Diagnosis of Tropical Cyclone Intensity and Structure ...

R. OYAMA et al.2018 3

1. Introduction

A tropical cyclone (TC) is generally identified by its low-pressure center and axisymmetric structure, which comprises eyewalls and spiral rainbands. The strong winds and heavy rain associated with the eyewalls and spiral rainbands cause disastrous destruction of human infrastructures. The area within a radial distance 2 – 3

Journal of the Meteorological Society of Japan, Vol. 96B, pp. 3−26, DOI:10.2151/jmsj.2017-024, 2018Special Issue on Meteorology and Climate Change Studies by Using the Geostationary Meteorological Satellite Himawari-8

Diagnosis of Tropical Cyclone Intensity and Structure Using Upper Tropospheric Atmospheric Motion Vectors

Ryo OYAMA, Masahiro SAWADA

Meteorological Research Institute, Japan Meteorological Agency, Tsukuba, Japan

and

Kazuki SHIMOJI

Numerical Prediction Division, Japan Meteorological Agency, Tokyo, Japan

(Manuscript received 14 April 2017, in final form 19 July 2017)

Abstract

The high temporal and spatial resolutions of geostationary satellite observations achieved by recent techno-logical advancements have facilitated the derivation of atmospheric motion vectors (AMVs), even in a tropical cyclone (TC) wherein the winds abruptly change. This study used TCs in the western North Pacific basin to investigate the ability of upper tropospheric AMVs to estimate the TC intensity and structure. We first examined the relationships between the cloud-top wind fields captured by 6-hourly upper tropospheric AMVs derived from images of the Multi-functional Transport Satellite (MTSAT) and the surface maximum sustained wind (MSW) of the Japan Meteorological Agency’s best-track data for 44 TCs during 2011–2014. The correlation between the maximum tangential winds of the upper tropospheric AMVs (UMaxWinds) and MSWs was high, approximately 0.73, suggesting that the cyclonic circulation near the cloud top was intensified by the upward transport of absolute angular momentum within the TC inner core. The upper tropospheric AMVs also revealed that the mean radii of UMaxWinds and the maximum radial outflows shifted inward as the TC intensification rate became large, imply-ing that the low-level inflow was strong for TCs undergoing rapid intensification. We further examined the pos-sibility of estimating the MSW using 30-min-interval UMaxWinds derived from Himawari-8 target observations, which have been used to track TCs throughout their lifetimes. A case study considering Typhoon Lionrock (1610) showed that the UMaxWinds captured the changes in the cyclonic circulation near the cloud top within the inner core on a timescale shorter than 1 day. It was apparent that the increase in the UMaxWind was associated with the intensification of the TC warm core and the shrinkage of UMaxWind radius. These results suggest that Himawari-8 AMVs include useful information about TC intensification and related structural changes to support the TC intensity analysis and structure monitoring.

Keywords tropical cyclone; nowcasting; atmospheric motion vector; satellite observation; Himawari

Corresponding author: Ryo Oyama, Meteorological Resear-ch Institute, Japan Meteorological Agency, 1-1, Nagamine, Tsukuba, Ibaraki 305-0052, JapanE-mail: [email protected] Advance Published Date: 10 August 2017

©The Author(s) 2018. This is an open access article published by the Meteorological Society of Japan under a Creative Commons Attribution 4.0 International (CC BY 4.0) license (http://creativecommons.org/license/by/4.0).

Journal of the Meteorological Society of Japan Vol. 96B4

times the radius of maximum wind (RMW) from the TC center is generally called the inner core, within which the tangential wind velocity reaches a maxi-mum and strong updrafts exist (Li and Wang 2012). The deep convections within a TC inner core convey the mass and absolute angular momentum upward, with the release of much latent heat. This process plays an essential role in TC development.

The wind structure of a TC can be described as a combination of cyclonic circulation (primary cir-culation) and radial–vertical circulation (secondary circulation). The primary circulation is intensified by low-level inflows, which convey large absolute angu-lar momentum toward the TC center. The low-level convergence, increased by the inflow in the boundary layer, enhances the inner core convections (Sawada and Iwasaki 2007; Rogers 2010). In contrast, the secondary circulation comprises low-level inflows, updrafts in the TC inner core, and radial outflows near the cloud top. The secondary circulation develops the TC cyclonic vortex (i.e., the primary circulation) ver-tically around the inner core and intensifies the warm core near the TC center by the latent heat release (Vigh and Schubert 2009; Houze 2010).

TC wind structures have been investigated in many numerical studies based on non-hydrostatic models. Bryan and Rotunno (2009) investigated the trajectory of an air parcel passing through the position of the maximum wind speed in a TC and showed that the air parcel moves upward along the isosurfaces of absolute angular momentum and entropy, which tilt outward. Stern and Nolan (2011) studied the vertical profiles of tangential winds with respect to the RMW and the maximum sustained wind (MSW) using the three- dimensional Doppler wind field data obtained from seven storms and simulated theoretical vortices. They found that the vertical decay rate of the maximum tangential wind normalized by the tangential wind at a height of 2 km was nearly constant. The results of these previous studies are consistent with the fact that TC wind fields in the middle and upper troposphere are physically related to the surface wind field.

The improved estimates of TC intensity through research and development will be necessary to prevent and mitigate disasters caused by TCs. The Dvorak technique (Dvorak 1975, 1984), which is based on the TC cloud pattern observed in satellite infrared images, is a method that has been used by TC forecast centers, such as the Japan Meteorological Agency (JMA), the Joint Typhoon Warning Center (JTWC), and the National Hurricane Center (NHC), to estimate the TC intensity. This technique is known to provide reliable

TC intensity estimates in the vast majority of TC cases (Velden et al. 2006). However, the Dvorak technique can struggle in certain situations wherein TC struc-tures may fluctuate under the central dense overcast (CDO) or in association with rapid intensity changes.

Recent technological advancements have enabled observations with high temporal and spatial resolu-tions to be made by geostationary satellites, such as the Multi-functional Transport Satellites (MTSATs), which was functional between 28 June 2005 and 07 July 2015, and Himawari-8, which is the first of new-generation geostationary satellites and has been in operation since 07 July 2015 (Bessho et al. 2016). In particular, high-temporal-resolution satellite images with time intervals shorter than approximately 15 min facilitate improved derivations of atmospheric motion vectors (AMVs; Velden et al. 2005; Oyama 2015), even in the case of mesoscale phenomena character-ized by winds that change abruptly. It should be noted that Himawari-8 can provide limited-domain imagery for targeted TCs within view at sampling intervals of 2.5 min.

AMVs are wind products that are derived by tracking clouds and water vapor patterns in successive satellite images. They have been used not only for numerical weather prediction (NWP) (Warrick 2016) but also for atmospheric wind analysis (Molinari and Vollaro 1989; Apke et al. 2016). For NWP, centers such as the European Centre for Medium-Range Weather Forecasts (ECMWF), JMA, and the National Centers for Environmental Prediction (NCEP) of the National Oceanic and Atmospheric Administration (NOAA) have improved their NWP products using AMV data for the initial analysis (Langland et al. 2009; Yamashita 2012; Salonen and Bormann 2014; Wu et al. 2014). For the TC analysis, Oyama et al. (2016b) demonstrated the capability of MTSAT upper tropospheric AMVs to detect the changes in the upper tropospheric wind fields within Typhoon Danas (1324) during its intensification, i.e., the increasing radial outflows and tangential winds around the cloud top, which were associated with convective bursts (CBs; Riehl and Malkus 1961; Steranka et al. 1986).

AMVs have a great advantage compared to in situ observations in terms of temporal resolution and data coverage. In particular, the upper tropospheric AMVs, which can be obtained throughout the TC lifetimes, are expected to contribute to TC analysis and mon-itoring. The purpose of this study was to identify the characteristics of upper tropospheric winds that presage TC intensification by the analysis of the upper tropospheric AMVs of TCs that have occurred

R. OYAMA et al.2018 5

in the western North Pacific basin. For this analysis, we investigated the relationship between the upper tropospheric wind field and the MSW. The results are expected to contribute to the elucidation of the TC intensification process and verifications of numerical studies. In addition, we examined the possibility to extract information about TC intensification and related structural changes from the upper tropospheric AMVs, aiming to propose a method that supports the operational TC intensity analysis and structure moni-toring. This new approach, which is different from the Dvorak technique, will provide new information that will help to improve the TC analysis.

This paper consists of seven sections. Following this introduction, Sections 2 and 3 describe TCs, the observational/numerical data, and the methods used herein. In Section 4, the characteristics of TC wind fields near the cloud top are investigated using MTSAT AMVs with respect to TC intensity and intensification rate. Section 5 compares MTSAT and Himawari-8 AMVs with respect to the quality and amount of data. Section 6 examines the feasibility of estimating MSWs using the upper tropospheric AMVs derived from the Himawari-8 target observations, fo-cusing on the changes in the TC wind and temperature structures. Finally, Section 7 provides a summary and discussion.

2. Studied TCs

This study examined TCs that occurred in the western North Pacific basin during 2011 – 2016 and were observed by MTSAT and/or Himawari-8. TCs in 2011 – 2014 were observed by MTSAT because Himawari-8 did not supersede MTSAT until 07 July 2015. To exclude considerably weak TCs with a short lifetime, we chose TCs with the lifetime maximum MSW greater than 20 m s−1. In addition to this screen-ing, TCs for which there were less AMV data near the TC center were excluded in advance. Specifically, only TCs for which the AMV data covered more than 50 % of a circular area with a radius of 200 km from the TC center were used. The possible reasons behind why AMV data were sparse near the center of some TCs are as follows: (i) the vague texture of clouds to be tracked in successive satellite images, e.g., due to the presence of distinct central dense overcast, (ii) the occurrence of image navigation errors, and (iii) the abrupt changes in atmospheric flows near the TC center. Consequently, 44 of 100 TCs occurring during 2011 – 2014 were studied using MTSAT AMVs (Table 1).

The derivations of AMVs using the Himawari-8

target observations were conducted based on the requirements of this study (Table 1). That is, we identified the differences between the MTSAT and Himawari-8 AMVs with respect to the quality and amount of data considering Typhoon Goni (1515), which existed in the period when both MTSAT and Himawari-8 observations were made, i.e., from July 2015 to March 2016. Typhoon Goni (1515) was a TC that occurred from 15 August 2015 to 25 August 2015 (see Section 5). We examined Typhoon Goni because (i) it was traced throughout its lifetime by the Himawari-8 target observations and (ii) its lifetime was relatively long, i.e., approximately 11 days. In addition to Typhoon Goni, Section 6 includes a case study of Typhoon Lionrock (1610), which was cap-tured by the Himawari-8 target observations in August 2016. Table 1 lists the 46 TCs that were selected for this study, and Fig. 1 shows their tracks.

3. Data and methods

3.1 DataTable 2 lists the observational/numerical data used

in this study. To locate the TC center positions, we used the metrics of intensity [i.e., the 10-min average MSW and the minimum sea-level pressure (MSLP) (also referred to as the central pressure)], and the smallest radius of a 30-kt wind (R30) based on the best-track data from the Regional Specialized Mete-orological Center Tokyo (RSMC Tokyo) – Typhoon Center of JMA (http://www.jma.go.jp/jma/jma-eng/jma-center/rsmc-hp-pub-eg/besttrack.html). For the analysis of the wind components of the AMVs (Sub-section 3.4), hourly TC positions were estimated by linear interpolation of the 6-hourly best-track data. The best-track analysis used the following intensity categories: (i) Tropical Depression [TD; MSW < 34 kt (17 m s−1)], (ii) Tropical Storm [TS; MSW = 34 – 48 kt (25 m s−1)], (iii) Severe Tropical Storm [STS; MSW = 48 – 64 kt (33 m s−1)], and (iv) Typhoon (TY; MSW > 64 kt). In this study, the MSW change per day was defined as the TC intensification rate.

The TC intensities estimated using the Dvorak tech-nique, i.e., the Current Intensity number (CI number), for the researched TCs were used to verify the results in Sections 5 and 6. The RSMC Tokyo converts the CI number into MSLP and MSW using a reference table proposed by Koba et al. (1990).

In this study, we used the images recorded by the MTSAT Japanese Advanced Meteorological Imager (JAMI) and Himawari-8 Advanced Himawari Imager (AHI). MTSAT has five bands: a visible band (VIS: wavelength, 0.63 µm), a shortwave infrared band (IR4:

Journal of the Meteorological Society of Japan Vol. 96B6

Table 1. Configurations of the tropical cyclones investigated in this study: (a) 2011, (b) 2012, (c) 2013, (d) 2014, and (e) 2015 and 2016. MT/H8 indicates that the AMVs were derived using MTSAT (Himawari-8) images.

(a)JMA typhoon

number Typhoon name Analysis Period Lifetime Max MSW (m s−1)

Satellite for AMVs

11021104110511081112111511171119

SongdaHaimaMeari

Nock-tenTalasRokeNesat

Nalgae

1200 UTC 21 May1200 UTC 21 Jun0000 UTC 22 Jun0000 UTC 26 Jul0000 UTC 25 Aug0600 UTC 13 Sep0000 UTC 24 Sep1800 UTC 27 Sep

––––––––

0600 UTC 29 May1800 UTC 24 Jun0600 UTC 27 Jun0000 UTC 31 Jul0600 UTC 05 Sep0600 UTC 22 Sep1800 UTC 30 Sep1800 UTC 04 Oct

54.120.630.925.825.843.841.248.9

MT

(b)12021203120412051207120812091210121112161217121912211223

SanvuMawarGucholTalim

KhanunVicenteSaola

DamreyHaikuiSanba

JelawatMaliksi

PrapiroonSon-tinh

0600 UTC 22 May1800 UTC 01 Jun1200 UTC 13 Jun0600 UTC 17 Jun0600 UTC 16 Jul1200 UTC 21 Jul0000 UTC 28 Jul1200 UTC 28 Jul0000 UTC 03 Aug0000 UTC 11 Sep1800 UTC 20 Sep0600 UTC 01 Oct1200 UTC 07 Oct1200 UTC 23 Oct

––––––––––––––

1800 UTC 27 May0600 UTC 06 Jun0000 UTC 20 Jun1800 UTC 20 Jun0000 UTC 19 Jul1800 UTC 24 Jul0600 UTC 03 Aug1200 UTC 03 Aug1200 UTC 09 Aug0000 UTC 18 Sep1200 UTC 01 Oct0600 UTC 04 Oct1200 UTC 19 Oct0600 UTC 29 Oct

30.938.651.525.825.841.236.136.133.556.756.725.846.443.8

MT

(c)1311131213151317131813191320132113231324132713281329

UtorTrami

Kong-reyToraji

Man-yiUsagiPabukWutipFitowDanas

FranciscoLekimaKrosa

1800 UTC 09 Aug0000 UTC 18 Aug0600 UTC 26 Aug1800 UTC 01 Sep0000 UTC 13 Sep1800 UTC 16 Sep0600 UTC 21 Sep0600 UTC 27 Sep1800 UTC 30 Sep0600 UTC 04 Oct0600 UTC 16 Oct1800 UTC 20 Oct1800 UTC 29 Oct

–––––––––––––

1200 UTC 15 Aug1800 UTC 22 Aug0000 UTC 30 Aug0000 UTC 04 Sep1200 UTC 16 Sep0600 UTC 23 Sep0000 UTC 27 Sep0000 UTC 01 Oct0600 UTC 07 Oct0000 UTC 09 Oct0600 UTC 26 Oct1200 UTC 26 Oct0600 UTC 04 Nov

54.130.928.325.833.556.730.933.538.646.454.159.238.6

MT

(d)140314051408141014141415141614171420

FaxaiTapah

NeoguriMatmo

FengshenKalmaegi

Fung-wongKammuri

Nuri

1200 UTC 28 Feb0000 UTC 28 Apr1800 UTC 03 Jul1200 UTC 17 Jul1800 UTC 06 Sep0600 UTC 12 Sep1200 UTC 17 Sep1200 UTC 24 Sep0000 UTC 31 Oct

–––––––––

1800 UTC 05 Mar0000 UTC 01 May0000 UTC 11 Jul0600 UTC 25 Jul1800 UTC 10 Sep1200 UTC 17 Sep0000 UTC 24 Sep0600 UTC 30 Sep1800 UTC 06 Nov

33.525.851.536.830.938.623.225.856.7

MT

(e)15151610

GoniLionrock

0000 UTC 15 Aug0400 UTC 23 Aug

––

1200 UTC 25 Aug1400 UTC 30 Aug

48.946.4

MT/H8H8

R. OYAMA et al.2018 7

Fig. 1. Tracks of tropical cyclones investigated in this study: (a) 2011, (b) 2012, (c) 2013, (d) 2014, and (e) 2015 and 2016. The black, blue, green, and red lines indicate Tropical Depressions (TDs), Tropical Storms (TSs), Severe Tropical Storms (STSs), and Typhoons (TYs), respectively.

Journal of the Meteorological Society of Japan Vol. 96B8

3.8 µm), two infrared bands (IR1: 10.8 µm and IR2: 12.0 µm), and a water vapor band (WV: 6.8 µm). The spatial resolution of the MTSAT images was 1 km for the VIS band and 4 km for the IR1, IR2, IR4, and WV bands. The interval of MTSAT observations was usually 30 min for the Northern Hemisphere (NH); however, it was 15 min for times around 0000, 0600, 1200, and 1800 UTC for the derivation of AMVs (Imai 2006).

The Himawari-8 AHI has 16 bands. These include three visible bands (Band01: wavelength, 0.47 μm; Band02: 0.51 μm; Band03: 0.64 μm), three near- infrared bands (Band04: 0.86 μm; Band05: 1.6 μm; Band06: 2.3 μm), a shortwave infrared band (Band07: 3.9 µm), three water vapor bands (Band08: 6.2 µm; Band09: 7.0 µm; Band10: 7.3 µm), an SO2 band (Band11: 8.6 µm), an O3 band (Band12: 9.6 µm), three infrared bands (Band13: 10.4 µm, Band14: 11.2 µm; Band15: 12.4 µm), and a CO2 band (Band16: 13.3 µm). The spatial resolution of Himawari-8 images is 0.5 or 1 km for the visible bands and 2 km for the other bands. The image intervals of Himawari-8 are 10

min for the full disk and 2.5 min for the two regions of Japan (referred to as “Region 1” and “Region 2” by JMA) and the target observation area (referred to as “Region 3”). The target observations were used to track a TC from its formation to its dissipation or its transition to an extratropical cyclone using the image window of a square with 1000-km sides (for details of AHI’s functions, see Bessho et al. 2016).

For this study, IR and WV AMVs were derived using the successive images of the infrared and water vapor bands of the MTSAT NH and Himawari-8 target observations, respectively. The MTSAT and Himawari-8 algorithms that were used to derive the AMVs are described in Subsections 3.2 and 3.3, re-spectively.

The infrared brightness temperature (TB) data from MTSAT IR1 and Himawari-8 Band13, which represent the cloud-top temperature of opaque clouds, such as cumulonimbus clouds and dense cirrus clouds, were used as a metric of the deepness of convection within a TC in Section 6.

The vertical wind shear (VWS) in the vicinity of a

Table 2. List of the data used in this study.

Name Explanation Spatial resolution Temporal resolutionMTSAT AMVs AMVs at pressure levels of 100–300 hPa

derived by using MTSAT-2 Northern Hemisphere imagery acquired at 15-min intervals, operationally taken at 0000, 0600, 1200, and 1800 UTC

0.25° latitude and longitude

6 hourly (0000, 0600, 1200, 1800 UTC)

Himawari-8 AMVs AMVs at pressure levels of 100–300 hPa derived by using imagery at 5-min intervals, from the Himawari-8 target observations

0.02° latitude and longitude

30 min

Cloud top temperature of opaque clouds

Brightness temperatures (TBs) of MTSAT IR1 band (10.8 μm) and Himawari-8 Band13 (10.4 μm)

4 km (MTSAT) or 2 km (Himawari-8)

Hourly

JMA best track data TC best-track data from RSMC Tokyo- Typhoon Center (TC position, central pressure, and maximum sustained wind, and the smallest radius of 30-kt wind (R30))

− 6 hourly

Dvorak analysis data Central pressure and maximum sustained wind estimated by Dvorak technique at RSMC Tokyo-Typhoon Center

– 6 hourly

JRA-55 Japanese 55-year Reanalysis data (Kobayashi et al. 2015)

37 levels in 1 to 1000 hPa, 1.25° latitude and longitude

6 hourly

Central pressure estimate by AMSU technique

Central pressure estimates from TC warm core intensity observed by 55-GHz bands of Advanced Microwave Sounding Unit -A (AMSU-A) (Oyama 2014)

48 km at sub-satellite point

Twice daily per satellite at maximum, four satellites, MetOp-B and NOAA-15, 18, and 19 used

R. OYAMA et al.2018 9

TC is known to tilt the TC vortex and consequently weaken the TC intensity (Gallina and Velden 2002; Kaplan and DeMaria 2003; Wong and Chan 2004; Paterson et al. 2005). To consider the influence of VWS on TC vortices, we computed the VWS using the Japanese 55-year reanalysis dataset (JRA-55; Kobayashi et al. 2015). In this study, the VWS is defined as the difference of the wind speeds between pressure levels of 200 and 850 hPa. The differences were averaged over an area within a radius of 600 km from the TC center. The grid size of the JRA-55 dataset is 1.25° in both latitude and longitude.

3.2 MTSAT AMVsFor MTSAT AMVs, the IR and WV AMVs were

derived from the NH IR1- and WV-band images, respectively, using the AMV derivation scheme of the JMA’s Meteorological Satellite Center (MSC; Oyama 2010). IR1 is the infrared band that is known to be the most favorable band for tracking clouds, and WV is a band that is well suited for tracking both clouds and water vapor patterns, which are mainly present in the upper troposphere.

The wind vectors of the AMVs were computed at around 0000, 0600, 1200, and 1800 UTC by tracking high-level clouds, such as cirrus clouds, in the three successive images at intervals of 15 min. The tracking process was conducted using a cross-correlation- matching technique that produced two displacement vectors from the first and second images and from the second and third images. The second vector was used as the final vector. For the derivation of a displace-ment vector, the subpixel displacement of clouds is considered by identifying the position with the maxi-mum correlation in the correlation-matching surface.

The target box (i.e., the image segment used for cloud tracking) was a square of 16 image pixels (the image pixel size for the IR and WV bands is 4 km at nadir) on a 0.25° latitude/longitude grid. The AMVs were assigned to a cloud-top height, which was estimated using the IR1 TB data in the target box, by referring to the first-guess field (6-h forecast) of the JMA’s Global Spectral Model (GSM), the horizontal resolution of which is 20 km. To consider the cloud semi-transparency, the IR1 TB data of the clouds were corrected by the IR-WV intercept method based on the relationship between IR1 TBs and WV TBs in the target box (Nieman et al. 1993; Schmetz et al. 1993).

For the quality control of the AMVs, we used the Quality Indicator (QI) developed by the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT; Holmlund 1998). The QI was

computed based on a temporal consistency check of the difference between the first and second vectors, a spatial consistency check based on a comparison of neighbored vectors, and the consistency with GSM forecast winds.

3.3 Himawari-8 AMVsHimawari-8 AMVs were derived using the images

from Himawari-8 target observations, which were made at intervals of 2.5 min (Subsection 3.1). Because the 2-km resolution of the Himawari-8 images of the infrared and water vapor bands is equivalent to 13.3 m s−1 for 2.5 min, it is technically very challenging to capture the displacement of clouds within the TC areas. Therefore, we decided to derive the IR and WV AMVs using the images at intervals of 5 min.

The IR and WV AMVs were computed at an interval of 30 min using the images of Band13 and Band10, respectively. Band13 is an infrared band with a wavelength similar to that of the IR1 band of MTSAT, and Band10, one of the Himawari-8 water vapor bands, was selected because it yields a larger number of AMV data compared to Band08 or Band09.

The derivations of Himawari-8 AMVs were per-formed using a new scheme that has been developed by the JMA/MSC (Shimoji 2014) to take advantage of Himawari-8’s upgraded functions relative to MTSAT. The new scheme is similar to that used to derive MTSAT AMVs (Subsection 3.2); however, it includes several new techniques and optimizations of parameters for AMV derivations. Herein, only the dif-ferences between the MTSAT AMV and Himawari-8 AMV schemes have been explained (for details of the new scheme for the Himawari-8 AMV, see Shimoji 2014). First, the target box size used for tracking clouds was set to five image pixels (the image pixel size for the infrared and water vapor bands is 2 km at nadir), which is considered sufficient for tracking the meteorological phenomena with horizontal scales as small as 10 km. The grid interval for the AMV deriva-tion was set to approximately 0.02° for both latitude and longitude. With this interval, the target box with five pixel sides overlapped the adjacent boxes, which were as large as half the target box size. Second, the displacement vector for the tracked clouds was com-puted using the average of two correlation-matching surfaces that were derived from the first and second images and from the second and third images, re-spectively, in the cross-correlation matching. The use of the average of two correlation-matching surfaces can lead to a reduction in the noise that may remain in the cloud-tracking process. Third, the height of

Journal of the Meteorological Society of Japan Vol. 96B10

an AMV was determined by a maximum likelihood method based on Himawari-8 observations and the outputs of GSM forecasts. With the introduction of an atmospheric model that includes three cloud layers allocated in the upper, middle, and lower troposphere, the height of the AMV could be optimally determined by solving the likelihood function, for which the displacement vector of an AMV and the observed TBs (Band08, Band09, Band10, Band13, Band15, and Band16), the GSM forecast wind fields, and the simulated TBs obtained by the forward calculation of the radiative transfer model were used.

3.4 Analysis of TC wind field using upper tropospheric AMVs

For the analysis of the wind field near the cloud tops within a TC area, we used complementary IR and WV AMVs that were computed between pressure levels of 100 and 300 hPa to maximize the spatial coverage of the AMV data. The AMV data were only screened using the AMV data with QI greater than 0.3. The QI threshold was determined by referring to the threshold that is imposed to the AMVs that are used to produce the EUMETSAT divergence product (EUMETSAT 2015).

We used the calculated tangential and radial com-ponents of the upper tropospheric AMVs to represent the primary and secondary circulations near the cloud top of a TC. The maximum tangential and radial wind speeds, hereinafter referred to as UMaxWind and UMaxOutflow, respectively (Table 3), were equated

to the maxima of the corresponding azimuthal means in six annuli at radii of 50, 100, 150, 200, 250, and 300 km from the best-track TC center (Fig. 2). We refer to the radii of UMaxWind and UMaxOutflow as Rumw and Ruout, respectively (Table 3). It is possible that in some cases, the TC center (i.e., the location of the MSLP), which is usually collocated with the rotation center near the surface, might be different from the rotation center around the cloud top. This situation might occur because of the tilting of the TC vortex caused by a large VWS or the inaccuracy of the hourly TC center in the best-track data. To evaluate the position error, the best-track TC center was com-pared with the rotation center around the cloud top, which was equated to the position of the rotation peak from the MTSAT upper tropospheric AMVs, that was spatially interpolated on a 0.125° latitude × 0.125° longitude grid for the 44 TCs during 2011 – 2014. This comparison revealed that the mean position difference during the TC intensification phase is 0.5 – 0.6°. The error in the position of the rotation center estimated using the AMVs could increase (i) during the TC

Table 3. Definitions of the parameters used in this study.

Parameter ExplanationUMaxWind The maximum of the azimuthal means

of tangential wind components of AMVs at pressure levels of 100–300 hPa in six annuli within a radius of 300 km (i.e., 50, 100, 150, 200, 250, and 300 km) from the best-track TC center

UMaxOutflow The maximum of the azimuthal means of radial wind components of AMVs at pressure levels of 100–300 hPa in six annuli within a radius of 300 km (i.e., 50, 100, 150, 200, 250, and 300 km) from the best-track TC center

Rumw Radius of UMaxWindRuout Radius of UMaxOutflowIRTB_R200 Averaged infrared-band brightness

temperature (TB) within a radius of 200 km from the TC center

Fig. 2. Illustration of derivations of the maximum tangential wind speed (UMaxWind) and the maximum radial wind speed (UMaxOutflow). UMaxWind (UMaxOutflow) is defined as the maximum azimuthally averaged tangential (radi-al) wind within annuli (±50 km) at the radii of 50, 100, 150, 200, 250, and 300 km (red circles) from the TC center (cross). MTSAT upper tro-pospheric IR AMVs (blue arrows) are displayed in the background of the MTSAT IR1-band brightness temperature (TB) for Typhoon Nuri (1420) at 0600 UTC on 31 October 2014.

R. OYAMA et al.2018 11

formation phase when the cyclonic circulation of the TC was considerably weak, thereby making it difficult to determine the position of the rotation center, and (ii) when the TC was in a large VWS caused by strong environment winds (e.g., jets during the TC decay phase). In fact, the mean position difference during TC intensification slightly depended on the VWS for the 44 TCs. That is, the average of the position differences for the TCs with VWS less than 6 m s−1 was 0.60°, whereas that for the TCs with VWS greater than 6 m s−1 was 0.64°. This result implies that the derivation accuracy of UMaxWind and UMaxOutflow tends to be high when the VWS is small.

3.5 Central pressure estimated from the TC warm core intensity observed by the Advanced Microwave Sounding Unit-A

To verify the TC intensity that was estimated using the upper tropospheric AMVs in Section 6, we used MSLPs estimated on the basis of TC warm core inten-sities observed by the Advanced Microwave Sounding Unit-A (AMSU-A) of the NOAA and MetOp polar- orbiting satellites (Oyama 2014). The MSLP was es-timated from the maximum TB anomaly of AMSU-A 55-GHz-band TB values near the TC center based on hydrostatic equilibrium approximation. The spatial resolution of the AMSU-A data is 48 km at nadir, and the frequency of observations was at most twice daily per satellite. This study used the MetOp-B and

NOAA-15, -18, and -19 satellites.

4.   Characteristics  of  TC  wind  fields  near  the cloud top determined from MTSAT AMVs

This section examines the characteristics of TC wind fields near the cloud tops, which were obtained by using the 6-hourly MTSAT AMVs, with reference to the best-track data. We studied the averaged tan-gential and radial wind components of the upper tro-pospheric AMVs for the 44 TCs that occurred during 2011 – 2014 (Tables 1a – d) with reference to the MSW, TC intensification rate, and R30 of best-track data.

4.1 Relationship between R30 and the radii of UMaxWind and UMaxOutflow

To find the average structural changes of TC cloud-top wind fields associated with the TC intensity and the intensification rate, we examined Rumw and Ruout as well as R30 as functions of the MSW and TC intensification rate (Fig. 3). With respect to the de-pendency of Rumw and Ruout on the MSW (Fig. 3a), Ruout tended to be relatively small for TCs with large MSWs despite the R30’s being relatively large. Spe-cifically, the average Ruouts for MSWs < 20 m s−1 and > 50 m s−1 were 218 and 196 km, respectively. This Ruout pattern may indicate that eyewalls and inner rainbands were formed more robustly at the radii near the TC center for strong TCs. However, the depen-dence of Rumw on MSW was unclear. Furthermore, it

Fig. 3. Dependences of the number of AMV observations, Rumw, Ruout, R30, and Ruout–Rumw, on (a) the best-track maximum sustained wind (MSW) and (b) the TC intensification rate (referred to as IRATE in the panel), averaged for the 44 TCs during 2011 – 2014. Rumw and Ruout were derived using 6-hourly MTSAT up-per tropospheric AMVs.

Journal of the Meteorological Society of Japan Vol. 96B12

was apparent that Ruout and Rumw were larger than the mean radius of the MSW at the surface, 20 – 100 km (Kepert 2010), indicating that the isosurface of the RMW generally tilted outward (Sawada and Iwasaki 2007; Bryan and Rotunno 2009). In addition, it was remarkable that Ruout tended to depart from Rumw for weak MSWs. Specifically, the difference between Ruout and Rumw was 48 km for MSWs < 20 m s−1 and 17 km for MSWs > 50 m s−1. This result may also suggest that a strong TC tends to have a robust inner core comprising eyewalls, which may be true until the vortex builds upward to the tropopause, usually after the TC eye appears in the infrared TB image.

With regard to the dependency of Rumw and Ruout on the TC intensification rate (Fig. 3b), both Rumw and Ruout tended to be small for TCs with high inten-sification rates despite the R30s being relatively small. These characteristics of Rumw and Ruout indicated that strong convection occurred near the TC center for TCs with a high intensification rate (Harnos and Nesbitt 2011; Rogers et al. 2013), indicating that the inflow toward the TC center in the boundary layer was strong for TCs with high intensification rates (Rogers 2010; Wang and Wang 2014).

4.2 Relationship between MSWs and upper tropo-spheric tangential winds

This subsection examines how the cyclonic circula-tion around the TC cloud tops observed with MTSAT AMVs was correlated with surface winds using the 44 TCs that occurred during 2011 – 2014. Table 4 lists the correlation coefficients between the best-track MSWs and several parameters from the azimuthally averaged tangential winds of the upper tropospheric AMVs at radii of 50, 100, 150, 200, 250, and 300

km from the TC center. The positive correlations between the MSWs and the tangential winds of the upper tropospheric AMVs imply that the cyclonic circulation around cloud tops is generally intensified in response to the strength of the surface winds via the upward transport of absolute angular momentum from the surface to the upper troposphere within the TC inner core. The correlation coefficient between the MSW and the maximum tangential wind of upper tropospheric AMVs, i.e., UMaxWind, was larger than the correlation coefficient between the MSW and the tangential wind at each radius, indicating that the radius of UMaxWind could change depending on the distance between the TC center and the eyewalls (i.e., eyewall radius), which can vary with the TC size and intensity, and the inflows in the boundary layer.

4.3 Relationship between the MSW and UMaxWindThis subsection examines the feasibility of estimat-

ing the MSW using UMaxWind, which is a parameter that is best correlated with the MSW (Subsection 4.2). Figure 4 shows the scatter plots between the UMax-Winds of 6-hourly MTSAT AMVs and the best-track MSWs for the 44 TCs during 2011 – 2014. As noted in Subsection 4.2, the correlation coefficient between UMaxWind and MSW was high, approximately 0.73, indicating that the MSW might be estimated from

Fig. 4. Scatter plots between the 6-hourly UMax-Winds of MTSAT AMVs (x-axis) and JMA best-track MSWs (y-axis) for the 44 TCs during 2011 – 2014. The number of data is 1064. The dotted and solid lines represent the first- and second-order polynomial regression equations, respectively. R2 and r denote the coefficient of determination and the correlation coefficient, respectively.

Table 4. Correlation between JMA best-track maximum sustained winds (MSWs) and parameters from the az-imuthally averaged tangential wind components of the upper tropospheric MTSAT AMVs within the annuli at six radii (Fig. 2) from 1016 observations for the 44 TCs that occurred during 2011 – 2014 (Table 1).

Parameter Correlation with best-track MSWs

UMaxWindAzimuthal average at radius of 50 kmAzimuthal average at radius of 100 kmAzimuthal average at radius of 150 kmAzimuthal average at radius of 200 kmAzimuthal average at radius of 250 kmAzimuthal average at radius of 300 km

0.730.240.580.690.690.650.56

R. OYAMA et al.2018 13

UMaxWind. We tried to approximate the relationship between the MSWs and UMaxWinds using two equations: the first- and a second-order polynomial regression equations (Fig. 4). The root-mean-square errors (RMSEs) of the two regression equations were similar, but the RMSE of the second-order polynomial regression (7.42 m s−1) was slightly smaller than that of the first-order regression (7.47 m s–1). The RMSE of the second-order polynomial regression was smaller because of the relatively weak UMaxWinds for large MSWs, which may be explained by the following scenarios: (i) deep convection was less active during the TC decay phase than during the intensification phase and (ii) the cyclonic flows near the cloud top were weakened by strong environment winds, such as high-level jets, when the TC approached mid-latitudes. The RMSEs were larger than that of the Dvorak esti-mate obtained by Koba et al. (1990), 3.6 – 6.2 m s−1, although they were comparable to the MSW estimates based on the TBs of the TRMM microwave imager (TMI), 6 – 8 m s−1 (Hoshino and Nakazawa 2007). The relatively large RMSE of MSW estimates from UMaxWind implies that it is necessary to consider TC structures, such as TC size and cloud-top height, which could depend on the deepness of convection within the TC inner core for precise MSW estimation.

5. Comparison between MTSAT and Himawari-8 AMVs for Typhoon Goni (1515)

In this section, we consider Typhoon Goni (1515) to examine the differences between the MTSAT and Himawari-8 AMVs used in this study with respect to the number of data, heights, and other elements.

Figure 5 shows the snapshots of the MTSAT and Himawari-8 AMVs for Typhoon Goni, which were obtained at 1800 UTC on 23 August 2015. It is re-markable that Himawari-8 AMVs were derived over areas wider than those for MTSAT AMVs despite the much larger grid size (0.25° vs. 0.02°) of the MTSAT AMVs. To identify the general difference in the data coverage between the MTSAT and Himawari-8 AMVs, we computed the frequency of the AMV data within 0.25° squares at 0.25° latitude/longitude intervals for Typhoon Goni from 0000 UTC on 15 August to 1200 UTC on 25 August in 2015 (Fig. 6). Considering the grid size of Himawari-8 data, which is smaller than that of the MTSAT AMV data, the frequencies of Himawari-8 AMVs were summed within squares only once per observation time. We found that the frequency of Himawari-8 AMVs within 0.25° squares was larger than that of MTSAT AMVs for almost all the grids, suggesting that the coverage of the Himawari-8 AMV data was wider than that of the MTSAT AMV data within the TC area. These results imply that Himawari-8 AMVs can capture smaller-scale winds in a TC area more efficiently compared to MTSAT AMVs. It should be noted that because the QI was computed based on a spatial consistency check, the QI is a function of the grid size of the AMV data. The additional comparison of the frequency of data for AMVs with QI greater than 0 also revealed that the coverage of Himawari-8 AMVs was larger than that of MTSAT AMVs. We inferred that the difference in the data coverage between the Himawari-8 and MTSAT AMVs resulted from (i) the higher spatial and temporal resolutions of Himawari-8

Fig. 5. Spatial distributions of MTSAT WV AMVs (left, blue arrows) and Himawari-8 WV AMVs (right, pink arrows) for Typhoon Goni (1515) at 1800 UTC on 23 August 2015.

Journal of the Meteorological Society of Japan Vol. 96B14

images (Subsection 3.1), (ii) the use of a small target box for the derivation of Himawari-8 AMVs, and (iii) the use of averaged matching correlation surfaces of the first–second images and the second–third images for deriving the displacement vector (Subsections 3.2 and 3.3).

Table 5 summarizes a comparison between the MTSAT and Himawari-8 AMVs at common locations in terms of wind speed, tangential and radial winds, and the number of data between pressure levels of 100 and 300 hPa throughout the Typhoon Goni ob-servation period. Note that the total number of AMV data within a given radial interval from the TC center slightly differed between the MTSAT and Himawari-8 AMVs because in some locations, only MTSAT AMVs or Himawari-8 AMVs were assigned below the 300-hPa level. With respect to the WV AMVs (Table 5a), the wind speeds of Himawari-8 AMVs were slightly stronger than those of MTSAT AMVs between pressure levels of 200 and 300 hPa but slightly weaker than those of MTSAT AMVs between pressure levels of 100 and 150 hPa for radii of 100 – 300 km. These differences in the wind speeds were similar to those apparent in the tangential and radial winds.

Some differences in the number of data between the Himawari-8 and MTSAT AMVs were also apparent for WV AMVs (Table 5a). For instance, as the AMV height became lower (higher) and the radius became

smaller (larger), the number of AMVs tended to be larger for Himawari-8 AMVs than for MTSAT AMVs. These differences in the number of data may have resulted from the differences in height assignments (Subsections 3.2 and 3.3). In fact, the cloud type and the optical depth can vary as a function of the radial distance from the TC center because the deep convec-tive clouds usually exist within the eyewalls and inner rainbands near the TC center. This dependence of the cloud type on the radius may explain the differences associated with the amount of AMV data.

For IR AMVs (Table 5b), the wind speed, wind components, and the number of data were similar to those for WV AMVs (Table 5a). The differences in the wind speed, wind components, and the number of data between the Himawari-8 and MTSAT AMVs were also similar to those for WV AMVs. A common dif-ference between the Himawari-8 and MTSAT AMVs, i.e., the larger tangential winds of the Himawari-8 vs. MTSAT AMVs for the radii of 0 – 100 km, implies that the smaller target box size for Himawari-8 AMV derivation and the high temporal and spatial resolution of Himawari-8 images facilitated the derivation of AMVs in the abruptly changing wind field near the TC center.

The 30-min-interval Himawari-8 AMVs calculated from the Himawari-8 target observations were ex-pected to capture the changes in the TC wind fields

Fig. 6. Frequencies of AMV data (0.3 < QI) over a square with sides of 6° latitude (y-axis)/longitude (x-axis) centered on Typhoon Goni (1515) at 0000, 0600, 1200, and 1800 UTC from 15 August 2015 to 25 August 2015. The value of each grid denotes the occurrence percentage of data in a square with sides of 0.25° (a value of 100 % corresponds to 44 observations.).

R. OYAMA et al.2018 15

near the cloud top more precisely than the 6-hourly MTSAT AMVs. The radius–time cross sections of the tangential and radial winds for Typhoon Goni (Figs. 7, 8) revealed that there were clear differences in the radial and temporal variations between the MTSAT and Himawari-8 AMVs. The tangential winds were more finely resolved for the Himawari-8

AMVs than for the MTSAT AMVs and tended to be larger for the Himawari-8 AMVs than for the MTSAT AMVs, especially within a radius of 100 km (Table 5). For the radial wind (Fig. 8), the outward shift of the outflow peak radius was captured more precisely by Himawari-8 AMVs than by MTSAT AMVs. These significant differences in the spatial and temporal

Table 5. Comparison of wind speed, tangential and radial wind components, and the number of data between MTSAT and Himawari-8 AMVs at common locations for Typhoon Goni (1515) from 0000 UTC on 15 August 2015 to 1200 UTC on 25 August 2015: (a) WV AMV and (b) IR AMV.

(a) WV-AMV

Himawari-8 AMV MTSAT AMV Difference (Himawari-8 – MTSAT)Wind speed (m s−1) Wind speed (m s−1) Wind speed (m s−1)

height (hPa)

radius (km) height (hPa)

radius (km) height (hPa)

radius (km)0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300

100 – 150150 – 200200 – 250250 – 300

11.012.515.413.6

12.213.814.614.7

12.312.812.513.0

100 – 150150 – 200200 – 250250 – 300

10.411.411.79.6

13.613.714.111.6

14.213.012.111.9

100 – 150150 – 200200 – 250250 – 300

0.61.13.74.0

−1.40.20.43.0

−1.9−0.3

0.41.0

Total 12.5 13.7 12.6 Total 10.7 13.6 12.6 Total 1.8 0.0 0.0

Tangential wind (m s−1) Tangential wind (m s−1) Tangential wind (m s−1)

height (hPa)

radius (km) height (hPa)

radius (km) height (hPa)

radius (km)0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300

100 – 150150 – 200200 – 250250 – 300

6.69.5

11.76.4

6.810.310.510.5

4.97.27.27.4

100 – 150150 – 200200 – 250250 – 300

6.07.22.43.3

7.89.59.64.9

7.46.46.15.2

100 – 150150 – 200200 – 250250 – 300

0.72.39.33.2

−1.00.70.85.6

−2.60.81.12.2

Total 8.5 9.5 6.8 Total 6.0 9.0 6.2 Total 2.5 0.5 0.5

Radial wind (m s−1) Radial wind (m s−1) Radial wind (m s−1)

height (hPa)

radius (km) height (hPa)

radius (km) height (hPa)

radius (km)0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300

100 – 150150 – 200200 – 250250 – 300

4.34.64.45.2

6.86.45.95.7

8.48.06.56.0

100 – 150150 – 200200 – 250250 – 300

4.04.23.53.7

8.66.34.95.1

9.88.16.96.5

100 – 150150 – 200200 – 250250 – 300

0.40.40.81.4

−1.80.11.00.6

−1.4−0.1−0.4−0.5

Total 4.5 6.3 7.4 Total 4.0 6.5 7.6 Total 0.5 −0.2 −0.2

Number Number Number

height (hPa)

radius (km) height (hPa)

radius (km) height (hPa)

radius (km)0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300

100 – 150150 – 200200 – 250250 – 300

39942218489

115018181074 548

1945415726161463

100 – 150150 – 200200 – 250250 – 300

648350889

10402425 924 194

701457838341073

100 – 150150 – 200200 – 250250 – 300

−249729680

110−607

150354

1244−421

−1218390

Total 1094 4590 10181 Total 1095 4583 10186 Total −1 7 −5

Journal of the Meteorological Society of Japan Vol. 96B16

distributions of AMVs indicated the advantages of Himawari-8 AMVs over MTSAT AMVs with respect to capturing changes of the TC’s primary and second-ary circulations.

To examine the impact of using Himawari-8 AMVs on the diagnosis of the TC wind field, the UMaxWind and UMaxOutflow of Himawari-8 AMVs obtained at intervals of 30 min were compared with those of

MTSAT AMVs obtained at intervals of 6 h (Fig. 9). The UMaxWind of Himawari-8 AMVs tended to be slightly larger than that of MTSAT AMVs and cap-tured detailed fluctuations in the cyclonic flows (Fig. 9a), implying that Himawari-8 AMVs could capture the cyclonic flows near the cloud top within the TC inner core more precisely than MTSAT AMVs (Fig. 7). The local maximum of UMaxWinds from both AMVs

Table 5. Continued.

(b) IR-AMV

Himawari-8 AMV MTSAT AMV Difference (Himawari-8 – MTSAT)Wind speed (m s−1) Wind speed (m s−1) Wind speed (m s−1)

height (hPa)

radius (km) height (hPa)

radius (km) height (hPa)

radius (km)0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300

100 – 150150 – 200200 – 250250 – 300

11.212.815.713.8

13.014.214.715.2

12.513.113.113.3

100 – 150150 – 200200 – 250250 – 300

10.611.111.6

−

13.614.315.212.8

14.013.112.812.6

100 – 150150 – 200200 – 250250 – 300

0.61.74.1−

−0.7−0.2−0.5

2.4

−1.50.00.20.6

Total 12.8 14.1 13.0 Total 10.8 14.1 13.1 Total 2.0 0.0 −0.1

Tangential wind (m s−1) Tangential wind (m s−1) Tangential wind (m s−1)

height (hPa)

radius (km) height (hPa)

radius (km) height (hPa)

radius (km)0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300

100 – 150150 – 200200 – 250250 – 300

7.810.012.07.3

7.410.810.711.0

5.27.77.77.8

100 – 150150 – 200200 – 250250 – 300

6.56.83.1−

8.110.911.16.0

6.77.08.36.4

100 – 150150 – 200200 – 250250 – 300

1.43.38.9−

−0.7−0.1−0.4

5.0

−1.50.6

−0.61.4

Total 9.3 9.9 7.2 Total 6.4 9.8 7.3 Total 2.9 0.2 −0.1

Radial wind (m s−1) Radial wind (m s−1) Radial wind (m s−1)

height (hPa)

radius (km) height (hPa)

radius (km) height (hPa)

radius (km)0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300

100 – 150150 – 200200 – 250250 – 300

4.54.55.54.5

7.66.35.85.6

8.78.06.65.5

100 – 150150 – 200200 – 250250 – 300

4.14.03.2−

8.45.34.53.5

10.07.35.54.9

100 – 150150 – 200200 – 250250 – 300

0.40.52.3−

−0.71.01.32.1

−1.40.71.10.6

Total 4.7 6.4 7.3 Total 4.1 6.4 7.0 Total 0.6 0.1 0.4

Number Number Number

height (hPa)

radius (km) height (hPa)

radius (km) height (hPa)

radius (km)0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300 0 – 100 100 – 200 200 – 300

100 – 150150 – 200200 – 250250 – 300

38941517590

113418441053 530

1672359024141485

100 – 150150 – 200200 – 250250 – 300

753285370

1738219853672

113249482552535

100 – 150150 – 200200 – 250250 – 300

−36413013890

−604−354

517458

540−1358−138

950Total 1069 4561 9161 Total 1075 4544 9167 Total −6 17 −6

R. OYAMA et al.2018 17

Fig. 7. The radius–time cross sections for the azimuthally averaged tangential wind com-ponent of the upper tropospheric AMV for (a) 6-hourly MTSAT AMVs and (b) 30-min-in-terval Himawari-8 AMVs for Typhoon Goni (1515) from 0000 UTC on 15 August 2015 to 1800 UTC on 24 August 2015.

Fig. 8. Same as Fig. 7 but shows radius–time cross sections for the azimuthally averaged radial wind component.

Journal of the Meteorological Society of Japan Vol. 96B18

roughly corresponded to those of the best-track and Dvorak MSWs; however, there were some differenc-es, e.g., from 1200 UTC on 19 August 2015 to 0000 UTC on 21 August 2015 (Fig. 9a). The difference between the UMaxWinds and Dvorak MSW implies that the tangential winds of upper tropospheric AMVs contain information on the TC intensity change, which is independent on the Dvorak TC intensity estimate.

The UMaxOutflow of Himawari-8 AMVs captured detailed fluctuations of the radial outflow near the cloud top compared to that of MTSAT AMVs. It should be noted that the local maximum of UMaxOut-flows from both AMVs roughly corresponded to the local minimum of the averaged infrared TB within a radius of 200 km from the TC center (IRTB_R200 in Fig. 9b), indicating that the increase in UMaxOutflow represented the outward movement of upper tropo-spheric clouds such as anvils from deep convection in the TC inner core.

6. Possibility of estimating the TC maximum surface wind from the maximum tangential wind of Himawari-8 upper tropospheric AMVs

In this section, we use the Himawari-8 target ob-servations of Typhoon Lionrock (1610) to examine the possibility of using UMaxWind for estimating the MSW. Typhoon Lionrock formed as a TD around the southeast part of a monsoon gyre located over Japan on August 17, 2016. The TD then moved counterclockwise with rapid intensification in an extra cyclonic wind field with weak VWS (average = 4.3 m s−1) and turned back near the Minamidaitojima Island because of the approach of a southwesterly jet.

The RMSE and bias of the MSW estimated from Himawari-8’s UMaxWinds via the second-order polynomial regression (Fig. 4) to the best-track data for Lionrock were 10.9 and −1.4 m s−1, respectively. A further investigation of the estimated MSWs at in-tervals of 30 min for Typhoon Lionrock revealed that the time series of UMaxWinds included clear varia-tions with timescales shorter than 1 day (Fig. 10a). It should be noted that the timescale variation apparent

Fig. 9. Analysis results for Typhoon Goni (1515). (a) Time series of UMaxWinds obtained from MTSAT (MT, blue circles and solid line) and Himawari-8 AMVs (H8, orange circles and solid line), and MSWs of the best-track data (black solid line) and Dvorak analysis (black dotted line). (b) Time series of UMaxOutflows obtained from MTSAT and Himawari-8 AMVs; Himawari-8 infrared (Band13) TB averaged within a radius of 200 km from the TC center (IRTB_R200, black solid line).

R. OYAMA et al.2018 19

in UMaxWind was not clear in the time series of the best-track analysis (Fig. 10a), which was quite similar to that of the Dvorak analysis (the graphs are not shown).

A comparison between the MSW estimated from UMaxWind and the central pressure via the AMSU technique (Subsection 3.5) revealed that the local

maximum in the time series of UMaxWind roughly corresponded to the deepening of the central pressure (Fig. 10a), indicating that the warm core was devel-oped around at that time. The spatial distributions of the upper tropospheric AMVs and AMSU-A TB anomalies relative to the average in the annulus be-tween the radii of 600 and 700 km from the TC center

Fig. 10. Analysis results for Typhoon Lionrock (1610). (a) Time series of the MSW estimated from Himawari-8 UMaxWind with the second-order polynomial regression equation (Fig. 4) (blue diamonds), JMA best-track MSW (black dotted line) and MSLP (black solid line), and the MSLP estimated using the Advanced Micro-wave Sounding Unit-A (AMSU-A) data (red circles and line). (b) Time series of UMaxWind (blue diamonds), 4-h running mean of Rumw (green solid line), and R30 of the best-track data (black solid line). (c) Time series of UMaxWind (blue diamonds) and UMaxOutflow (pink diamonds); the Himawari-8 infrared (Band13) TB av-eraged within a radius of 200 km from the TC center (IRTB_R200, orange solid line). Period-A is the period that is investigated in Figs. 11, 12, and 13. The black arrows in (c) denote the local minimum of IRTB_R200 between 1500 UTC and 2300 UTC, as explained in Section 6.

Journal of the Meteorological Society of Japan Vol. 96B20

from 1400 UTC on 25 August 2016 to 1700 UTC on 26 August 2016 (shown as “Period-A” in Fig. 10) indicated that the positive temperature anomaly of the TC warm core at a pressure level of approximately 250 hPa increased when the tangential winds near the TC center increased (Figs. 11a–c, 12a, b). This syn-chronism between the UMaxWind increase and the warm core development suggests that the primary circulation near the cloud top was increased via the upward transport of a large amount of absolute angular

momentum from the surface to the upper troposphere within the inner core, together with the development of the warm core due to the latent heat release and adiabatic warming within the eye (Willoughby 1998; Ohno and Satoh 2015). The warm infrared TB area near the TC center around at the time when the AMSU MSLP reached the minimum implied the formation of the eye (Figs. 13b, c).

In addition to these changes apparent within the TC warm core, it was relatively clear that Rumw

Fig. 11. Spatial distributions of the upper tropospheric winds of the IR and WV AMVs (0.3 < QI) from the Himawari-8 target observations for Typhoon Lionrock (1610) during the period-A shown in Fig. 10. UMax-Wind reached its local maximum at 0900 UTC on 26 August 2016 (c).

R. OYAMA et al.2018 21

computed using the upper tropospheric AMVs reached a local minimum when UMaxWind reached its local maximum (Fig. 10b), indicating that the RMW at the surface shrunk due to the increase of inflow in the boundary layer at such times (Subsection 4.1). Fur-thermore, it was noteworthy that the local minimum and maximum of the cloud-top temperature, which were represented by infrared TB, of the convective and dense cirrus clouds near the TC center roughly corresponded to the local minimum and maximum,

respectively, of UMaxWind (Fig. 10c). Additionally, it should be noted that some of the occurrence times when the cloud-top temperature within a radius of 200 km from the TC center (IRTB_R200 in Fig. 10c) reached its local minimum were between 1500 UTC (0300 LST) and 2300 UTC (0800 LST), which are indicated by black arrows in Fig. 10c. This time inter-val corresponds to the occurrence time of convection peaks associated with the TC diurnal cycle (Muramatsu 1983; Kossin 2002; Dunion et al. 2014). The syn-

Fig. 12. Spatial distributions of the AMSU-A channel 7 (55-GHz band, observation of approximately 250 hPa level) TB anomaly relative to the average within the annulus between 600 and 700 km from the center of Ty-phoon Lionrock (1610) during the period-A shown in Fig. 10.

Journal of the Meteorological Society of Japan Vol. 96B22

chronism between the decrease in the cloud-top tem-perature and the increase in UMaxOutflow (Fig. 10c) implied that the convection deepening within the inner core drove the vertical motion and intensified the TC secondary circulation. This situation was also implied by the outward shift of the low-TB cloud area from the TC center, which was identified by the difference between Figs. 13a and 13c. Based on these results, it was possible that the local maximum of UMaxWind following convection deepening indicated that the

cyclonic circulation near the TC center in the upper troposphere was increased by the vertical develop-ment of the TC primary circulation via the preceding consecutive deep convections (Miyamoto and Takemi 2013; Oyama et al. 2016b).

The results obtained from the case study consider-ing Typhoon Lionrock indicate that the UMaxWind of Himawari-8 AMVs contains information about the TC intensity change, which is independent of the Dvorak TC intensity estimate, suggesting that UMaxWind

Fig. 13. Spatial distributions of the infrared TB of Himawari-8 (Band 13) for Typhoon Lionrock (1610) during the period-A shown in Fig. 10. The observation times of panels (a) – (d) correspond to those of Figs. 11a–d, re-spectively. UMaxWind reached its local maximum at 0900 UTC on 26 August 2016 (c).

R. OYAMA et al.2018 23

could contribute to the operational TC intensity anal-ysis, e.g., as a member of the consensus TC intensity estimate (Velden et al. 2007; Oyama et al. 2016a). In addition, the results suggest that Himawari-8 AMVs could capture the TC structural change related to TC intensification, e.g., the shrinkage of the eyewall radius and the increase in the TC secondary circula-tion enhanced by a deep convection within the TC inner core.

7. Summary and discussion

To investigate the possibility of estimating the TC intensity and characteristic structures from the upper tropospheric AMVs, which represent the winds near the cloud top, this study examined the relationships between the upper tropospheric AMVs derived from the MTSAT and Himawari-8 images and the MSWs of JMA’s best-track data for TCs that occurred in the western North Pacific basin during 2011 – 2016.

Statistical investigations using 6-hourly MTSAT upper tropospheric AMVs for the 44 TCs during 2011 – 2014 revealed that the tangential winds near the cloud top were well correlated with the best-track MSWs. It was apparent that the maximum tangential winds, UMaxWinds, were most correlated with MSWs, indicating that the radius of UMaxWind could change depending on the eyewall radius, which could vary with the TC size and the inflow in the boundary layer. The RMSE of the regressions between UMaxWinds and MSWs was approximately 7.4 m s−1 and larger than the RMSE of the Dvorak estimate, 3.6 – 6.2 m s−1, as evaluated by Koba et al. (1990). It is also noted that the RMSE of the second-order poly-nomial regression equation is slightly smaller than that of the first-order polynomial regression equation, which is considered to result from the relatively weak UMaxWinds during the TC decay phase. It is noteworthy that for TCs that rapidly intensified, the radii of UMaxWind and the maximum radial outflow, UMaxOutflow, tended to be small. These results imply that the inflow in the boundary layer increased and the RMW at the surface and the eyewall radius shrunk when the TC was rapidly intensified.

Finally, the possibility of estimating the MSW using UMaxWind derived from Himawari-8 target observa-tions was examined. It is noteworthy that UMaxWinds at intervals of 30 min captured the changes in the TC structure on a timescale shorter than 1 day; these were the changes that were not clearly recognized in the Dvorak analysis. A case study of Typhoon Lion-rock (1610) revealed several pieces of evidence that indicated a change in the TC intensity: (i) the devel-

opment of the upper tropospheric warm core and (ii) the shrinkage of the radius of UMaxWind when the local maximum of UMaxWind appeared, and (iii) the occurrence of consecutive deep convections within the inner core prior to the local maximum of UMaxWind. It is possible that the temporal variations in the tan-gential winds near the cloud top of Typhoon Lionrock were clearly observed by Himawari-8 AMVs because the VWS around the TC was small throughout its lifetime.

This study revealed several fundamental relation-ships between a TC’s upper tropospheric wind and intensity, indicating that the upper tropospheric AMVs should include information about the TC intensity change and related structural changes. These findings suggest that the upper tropospheric AMVs could con-tribute to the TC intensity analysis and TC structure monitoring. However, several fundamental issues still remain unresolved. It is important to find how the tem-poral variation, which was apparent in UMaxWinds, is related to the TC’s diurnal cycle and others cycles with periods shorter than 1 day (Kossin 2002; Takeda and Oyama 2003). Another necessary task is to inves-tigate the response of the upper tropospheric winds to the lower tropospheric winds, which could vary depending on the robustness of the TC inner core, by evaluating the budgets of absolute angular momentum within a TC (Rozoff et al. 2012). We expect that the numerical studies based on a high-resolution non- hydrostatic model will help in elucidating the pro-cesses and mechanisms that enable the linking of the boundary layer to the upper troposphere for the TC’s diurnal cycle (Navarro and Hakim 2016) and other phenomena. These studies will also provide informa-tion and ideas essential for improving the methods used to estimate the TC intensity and structure via AMVs. In addition to studies focusing on TC intensity estimation, the relationship between the TC intensi-fication rate and the secondary circulation, which is represented by the radial outflow obtained from upper tropospheric AMVs, is another interesting research topic to tackle.

To obtain more insights about the TC intensity change and related structural changes from Himawari-8 AMVs, it is necessary to investigate more TCs that were traced by Himawari-8 target observations. This investigation should include weak TCs with the life-time maximum MSW less than 20 m s−1, which were not examined in this study. In addition, for obtaining more precise estimation of TC intensity from the AMVs, it is necessary to study the dependencies of UMaxWinds on the TC size and intensity as well as

Journal of the Meteorological Society of Japan Vol. 96B24

the cloud-top height. For future improvements of Himawari-8 AMVs, it is important to continue efforts in identifying optimal parameter values for the der-ivation of AMVs, such as the grid size and the time interval (Sohn and Borde 2008; Bresky et al. 2012; Shimoji 2012). Furthermore, it is important to identify the optimal QI threshold for screening the AMV data and seek for the possibility of using the QI without forecast check for this study. Such tasks will enhance the usefulness of Himawari-8 AMVs in the estima-tions of the TC intensity and structure.

Acknowledgments

This study used the MTSAT imagery and ancillary data obtained from the JMA/MSC for computing AMVs. The AMSU-A data were obtained from NOAA’s Comprehensive Large Array-data Steward-ship System (CLASS) website (http://www.class.ncdc.noaa.gov/saa/products/welcome). The authors thank the reviewers and Christopher Velden of the Coop-erative Institute for Meteorological Satellite Studies (CIMSS), University of Wisconsin for their many valuable comments that helped to improve this paper.

References

Apke, J. M., J. R. Mecikalski, and C. P. Jewett, 2016: Analy-sis of mesoscale atmospheric flows above mature deep convection using super rapid scan geostationary satel-lite data. J. Appl. Meteor. Climatol., 55, 1859–1887.

Bessho, K., K. Date, M. Hayashi, A. Ikeda, T. Imai, H. Inoue, Y. Kumagai, T. Miyakawa, H. Murata, T. Ohno, A. Okuyama, R. Oyama, Y. Sasaki, Y. Shimazu, K. Shimoji, Y. Sumida, M. Suzuki, H. Taniguchi, H. Tsuchiyama, D. Uesawa, H. Yokota, and R. Yoshida, 2016: An introduction to Himawari-8/9 – Japan’s new- generation geostationary meteorological satellites. J. Meteor. Soc. Japan, 94, 151–183.

Bresky, W. C., J. M. Daniels, A. A. Bailey, and S. T. Wan-zong, 2012: New methods toward minimizing the slow speed bias associated with atmospheric motion vectors. J. Appl. Meteor. Climatol., 51, 2137–2151.

Bryan, G. H., and R. Rotunno, 2009: Evaluation of an an-alytical model for the maximum intensity of tropical cyclones. J. Atmos. Sci., 66, 3042–3060.

Dunion, J. P., C. D. Thorncroft, and C. S. Velden, 2014: The tropical cyclone diurnal cycle of mature hurricanes. Mon. Wea. Rev., 142, 3900–3919.

Dvorak, V. F., 1975: Tropical cyclone intensity analysis and forecasting from satellite imagery. Mon. Wea. Rev., 103, 420–430.

Dvorak, V. F., 1984: Tropical cyclone intensity analysis using satellite data. NOAA Technical Report NESDIS, 11, 47 pp.

EUMETSAT, 2015: Divergence product: Product guide.

[Available at http://www.eumetsat.int/website/wcm/ idc/idcplg?IdcService=GET_FILE&dDocName=PDF _DIV_FACTSHEET&RevisionSelectionMethod= LatestReleased&Rendition=Web.]

Gallina, G. M., and C. S. Velden, 2002: Environmental ver-tical wind shear and tropical cyclone intensity change utilizing enhanced satellite derived wind information. Proc. 25th Conference on Hurricanes and Tropical Meteorology, Amer. Meteor. Soc., San Diego, CA, 172–173. [Available at https://ams.confex.com/ams/pdfpapers/35650.pdf.]

Harnos, D. S., and S. W. Nesbitt, 2011: Convective structure in rapidly intensifying tropical cyclones as depicted by passive microwave measurements. Geophys. Res. Lett., 38, L07805, doi:10.1029/2011GL047010.

Holmlund, K., 1998: The utilization of statistical properties of satellite-derived Atmospheric Motion Vectors to derive quality indicators. Wea. Forecasting, 13, 1093– 1104.

Hoshino, S., and T. Nakazawa, 2007: Estimation of tropical cyclone’s intensity using TRMM/TMI brightness temperature data. J. Meteor. Soc. Japan, 85, 437–454.

Houze, Jr., R. A., 2010: Clouds in tropical cyclones. Mon. Wea. Rev., 138, 293–344.

Imai, T., 2006: Status of atmospheric motion vector in JMA. Proc. 8th International Winds Workshop. [Available at https://www.eumetsat.int/website/home/News/ ConferencesandEvents/DAT_2043687.html.]

Kaplan, J., and M. DeMaria, 2003: Large-scale character-istics of rapidly intensifying tropical cyclones in the North Atlantic basin. Wea. Forecasting, 18, 1093– 1108.

Kepert, J. D., 2010: Tropical cyclone structure and dynam-ics. Global Perspectives on Tropical Cyclones. World Scientific Series on Asia-Pacific Weather and Climate, vol. 4, Chan, J. C. L., and J. D. Kepert (eds.), World Scientific Publishing, 3–53.

Koba, H., T. Hagiwara, S. Osano, and S. Akashi, 1990: Relationship between the CI-number and central pressure and maximum wind speed in typhoons. J. Meteor. Res., 42, 59–67 (in Japanese).

Kobayashi, S., Y. Ota, Y. Harada, A. Ebita, M. Moriya, H. Onoda, K. Onogi, H. Kamahori, C. Kobayashi, H. Endo, K. Miyaoka, and K. Takahashi, 2015: The JRA-55 Reanalysis: General specifications and basic characteristics. J. Meteor. Soc. Japan, 93, 5–48.

Kossin, J. P., 2002: Daily hurricane variability inferred from GOES infrared imagery. Mon. Wea. Rev., 130, 2260–2270.

Langland, R. H., C. Velden, P. M. Pauley, and H. Berger, 2009: Impact of satellite-derived rapid-scan wind ob-servations on numerical model forecasts of Hurricane Katrina. Mon. Wea. Rev., 137, 1615–1622.

Li, Q.-Q., and Y. Wang, 2012: A comparison of inner and outer spiral rainbands in a numerically simulated tropical cyclone. Mon. Wea. Rev., 140, 2782–2805.

R. OYAMA et al.2018 25

Miyamoto, Y., and T. Takemi, 2013: A transition mechanism for the spontaneous axisymmetric intensification of tropical cyclone. J. Atmos. Sci., 70, 112–129.

Molinari, J., and D. Vollaro, 1989: External influences on hurricane intensity. Part I: Outflow layer eddy angular momentum fluxes. J. Atmos. Sci., 46, 1093–1105.

Muramatsu, T., 1983: Diurnal variations of satellite-measured TBB areal distribution and eye diameter of mature typhoons. J. Meteor. Soc. Japan, 61, 77–90.

Navarro, E. L., and G. J. Hakim, 2016: Idealized numerical modeling of the diurnal cycle of tropical cyclones. J. Atmos. Sci., 73, 4189–4201.

Nieman, S. J., J. Schmetz, and W. P. Menzel, 1993: A comparison of several techniques to assign heights to cloud tracers. J. Appl. Meteor., 32, 1559–1568.

Ohno, T., and M. Satoh, 2015: On the warm core of a trop-ical cyclone formed near the tropopause. J. Atmos. Sci., 72, 551–571.

Oyama, R., 2010: Upgrade of Atmospheric Motion Vector derivation algorithms at JMA/MSC. Meteorological Satellite Center Tech. Note, 54, 1–31. [Available at http://www.data.jma.go.jp/mscweb/technotes/msctech rep54-1.pdf.]

Oyama, R., 2014: Estimation of tropical cyclone central pressure from warm core intensity observed by the Advanced Microwave Sounding Unit-A (AMSU-A). Pap. Meteor. Geophys., 65, 35–56.

Oyama, R., 2015: Characteristics of upper-tropospheric At-mospheric Motion Vectors (AMV) in tropical cyclone areas derived using MTSAT rapid-scan observation data. Tenki, 62, 881–894 (in Japanese with English abstract).

Oyama, R., K. Nagata, H. Kawada, and N. Koide, 2016a: Development of a product based on consensus be-tween Dvorak and AMSU tropical cyclone central pressure estimates at JMA. RSMC Tokyo-Typhoon Center Technical Review, 18, 8 pp. [Available at http://www.jma.go.jp/jma/jma-eng/jma-center/rsmc-hp-pub-eg/techrev/text18-1.pdf.]

Oyama, R., A. Wada, and M. Sawada, 2016b: Intensification of Typhoon Danas (1324) captured by MTSAT upper tropospheric Atmospheric Motion Vectors. SOLA, 12, 135−139.

Paterson, L. A., B. N. Hanstrum, N. E. Davidson, and H. C. Weber, 2005: Influence of environmental vertical wind shear on the intensity of hurricane-strength tropical cyclones in the Australian region. Mon. Wea. Rev., 133, 3644–3660.

Riehl, H., and J. Malkus, 1961: Some aspects of hurricane Daisy, 1958. Tellus, 13, 181–213.

Rogers, R., 2010: Convective-scale structure and evolution during a high-resolution simulation of tropical cy-clone rapid intensification. J. Atmos. Sci., 67, 44–70.

Rogers, R., P. Reasor, and S. Lorsolo, 2013: Airborne Dop-pler observations of the inner-core structural differ-ences between intensifying and steady-state tropical

cyclones. Mon. Wea. Rev., 141, 2970–2991.Rozoff, C. M., D. S. Nolan, J. P. Kossin, F. Zhang, and J.

Fang, 2012: The role of an expanding wind field and inertial stability in tropical cyclone secondary eyewall formation. J. Atmos. Sci., 69, 2621–2643.

Salonen, K., and N. Bormann, 2014: AMVs in the oper-ational ECMWF system. Proc. 12th International Winds Workshop. [Available at http://cimss.ssec.wisc.edu/iwwg/iww12/talks/02_Tuesday/1110_AMVs_in_ECMWF_operational_system_Salonen_v2.pdf.]

Sawada, M., and T. Iwasaki, 2007: Impacts of ice phase processes on tropical cyclone development. J. Meteor. Soc. Japan, 85, 479–494.

Schmetz, J., K. Holmlund, J. Hoffman, B. Strauss, B. Mason, V. Gartner, A. Koch, and L. van de Berg, 1993: Op-erational cloud-motion winds from Meteosat infrared images. J. Appl. Meteor., 32, 1206–1225.

Shimoji, K., 2012: A study on the relationship between spa-tial and temporal image resolutions for AMV deriva-tion with next-generation satellites. Proc. 11th Inter-national Winds Workshop. [Available at http://www. eumetsat.int/website/wcm/idc/idcplg?IdcService= GET_FILE&dDocName=PDF_CONF_P60_S2_07_SHIMOJI_V&RevisionSelectionMethod=Latest Released&Rendition=Web.]

Shimoji, K., 2014: Motion tracking and cloud height assign-ment methods for Himawari-8 AMV. Proc. 12th Inter-national Winds Workshop. [Available at http://www. eumetsat.int/website/wcm/idc/idcplg?IdcService= GET_FILE&dDocName=PDF_CONF_P61_S2_06_SHIMOJI_V&RevisionSelectionMethod=Latest Released&Rendition=Web.]

Sohn, E., and R. Borde, 2008: The impact of window size on AMV. Proc. 9th International Winds Workshop. [Available at http://www.eumetsat.int/website/wcm/ idc/idcplg?IdcService=GET_FILE&dDocName=PDF _CONF_P51_S4_18_SOHN_V&RevisionSelection Method=LatestReleased&Rendition=Web.]

Steranka, J., E. B. Rodgers, and R. C. Gentry, 1986: The relationship between satellite measured convective bursts and tropical cyclone intensification. Mon. Wea. Rev., 114, 1539–1546.

Stern, D. P., and D. S. Nolan, 2011: On the vertical decay rate of the maximum tangential winds in tropical cyclones. J. Atmos. Sci., 68, 2073–2094.

Takeda, T., and R. Oyama, 2003: Periodic time variation of low-TBB cloud area in typhoon. J. Meteor. Soc. Japan, 81, 1497–1503.

Velden, C., J. Daniels, D. Stettner, D. Santek, J. Key, J. Dunion, K. Holmlund, G. Dengel, W. Breskey, and P. Menzel, 2005: Recent innovations in deriving tro-pospheric winds from meteorological satellites. Bull. Amer. Meteor. Soc., 86, 205–223.

Velden, C., B. Harper, F. Wells, J. L. Beven II, R. Zehr, T. Olander, M. Mayfield, C. C. Guard, M. Lander, R. Edson, L. Avila, A. Burton, M. Turk, A. Kikuchi,

Journal of the Meteorological Society of Japan Vol. 96B26

A. Christian, P. Caroff, and P. McCrone, 2006: The Dvorak tropical cyclone intensity estimation tech-nique: A satellite-based method that has endured for over 30 years. Bull. Amer. Meteor. Soc., 87, 1195– 1210.

Velden, C., D. C. Herndon, J. Kossin, J. Hawkins, and M. DeMaria, 2007: Consensus estimates of tropical cy-clone (TC) intensity using integrated multispectral (IR and MW) satellite observations. Proc. 27th Confer-ence on Hurricanes and Tropical Meteorology. [Avail-able at https://ams.confex.com/ams/27Hurricanes/ techprogram/programexpanded_339.htm.]

Vigh, J. L., and W. H. Schubert, 2009: Rapid development of the tropical cyclone warm core. J. Atmos. Sci., 66, 3335–3350.

Wang, H., and Y. Wang, 2014: A numerical study of Ty-phoon Megi (2010). Part I: Rapid intensification. Mon. Wea. Rev., 142, 29–48.

Warrick, F., 2016: NWP SAF AMV monitoring: The 7th

analysis report (AR7). the EUMETSAT Network of Satellite Application Facilities, 35 pp. [Available at http://nwpsaf.eu/site/monitoring/winds-quality- evaluation/amv/amv-analysis-reports/.]

Willoughby, H. E., 1998: Tropical cyclone eye thermody-namics. Mon. Wea. Rev., 126, 3053–3067.

Wong, M. L., and J. C. L. Chan, 2004: Tropical cyclone intensity in vertical wind shear. J. Atmos. Sci., 61, 1859–1876.

Wu, T.-C., H. Liu, S. J. Majumdar, C. S. Velden, and J. L. Anderson, 2014: Influence of assimilating satellite- derived atmospheric motion vector observations on numerical analyses and forecasts of tropical cyclone track and intensity. Mon. Wea. Rev., 142, 49–71.

Yamashita, K., 2012: An observing system experiment of MTSAT rapid scan AMV using JMA meso-scale oper-ational NWP system. Proc. 11th International Winds Workshop. [Available at http://cimss.ssec.wisc.edu/iwwg/iww11/talks/Session4_Yamashita2.pdf.]