Influence of Warm SST in the Oyashio Region on Rainfall Distribution of Typhoon Hagibis ( 2019 )

Typhoon Hagibis (2019) caused widespread flooding and damage over eastern Japan. The associated rainfall maxima were primarily observed on the windward mountain slopes along with the west of the leading edge of a low-level front. Concomitantly, a significant positive value in sea surface temperature anomalies (SSTAs) was observed in association with an ocean eddy over the Oyashio region, together with anomalous warmth over the entire western North Pacific. The present study examines the role of the SSTAs in the rainfall distribution associated with Hagibis, to deepen our understanding of the influence of the midlatitude ocean on tropical cyclones and associated rainfall. Our sensitivity experiments demonstrate that the observed warm SSTAs had the potential to displace the rainfall caused by Hagibis inland and thereby acted to increase precipitation along the Pacific coast of northeastern Japan. Our results suggest that midlatitude SSTAs on ocean-eddy scales can also influence the synoptic-scale atmospheric front and associated heavy rainfall. (Citation: Iizuka, S., R. Kawamura, H. Nakamura, and T. Miyama, 2021: Influence of warm SST in the Oyashio region on rainfall distribution of Typhoon Hagibis (2019). SOLA, 17A, 21− 28, doi:10.2151/sola.17A-004.)


Introduction
Typhoon Hagibis (2019) made landfall in the main island of Japan on October 12, 2019. The highest daily precipitation recorded, 922.5 mm, was observed on the day of the typhoon's landfall by the Hakone Automatic Meteorological Data Acquisition System (AMeDAS) station of the Japan Meteorological Agency (JMA). Furthermore, nearly 30% of the climatological annual rainfall was observed only within 24 h in some municipalities within the Gunma, Saitama, Tokyo, Miyagi, and Iwate prefectures (Fig. S1) (JMA 2019). The heavy rainfall caused widespread flooding and damage over eastern Japan. This rainfall event urged the ongoing need to deepen our understanding for further improvement of prediction skill of the impact of tropical cyclone (TC) landfall.
The importance of high sea surface temperature (SST) in the formation and intensification of TCs has been well known (Gray 1968;Emanuel 1986). It has also been suggested that TC rainfall rate and area both increase with SST in the tropics (Lin et al. 2015). In midlatitudes, however, environmental conditions are different as characterized by cool SSTs and strong baroclinicity. Kim et al. (2018) pointed out that stronger vertical wind shear significantly contributes to an increase in TC rainfall area in rendering it more asymmetric, while SST rarely affects the TC rainfall area. Figure 1a shows SST distribution on October 11, before Hagibis made landfall in Japan. Ito and Ichikawa (2021) demonstrated that the warm SST anomalies (SSTAs) over the western North Pacific (WNP) accelerated the movement of Hagibis toward Japan. Kawase et al. (2021) documented that the historical warming of both the atmosphere and ocean over the WNP sustained the intensity of Hagibis and increased the associated rainfall. In addition to the anomalous warmth observed uniformly over the WNP, marked positive SSTAs were observed over the Kuroshio off the southern coast of the main island of Japan and over the Kuroshio-Oyashio Extension (KOE) off the east coast of Japan. The former was associated with the Kuroshio large meander (Sugimoto et al. 2020), whereas the latter with a warm-core eddy over the Oyashio (Fig.  1c). Fujiwara et al. (2020) pointed out that moisture supply from the warm Kuroshio can influence TC intensity under a particular synoptic condition. Unlike most of the studies that focus on the influence of warm (sub-) tropical SSTs on TC development, this study examines the influence of midlatitude SST on the rainfall induced by Hagibis through numerical experiments to deepen our understanding of the impacts of the midlatitude ocean on TCs.

Data and model
The analysis data from the JMA Meso Scale Model (MSM) were used to describe synoptic features of the heavy rainfall event due to Hagibis. To validate the simulations, the observations were derived from the best track of the Regional Specialized Meteorological Center Tokyo and radar/rain gauge-analyzed precipitation data provided by JMA.
This study uses the Weather Research and Forecasting (WRF) model (version 4.1.1) (Skamarock et al. 2019). The configuration of the simulations used in the present study are described in the Supplement.
The meteorological initial and lateral boundary conditions were sourced from the Global Forecast System of the National Centers for Environmental Prediction at a grid resolution of 0.25° at 6 h interval. Merged satellite and in situ data Global Daily Sea Surface Temperature (MGDSST), provided by JMA (Sakurai et al. 2005), was used as the lower boundary condition for the control simulation (referred to as CTL-simulation) (Table S1). To assess the response of the Hagibis-induced rainfall to the midlatitude SST, the simulations were performed for 42 h from 12UTC on October 11 before eastern Japan started to receive the rainfall. The SST given at the initial time was not updated throughout the simulations.
To examine the impacts of SSTAs on the precipitation associated with Hagibis, another WRF experiment was carried out under the same meteorological initial and lateral boundary conditions as used for CTL-simulation, but the climatological-mean MGDSST over the 30-y period, from 1982 to 2011, was prescribed as the lower boundary condition (referred to as CLM-simulation). Another experiment was conducted using SST that was eliminated warm SSTAs over the Kuroshio (Fig. S2a) to examine the impacts (referred to as KC-simulation). Additionally, to investigate the impacts of the SSTAs in the Oyashio region on precipitation, we conducted the other simulation prescribing the SST whose gradient had been artificially smoothed only over the specific domain, as indicated by a green box in Fig. 1b (referred to as SMTHsimulation).

Influence of Warm SST in the Oyashio Region on Rainfall Distribution of Typhoon Hagibis (2019)
windward side of the mountains along the southern coast of eastern Japan (Figs. 2a and 2b) due to the orographic effect acting on the southeasterlies (Fig. S1). After landfall, heavy rainfall spread northward along the east coast of Japan (

Overview of Typhoon Hagibis
Figures 2a−2d show the time evolution of the 6 h accumulated rainfall from 00UTC to 18UTC on October 12. As Hagibis approached Japan, large amounts of rainfall were observed on the  -h accumulated rainfall (mm) measured by JMA radar-AMeDAS from 00, 06, 12, and 18UTC, respectively, on 12 October. Black line with cross mark indicates the JMA best track for Hagibis. (e)−(h) Same as in (a)−(d), respectively, but for vertically integrated water vapor (kg m −2 ) (shading) and its transport (kg m −1 s −1 ) (arrows) from the surface to 300 hPa, and geopotential height at 500 hPa (contours) of the analysis data. Contour interval is 50 m. (i)−(l) Same as in (a)−(d), respectively, but for air-sea thermal contrast (°C) (shading), surface wind at 10 m (m s −1 ) (arrows), magnitude of 975 hPa θ gradient (K 100-km −1 ) (thick green lines) more than 12 K 100-km −1 , and sea level pressure (contour) from the analysis data. Contour interval is 10 hPa.
water accompanied by high relative humidity extending throughout the depth of the troposphere and a lower-troposphere lapse rate exceeding the moist adiabatic lapse rate contributed to the heavy rainfall associated with Hagibis. Figures 2e−2h show the time evolution of water vapor and its transport both integrated from the surface to 300 hPa as well as geopotential height at 500 hPa derived from the MSM analysis. Humid air was observed almost uniformly around the center of Hagibis before its landfall (Figs. 2e and 2f), but the distribution gradually became axially asymmetric (Figs. 2g and 2h) as an indication of its transition into an extratropical cyclone.
The interaction of a TC with a pre-existing baroclinic zone has been known to result in its extratropical transition (Harr and Elsberry 2000;Jones et al. 2003;Kitabatake 2008). Before the landfall of Hagibis, a near-surface front was observed along the Pacific coast of eastern Japan, extending northeastward into the Oyashio region (Figs. 2i−2l). The front over the Oyashio region had been associated with a pre-existing extratropical cyclone (JMA 2019). To the south of this front, negative values of air-sea thermal contrast were observed as an indication that the moist southerlies associated with Hagibis were warmer than the underlying ocean. Above the stable boundary layer, deep convection is likely to occur in a conditionally unstable warm air mass in the mid-troposphere (Kitabatake 2008). On the contrary, positive values of air-sea thermal contrast were observed both on the western side of Hagibis and to the north of the LLF under the cool low-level northerlies. Figure 3 compares the accumulated rainfall for 24 h from 00UTC on 12 October between the observations (Fig. 3a) and CTL-simulation (Fig. 3b). The comparison indicates that the While the rainfall maxima south of 38°N were observed mostly on the windward slope of the mountains, a swath of heavy rainfall north of 38°N was located to the west of the LLF (Fig. 4). As Hagibis moved northward, the LLF developed just inland of the east coast (Fig. 4e), extending along the coast up to approximately 40°N (Fig. 4f). Along this front, low-level southeasterlies associated with Hagibis and cool northeasterlies are confluent to reinforce the cross-frontal thermal contrast. Similar frontogenetical features have been reported for heavy rainfall during the extratropical transition of Hurricane Floyd (1999) and Matthew (2016) (Atallah and Bosart 2003;Colle 2003;Powell and Bell 2019).

Control simulation
To understand the frontogenetical processes associated with the LLF, the following frontogenesis equation is evaluated at pressure levels: where u, v, and w are the zonal, meridional, and vertical wind velocities, respectively. θ is potential temperature and Ñ h represents horizontal derivative. In (1), LHS is the frontogenesis function (F ), while the first, second, third, and forth terms with square brackets on RHS represent the confluence, shearing, tilting, and diabatic terms, respectively. They are shown in Figs. 5a−5e for the CTL-simulation averaged from 12UTC to 18UTC on 12 October, at the 975-hPa level, where the horizontal θ gradient was sharpest (Fig. S5a). The southeasterlies and northeasterlies are frontogenetical as they are confluent over the LLF indicated by the pronounced horizontal θ gradient (Figs. 5c and 5f). Along the LLF, the offshore northeasterlies veer to northerlies along the Pacific coast of northeastern Japan (Fig. 5c). These northerlies confluent with the southeasterlies lead to the frontogenetical shear over the LLF along the coast (Figs. 5b and 5f). The cool air mass advected by the northeasterlies is warmed up due to turbulent mixing within the atmospheric boundary layer around the LLF (Figs. S5b− S5d), inducing the frontolysis over the LLF (Figs. 5e and 5f). The tilting term linked with updraft ( Fig. S5d) acts as frontolysis on the onshore side but frontogenesis on the offshore side (Figs. 5d and 5f), as the warm, moist airflow from the southeast is lifted up over the cool, dry airflow from the northeast over the frontal zone, in yielding latent heat release. In the coastal region (Figs. 5a and 5f), the frontogenetic contributions of the confluence, shear and diabatic terms are largely offset by the strong frontolytic effect by tilting term. On the offshore side of the front, by contrast, all the terms but the diabatic term contribute positively to frontogenesis. Figure 3d presents the accumulated rainfall for 24 h from 00UTC on 12 October in the CLM-simulation to reveal the impacts of SSTAs over the WNP on the precipitation associated with Hagibis. Its difference between the CTL-and CLM-simulations is widespread over eastern Japan (Fig. 3e), including a notable rainfall difference resulting from the onshore shift of the heavy precipitation area along the east coast north of 38°N. This onshore shift of the heavy precipitation area is related to the corresponding shift of the LLF (Figs. 3c and 3f). We argue that the warm SSTAs over the WNP (Fig. 1a) had the potential to push the rainfall caused by Hagibis inland. Figures 6a−6i show time evolution of the differences in surface heat fluxes (SHF) as the sum of sensible and latent heat flux, θ and moisture at the 975-hPa level between the CTL-and CLM-simulations. At 00UTC on 12 October, the differences in θ and moisture as well as SHF are accompanied by the cyclonic circulation associated with Hagibis (Figs. 6a, 6d, and 6g), which is attributable to its stronger intensity simulated under the abovenormal SST over the WNP (Fig. S4). Furthermore, the warmer SST over the WNP acts to increase near-surface temperature and moisture associated with the southeasterlies off eastern Japan (Figs. 6f and 6i). The associated differences in southeasterly moisture flux off the east coast of Japan eventually yield the onshore shift of LLF in the CTL-simulation relative to CLM-simulation (Figs. 5f, 5g and 5i), leading to an increase in rainfall along the Pacific coast of eastern Japan relative to the CLM-simulation. The warm positive SSTAs over the Kuroshio also contribute partially to an increase in rainfall along the coast (Fig. S2), by acting to sustain the intensity of Hagibis (Fig. S4). However, the corresponding differences from the KC-simulation are much less compared with those from the CLM-simulation, suggesting that the positive SSTAs over the WNP south of the Kuroshio could be more influential in the intensity.

Sensitivity experiment
Meanwhile, at 06UTC on 12 October, remarkable θ and mois-ture differences are also evident offshore of the east coast north of 38°N (Figs. 6e and 6h), in association with a significant increase in SHF around [42°N, 144°E] over the Oyashio region in the CTL-simulation relative to the CLM-simulation (Fig. 6b). The SHF difference is not significant at 00UTC on 12 October (Fig.  6a), when the northeasterlies are relatively weak (Fig. 6d). Until 12UTC on 12 October, the SHF difference is still significant (Fig.  6c), while both θ and moisture increase further under the enhanced northeasterlies along the east coast north of 38°N (Figs. 6f and 6i). The anomalous θ is also accompanied by the stronger southerlies and onshore winds (Fig. 5i), yielding onshore shift of the LLF. This suggests that the onshore LLF shift resulted from both the warm southeasterlies associated with Hagibis and the warmer northeasterlies in response to the higher SST over the Oyashio region.
To further investigate the impacts of the SSTAs in the Oyashio region on the coastal precipitation, we present the accumulated rainfall for 24 h from 00UTC on 12 October in the SMTH-simulation (Fig. 3g), in addition to its difference in the CTL-simulation relative to the SMTH-simulation (Fig. 3h). Though less pronounced, the onshore shift of LLF (Fig. 3i) and the corresponding increase in the accumulated rainfall (Fig. 3h) along the east coast north of 38°N are similar to their counterpart from the CLMsimulation (Figs. 3e and 3f). However, the precipitation difference south of 38°N is not significant. The time evolution of the θ and moisture differences between CTL-and SMTH-simulations and the corresponding SHF difference (Figs. 6j−6r) demonstrate that the warmer northeasterlies in response to the higher SST over the Oyashio region are influential along the east coast (Fig. 5j), leading to the onshore shift of the LLF (Fig. 3i) through the slight onshore shift of the maxima of the net frontogenesis, although the relative importance among the individual terms does not significantly change between the CTL-and SMTH-simulations (Figs. 5f and 5h).
Furthermore, we have conducted additional sensitivity experiments where the amplitude of the positive SSTAs observed in the Oyashio region (Fig. 1b) was artificially increased or decreased while flipping its sign. The simulated anomalies in the accumulated rainfall in the coastal region of northeastern Japan north of 38°N are positive (negative) in response to the positive (negative) SSTAs, and the amplitude of the rainfall anomaly tends to increase with that of the SSTAs (Fig. 7). The accumulated precipitation over the region marked with the green box in Fig. 7 for each of the simulations tends to increase with the SSTA over the Oyashio region (Table S1). The fraction over land also shows a similar increasing tendency. However, the relationship is not necessarily linear because the orographic effect acting on the southeasterlies is sensitive to the shift of LLF. These experiments demonstrate the high sensitivity of the rainfall anomaly in the coastal region to SST over the Oyashio region, seemingly through the corresponding sensitivity of the position of the LLF (not shown). We therefore argue that the anomalous warmth of the Oyashio may contribute, at least in part, to the extreme 24 h precipitation by Hagibis observed along the coast of northeastern Japan, which was even in excess of the climatological monthly rainfall in October.
We have assessed the impacts of the SSTAs on the rainfall distribution associated with Hagibis through numerical experiments. Our experiments have demonstrated that the warm SSTAs around the Oyashio were likely to have the potential to push the LLF inland and thereby act to enhance the rainfall along the Pacific coast of northeastern Japan. In fact, our additional sensitivity experiments have also demonstrated a clear tendency for the LLF to shift inland (offshore) under warmer (cooler) SST around the Oyashio region, influential the coastal rainfall. We therefore consider that the anomalous warmth of the Oyashio as well as of the WNP during the autumn of 2019 was likely to contribute to the observed extreme rainfall.
Despite substantial improvement of TC track forecast in recent decades, forecasting local precipitation associated with the landfall of a TC system still remains challenging. This may be attributable partially to difficulty in forecasting the position of an associated front that forms during the transformation of the TC into an extratropical cyclone. Additionally, the currently available satellite-based SST datasets are unable to accurately reproduce the observed fine-scale spatiotemporal SST distributions over the KOE (Kawai et al. 2015). Thus, uncertainties in SST data may also lead to forecast errors of precipitation distribution associated with TCs, including their transformation stage into extratropical cyclones.  Table S1.