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

Diagnosis of 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, , 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 . We further examined- thepos 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 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

1. Introduction A tropical cyclone (TC) is generally identified by its low-pressure center and axisymmetric structure, which Corresponding author: Ryo Oyama, Meteorological Resear­ comprises eyewalls and spiral rainbands. The strong ch Institute, Japan Meteorological Agency, 1-1, Nagamine, Tsukuba, Ibaraki 305-0052, Japan winds and heavy rain associated with the eyewalls and E-mail: [email protected] spiral rainbands cause disastrous destruction of human J-stage Advance Published Date: 10 August 2017 infrastructures. The area within a radial distance 2 – 3 ©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). 4 Journal of the Meteorological Society of Japan Vol. 96B times the radius of maximum wind (RMW) from the TC intensity estimates in the vast majority of TC cases TC center is generally called the inner core, within (Velden et al. 2006). However, the Dvorak technique which the tangential wind velocity reaches a maxi- can struggle in certain situations wherein TC struc- mum and strong updrafts exist (Li and Wang 2012). tures may fluctuate under the The deep convections within a TC inner core convey (CDO) or in association with rapid intensity changes. the mass and absolute angular momentum upward, Recent technological advancements have enabled with the release of much latent heat. This process observations with high temporal and spatial resolu- plays an essential role in TC development. tions to be made by geostationary satellites, such as The wind structure of a TC can be described as the Multi-functional Transport Satellites (MTSATs), a combination of cyclonic circulation (primary cir- which was functional between 28 June 2005 and culation) and radial–vertical circulation (secondary 07 July 2015, and Himawari-8, which is the first of circulation). The primary circulation is intensified by new-generation geostationary satellites and has been low-level inflows, which convey large absolute angu- in operation since 07 July 2015 (Bessho et al. 2016). lar momentum toward the TC center. The low-level In particular, high-temporal-resolution satellite images convergence, increased by the inflow in the boundary with time intervals shorter than approximately 15 min layer, enhances the inner core convections (Sawada facilitate improved derivations of atmospheric motion and Iwasaki 2007; Rogers 2010). In contrast, the vectors (AMVs; Velden et al. 2005; Oyama 2015), secondary circulation comprises low-level inflows, even in the case of mesoscale phenomena character- updrafts in the TC inner core, and radial outflows near ized by winds that change abruptly. It should be noted the cloud top. The secondary circulation develops the that Himawari-8 can provide limited-domain imagery TC cyclonic vortex (i.e., the primary circulation) ver- for targeted TCs within view at sampling intervals of tically around the inner core and intensifies the warm 2.5 min. core near the TC center by the latent heat release (Vigh AMVs are wind products that are derived by and Schubert 2009; Houze 2010). tracking clouds and water vapor patterns in successive TC wind structures have been investigated in many satellite images. They have been used not only for numerical studies based on non-hydrostatic models. numerical weather prediction (NWP) (Warrick 2016) Bryan and Rotunno (2009) investigated the trajectory but also for atmospheric wind analysis (Molinari and of an air parcel passing through the position of the Vollaro 1989; Apke et al. 2016). For NWP, centers maximum wind speed in a TC and showed that the air such as the European Centre for Medium-Range parcel moves upward along the isosurfaces of absolute Weather Forecasts (ECMWF), JMA, and the National angular momentum and entropy, which tilt outward. Centers for Environmental Prediction (NCEP) of the Stern and Nolan (2011) studied the vertical profiles National Oceanic and Atmospheric Administration of tangential winds with respect to the RMW and the (NOAA) have improved their NWP products using maximum sustained wind (MSW) using the three- AMV data for the initial analysis (Langland et al. dimensional Doppler wind field data obtained from 2009; Yamashita 2012; Salonen and Bormann 2014; seven storms and simulated theoretical vortices. They Wu et al. 2014). For the TC analysis, Oyama et al. found that the vertical decay rate of the maximum (2016b) demonstrated the capability of MTSAT upper tangential wind normalized by the tangential wind at tropospheric AMVs to detect the changes in the upper a height of 2 km was nearly constant. The results of tropospheric wind fields within Typhoon Danas (1324) these previous studies are consistent with the fact that during its intensification, i.e., the increasing radial TC wind fields in the middle and upper troposphere outflows and tangential winds around the cloud top, are physically related to the surface wind field. which were associated with convective bursts (CBs; The improved estimates of TC intensity through Riehl and Malkus 1961; Steranka et al. 1986). research and development will be necessary to prevent AMVs have a great advantage compared to in situ and mitigate disasters caused by TCs. The Dvorak observations in terms of temporal resolution and data technique (Dvorak 1975, 1984), which is based on the coverage. In particular, the upper tropospheric AMVs, TC cloud pattern observed in satellite infrared images, which can be obtained throughout the TC lifetimes, is a method that has been used by TC forecast centers, are expected to contribute to TC analysis and mon- such as the Japan Meteorological Agency (JMA), itoring. The purpose of this study was to identify the Joint Typhoon Warning Center (JTWC), and the the characteristics of upper tropospheric winds that National Hurricane Center (NHC), to estimate the TC presage TC intensification by the analysis of the intensity. This technique is known to provide reliable upper tropospheric AMVs of TCs that have occurred 2018 R. OYAMA et al. 5 in the western North Pacific basin. For this analysis, target observations were conducted based on the we investigated the relationship between the upper requirements of this study (Table 1). That is, we tropospheric wind field and the MSW. The results are identified the differences between the MTSAT and expected to contribute to the elucidation of the TC Himawari-8 AMVs with respect to the quality and intensification process and verifications of numerical amount of data considering Typhoon Goni (1515), studies. In addition, we examined the possibility which existed in the period when both MTSAT and to extract information about TC intensification and Himawari-8 observations were made, i.e., from July related structural changes from the upper tropospheric 2015 to March 2016. Typhoon Goni (1515) was a TC AMVs, aiming to propose a method that supports the that occurred from 15 August 2015 to 25 August 2015 operational TC intensity analysis and structure moni- (see Section 5). We examined Typhoon Goni because toring. This new approach, which is different from the (i) it was traced throughout its lifetime by the Dvorak technique, will provide new information that Himawari-8 target observations and (ii) its lifetime will help to improve the TC analysis. was relatively long, i.e., approximately 11 days. In This paper consists of seven sections. Following addition to Typhoon Goni, Section 6 includes a case this introduction, Sections 2 and 3 describe TCs, study of (1610), which was cap- the observational/numerical data, and the methods tured by the Himawari-8 target observations in August used herein. In Section 4, the characteristics of TC 2016. Table 1 lists the 46 TCs that were selected for wind fields near the cloud top are investigated using this study, and Fig. 1 shows their tracks. MTSAT AMVs with respect to TC intensity and 3. Data and methods intensification rate. Section 5 compares MTSAT and Himawari-8 AMVs with respect to the quality and 3.1 Data amount of data. Section 6 examines the feasibility of Table 2 lists the observational/numerical data used estimating MSWs using the upper tropospheric AMVs in this study. To locate the TC center positions, we derived from the Himawari-8 target observations, fo- used the metrics of intensity [i.e., the 10-min average cusing on the changes in the TC wind and temperature MSW and the minimum sea-level pressure (MSLP) structures. Finally, Section 7 provides a summary and (also referred to as the central pressure)], and the discussion. smallest radius of a 30-kt wind (R30) based on the best-track data from the Regional Specialized Mete- 2. Studied TCs orological Center Tokyo (RSMC Tokyo) – Typhoon This study examined TCs that occurred in the Center of JMA (http://www.jma.go.jp/jma/jma-eng/ western North Pacific basin during 2011 – 2016 and jma-center/rsmc-hp-pub-eg/besttrack.html). For the were observed by MTSAT and/or Himawari-8. TCs analysis of the wind components of the AMVs (Sub- in 2011 – 2014 were observed by MTSAT because section 3.4), hourly TC positions were estimated by Himawari-8 did not supersede MTSAT until 07 July linear interpolation of the 6-hourly best-track data. 2015. To exclude considerably weak TCs with a short The best-track analysis used the following intensity lifetime, we chose TCs with the lifetime maximum categories: (i) Tropical Depression [TD; MSW < 34 kt MSW greater than 20 m s−1. In addition to this screen- (17 m s−1)], (ii) Tropical Storm [TS; MSW = 34 – 48 kt ing, TCs for which there were less AMV data near (25 m s−1)], (iii) Severe Tropical Storm [STS; MSW the TC center were excluded in advance. Specifically, = 48 – 64 kt (33 m s−1)], and (iv) Typhoon (TY; MSW only TCs for which the AMV data covered more than > 64 kt). In this study, the MSW change per day was 50 % of a circular area with a radius of 200 km from defined as the TC intensification rate. the TC center were used. The possible reasons behind The TC intensities estimated using the Dvorak tech- why AMV data were sparse near the center of some nique, i.e., the Current Intensity number (CI number), TCs are as follows: (i) the vague texture of clouds to for the researched TCs were used to verify the results be tracked in successive satellite images, e.g., due to in Sections 5 and 6. The RSMC Tokyo converts the CI the presence of distinct central dense overcast, (ii) number into MSLP and MSW using a reference table the occurrence of image navigation errors, and (iii) proposed by Koba et al. (1990). the abrupt changes in atmospheric flows near the In this study, we used the images recorded by the TC center. Consequently, 44 of 100 TCs occurring MTSAT Japanese Advanced Meteorological Imager during 2011 – 2014 were studied using MTSAT AMVs (JAMI) and Himawari-8 Advanced Himawari Imager (Table 1). (AHI). MTSAT has five bands: a visible band (VIS: The derivations of AMVs using the Himawari-8 wavelength, 0.63 µm), a shortwave infrared band (IR4: 6 Journal of the Meteorological Society of Japan Vol. 96B

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 Lifetime Max MSW Satellite for Typhoon name Analysis Period −1 number (m s ) AMVs 1102 Songda 1200 UTC 21 May – 0600 UTC 29 May 54.1 MT 1104 Haima 1200 UTC 21 Jun – 1800 UTC 24 Jun 20.6 1105 Meari 0000 UTC 22 Jun – 0600 UTC 27 Jun 30.9 1108 Nock-ten 0000 UTC 26 Jul – 0000 UTC 31 Jul 25.8 1112 Talas 0000 UTC 25 Aug – 0600 UTC 05 Sep 25.8 1115 Roke 0600 UTC 13 Sep – 0600 UTC 22 Sep 43.8 1117 Nesat 0000 UTC 24 Sep – 1800 UTC 30 Sep 41.2 1119 Nalgae 1800 UTC 27 Sep – 1800 UTC 04 Oct 48.9 (b) 1202 Sanvu 0600 UTC 22 May – 1800 UTC 27 May 30.9 MT 1203 Mawar 1800 UTC 01 Jun – 0600 UTC 06 Jun 38.6 1204 Guchol 1200 UTC 13 Jun – 0000 UTC 20 Jun 51.5 1205 Talim 0600 UTC 17 Jun – 1800 UTC 20 Jun 25.8 1207 Khanun 0600 UTC 16 Jul – 0000 UTC 19 Jul 25.8 1208 Vicente 1200 UTC 21 Jul – 1800 UTC 24 Jul 41.2 1209 Saola 0000 UTC 28 Jul – 0600 UTC 03 Aug 36.1 1210 Damrey 1200 UTC 28 Jul – 1200 UTC 03 Aug 36.1 1211 Haikui 0000 UTC 03 Aug – 1200 UTC 09 Aug 33.5 1216 Sanba 0000 UTC 11 Sep – 0000 UTC 18 Sep 56.7 1217 Jelawat 1800 UTC 20 Sep – 1200 UTC 01 Oct 56.7 1219 Maliksi 0600 UTC 01 Oct – 0600 UTC 04 Oct 25.8 1221 Prapiroon 1200 UTC 07 Oct – 1200 UTC 19 Oct 46.4 1223 Son-tinh 1200 UTC 23 Oct – 0600 UTC 29 Oct 43.8 (c) 1311 Utor 1800 UTC 09 Aug – 1200 UTC 15 Aug 54.1 MT 1312 Trami 0000 UTC 18 Aug – 1800 UTC 22 Aug 30.9 1315 Kong-rey 0600 UTC 26 Aug – 0000 UTC 30 Aug 28.3 1317 Toraji 1800 UTC 01 Sep – 0000 UTC 04 Sep 25.8 1318 Man-yi 0000 UTC 13 Sep – 1200 UTC 16 Sep 33.5 1319 Usagi 1800 UTC 16 Sep – 0600 UTC 23 Sep 56.7 1320 Pabuk 0600 UTC 21 Sep – 0000 UTC 27 Sep 30.9 1321 Wutip 0600 UTC 27 Sep – 0000 UTC 01 Oct 33.5 1323 Fitow 1800 UTC 30 Sep – 0600 UTC 07 Oct 38.6 1324 Danas 0600 UTC 04 Oct – 0000 UTC 09 Oct 46.4 1327 Francisco 0600 UTC 16 Oct – 0600 UTC 26 Oct 54.1 1328 Lekima 1800 UTC 20 Oct – 1200 UTC 26 Oct 59.2 1329 Krosa 1800 UTC 29 Oct – 0600 UTC 04 Nov 38.6 (d) 1403 Faxai 1200 UTC 28 Feb – 1800 UTC 05 Mar 33.5 MT 1405 Tapah 0000 UTC 28 Apr – 0000 UTC 01 May 25.8 1408 Neoguri 1800 UTC 03 Jul – 0000 UTC 11 Jul 51.5 1410 Matmo 1200 UTC 17 Jul – 0600 UTC 25 Jul 36.8 1414 Fengshen 1800 UTC 06 Sep – 1800 UTC 10 Sep 30.9 1415 Kalmaegi 0600 UTC 12 Sep – 1200 UTC 17 Sep 38.6 1416 Fung-wong 1200 UTC 17 Sep – 0000 UTC 24 Sep 23.2 1417 Kammuri 1200 UTC 24 Sep – 0600 UTC 30 Sep 25.8 1420 Nuri 0000 UTC 31 Oct – 1800 UTC 06 Nov 56.7 (e) 1515 Goni 0000 UTC 15 Aug – 1200 UTC 25 Aug 48.9 MT/H8 1610 Lionrock 0400 UTC 23 Aug – 1400 UTC 30 Aug 46.4 H8 2018 R. OYAMA et al. 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 (TYs), respectively. 8 Journal of the Meteorological Society of Japan Vol. 96B

Table 2. List of the data used in this study. Name Explanation Spatial resolution Temporal resolution MTSAT AMVs AMVs at pressure levels of 100–300 hPa 0.25° latitude and 6 hourly (0000, 0600, derived by using MTSAT-2 Northern longitude 1200, 1800 UTC) Hemisphere imagery acquired at 15-min intervals, operationally taken at 0000, 0600, 1200, and 1800 UTC Himawari-8 AMVs AMVs at pressure levels of 100–300 hPa 0.02° latitude and 30 min derived by using imagery at 5-min intervals, longitude from the Himawari-8 target observations Cloud top temperature Brightness temperatures (TBs) of MTSAT 4 km (MTSAT) or Hourly of opaque clouds IR1 band (10.8 μm) and Himawari-8 2 km (Himawari-8) Band13 (10.4 μm) JMA best track data TC best-track data from RSMC Tokyo- − 6 hourly Typhoon Center (TC position, central pressure, and maximum sustained wind, and the smallest radius of 30-kt wind (R30)) Dvorak analysis data Central pressure and maximum sustained – 6 hourly wind estimated by Dvorak technique at RSMC Tokyo-Typhoon Center JRA-55 Japanese 55-year Reanalysis data 37 levels in 1 to 1000 6 hourly (Kobayashi et al. 2015) hPa, 1.25° latitude and longitude Central pressure Central pressure estimates from TC warm 48 km at sub-satellite Twice daily per satellite at estimate by AMSU core intensity observed by 55-GHz bands point maximum, four satellites, technique of Advanced Microwave Sounding Unit -A MetOp-B and NOAA-15, (AMSU-A) (Oyama 2014) 18, and 19 used

3.8 µm), two infrared bands (IR1: 10.8 µm and IR2: min for the full disk and 2.5 min for the two regions 12.0 µm), and a water vapor band (WV: 6.8 µm). The of Japan (referred to as “Region 1” and “Region 2” spatial resolution of the MTSAT images was 1 km by JMA) and the target observation area (referred to for the VIS band and 4 km for the IR1, IR2, IR4, and as “Region 3”). The target observations were used to WV bands. The interval of MTSAT observations was track a TC from its formation to its dissipation or its usually 30 min for the Northern Hemisphere (NH); transition to an using the image however, it was 15 min for times around 0000, 0600, window of a square with 1000-km sides (for details of 1200, and 1800 UTC for the derivation of AMVs (Imai AHI’s functions, see Bessho et al. 2016). 2006). For this study, IR and WV AMVs were derived The Himawari-8 AHI has 16 bands. These include using the successive images of the infrared and water three visible bands (Band01: wavelength, 0.47 μm; vapor bands of the MTSAT NH and Himawari-8 Band02: 0.51 μm; Band03: 0.64 μm), three near- target observations, respectively. The MTSAT and infrared bands (Band04: 0.86 μm; Band05: 1.6 μm; Himawari-8 algorithms that were used to derive the Band06: 2.3 μm), a shortwave infrared band (Band07: AMVs are described in Subsections 3.2 and 3.3, re- 3.9 µm), three water vapor bands (Band08: 6.2 µm; spectively.

Band09: 7.0 µm; Band10: 7.3 µm), an SO2 band The infrared brightness temperature (TB) data

(Band11: 8.6 µm), an O3 band (Band12: 9.6 µm), three from MTSAT IR1 and Himawari-8 Band13, which infrared bands (Band13: 10.4 µm, Band14: 11.2 µm; represent the cloud-top temperature of opaque clouds,

Band15: 12.4 µm), and a CO2 band (Band16: 13.3 such as cumulonimbus clouds and dense cirrus clouds, µm). The spatial resolution of Himawari-8 images were used as a metric of the deepness of convection is 0.5 or 1 km for the visible bands and 2 km for the within a TC in Section 6. other bands. The image intervals of Himawari-8 are 10 The vertical wind shear (VWS) in the vicinity of a 2018 R. OYAMA et al. 9

TC is known to tilt the TC vortex and consequently computed based on a temporal consistency check of weaken the TC intensity (Gallina and Velden 2002; the difference between the first and second vectors, a Kaplan and DeMaria 2003; Wong and Chan 2004; spatial consistency check based on a comparison of Paterson et al. 2005). To consider the influence of neighbored vectors, and the consistency with GSM VWS on TC vortices, we computed the VWS using forecast winds. the Japanese 55-year reanalysis dataset (JRA-55; Kobayashi et al. 2015). In this study, the VWS is 3.3 Himawari-8 AMVs defined as the difference of the wind speeds between Himawari-8 AMVs were derived using the images pressure levels of 200 and 850 hPa. The differences from Himawari-8 target observations, which were were averaged over an area within a radius of 600 made at intervals of 2.5 min (Subsection 3.1). Because km from the TC center. The grid size of the JRA-55 the 2-km resolution of the Himawari-8 images of the dataset is 1.25° in both latitude and longitude. infrared and water vapor bands is equivalent to 13.3 m s−1 for 2.5 min, it is technically very challenging 3.2 MTSAT AMVs to capture the displacement of clouds within the TC For MTSAT AMVs, the IR and WV AMVs were areas. Therefore, we decided to derive the IR and WV derived from the NH IR1- and WV-band images, AMVs using the images at intervals of 5 min. respectively, using the AMV derivation scheme of the The IR and WV AMVs were computed at an JMA’s Meteorological Satellite Center (MSC; Oyama interval of 30 min using the images of Band13 and 2010). IR1 is the infrared band that is known to be the Band10, respectively. Band13 is an infrared band most favorable band for tracking clouds, and WV is a with a wavelength similar to that of the IR1 band of band that is well suited for tracking both clouds and MTSAT, and Band10, one of the Himawari-8 water water vapor patterns, which are mainly present in the vapor bands, was selected because it yields a larger upper troposphere. number of AMV data compared to Band08 or Band09. The wind vectors of the AMVs were computed at The derivations of Himawari-8 AMVs were per- around 0000, 0600, 1200, and 1800 UTC by tracking formed using a new scheme that has been developed high-level clouds, such as cirrus clouds, in the three by the JMA/MSC (Shimoji 2014) to take advantage successive images at intervals of 15 min. The tracking of Himawari-8’s upgraded functions relative to process was conducted using a cross-correlation- MTSAT. The new scheme is similar to that used to matching technique that produced two displacement derive MTSAT AMVs (Subsection 3.2); however, it vectors from the first and second images and from the includes several new techniques and optimizations of second and third images. The second vector was used parameters for AMV derivations. Herein, only the dif- as the final vector. For the derivation of a displace- ferences between the MTSAT AMV and Himawari-8 ment vector, the subpixel displacement of clouds is AMV schemes have been explained (for details of the considered by identifying the position with the maxi- new scheme for the Himawari-8 AMV, see Shimoji mum correlation in the correlation-matching surface. 2014). First, the target box size used for tracking The target box (i.e., the image segment used for clouds was set to five image pixels (the image pixel cloud tracking) was a square of 16 image pixels (the size for the infrared and water vapor bands is 2 km at image pixel size for the IR and WV bands is 4 km at nadir), which is considered sufficient for tracking the nadir) on a 0.25° latitude/longitude grid. The AMVs meteorological phenomena with horizontal scales as were assigned to a cloud-top height, which was small as 10 km. The grid interval for the AMV deriva- estimated using the IR1 TB data in the target box, by tion was set to approximately 0.02° for both latitude referring to the first-guess field (6-h forecast) of the and longitude. With this interval, the target box with JMA’s Global Spectral Model (GSM), the horizontal five pixel sides overlapped the adjacent boxes, which resolution of which is 20 km. To consider the cloud were as large as half the target box size. Second, the semi-transparency, the IR1 TB data of the clouds were displacement vector for the tracked clouds was com- corrected by the IR-WV intercept method based on puted using the average of two correlation-matching the relationship between IR1 TBs and WV TBs in the surfaces that were derived from the first and second target box (Nieman et al. 1993; Schmetz et al. 1993). images and from the second and third images, re- For the quality control of the AMVs, we used the spectively, in the cross-correlation matching. The use Quality Indicator (QI) developed by the European of the average of two correlation-matching surfaces Organization for the Exploitation of Meteorological can lead to a reduction in the noise that may remain Satellites (EUMETSAT; Holmlund 1998). The QI was in the cloud-tracking process. Third, the height of 10 Journal of the Meteorological Society of Japan Vol. 96B an AMV was determined by a maximum likelihood to the maxima of the corresponding azimuthal means method based on Himawari-8 observations and the in six annuli at radii of 50, 100, 150, 200, 250, and outputs of GSM forecasts. With the introduction of 300 km from the best-track TC center (Fig. 2). We an atmospheric model that includes three cloud layers refer to the radii of UMaxWind and UMaxOutflow as allocated in the upper, middle, and lower troposphere, Rumw and Ruout, respectively (Table 3). It is possible the height of the AMV could be optimally determined that in some cases, the TC center (i.e., the location by solving the likelihood function, for which the of the MSLP), which is usually collocated with the displacement vector of an AMV and the observed rotation center near the surface, might be different TBs (Band08, Band09, Band10, Band13, Band15, from the rotation center around the cloud top. This and Band16), the GSM forecast wind fields, and the situation might occur because of the tilting of the TC simulated TBs obtained by the forward calculation of vortex caused by a large VWS or the inaccuracy of the the radiative transfer model were used. hourly TC center in the best-track data. To evaluate the position error, the best-track TC center was com- 3.4 Analysis of TC wind field using upper pared with the rotation center around the cloud top, tropospheric AMVs which was equated to the position of the rotation peak For the analysis of the wind field near the cloud tops from the MTSAT upper tropospheric AMVs, that was within a TC area, we used complementary IR and WV spatially interpolated on a 0.125° latitude × 0.125° AMVs that were computed between pressure levels longitude grid for the 44 TCs during 2011 – 2014. This of 100 and 300 hPa to maximize the spatial coverage comparison revealed that the mean position difference of the AMV data. The AMV data were only screened during the TC intensification phase is 0.5 – 0.6°. The using the AMV data with QI greater than 0.3. The QI error in the position of the rotation center estimated threshold was determined by referring to the threshold using the AMVs could increase (i) during the TC 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

Table 3. Definitions of the parameters used in this study. Parameter Explanation UMaxWind 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 Fig. 2. Illustration of derivations of the maximum of radial wind components of AMVs at tangential wind speed (UMaxWind) and the pressure levels of 100–300 hPa in six maximum radial wind speed (UMaxOutflow). annuli within a radius of 300 km (i.e., UMaxWind (UMaxOutflow) is defined as the 50, 100, 150, 200, 250, and 300 km) maximum azimuthally averaged tangential (radi- from the best-track TC center al) wind within annuli (±50 km) at the radii of Rumw Radius of UMaxWind 50, 100, 150, 200, 250, and 300 km (red circles) from the TC center (cross). MTSAT upper tro- Ruout Radius of UMaxOutflow pospheric IR AMVs (blue arrows) are displayed IRTB_R200 Averaged infrared-band brightness in the background of the MTSAT IR1-band temperature (TB) within a radius of 200 brightness temperature (TB) for Typhoon Nuri km from the TC center (1420) at 0600 UTC on 31 October 2014. 2018 R. OYAMA et al. 11 formation phase when the cyclonic circulation of the NOAA-15, -18, and -19 satellites. TC was considerably weak, thereby making it difficult 4. Characteristics of TC wind fields near the to determine the position of the rotation center, and cloud top determined from MTSAT AMVs (ii) when the TC was in a large VWS caused by strong environment winds (e.g., jets during the TC decay This section examines the characteristics of TC phase). In fact, the mean position difference during wind fields near the cloud tops, which were obtained TC intensification slightly depended on the VWS by using the 6-hourly MTSAT AMVs, with reference for the 44 TCs. That is, the average of the position to the best-track data. We studied the averaged tan- differences for the TCs with VWS less than 6 m s−1 gential and radial wind components of the upper tro- was 0.60°, whereas that for the TCs with VWS greater pospheric AMVs for the 44 TCs that occurred during than 6 m s−1 was 0.64°. This result implies that the 2011 – 2014 (Tables 1a – d) with reference to the MSW, derivation accuracy of UMaxWind and UMaxOutflow TC intensification rate, and R30 of best-track data. tends to be high when the VWS is small. 4.1 Relationship between R30 and the radii of 3.5 Central pressure estimated from the TC warm UMaxWind and UMaxOutflow core intensity observed by the Advanced To find the average structural changes of TC cloud- Microwave Sounding Unit-A top wind fields associated with the TC intensity and To verify the TC intensity that was estimated using the intensification rate, we examined Rumw and the upper tropospheric AMVs in Section 6, we used Ruout as well as R30 as functions of the MSW and MSLPs estimated on the basis of TC warm core inten- TC intensification rate (Fig. 3). With respect to the de- sities observed by the Advanced Microwave Sounding pendency of Rumw and Ruout on the MSW (Fig. 3a), Unit-A (AMSU-A) of the NOAA and MetOp polar- Ruout tended to be relatively small for TCs with large orbiting satellites (Oyama 2014). The MSLP was es- MSWs despite the R30’s being relatively large. Spe- timated from the maximum TB anomaly of AMSU-A cifically, the average Ruouts for MSWs < 20 m s−1 and 55-GHz-band TB values near the TC center based on > 50 m s−1 were 218 and 196 km, respectively. This hydrostatic equilibrium approximation. The spatial Ruout pattern may indicate that eyewalls and inner resolution of the AMSU-A data is 48 km at nadir, rainbands were formed more robustly at the radii near and the frequency of observations was at most twice the TC center for strong TCs. However, the depen- daily per satellite. This study used the MetOp-B and 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. 12 Journal of the Meteorological Society of Japan Vol. 96B was apparent that Ruout and Rumw were larger than km from the TC center. The positive correlations the mean radius of the MSW at the surface, 20 – 100 between the MSWs and the tangential winds of the km (Kepert 2010), indicating that the isosurface of the upper tropospheric AMVs imply that the cyclonic RMW generally tilted outward (Sawada and Iwasaki circulation around cloud tops is generally intensified 2007; Bryan and Rotunno 2009). In addition, it was in response to the strength of the surface winds via remarkable that Ruout tended to depart from Rumw the upward transport of absolute angular momentum for weak MSWs. Specifically, the difference between from the surface to the upper troposphere within the Ruout and Rumw was 48 km for MSWs < 20 m s−1 TC inner core. The correlation coefficient between and 17 km for MSWs > 50 m s−1. This result may also the MSW and the maximum tangential wind of upper suggest that a strong TC tends to have a robust inner tropospheric AMVs, i.e., UMaxWind, was larger than core comprising eyewalls, which may be true until the the correlation coefficient between the MSW and the vortex builds upward to the tropopause, usually after tangential wind at each radius, indicating that the the TC appears in the infrared TB image. radius of UMaxWind could change depending on the With regard to the dependency of Rumw and Ruout distance between the TC center and the eyewalls (i.e., on the TC intensification rate (Fig. 3b), both Rumw eyewall radius), which can vary with the TC size and and Ruout tended to be small for TCs with high inten- intensity, and the inflows in the boundary layer. sification rates despite the R30s being relatively small. These characteristics of Rumw and Ruout indicated 4.3 Relationship between the MSW and UMaxWind that strong convection occurred near the TC center This subsection examines the feasibility of estimat- for TCs with a high intensification rate (Harnos and ing the MSW using UMaxWind, which is a parameter Nesbitt 2011; Rogers et al. 2013), indicating that the that is best correlated with the MSW (Subsection 4.2). inflow toward the TC center in the boundary layer was Figure 4 shows the scatter plots between the UMax- strong for TCs with high intensification rates (Rogers Winds of 6-hourly MTSAT AMVs and the best-track 2010; Wang and Wang 2014). MSWs for the 44 TCs during 2011 – 2014. As noted in Subsection 4.2, the correlation coefficient between 4.2 Relationship between MSWs and upper tropo- UMaxWind and MSW was high, approximately 0.73, spheric tangential winds indicating that the MSW might be estimated from 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

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). Correlation with Parameter best-track MSWs Fig. 4. Scatter plots between the 6-hourly UMax­ Winds of MTSAT AMVs (x-axis) and JMA UMaxWind 0.73 best-track MSWs (y-axis) for the 44 TCs during Azimuthal average at radius of 50 km 0.24 2011 – 2014. The number of data is 1064. The Azimuthal average at radius of 100 km 0.58 dotted and solid lines represent the first- and Azimuthal average at radius of 150 km 0.69 second-order polynomial regression equations, Azimuthal average at radius of 200 km 0.69 respectively. R2 and r denote the coefficient of Azimuthal average at radius of 250 km 0.65 determination and the correlation coefficient, Azimuthal average at radius of 300 km 0.56 respectively. 2018 R. OYAMA et al. 13

UMaxWind. We tried to approximate the relationship Figure 5 shows the snapshots of the MTSAT and between the MSWs and UMaxWinds using two Himawari-8 AMVs for Typhoon Goni, which were equations: the first- and a second-order polynomial obtained at 1800 UTC on 23 August 2015. It is re- regression equations (Fig. 4). The root-mean-square markable that Himawari-8 AMVs were derived over errors (RMSEs) of the two regression equations were areas wider than those for MTSAT AMVs despite the similar, but the RMSE of the second-order polynomial much larger grid size (0.25° vs. 0.02°) of the MTSAT regression (7.42 m s−1) was slightly smaller than that AMVs. To identify the general difference in the of the first-order regression (7.47 m s–1). The RMSE of data coverage between the MTSAT and Himawari-8 the second-order polynomial regression was smaller AMVs, we computed the frequency of the AMV because of the relatively weak UMaxWinds for large data within 0.25° squares at 0.25° latitude/longitude MSWs, which may be explained by the following intervals for Typhoon Goni from 0000 UTC on 15 scenarios: (i) deep convection was less active during August to 1200 UTC on 25 August in 2015 (Fig. 6). the TC decay phase than during the intensification Considering the grid size of Himawari-8 data, which phase and (ii) the cyclonic flows near the cloud top is smaller than that of the MTSAT AMV data, the were weakened by strong environment winds, such as frequencies of Himawari-8 AMVs were summed high-level jets, when the TC approached mid-latitudes. within squares only once per observation time. We The RMSEs were larger than that of the Dvorak esti- found that the frequency of Himawari-8 AMVs within mate obtained by Koba et al. (1990), 3.6 – 6.2 m s−1, 0.25° squares was larger than that of MTSAT AMVs although they were comparable to the MSW estimates for almost all the grids, suggesting that the coverage based on the TBs of the TRMM microwave imager of the Himawari-8 AMV data was wider than that (TMI), 6 – 8 m s−1 (Hoshino and Nakazawa 2007). of the MTSAT AMV data within the TC area. These The relatively large RMSE of MSW estimates from results imply that Himawari-8 AMVs can capture UMaxWind implies that it is necessary to consider smaller-scale winds in a TC area more efficiently TC structures, such as TC size and cloud-top height, compared to MTSAT AMVs. It should be noted that which could depend on the deepness of convection because the QI was computed based on a spatial within the TC inner core for precise MSW estimation. consistency check, the QI is a function of the grid size of the AMV data. The additional comparison of the 5. Comparison between MTSAT and Himawari-8 frequency of data for AMVs with QI greater than 0 AMVs for Typhoon Goni (1515) also revealed that the coverage of Himawari-8 AMVs In this section, we consider Typhoon Goni (1515) was larger than that of MTSAT AMVs. We inferred to examine the differences between the MTSAT and that the difference in the data coverage between the Himawari-8 AMVs used in this study with respect to Himawari-8 and MTSAT AMVs resulted from (i) the the number of data, heights, and other elements. 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. 14 Journal of the Meteorological Society of Japan Vol. 96B

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.).

images (Subsection 3.1), (ii) the use of a small target smaller (larger), the number of AMVs tended to be box for the derivation of Himawari-8 AMVs, and (iii) larger for Himawari-8 AMVs than for MTSAT AMVs. the use of averaged matching correlation surfaces of These differences in the number of data may have the first–second images and the second–third images resulted from the differences in height assignments for deriving the displacement vector (Subsections 3.2 (Subsections 3.2 and 3.3). In fact, the cloud type and and 3.3). the optical depth can vary as a function of the radial Table 5 summarizes a comparison between the distance from the TC center because the deep convec- MTSAT and Himawari-8 AMVs at common locations tive clouds usually exist within the eyewalls and inner in terms of wind speed, tangential and radial winds, rainbands near the TC center. This dependence of the and the number of data between pressure levels of cloud type on the radius may explain the differences 100 and 300 hPa throughout the Typhoon Goni ob- associated with the amount of AMV data. servation period. Note that the total number of AMV For IR AMVs (Table 5b), the wind speed, wind data within a given radial interval from the TC center components, and the number of data were similar to slightly differed between the MTSAT and Himawari-8 those for WV AMVs (Table 5a). The differences in the AMVs because in some locations, only MTSAT wind speed, wind components, and the number of data AMVs or Himawari-8 AMVs were assigned below the between the Himawari-8 and MTSAT AMVs were 300-hPa level. With respect to the WV AMVs (Table also similar to those for WV AMVs. A common dif- 5a), the wind speeds of Himawari-8 AMVs were ference between the Himawari-8 and MTSAT AMVs, slightly stronger than those of MTSAT AMVs between i.e., the larger tangential winds of the Himawari-8 vs. pressure levels of 200 and 300 hPa but slightly weaker MTSAT AMVs for the radii of 0 – 100 km, implies than those of MTSAT AMVs between pressure levels that the smaller target box size for Himawari-8 AMV of 100 and 150 hPa for radii of 100 – 300 km. These derivation and the high temporal and spatial resolution differences in the wind speeds were similar to those of Himawari-8 images facilitated the derivation of apparent in the tangential and radial winds. AMVs in the abruptly changing wind field near the Some differences in the number of data between the TC center. Himawari-8 and MTSAT AMVs were also apparent The 30-min-interval Himawari-8 AMVs calculated for WV AMVs (Table 5a). For instance, as the AMV from the Himawari-8 target observations were ex- height became lower (higher) and the radius became pected to capture the changes in the TC wind fields 2018 R. OYAMA et al. 15

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 radius (km) height radius (km) height radius (km) (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 100 – 150 11.0 12.2 12.3 100 – 150 10.4 13.6 14.2 100 – 150 0.6 −1.4 −1.9 150 – 200 12.5 13.8 12.8 150 – 200 11.4 13.7 13.0 150 – 200 1.1 0.2 −0.3 200 – 250 15.4 14.6 12.5 200 – 250 11.7 14.1 12.1 200 – 250 3.7 0.4 0.4 250 – 300 13.6 14.7 13.0 250 – 300 9.6 11.6 11.9 250 – 300 4.0 3.0 1.0 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 radius (km) height radius (km) height radius (km) (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 100 – 150 6.6 6.8 4.9 100 – 150 6.0 7.8 7.4 100 – 150 0.7 −1.0 −2.6 150 – 200 9.5 10.3 7.2 150 – 200 7.2 9.5 6.4 150 – 200 2.3 0.7 0.8 200 – 250 11.7 10.5 7.2 200 – 250 2.4 9.6 6.1 200 – 250 9.3 0.8 1.1 250 – 300 6.4 10.5 7.4 250 – 300 3.3 4.9 5.2 250 – 300 3.2 5.6 2.2 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 radius (km) height radius (km) height radius (km) (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 100 – 150 4.3 6.8 8.4 100 – 150 4.0 8.6 9.8 100 – 150 0.4 −1.8 −1.4 150 – 200 4.6 6.4 8.0 150 – 200 4.2 6.3 8.1 150 – 200 0.4 0.1 −0.1 200 – 250 4.4 5.9 6.5 200 – 250 3.5 4.9 6.9 200 – 250 0.8 1.0 −0.4 250 – 300 5.2 5.7 6.0 250 – 300 3.7 5.1 6.5 250 – 300 1.4 0.6 −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 radius (km) height radius (km) height radius (km) (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 100 – 150 399 1150 1945 100 – 150 648 1040 701 100 – 150 −249 110 1244 150 – 200 422 1818 4157 150 – 200 350 2425 4578 150 – 200 72 −607 −421 200 – 250 184 1074 2616 200 – 250 88 924 3834 200 – 250 96 150 −1218 250 – 300 89 548 1463 250 – 300 9 194 1073 250 – 300 80 354 390 Total 1094 4590 10181 Total 1095 4583 10186 Total −1 7 −5 near the cloud top more precisely than the 6-hourly AMVs than for the MTSAT AMVs and tended to be MTSAT AMVs. The radius–time cross sections of larger for the Himawari-8 AMVs than for the MTSAT the tangential and radial winds for Typhoon Goni AMVs, especially within a radius of 100 km (Table (Figs. 7, 8) revealed that there were clear differences 5). For the radial wind (Fig. 8), the outward shift of in the radial and temporal variations between the the outflow peak radius was captured more precisely MTSAT and Himawari-8 AMVs. The tangential by Himawari-8 AMVs than by MTSAT AMVs. These winds were more finely resolved for the Himawari-8 significant differences in the spatial and temporal 16 Journal of the Meteorological Society of Japan Vol. 96B

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 radius (km) height radius (km) height radius (km) (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 100 – 150 11.2 13.0 12.5 100 – 150 10.6 13.6 14.0 100 – 150 0.6 −0.7 −1.5 150 – 200 12.8 14.2 13.1 150 – 200 11.1 14.3 13.1 150 – 200 1.7 −0.2 0.0 200 – 250 15.7 14.7 13.1 200 – 250 11.6 15.2 12.8 200 – 250 4.1 −0.5 0.2 250 – 300 13.8 15.2 13.3 250 – 300 − 12.8 12.6 250 – 300 − 2.4 0.6 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 radius (km) height radius (km) height radius (km) (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 100 – 150 7.8 7.4 5.2 100 – 150 6.5 8.1 6.7 100 – 150 1.4 −0.7 −1.5 150 – 200 10.0 10.8 7.7 150 – 200 6.8 10.9 7.0 150 – 200 3.3 −0.1 0.6 200 – 250 12.0 10.7 7.7 200 – 250 3.1 11.1 8.3 200 – 250 8.9 −0.4 −0.6 250 – 300 7.3 11.0 7.8 250 – 300 − 6.0 6.4 250 – 300 − 5.0 1.4 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 radius (km) height radius (km) height radius (km) (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 100 – 150 4.5 7.6 8.7 100 – 150 4.1 8.4 10.0 100 – 150 0.4 −0.7 −1.4 150 – 200 4.5 6.3 8.0 150 – 200 4.0 5.3 7.3 150 – 200 0.5 1.0 0.7 200 – 250 5.5 5.8 6.6 200 – 250 3.2 4.5 5.5 200 – 250 2.3 1.3 1.1 250 – 300 4.5 5.6 5.5 250 – 300 − 3.5 4.9 250 – 300 − 2.1 0.6 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 radius (km) height radius (km) height radius (km) (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 (hPa) 0 – 100 100 – 200 200 – 300 100 – 150 389 1134 1672 100 – 150 753 1738 1132 100 – 150 −364 −604 540 150 – 200 415 1844 3590 150 – 200 285 2198 4948 150 – 200 130 −354 −1358 200 – 250 175 1053 2414 200 – 250 37 536 2552 200 – 250 138 517 −138 250 – 300 90 530 1485 250 – 300 0 72 535 250 – 300 90 458 950 Total 1069 4561 9161 Total 1075 4544 9167 Total −6 17 −6

distributions of AMVs indicated the advantages of MTSAT AMVs obtained at intervals of 6 h (Fig. 9). Himawari-8 AMVs over MTSAT AMVs with respect The UMaxWind of Himawari-8 AMVs tended to be to capturing changes of the TC’s primary and second- slightly larger than that of MTSAT AMVs and cap- ary circulations. tured detailed fluctuations in the cyclonic flows (Fig. To examine the impact of using Himawari-8 AMVs 9a), implying that Himawari-8 AMVs could capture on the diagnosis of the TC wind field, the UMaxWind the cyclonic flows near the cloud top within the TC and UMaxOutflow of Himawari-8 AMVs obtained inner core more precisely than MTSAT AMVs (Fig. 7). at intervals of 30 min were compared with those of The local maximum of UMaxWinds from both AMVs 2018 R. OYAMA et al. 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. 18 Journal of the Meteorological Society of Japan Vol. 96B

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).

roughly corresponded to those of the best-track and 6. Possibility of estimating the TC maximum Dvorak MSWs; however, there were some differenc- surface wind from the maximum tangential es, e.g., from 1200 UTC on 19 August 2015 to 0000 wind of Himawari-8 upper tropospheric AMVs UTC on 21 August 2015 (Fig. 9a). The difference between the UMaxWinds and Dvorak MSW implies In this section, we use the Himawari-8 target ob- that the tangential winds of upper tropospheric AMVs servations of Typhoon Lionrock (1610) to examine contain information on the TC intensity change, which the possibility of using UMaxWind for estimating is independent on the Dvorak TC intensity estimate. the MSW. Typhoon Lionrock formed as a TD around The UMaxOutflow of Himawari-8 AMVs captured the southeast part of a monsoon gyre located over detailed fluctuations of the radial outflow near the Japan on August 17, 2016. The TD then moved cloud top compared to that of MTSAT AMVs. It counterclockwise with rapid intensification in an should be noted that the local maximum of UMaxOut- extra cyclonic wind field with weak VWS (average = flows from both AMVs roughly corresponded to the 4.3 m s−1) and turned back near the Minamidaitojima local minimum of the averaged infrared TB within a Island because of the approach of a southwesterly jet. radius of 200 km from the TC center (IRTB_R200 in The RMSE and bias of the MSW estimated from Fig. 9b), indicating that the increase in UMaxOutflow Himawari-8’s UMaxWinds via the second-order represented the outward movement of upper tropo- polynomial regression (Fig. 4) to the best-track data spheric clouds such as anvils from deep convection in for Lionrock were 10.9 and −1.4 m s−1, respectively. the TC inner core. 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 2018 R. OYAMA et al. 19

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.

in UMaxWind was not clear in the time series of the maximum in the time series of UMaxWind roughly best-track analysis (Fig. 10a), which was quite similar corresponded to the deepening of the central pressure to that of the Dvorak analysis (the graphs are not (Fig. 10a), indicating that the warm core was devel- shown). oped around at that time. The spatial distributions A comparison between the MSW estimated from of the upper tropospheric AMVs and AMSU-A TB UMaxWind and the central pressure via the AMSU anomalies relative to the average in the annulus be- technique (Subsection 3.5) revealed that the local tween the radii of 600 and 700 km from the TC center 20 Journal of the Meteorological Society of Japan Vol. 96B from 1400 UTC on 25 August 2016 to 1700 UTC on momentum from the surface to the upper troposphere 26 August 2016 (shown as “Period-A” in Fig. 10) within the inner core, together with the development indicated that the positive temperature anomaly of the of the warm core due to the latent heat release and TC warm core at a pressure level of approximately adiabatic warming within the eye (Willoughby 1998; 250 hPa increased when the tangential winds near the Ohno and Satoh 2015). The warm infrared TB area TC center increased (Figs. 11a–c, 12a, b). This syn- near the TC center around at the time when the AMSU chronism between the UMaxWind increase and the MSLP reached the minimum implied the formation of warm core development suggests that the primary the eye (Figs. 13b, c). circulation near the cloud top was increased via the In addition to these changes apparent within the upward transport of a large amount of absolute angular 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). 2018 R. OYAMA et al. 21 computed using the upper tropospheric AMVs reached respectively, of UMaxWind (Fig. 10c). Additionally, a local minimum when UMaxWind reached its local it should be noted that some of the occurrence times maximum (Fig. 10b), indicating that the RMW at the when the cloud-top temperature within a radius of surface shrunk due to the increase of inflow in the 200 km from the TC center (IRTB_R200 in Fig. 10c) boundary layer at such times (Subsection 4.1). Fur- reached its local minimum were between 1500 UTC thermore, it was noteworthy that the local minimum (0300 LST) and 2300 UTC (0800 LST), which are and maximum of the cloud-top temperature, which indicated by black arrows in Fig. 10c. This time inter- were represented by infrared TB, of the convective val corresponds to the occurrence time of convection and dense cirrus clouds near the TC center roughly peaks associated with the TC diurnal cycle (Muramatsu corresponded to the local minimum and maximum, 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. 22 Journal of the Meteorological Society of Japan Vol. 96B

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). chronism between the decrease in the cloud-top tem- cyclonic circulation near the TC center in the upper perature and the increase in UMaxOutflow (Fig. 10c) troposphere was increased by the vertical develop- implied that the convection deepening within the inner ment of the TC primary circulation via the preceding core drove the vertical motion and intensified the TC consecutive deep convections (Miyamoto and Takemi secondary circulation. This situation was also implied 2013; Oyama et al. 2016b). by the outward shift of the low-TB cloud area from The results obtained from the case study consider- the TC center, which was identified by the difference ing Typhoon Lionrock indicate that the UMaxWind of between Figs. 13a and 13c. Based on these results, it Himawari-8 AMVs contains information about the TC was possible that the local maximum of UMaxWind intensity change, which is independent of the Dvorak following convection deepening indicated that the TC intensity estimate, suggesting that UMaxWind 2018 R. OYAMA et al. 23 could contribute to the operational TC intensity anal- opment of the upper tropospheric warm core and (ii) ysis, e.g., as a member of the consensus TC intensity the shrinkage of the radius of UMaxWind when the estimate (Velden et al. 2007; Oyama et al. 2016a). In local maximum of UMaxWind appeared, and (iii) the addition, the results suggest that Himawari-8 AMVs occurrence of consecutive deep convections within the could capture the TC structural change related to inner core prior to the local maximum of UMaxWind. TC intensification, e.g., the shrinkage of the eyewall It is possible that the temporal variations in the tan- radius and the increase in the TC secondary circula- gential winds near the cloud top of Typhoon Lionrock tion enhanced by a deep convection within the TC were clearly observed by Himawari-8 AMVs because inner core. the VWS around the TC was small throughout its lifetime. 7. Summary and discussion This study revealed several fundamental relation- To investigate the possibility of estimating the TC ships between a TC’s upper tropospheric wind and intensity and characteristic structures from the upper intensity, indicating that the upper tropospheric AMVs tropospheric AMVs, which represent the winds near should include information about the TC intensity the cloud top, this study examined the relationships change and related structural changes. These findings between the upper tropospheric AMVs derived from suggest that the upper tropospheric AMVs could con- the MTSAT and Himawari-8 images and the MSWs tribute to the TC intensity analysis and TC structure of JMA’s best-track data for TCs that occurred in the monitoring. However, several fundamental issues still western North Pacific basin during 2011 – 2016. remain unresolved. It is important to find how the tem- Statistical investigations using 6-hourly MTSAT poral variation, which was apparent in UMaxWinds, upper tropospheric AMVs for the 44 TCs during is related to the TC’s diurnal cycle and others cycles 2011 – 2014 revealed that the tangential winds near with periods shorter than 1 day (Kossin 2002; Takeda the cloud top were well correlated with the best- and Oyama 2003). Another necessary task is to inves- track MSWs. It was apparent that the maximum tigate the response of the upper tropospheric winds tangential winds, UMaxWinds, were most correlated to the lower tropospheric winds, which could vary with MSWs, indicating that the radius of UMaxWind depending on the robustness of the TC inner core, by could change depending on the eyewall radius, which evaluating the budgets of absolute angular momentum could vary with the TC size and the inflow in the within a TC (Rozoff et al. 2012). We expect that the boundary layer. The RMSE of the regressions between numerical studies based on a high-resolution non- UMaxWinds and MSWs was approximately 7.4 m s−1 hydrostatic model will help in elucidating the pro- and larger than the RMSE of the Dvorak estimate, cesses and mechanisms that enable the linking of the 3.6 – 6.2 m s−1, as evaluated by Koba et al. (1990). It boundary layer to the upper troposphere for the TC’s is also noted that the RMSE of the second-order poly- diurnal cycle (Navarro and Hakim 2016) and other nomial regression equation is slightly smaller than phenomena. These studies will also provide informa- that of the first-order polynomial regression equation, tion and ideas essential for improving the methods which is considered to result from the relatively used to estimate the TC intensity and structure via weak UMaxWinds during the TC decay phase. It is AMVs. In addition to studies focusing on TC intensity noteworthy that for TCs that rapidly intensified, the estimation, the relationship between the TC intensi- radii of UMaxWind and the maximum radial outflow, fication rate and the secondary circulation, which is UMaxOutflow, tended to be small. These results imply represented by the radial outflow obtained from upper that the inflow in the boundary layer increased and tropospheric AMVs, is another interesting research the RMW at the surface and the eyewall radius shrunk topic to tackle. when the TC was rapidly intensified. To obtain more insights about the TC intensity Finally, the possibility of estimating the MSW using change and related structural changes from Himawari-8 UMaxWind derived from Himawari-8 target observa- AMVs, it is necessary to investigate more TCs that tions was examined. It is noteworthy that UMaxWinds were traced by Himawari-8 target observations. This at intervals of 30 min captured the changes in the investigation should include weak TCs with the life- TC structure on a timescale shorter than 1 day; these time maximum MSW less than 20 m s−1, which were were the changes that were not clearly recognized in not examined in this study. In addition, for obtaining the Dvorak analysis. A case study of Typhoon Lion- more precise estimation of TC intensity from the rock (1610) revealed several pieces of evidence that AMVs, it is necessary to study the dependencies of indicated a change in the TC intensity: (i) the devel- UMaxWinds on the TC size and intensity as well as 24 Journal of the Meteorological Society of Japan Vol. 96B the cloud-top height. For future improvements of [Available at http://www.eumetsat.int/website/wcm/ Himawari-8 AMVs, it is important to continue efforts idc/idcplg?IdcService=GET_FILE&dDocName=PDF in identifying optimal parameter values for the der- _DIV_FACTSHEET&RevisionSelectionMethod= ivation of AMVs, such as the grid size and the time LatestReleased&Rendition=Web.] interval (Sohn and Borde 2008; Bresky et al. 2012; Gallina, G. M., and C. S. Velden, 2002: Environmental ver- tical wind shear and tropical cyclone intensity change Shimoji 2012). Furthermore, it is important to identify utilizing enhanced satellite derived wind information. the optimal QI threshold for screening the AMV data Proc. 25th Conference on Hurricanes and Tropical and seek for the possibility of using the QI without Meteorology, Amer. Meteor. Soc., San Diego, CA, forecast check for this study. Such tasks will enhance 172–173. [Available at https://ams.confex.com/ams/ the usefulness of Himawari-8 AMVs in the estima- pdfpapers/35650.pdf.] tions of the TC intensity and structure. Harnos, D. S., and S. W. Nesbitt, 2011: Convective structure in rapidly intensifying tropical cyclones as depicted Acknowledgments by passive microwave measurements. Geophys. Res. This study used the MTSAT imagery and ancillary Lett., 38, L07805, doi:10.1029/2011GL047010. data obtained from the JMA/MSC for computing Holmlund, K., 1998: The utilization of statistical properties AMVs. The AMSU-A data were obtained from of satellite-derived Atmospheric Motion Vectors to NOAA’s Comprehensive Large Array-data Steward- derive quality indicators. Wea. Forecasting, 13, 1093– ship System (CLASS) website (http://www.class.ncdc. 1104. Hoshino, S., and T. Nakazawa, 2007: Estimation of tropical noaa.gov/saa/products/welcome). The authors thank cyclone’s intensity using TRMM/TMI brightness the reviewers and Christopher Velden of the Coop- temperature data. J. Meteor. Soc. Japan, 85, 437–454. erative Institute for Meteorological Satellite Studies Houze, Jr., R. A., 2010: Clouds in tropical cyclones. Mon. (CIMSS), University of Wisconsin for their many Wea. Rev., 138, 293–344. valuable comments that helped to improve this paper. Imai, T., 2006: Status of atmospheric motion vector in JMA. References Proc. 8th International Winds Workshop. [Available at https://www.eumetsat.int/website/home/News/ Apke, J. M., J. R. Mecikalski, and C. P. Jewett, 2016: Analy- ConferencesandEvents/DAT_2043687.html.] sis of mesoscale atmospheric flows above mature deep Kaplan, J., and M. DeMaria, 2003: Large-scale character- convection using super rapid scan geostationary satel- istics of rapidly intensifying tropical cyclones in the lite data. J. Appl. Meteor. Climatol., 55, 1859–1887. North Atlantic basin. Wea. Forecasting, 18, 1093– Bessho, K., K. Date, M. Hayashi, A. Ikeda, T. Imai, H. 1108. Inoue, Y. Kumagai, T. Miyakawa, H. Murata, T. Kepert, J. D., 2010: Tropical cyclone structure and dynam- Ohno, A. Okuyama, R. Oyama, Y. Sasaki, Y. Shimazu, ics. Global Perspectives on Tropical Cyclones. World K. Shimoji, Y. Sumida, M. Suzuki, H. Taniguchi, H. Scientific Series on Asia-Pacific Weather and Climate, Tsuchiyama, D. Uesawa, H. Yokota, and R. Yoshida, vol. 4, Chan, J. C. L., and J. D. Kepert (eds.), World 2016: An introduction to Himawari-8/9 – Japan’s new- Scientific Publishing, 3–53. generation geostationary meteorological satellites. J. Koba, H., T. Hagiwara, S. Osano, and S. Akashi, 1990: Meteor. Soc. Japan, 94, 151–183. Relationship between the CI-number and central Bresky, W. C., J. M. Daniels, A. A. Bailey, and S. T. Wan- pressure and maximum wind speed in typhoons. J. zong, 2012: New methods toward minimizing the Meteor. Res., 42, 59–67 (in Japanese). slow speed bias associated with atmospheric motion Kobayashi, S., Y. Ota, Y. Harada, A. Ebita, M. Moriya, H. vectors. J. Appl. Meteor. Climatol., 51, 2137–2151. Onoda, K. Onogi, H. Kamahori, C. Kobayashi, H. Bryan, G. H., and R. Rotunno, 2009: Evaluation of an an- Endo, K. Miyaoka, and K. Takahashi, 2015: The alytical model for the maximum intensity of tropical JRA-55 Reanalysis: General specifications and basic cyclones. J. Atmos. Sci., 66, 3042–3060. characteristics. J. Meteor. Soc. Japan, 93, 5–48. Dunion, J. P., C. D. Thorncroft, and C. S. Velden, 2014: The Kossin, J. P., 2002: Daily hurricane variability inferred tropical cyclone diurnal cycle of mature hurricanes. from GOES infrared imagery. Mon. Wea. Rev., 130, Mon. Wea. Rev., 142, 3900–3919. 2260–2270. Dvorak, V. F., 1975: Tropical cyclone intensity analysis and Langland, R. H., C. Velden, P. M. Pauley, and H. Berger, forecasting from satellite imagery. Mon. Wea. Rev., 2009: Impact of satellite-derived rapid-scan wind ob- 103, 420–430. servations on numerical model forecasts of Hurricane Dvorak, V. F., 1984: Tropical cyclone intensity analysis Katrina. Mon. Wea. Rev., 137, 1615–1622. using satellite data. NOAA Technical Report Li, Q.-Q., and Y. Wang, 2012: A comparison of inner and NESDIS, 11, 47 pp. outer spiral rainbands in a numerically simulated EUMETSAT, 2015: Divergence product: Product guide. tropical cyclone. Mon. Wea. Rev., 140, 2782–2805. 2018 R. OYAMA et al. 25

Miyamoto, Y., and T. Takemi, 2013: A transition mechanism cyclones. Mon. Wea. Rev., 141, 2970–2991. for the spontaneous axisymmetric intensification of Rozoff, C. M., D. S. Nolan, J. P. Kossin, F. Zhang, and J. tropical cyclone. J. Atmos. Sci., 70, 112–129. Fang, 2012: The role of an expanding wind field and Molinari, J., and D. Vollaro, 1989: External influences on inertial stability in tropical cyclone secondary eyewall hurricane intensity. Part I: Outflow layer eddy angular formation. J. Atmos. Sci., 69, 2621–2643. momentum fluxes.J. Atmos. Sci., 46, 1093–1105. Salonen, K., and N. Bormann, 2014: AMVs in the oper- Muramatsu, T., 1983: Diurnal variations of satellite-measured ational ECMWF system. Proc. 12th International TBB areal distribution and eye diameter of mature Winds Workshop. [Available at http://cimss.ssec.wisc. typhoons. J. Meteor. Soc. Japan, 61, 77–90. edu/iwwg/iww12/talks/02_Tuesday/1110_AMVs_in_ Navarro, E. L., and G. J. Hakim, 2016: Idealized numerical ECMWF_operational_system_Salonen_v2.pdf.] modeling of the diurnal cycle of tropical cyclones. J. Sawada, M., and T. Iwasaki, 2007: Impacts of ice phase Atmos. Sci., 73, 4189–4201. processes on tropical cyclone development. J. Meteor. Nieman, S. J., J. Schmetz, and W. P. Menzel, 1993: A Soc. Japan, 85, 479–494. comparison of several techniques to assign heights to Schmetz, J., K. Holmlund, J. Hoffman, B. Strauss, B. Mason, cloud tracers. J. Appl. Meteor., 32, 1559–1568. V. Gartner, A. Koch, and L. van de Berg, 1993: Op- Ohno, T., and M. Satoh, 2015: On the warm core of a trop- erational cloud-motion winds from Meteosat infrared ical cyclone formed near the tropopause. J. Atmos. images. J. Appl. Meteor., 32, 1206–1225. Sci., 72, 551–571. Shimoji, K., 2012: A study on the relationship between spa- Oyama, R., 2010: Upgrade of Atmospheric Motion Vector tial and temporal image resolutions for AMV deriva- derivation algorithms at JMA/MSC. Meteorological tion with next-generation satellites. Proc. 11th Inter­ Satellite Center Tech. Note, 54, 1–31. [Available at national Winds Workshop. [Available at http://www. http://www.data.jma.go.jp/mscweb/technotes/msctech eumetsat.int/website/wcm/idc/idcplg?IdcService= rep54-1.pdf.] GET_FILE&dDocName=PDF_CONF_P60_S2_07_ Oyama, R., 2014: Estimation of tropical cyclone central SHIMOJI_V&RevisionSelectionMethod=Latest pressure from warm core intensity observed by the Released&Rendition=Web.] Advanced Microwave Sounding Unit-A (AMSU-A). Shimoji, K., 2014: Motion tracking and cloud height assign- Pap. Meteor. Geophys., 65, 35–56. ment methods for Himawari-8 AMV. Proc. 12th Inter- Oyama, R., 2015: Characteristics of upper-tropospheric At- national Winds Workshop. [Available at http://www. mospheric Motion Vectors (AMV) in tropical cyclone eumetsat.int/website/wcm/idc/idcplg?IdcService= areas derived using MTSAT rapid-scan observation GET_FILE&dDocName=PDF_CONF_P61_S2_06_ data. Tenki, 62, 881–894 (in Japanese with English SHIMOJI_V&RevisionSelectionMethod=Latest abstract). Released&Rendition=Web.] Oyama, R., K. Nagata, H. Kawada, and N. Koide, 2016a: Sohn, E., and R. Borde, 2008: The impact of window size Development of a product based on consensus be- on AMV. Proc. 9th International Winds Workshop. tween Dvorak and AMSU tropical cyclone central [Available at http://www.eumetsat.int/website/wcm/ pressure estimates at JMA. RSMC Tokyo-Typhoon idc/idcplg?IdcService=GET_FILE&dDocName=PDF Center Technical Review, 18, 8 pp. [Available at _CONF_P51_S4_18_SOHN_V&RevisionSelection http://www.jma.go.jp/jma/jma-eng/jma-center/rsmc- Method=LatestReleased&Rendition=Web.] hp-pub-eg/techrev/text18-1.pdf.] Steranka, J., E. B. Rodgers, and R. C. Gentry, 1986: The Oyama, R., A. Wada, and M. Sawada, 2016b: Intensification relationship between satellite measured convective of Typhoon Danas (1324) captured by MTSAT upper bursts and tropical cyclone intensification. Mon. Wea. tropospheric Atmospheric Motion Vectors. SOLA, 12, Rev., 114, 1539–1546. 135−139. Stern, D. P., and D. S. Nolan, 2011: On the vertical decay Paterson, L. A., B. N. Hanstrum, N. E. Davidson, and H. rate of the maximum tangential winds in tropical C. Weber, 2005: Influence of environmental vertical cyclones. J. Atmos. Sci., 68, 2073–2094. wind shear on the intensity of hurricane-strength Takeda, T., and R. Oyama, 2003: Periodic time variation of

tropical cyclones in the Australian region. Mon. Wea. low-TBB cloud area in typhoon. J. Meteor. Soc. Japan, Rev., 133, 3644–3660. 81, 1497–1503. Riehl, H., and J. Malkus, 1961: Some aspects of hurricane Velden, C., J. Daniels, D. Stettner, D. Santek, J. Key, J. Daisy, 1958. Tellus, 13, 181–213. Dunion, K. Holmlund, G. Dengel, W. Breskey, and Rogers, R., 2010: Convective-scale structure and evolution P. Menzel, 2005: Recent innovations in deriving tro- during a high-resolution simulation of tropical cy- pospheric winds from meteorological satellites. Bull. clone rapid intensification.J. Atmos. Sci., 67, 44–70. Amer. Meteor. Soc., 86, 205–223. Rogers, R., P. Reasor, and S. Lorsolo, 2013: Airborne Dop- Velden, C., B. Harper, F. Wells, J. L. Beven II, R. Zehr, T. pler observations of the inner-core structural differ- Olander, M. Mayfield, C. C. Guard, M. Lander, R. ences between intensifying and steady-state tropical Edson, L. Avila, A. Burton, M. Turk, A. Kikuchi, 26 Journal of the Meteorological Society of Japan Vol. 96B

A. Christian, P. Caroff, and P. McCrone, 2006: The analysis report (AR7). the EUMETSAT Network of Dvorak tropical cyclone intensity estimation tech- Satellite Application Facilities, 35 pp. [Available at nique: A satellite-based method that has endured for http://nwpsaf.eu/site/monitoring/winds-quality- over 30 years. Bull. Amer. Meteor. Soc., 87, 1195– evaluation/amv/amv-analysis-reports/.] 1210. Willoughby, H. E., 1998: Tropical cyclone eye thermody- Velden, C., D. C. Herndon, J. Kossin, J. Hawkins, and M. namics. Mon. Wea. Rev., 126, 3053–3067. DeMaria, 2007: Consensus estimates of tropical cy- Wong, M. L., and J. C. L. Chan, 2004: Tropical cyclone clone (TC) intensity using integrated multispectral (IR intensity in vertical wind shear. J. Atmos. Sci., 61, and MW) satellite observations. Proc. 27th Confer- 1859–1876. ence on Hurricanes and Tropical Meteorology. [Avail­ Wu, T.-C., H. Liu, S. J. Majumdar, C. S. Velden, and J. L. able at https://ams.confex.com/ams/27Hurricanes/ Anderson, 2014: Influence of assimilating satellite- techprogram/programexpanded_339.htm.] derived atmospheric motion vector observations on Vigh, J. L., and W. H. Schubert, 2009: Rapid development numerical analyses and forecasts of tropical cyclone of the tropical cyclone warm core. J. Atmos. Sci., 66, track and intensity. Mon. Wea. Rev., 142, 49–71. 3335–3350. Yamashita, K., 2012: An observing system experiment of Wang, H., and Y. Wang, 2014: A numerical study of Ty- MTSAT rapid scan AMV using JMA meso-scale oper- phoon Megi (2010). Part I: Rapid intensification. ational NWP system. Proc. 11th International Winds Mon. Wea. Rev., 142, 29–48. Workshop. [Available at http://cimss.ssec.wisc.edu/ Warrick, F., 2016: NWP SAF AMV monitoring: The 7th iwwg/iww11/talks/Session4_Yamashita2.pdf.]