology, Miami, FL, Amer. Meteor. Soc., 10A.2, https://ams.confex.com/ams/26HURR/techprogram/paper_75463.htm.
Emanuel, K. A., R. Sundararajan, and J. Williams, 2008: Hurricanes and global warming: Results from downscaling IPCC AR4 simulations. Bull. Amer. Meteor. Soc., 89, 347–368, https://doi.org/10.1175/BAMS-89-3-347.
Fu, B., M. S. Peng, T. Li, and D. E. Stevens, 2012: Developing versus nondeveloping disturbances for tropical cyclone formation. Part II: Western North Pacific. Mon. Wea. Rev., 140, 1067–1080, https://doi.org/10.1175/2011MWR3618.1.
Fudeyasu, H., and R. Yoshida, 2018: Western North Pacific tropical cyclone characteristics stratified by genesis environment. Mon. Wea. Rev., 146, 435–446, https://doi.org/10.1175/MWR-D-17-0110.1.
Gao, K., L. Harris, L. Zhou, M. Bender, and M. Morin, 2021: On the sensitivity of hurricane intensity and structure to horizontal tracer advection schemes in FV3. J. Atmos. Sci., 78, 3007–3021, https://doi.org/10.1175/JAS-D-20-0331.1.
Goldenberg, S. B., and L. J. Shapiro, 1996: Physical mechanisms for the association of El Niño and West African rainfall with Atlantic major hurricane activity. J. Climate, 9, 1169–1187, https://doi.org/10.1175/1520-0442(1996)009<1169:PMFTAO>2.0.CO;2.
Gray, W. M., 1968: Global view of the origin of tropical disturbances and storms. Mon. Wea. Rev., 96, 669–700, https://doi.org/10.1175/1520-0493(1968)096<0669:GVOTOO>2.0.CO;2.
Gray, W. M., 1998: The formation of tropical cyclones. Meteor. Atmos. Phys., 67, 37–69, https://doi.org/10.1007/BF01277501.
Halperin, D. J., H. E. Fuelberg, R. E. Hart, and J. H. Cossuth, 2016: Verification of tropical cyclone genesis forecasts from global numerical models: Comparisons between the North Atlantic and eastern North Pacific basins. Wea. Forecasting, 31, 947–955, https://doi.org/10.1175/WAF-D-15-0157.1.
Halperin, D. J., R. E. Hart, H. E. Fuelberg, and J. H. Cossuth, 2017: The development and evaluation of a statistical–dynamical tropical cyclone genesis guidance tool. Wea. Forecasting, 32, 27–46, https://doi.org/10.1175/WAF-D-16-0072.1.
Halperin, D. J., A. B. Penny, and R. E. Hart, 2020: A comparison of tropical cyclone genesis forecast verification from three Global Forecast System (GFS) operational configurations. Wea. Forecasting, 35, 1801–1815, https://doi.org/10.1175/WAF-D-20-0043.1.
Harris, L., and Coauthors, 2020: GFDL SHiELD: A unified system for weather‐to‐seasonal prediction. J. Adv. Model. Earth Syst., 12, e2020MS002223, https://doi.org/10.1029/2020MS002223.
Hazelton, A. T., L. Harris, and S.-J. Lin, 2018a: Evaluation of tropical cyclone structure forecasts in a high-resolution version of the multiscale GFDL fvGFS model. Wea. Forecasting, 33, 419–442, https://doi.org/10.1175/WAF-D-17-0140.1.
Hazelton, A. T., M. Bender, M. Morin, L. Harris, and S.-J. Lin, 2018b: 2017 Atlantic hurricane forecasts from a high-resolution version of the GFDL fvGFS model: Evaluation of track, intensity, and structure. Wea. Forecasting, 33, 1317–1337, https://doi.org/10.1175/WAF-D-18-0056.1.
Hazelton, A. T., X. Zhang, W. Ramstrom, S. Gopalakrishan, F. D. Marks, and J. A. Zhang, 2020: High-resolution ensemble HFV3 forecasts of Hurricane Michael (2018): Rapid intensification in shear. Mon. Wea. Rev., 148, 2009–2032, https://doi.org/10.1175/MWR-D-19-0275.1.
Hsieh, T.-L., W. Yang, G. A. Vecchi, and M. Zhao, 2022: Model spread in the tropical cyclone frequency and seed propensity index across global warming and ENSO-like perturbations. Geophys. Res. Lett., 49, e2021GL097157, https://doi.org/10.1029/2021GL097157.
Hsieh, Y.-H., C.-S. Lee, and C.-H. Sui, 2017: A study on the influences of low-frequency vorticity on tropical cyclone formation in the western North Pacific. Mon. Wea. Rev., 145, 4151–4169, https://doi.org/10.1175/MWR-D-17-0085.1.
Hsu, H.-H., 2012: East Asian monsoon. Intraseasonal Variability in the Atmosphere-Ocean-Climate System, W. K.-M. Lau and D. E. Waliser, Eds., Springer, 63–94.
Hsu, H.-H., Y.-L. Chen, A.-K. Lo, C.-H. Hung, W.-S. Kau, and C.-C. Wu, 2008: Coupling of the intraseasonal oscillation-tropical cyclone coupling in the western North Pacific during the 2004 typhoon season. Recent Progress in Atmospheric Sciences: Applications to the Asia-Pacific Region, K. N. Liou and M. D. Chou, Eds., World Scientific, 49–65.
Huang, C.-Y., T.-C. Juan, H.-C. Kuo, and J.-H. Chen, 2020: Track deflection of Typhoon Maria (2018) during a westbound passage offshore of northern Taiwan: Topographic influence. Mon. Wea. Rev., 148, 4519–4544, https://doi.org/10.1175/MWR-D-20-0117.1.
Huang, C.-Y., S.-H. Sha, and H.-C. Kuo, 2022: A modeling study of Typhoon Lekima (2019) with the topographic influence of Taiwan. Mon. Wea. Rev., 150, 1993–2011, https://doi.org/10.1175/MWR-D-21-0183.1.
Huang, P., C. Chou, and R. Huang, 2011: Seasonal modulation of tropical intraseasonal oscillations on tropical cyclone geneses in the western North Pacific. J. Climate, 24, 6339–6352, https://doi.org/10.1175/2011JCLI4200.1.
Ikehata, K., and M. Satoh, 2021: Climatology of tropical cyclone seed frequency and survival rate in tropical cyclones. Geophys. Res. Let., 48, e2021GL093626, https://doi.org/10.1029/2021GL093626.
Jiang, X., M. Zhao, and D. E. Waliser, 2012a: Modulation of tropical cyclones over the eastern Pacific by the intraseasonal variability simulated in an AGCM. J. Climate, 25, 6524–6538, https://doi.org/10.1175/JCLI-D-11-00531.1.
Kerns, B. W., and S. S. Chen, 2013: Cloud clusters and tropical cyclogenesis: Developing and nondeveloping systems and their large-scale environment. Mon. Wea. Rev., 141, 192–210, https://doi.org/10.1175/MWR-D-11-00239.1.
Kikuchi, K., 2021: The Boreal Summer Intraseasonal Oscillation (BSISO): A review. J. Meteor. Soc. Japan, 99, 933–972, https://doi.org/10.2151/jmsj.2021-045.
Knutson, T. R., and Coauthors, 2010: Tropical cyclones and climate change. Nat. Geosci., 3, 157–163, https://doi.org/10.1038/ngeo779.
Lee, C.-S., K. K. W. Cheung, J. S. N. Hui, and R. L. Elsberry, 2008: Mesoscale features associated with tropical cyclone formations in the western North Pacific. Mon. Wea. Rev., 136, 2006–2022, https://doi.org/10.1175/2007MWR2267.1.
Lee, C.-S., K. K. W. Cheung, W.-T. Fang, and R. L. Elsberry, 2010: Initial maintenance of tropical cyclone size in the western North Pacific. Mon. Wea. Rev., 138, 3207–3223, https://doi.org/10.1175/2010MWR3023.1.
Li, R. C. Y., W. Zhou, J. C. L. Chan, and P. Huang, 2012: Asymmetric modulation of western North Pacific cyclogenesis by the Madden–Julian oscillation under ENSO conditions. J. Climate, 25, 5374–5385, https://doi.org/10.1175/JCLI-D-11-00337.1.
Lin, Y.-L., and C.-S. Lee, 2011: An analysis of tropical cyclone formations in the South China Sea during the late season. Mon. Wea. Rev., 139, 2748–2760, https://doi.org/10.1175/2011MWR3495.1.
Magnusson, L., and Coauthors, 2022: Skill of medium-range forecast models using the same initial conditions. Bull. Amer. Meteor. Soc., 103, E2050–E2068, https://doi.org/10.1175/BAMS-D-21-0234.1.
McBride, J. L., and R. Zehr, 1981: Observational analysis of tropical cyclone formation. Part II: Comparison of non-developing versus non-developing systems. J. Atmos. Sci., 38, 1132–1151, https://doi.org/10.1175/1520-0469(1981)038<1132:OAOTCF>2.0.CO;2.
Merlis, T. M., M. Zhao, and I. M. Held, 2013: The sensitivity of hurricane frequency to ITCZ changes and radiatively forced warming in aquaplanet simulations. Geophys. Res. Lett., 40, 4109–4114, https://doi.org/10.1002/grl.50680.
Merlis, T. M., W. Zhou, I. M. Held, and M. Zhao, 2016: Surface temperature dependence of tropical cyclone-permitting simulations in a spherical model with uniform thermal forcing. Geophys. Res. Lett., 43, 2859–2865, https://doi.org/10.1002/2016GL067730.
Nakano, M., M. Sawada, T. Nasuno, and M. Satoh, 2015: Intraseasonal variability and tropical cyclogenesis in the western North Pacific simulated by a global nonhydrostatic atmospheric model. Geophys. Res. Lett., 42, 565–571, https://doi.org/10.1002/2014GL062479.
Nakano, M., and Coauthors, 2017: Global 7 km mesh nonhydrostatic Model Intercomparison Project for improving TYphoon forecast (TYMIP-G7): Experimental design and preliminary results. Geosci. Model Dev., 10, 1363–1381, https://doi.org/10.5194/gmd-10-1363-2017.
Ooyama, K. V., 1982: Conceptual evolution of the theory and modeling of the tropical cyclone. J. Meteor. Soc. Japan, 60, 369–380, https://doi.org/10.2151/jmsj1965.60.1_369.
Pratt, A. S., and J. L. Evans, 2009: Potential impacts of the Saharan air layer on numerical model forecasts of North Atlantic tropical cyclogenesis. Wea. Forecasting, 24, 420–435, https://doi.org/10.1175/2008WAF2007090.1.
Ritchie, E. A., and G. J. Holland, 1999: Large-scale patterns associated with tropical cyclogenesis in the western Pacific. Mon. Wea. Rev., 127, 2027–2043, https://doi.org/10.1175/1520-0493(1999)127<2027:LSPAWT>2.0.CO;2.
Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1058, https://doi.org/10.1175/2010BAMS3001.1.
Saha, S., and Coauthors, 2014: The NCEP Climate Forecast System version 2. J. Climate, 27, 2185–2208, https://doi.org/10.1175/JCLI-D-12-00823.1.
Simpson, J., E. Ritchie, G. J. Holland, J. Halverson, and S. Stewart, 1997: Mesoscale interactions in tropical cyclone genesis. Mon. Wea. Rev., 125, 2643–2661, https://doi.org/10.1175/1520-0493(1997)125<2643:MIITCG>2.0.CO;2.
Sobel, A. H., and S. J. Camargo, 2005: Influence of western North Pacific tropical cyclones on their large-scale environment. J. Atmos. Sci., 62, 3396–3407, https://doi.org/10.1175/JAS3539.1.
Teng, H.-F., C.-S. Lee, H.-H. Hsu, J. M. Done, and G. J. Holland, 2019: Tropical cloud cluster environments and their importance for tropical cyclone formation. J. Climate, 32, 4069–4088, https://doi.org/10.1175/JCLI-D-18-0679.1.
Teng, H.-F., J. M. Done, C.-S. Lee, H.-H. Hsu, and Y.-H. Kuo, 2020: Large-scale environmental influences on tropical cyclone formation processes and development time. J. Climate, 33, 9763–9782, https://doi.org/10.1175/JCLI-D-19-0709.1.
Teng, H.-F., Y.-H. Kuo, and J. M. Done, 2021: Importance of midlevel moisture for tropical cyclone formation in easterly and monsoon environments over the western North Pacific. Mon. Wea. Rev., 149, 2449–2469, https://doi.org/10.1175/MWR-D-20-0313.1.
Tory, K. J., and R. A. Dare, 2015: Sea surface temperature thresholds for tropical cyclone formation. J. Climate, 28, 8171–8183, https://doi.org/10.1175/JCLI-D-14-00637.1.
Tsai, H.-C., and R. L. Elsberry, 2019: Combined three-stage 7-day weighted analog intensity prediction technique for western North Pacific tropical cyclones: Demonstration of optimum performance. Wea. Forecasting, 34, 1979–1998, https://doi.org/10.1175/WAF-D-19-0130.1.
Tsai, H.-C., K.-C. Lu, R. L. Elsberry, M.-M. Lu, and C.-H. Sui, 2011: Tropical cyclone-like vortices detection in the NCEP 16-day ensemble system over the western North Pacific in 2008: Application and forecast evaluation. Wea. Forecasting, 26, 77–93, https://doi.org/10.1175/2010WAF2222415.1.
Vitart, F., and T. N. Stockdale, 2001: Seasonal forecasting of tropical storms using coupled GCM integrations. Mon. Wea. Rev., 129, 2521–2537, https://doi.org/10.1175/1520-0493(2001)129<2521:SFOTSU>2.0.CO;2.
Vitart, F., J. L. Anderson, and W. F. Stern, 1997: Simulation of the interannual variability of tropical storm frequency in an ensemble of GCM integrations. J. Climate, 10, 745–760, https://doi.org/10.1175/1520-0442(1997)010<0745:SOIVOT>2.0.CO;2.
Walsh, K. J. E., M. Fiorino, C. W. Landsea, and K. L. McInnes, 2007: Objectively determined resolution-dependent threshold criteria for the detection of tropical cyclones in climate models and reanalyses. J. Climate, 20, 2307–2314, https://doi.org/10.1175/JCLI4074.1.
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.
Weng, C.-H., and H.-H. Hsu, 2017: Intraseasonal oscillation enhancing C5 typhoon occurrence over the tropical western North Pacific. Geophys. Res. Lett., 44, 3339–3345, https://doi.org/10.1002/2017GL072743.
Wu, L., and J. Duan, 2015: Extended simulation of tropical cyclone formation in the western North Pacific monsoon trough. J. Atmos. Sci., 72, 4469–4485, https://doi.org/10.1175/JAS-D-14-0375.1.
Yamada, Y., and Coauthors, 2021: Evaluation of the contribution of tropical cyclone seeds to changes in tropical cyclone frequency due to global warming in high-resolution multi-model ensemble simulations. Prog. Earth Planet. Sci., 8, 11, https://doi.org/10.1186/s40645-020-00397-1.
Yamada, Y., and Coauthors, 2023: Large ensemble simulation for investigation predictability of precursor vortices of Typhoon Faxai in 2019 with a 14-km mesh global nonhydrostatic atmospheric model. Geophys. Res. Lett., 50, e2022GL100565, https://doi.org/10.1029/2022GL100565.
Yamaguchi, M., and N. Koide, 2017: Tropical cyclone genesis guidance using the early stage Dvorak analysis and global ensembles. Wea. Forecasting, 32, 2133–2141, https://doi.org/10.1175/WAF-D-17-0056.1.
Yang, W., T.-L. Hsieh, and G. A. Vecchi, 2021: Hurricane annual cycle controlled by both seeds and genesis probability. Proc. Natl. Acad. Sci. USA, 118, e2108397118, https://doi.org/10.1073/pnas.2108397118.
Yoshida, R., and H. Ishikawa, 2013: Environmental factors contributing to tropical cyclone genesis over the western North Pacific. Mon. Wea. Rev., 141, 451–467, https://doi.org/10.1175/MWR-D-11-00309.1.
Yoshida, R., and H. Fudeyasu, 2020: How significant are low-level flow patterns in tropical cyclone genesis over the western North Pacific? Mon. Wea. Rev., 148, 559–576, https://doi.org/10.1175/MWR-D-19-0023.1.
Zhang, Y., X. Chen, and Y. Lu, 2021: Structure and dynamics of ensemble correlations for satellite all-sky observations in an FV3-based global-to-regional nested convection-permitting ensemble forecast of Hurricane Harvey. Mon. Wea. Rev., 149, 2409–2430, https://doi.org/10.1175/MWR-D-20-0369.1.
Zhao, H., X. Jiang, and L. Wu, 2015a: Modulation of northwest Pacific tropical cyclone genesis by the intraseasonal variability. J. Meteor. Soc. Japan, 93, 81–97, https://doi.org/10.2151/jmsj.2015-006.
Zhao, H., R. Yoshida, and G. B. Raga, 2015b: Impact of the Madden–Julian oscillation on western North Pacific tropical cyclogenesis associated with large-scale patterns. J. Appl. Meteor. Climatol., 54, 1413–1429, https://doi.org/10.1175/JAMC-D-14-0254.1.
Zhao, H., S. Chen, P. J. Klotzbach, and G. B. Raga, 2018: Impact of the extended boreal summer intraseasonal oscillation on western North Pacific tropical cloud cluster genesis productivity. J. Climate, 31, 9175–9191, https://doi.org/10.1175/JCLI-D-18-0113.1.
Zhao, M., I. M. Held, S.-J. Lin, and G. A. Vecchi, 2009: Simulations of global hurricane climatology, interannual variability, and response to global warming using a 50-km resolution GCM. J. Climate, 22, 6653–6678, https://doi.org/10.1175/2009JCLI3049.1.
