Storm Surge and Wave in Miami

107年天氣分析與預報研討會 A7海象測報與應用(I) 會議室617 暴潮系集預報系統發展暨波浪與暴潮校驗系統建置 2019/1/4 吳祚任1、林君蔚1 、李奎模1 、莊美惠1 、蔡育霖1 、滕春慈2 、朱啟豪2 、葉天降2 1國立中央大學水文與海洋科學研究所 2交通部中央氣象局

水文與海洋科學研究所 Tsunami Science Laboratory https://udn.com/news/story/11314/3349037

2 https://udn.com/news/story/11314/3349037 3 https://udn.com/news/story/11314/3349037 4 https://zh.wikipedia.org/wiki/%E9%A2%B1%E9%A2%A8%E7 %91%AA%E8%8E%89%E4%BA%9E_(2018%E5%B9%B4) 5 6 2017 Hurricane Irma (USA) 2017.09.01 – 2017.09.13

Suomi NPP/VIIRS satellite view of Hurricane Irma on 03 September 2017. Resolution: 1px=1km.

7 Storm Surge and Wave in Miami

CNN Report https://www.facebook.com/cnn/videos/10157293993976509/?hc_ref=ARR5mV07Uvi7HUJj24o11m3wBF8N 8 wzTqeorWgjeOGoAhT5cKInnkDzro4bZ1Qb0z8WQ&fref=gs&dti=143043675792340&hc_location=group Philippines - 2013 Typhoon Haiyan (Yolanda) Typhoon Life Cycle: November 3rd –November 11th

Typhoon Haiyan: 'It was like the end of the world’. Typhoon Haiyan Track (Source: HKO)

San Juanico Bridge Tagpuro Barangay Rosal

Field Survey (Mas et al., 2015,NHESS) 9 Inundation induced by Storm Surges 2005 Hurricane Katrina (USA)

• Destroy of homes and business • Potential threat of coastal communities • Damages of roads and bridges

Inundation induced by 2005 Hurricane Katrina. Flooded by storm surge of Hurricane (http://www.stormsurge.noaa.gov/) Katrina (2005) in the northwest New Orleans. 10 Storm Surge and Wave in Coastal Regions

NOAA COMET Program 11 Storm Surge

• Storm surge is a abnormal water level rising induced by low pressure weather systems in open ocean or coastal regions :

 Tropical cyclones  Storms Sea Surface induced by typhoons (NHC, NOAA)  Typhoons  Hurricanes

• The two main meteorological factors contributing to a storm surge are:

 Pressure gradient  Wind shear stress Tidal Effect with Storm Surges (NHC, NOAA)

12 Hong Kong - Catogory-4 Typhoon Hagupit 2008.09.19 – 2008.09.25

Hong Kong Local News

Track of Typhoon Hagupit (HKO)

13 Records of Storm Surge at Victoria Harbour (Hong Kong)

Observed Water Level

Astronomical Tide

Storm Surge

香港天文台 (Hong Kong Observatory) http://www.weather.gov.hk/m/article_uc.htm?title=ele_00184

14 2018 Pacific typhoon season

This map shows the tracks of all tropical cyclones in the 2018 Pacific typhoon season. The points show the location of each storm at 6-hour intervals. The colour represents the storm's maximum sustained wind speeds as classified in the Saffir- Simpson Hurricane Scale (see below), and the shape of the data points represent the type of the storm. Map generation parameters: --res 4000 --extra 1 --dots 0.2 --lines 0.04 --xmin 100 --xmax 180 --ymin 0 --ymax 60 https://zh.wikipedia.org/wiki/2018%E5%B9%B4%E5%A4%AA%E5%B9%B3%E6%B4%8B%E9%A2%B1%E9%A2%A8%E5%AD%A3# /media/File:2018_Pacific_typhoon_season_summary.png 15 Tropical Cyclones in East Asia

Korea

Japan Taiwan China

Taiwan Hong Kong SCS Pacific Ocean

Philippines Philippines South China Sea

Tracks of all tropical cyclones in the northwestern Pacific Ocean between 1951 and 2014.

16 Goals for a Storm Surge Model

• Adopt large enough spherical computational domain to cover the complete typhoon life cycle and full storm surge propagation. • Include nonlinear calculation, bottom shear stresses and shoaling effects in near-shore regions. • Consider multi-scale storm surge propagation from Copyright@TTFRI (Taiwan) open ocean to coastal regions. Uncertainty of Storm Tracks • Calculate high-resolution storm surge inundation area for risk assessment. • Combine with the dynamic atmospheric model. • Combine with the global tidal model. • High-speed efficiency for the early-warning system.

Multi-Scale Storm Surge Propagation (NOAA)

17 COMCOT-SURGE Model (COrnell Multi-grid COupled Tsunami Model – Storm Surge)

Nonlinear Shallow Water Equations on the Spherical Coordinate with Meteorological Forcing Terms

 1 P  cosQ 0 tRcos 

2   s PPPQgHH11   b Pa F   fQ F tRcos     H R   H R cos ww R cos 

2 s QPQQgHH11   b Pa F   fP F tRcos  H R  H R  ww R 

• Solve nonlinear shallow water equations on both spherical and Cartesian coordinates. • Explicit leapfrog Finite Difference Method for stable and high speed calculation. • Multi/Nested-grid system for multiple shallow water wave scales. • Moving Boundary Scheme for inundation. • High-speed efficiency. 18 (1). NOAA Benchmark Problem Validation Compare with the Solitary Wave Run-up Experiments (Synolakis, 1986 and 1987).

(from NOAA Official Website)

soliton

Simulated by COMCOT (Wu, 2012)

19 (2). OpenMP Parallel Computing

COMCOT-SURGE Model can finish 48-hrs forecast and 48-hrs warm-up in 30 mins on PC-level computational resources.

Parallel Computing on Multi Cores.

Dynamic resources sharing.

The results has been published on Ocean Engineering (Simon C. Lin et al., 2015). 20 (3). Combine with the Atmospheric Model

TWRF (Typhoon Weather Research and Forecasting Model)

• TWRF model is an atmospheric model Wind Field adopted for operational forecasts by Central Weather Bureau in Taiwan.

• The TWRF model will start its simulation per 6 hours in a day at 00, 06, 12 and 18 UTC time respectively.

Pressure Field

WRF Computational Domain (CWB) 21 (4). Combine with Global Tide TPXO Model (USA OSU TOPEX/POSEIDON Global Tidal Model)

The tides are provided as complex amplitudes of earth-relative sea- surface elevation for eight primary (M2, S2, N2, K2, K1, O1, P1, Q1), two long period (Mf,Mm) and 3 non-linear (M4, MS4, MN4) harmonic constituents.

User Interface of TPXO

(Dushaw et al., 1997) TPXO can provide tidal information, like M2. 22 (5). High-Accuracy Tide Simulation The bias is smaller than 0.1 m and RMSE is smaller than 0.4 m.

Validated Gauge Locations at Taiwan

The observed data and harmonic data are provided by CWB (Taiwan). 23 (6). Model Validation Storm surge Induced by 2015 Typhoon Soudelor

• Typhoon Soudelor was the strongest typhoon in Western North Pacific regions at 2015. According to the brief analysis, more than 4,000 thousands families lost their electricity during typhoon period and accumulative rainfall is more than 1,000 mm. • Because of the destructive damages, economic loss and human casualties at Mariana Islands, Taiwan, and China, the name “Soudelor” was removed from the list of typhoon names and would not be used forever.

The flood in low-lying region at Ilan because of Typhoon Soudelor. (中央社記者沈如峰宜蘭縣) （資料來源：中央氣象局颱風資料庫） 24 Res = 1.0 3-Level Nested-Grid km Computational Domain

4.0 km/ 1.0 km/ 0.2 km

Res = 4.0 km

Res = 0.2 km

25 Storm Surge Triggered by Parametric and Typhoon WRF Fields

Wind Fields Storm Surge Propagation 26 Comparison with Observed Data 2015.08.06 00:00 -2015.08.09 06:00 (UTC)

The tide observed data are provided by our CWB in Taiwan. 27 Comparison with Other Operational Storm Surge Model

Nonlinear Tidal Model Country Resolution Coordinate Grid System Inundation Effect SLOSH USA 3 km Polar Structured Yes No USA/Chin ADCIRC 1 km Spherical Unstructured Yes Yes a JMA Japan 2 km Cartesian Structured No No POM Korea 10 km Cartesian Sigma Grid No Yes CWB Storm Spherical/ Nested-Grid Taiwan 1 km Yes Yes Surge Model Cartesian System

• Solve spherical nonlinear shallow water equation directly with Coriolis effect. • Cover complete storm surge propagation from open oceans to coastal regions. • Calculate high-resolution storm surge inundation. • Tide warm-up is stable and low-time consuming. • The resolution in coastal regions can be promoted easily and be separately calculated in nested- grid scheme. • Bathymetry and elevation data are easily to be input. • The resolution can be modified easily. 28 Track of Super 2013 Typhoon Haiyan

Source: Hong Kong Observatory

29 Nested Computational Domain for Haiyan Case

LAYER 01 (4 km)

LAYER 02 (1 km) • Layer 01 can cover the complete typhoon life cycle of Typhoon Haiyan and the full storm surge propagation.

• Layer 02 can include the offshore hydrodynamic progresses of storm surge on the fine mesh domain.

30 Near-shore Computational Domain Layer 03 (500 m)/ Layer 04 (120 m)

LAYER 04 (120 LAYER 03 (500 m) m)

Leyte Gulf

The computational domain of Layer 03 and Layer 04 could cover the storm surge propagations in offshore and nearshore regions.

31 Combine with the Atmospheric WRF Model

Pressure Field

Wind Field • Asymmetric effect • Topographic effect • Hydrodynamic Pressure

The WRF simulations are provided by Dr. Chuan- Yao Lin, AAR Modeling Laboratory (Sinica).

32 Storm Surges Induced by Typhoon Haiyan 2013.11.06 00:00 – 2013.11.09 00:00 (UTC+0)

Large computational domain to cover the complete storm surge propagation induced by Typhoon Haiyan with Coriolis effect.

33 Snapshots of Storm Surges in the Philippines

34 Target Area – Leyte Gulf

35 36 Snapshots of Storm Surges inside the Leyte Gulf

37 Time Series at Specified Gauge Stations

38 Effect of Wind Speed and Flow Angle

WRF Simulation

Parametric Model

39 • 中央氣象局現有建置之風暴潮預報模式，為單一氣象場輸入之決定性預報，惟臺灣所屬之西北太平洋區域為全球強烈颱風密集生成區域，採用決定性預報方式恐無法全面地掌握風暴潮之特徵。為因應未來潛在強烈颱風強度變動、路徑和相關物理因子之不確定因素，於本研究中發展暴潮系集預報作業系統，針對臺灣特有之地理環境建構網格計算系統，並且發展相關機率預報產品。此外，亦針對中央氣象局之需求，設計氣象局指定暴潮模式和波浪模式設計專屬統計校驗系統，以利預報員和資料分析員進行後續之模式模報能力校驗。

40 暴潮系集預報系統發展

41 系集成員建立方法

系集成員之建立，依據資源豐富程度可以大致區分為三種： 1. 採用CWB WEPS 系集預報之20組氣象場，可得到20組系集成員。 2. 採用CWB WEPS 系集預報之20組颱風路徑及其強度代入參數化風場模式計算氣象場，可得到20組系集成員。 3. 採用氣象局決定性預報之颱風路徑、颱風中心氣壓、七級風半徑、最大風速值等資料，利用過往預報各項參數之誤差之機率分布，產生不同參數組合之系集成員，可得系集成員數可自由決定。

42 Ensemble Forecasting Experiment – Super Typhoon Meranti in 2016

• Typhoon Meranti was one of the most intense tropical cyclones on record. Tracking to the west northwest, Meranti gradually intensified until September 11, at which point it began a period of rapid intensification. Continuing to rapidly intensify, it became a super typhoon early on September 12, as it passed through the Luzon Strait, ultimately reaching its peak intensity on September 13 with 1- minute sustained winds of 315 km/h (195 mph).

近中心最中心位置中心氣壓七級風半徑年月日時大風速 (經/緯) (Pa) (km) (m/s)

2016 9 12 00 130.4 18.0 940 180 45 2016 9 12 06 129.3 18.3 925 200 51 2016 9 12 12 128.2 18.9 910 200 55 2016 9 12 18 126.7 19.3 905 200 58 2016 9 13 00 125.5 19.6 905 200 58 2016 9 13 06 124.1 20.2 905 200 58 2016 9 13 12 122.9 20.4 900 220 60 2016 9 13 18 121.8 20.9 905 220 58 2016 9 14 00 120.8 21.5 905 220 58 2016 9 14 06 119.8 22.6 925 200 51 2016 9 14 12 118.9 23.4 930 200 48 2016 9 14 18 118.4 24.4 950 180 40 2016 9 15 00 117.6 25.2 970 150 33 Best-Track Parameters Storm Track of Typhoon Meranti 43 2-Level Nested-Grid Domains for Operational System LAYER 01 (Open Ocean) – 4 km LAYER 02 (Coastal Regions) – 2 km / 1 km D

Grid Number: 135,736 Grid Number: 56,192 1. The computational domains are divided to main Taiwan Island, Kinman, Penghou, and Mazhou by the nested-grid scheme. 2. The bathymetry and land elevation data are from ETOPO1 and 44 WEPS (Weather Ensemble Prediction System) WEPS can provide twenty ensemble members

1. CWB WEPS model is an atmospheric model 微物理參數法邊界層參數法積雲參數法地表參數化法 adopted for operational forecasts by Central (mp_physics) (bl_pbl_physics) (cu_physics) (sf_sfclay_physics) Weather Bureau in Taiwan. 1 GCE YSU Grell MM5 2 GCE YSU Tiedtke MM5 3 GCE MYJ Betts-Miller Eta 2. The CWB WEPS model will start its simulation per 6 hours in a day at 00, 06, 12 and 18 UTC time 4 GCE MYJ K-F Eta respectively. 5 GCE MYJ Tiedtke Eta 6 GCE MYJ Old SAS Eta 7 GCE MYJ New SAS Eta 8 GCE ACM2 Grell MM5 9 GCE ACM2 Tiedtke MM5 10 GCE ACM2 New SAS MM5 11 WSM5 YSU Tiedtke MM5 12 WSM5 MYJ Betts-Miller Eta 13 WSM5 MYJ K-F Eta 14 WSM5 MYJ Tiedtke Eta 15 WSM5 MYJ Old SAS Eta 16 WSM5 MYJ New SAS Eta 17 WSM5 ACM2 Grell MM5 18 WSM5 ACM2 Tiedtke MM5 19 WSM5 ACM2 New SAS MM5 20 WSM5 MYNN2 Old SAS MYNN 45 2016年莫蘭蒂颱風 (Typhoon Meranti)

強烈颱風莫蘭蒂（Severe Typhoon Meranti）為2016年太平洋颱風季第14個被命名的風暴；莫蘭蒂是2016年西北太平洋最強的熱帶氣旋，更是21世紀西北太平洋海域第三強風暴，僅次於2013年颱風海燕和2010年颱風梅姬。

近中心最中心位置中心氣壓七級風半徑年月日時大風速 (經/緯) (Pa) (km) (m/s) 2016 9 12 00 130.4 18.0 940 180 45 2016 9 12 06 129.3 18.3 925 200 51 2016 9 12 12 128.2 18.9 910 200 55 2016 9 12 18 126.7 19.3 905 200 58 2016 9 13 00 125.5 19.6 905 200 58 2016 9 13 06 124.1 20.2 905 200 58 2016 9 13 12 122.9 20.4 900 220 60 2016 9 13 18 121.8 20.9 905 220 58 2016 9 14 00 120.8 21.5 905 220 58 2016 9 14 06 119.8 22.6 925 200 51 2016 9 14 12 118.9 23.4 930 200 48 2016 9 14 18 118.4 24.4 950 180 40 2016 9 15 00 117.6 25.2 970 150 33 莫蘭蒂颱風參數

莫蘭蒂路徑圖 46 /43 WEPS Wind Fields – 2016.09.14 00:00 (init time = 2016.09.13 00:00)

47 WEPS Wind Fields – 2016.09.15 00:00 (init time = 2016.09.13 00:00)

48 Storm Surge Simulation – 2016.09.14 00:00 (init time = 2016.09.13 00:00)

49 暴潮系集預報結果總水位與觀測資料比對

2016.09.13 00:00 – 2016.09.15 00:00

觀測資料由中央氣象局提供。

50 /43 暴潮系集預報結果暴潮偏差與觀測資料比對

2016.09.13 00:00 – 2016.09.15 00:00

觀測資料由中央氣象局提供。

51 /43 Comparison with Observation 2016.09.13 – 2016.09.15

Blue – Observed Storm Surge.

Black – Ensemble Mean of 20 Members.

52 編號對應潮位測站最大總水位產品盒鬚圖 1 龍洞 2 基隆 2016.09.13 00:00 – 2016.09.15 00:00 3 福隆 4 頭城 5 蘇澳 6 花蓮 7 石梯 8 成功 9 台東 10 大武 11 綠島 12 蘭嶼 13 石門 14 淡水 15 桃園 16 新竹 17 後龍 18 梧棲 19 芳苑 20 麥寮 21 台西 22 東石 23 將軍 24 安平 25 永安 26 高雄 27 東港 28 嘉和 29 蟳廣嘴 30 南灣 31 小琉球 32 澎湖藍框邊界為上四分位數及下四分位數，紅線為中間值，黑色米字號為觀測值。 33 金門 34 馬祖53 /43 編號對應潮位測站 1 龍洞最大風暴潮產品盒鬚圖 2 基隆 2016.09.13 00:00 – 2016.09.15 00:00 3 福隆 4 頭城 5 蘇澳 6 花蓮 7 石梯 8 成功 9 台東 10 大武 11 綠島 12 蘭嶼 13 石門 14 淡水 15 桃園 16 新竹 17 後龍 18 梧棲 19 芳苑 20 麥寮 21 台西 22 東石 23 將軍 24 安平 25 永安 26 高雄 27 東港 28 嘉和 29 蟳廣嘴 30 南灣 31 小琉球 32 澎湖 33 金門藍框邊界為上四分位數及下四分位數，紅線為中間值，黑色米字號為觀測值。 34 馬祖54 /43 塔拉格倫排名長條圖及系集分歧度評估分析

Talagrand Rank Ensemble Member Histogram Analysis Equalikelihood

系集預報離散程度與真值相比，預報結果較透過系集成員相似性分析，系集成員05、09 容易落在偏大或偏小的分區中。和10之預報結果與真值接近程度較高。

55 /43 系集分歧度評估分析

1 分歧度 SPRD ̅ SPRD 1

方均根誤差 1 RMSE RMSE

1. 若SPRD值和RMSE值相近時，代表離散程度合理。 2016.09.13 00:00 UTC系集預報結果與 2. 若SPRD值高於RMSE值，表示過度離散。真值(觀測值)比較。 3. 若SPRD值低於RMSE值，代表離散程度較為不足。 56 /43 微物理參數法邊界層參數法積雲參數法地表參數化法 CWB WEPS區域系集預報系統 (mp_physics) (bl_pbl_physics) (cu_physics) (sf_sfclay_physics) 第1組 GCE YSU Grell MM5 1. CWB區域系集預報系統 (CWB WEPS) 使用10組擾動邊界條件、20組擾動初始場和20組不同之模式物理參數法第2組 GCE YSU Tiedtke MM5 設定，產生20組系集成員。第3組 GCE MYJ Betts-Miller Eta 2. 預報作業每日執行4次，分別為每日的00、06、12和18 第4組 GCE MYJ K-F Eta UTC，每次需產生20個系集預報成員之預報結果，即每第5組 GCE MYJ Tiedtke Eta 日產生80組預報結果。第6組 GCE MYJ Old SAS Eta 3. 風場向量 (B1020)、緯度風場向量 (B1021) 和氣壓場第組 (SSL01)預報場，導入氣象局現採用之COMCOT暴潮作 7 GCE MYJ New SAS Eta 業模式，並發展相關暴潮系集產品第8組 GCE ACM2 Grell MM5 第9組 GCE ACM2 Tiedtke MM5 第10組 GCE ACM2 New SAS MM5 第11組 WSM5 YSU Tiedtke MM5 第12組 WSM5 MYJ Betts-Miller Eta 第13組 WSM5 MYJ K-F Eta 第14組 WSM5 MYJ Tiedtke Eta 第15組 WSM5 MYJ Old SAS Eta 第16組 WSM5 MYJ New SAS Eta 第17組 WSM5 ACM2 Grell MM5 第18組 WSM5 ACM2 Tiedtke MM5 57 第19組 WSM5 ACM2 New SAS MM5 WEPS TRACKS 2017 TALIM (PS_17091012)

58 59 P-Surge - Vary Cross Track Error distributions are computed for cross track, along track and intensity by: Assuming a normal distribution Using a 5-year “mean absolute error” and getting the standard deviation (sigma) from:

Arthur A. Taylor, NOAA (Taiwan –April, 2018) 60 WEPS TRACKS 2017 NESAT 鄭等人（2016）：24/48/72小時颱風路徑預報誤差為60/106/170 公里。系集成員路徑由內至外依序為0.25、0.5、0.75、 1.0、1.25、1.5、1.75、2.0倍誤差。

61 62 Cross-Track Error and Along-Track Error

https://www.meted.ucar.edu/tropical/tc_wind_haz/errors/print.php 63 Probability density functions (PDFs) for forecast model errors

A Gaussian kernel

Bandwidths were selected using Scott (1992) with a factor of 1.06. 非對稱性峰度

引用自Davis et al. (2010)

64 Ensemble Member Generation 針對颱風轉向(登陸點)討論

Probability Density Function

1. 黑圈位置分別為平均值、1標準差和2標準差。

2. 颱風登陸前12小時、前24小時、前48小時及前72小時。

65 2016年9月13日 18:00時 (UTC) 莫蘭蒂颱風登陸前

66 2016年9月13日 18:00時 (UTC) 颱風登陸點之變異

67 Storm Surges Forecasting by the Track Variations

68 Results of Probabilistic Storm Surge Ensemble Simulation

福隆台東

大武南灣

潮位站位置

觀測資料由中央氣象局海象中心提供。