Prediction of Landfalling Tropical Cyclones over East Coast of India in the Global Warming Era

U. C. Mohanty

School of Earth, Ocean and Climate Sciences Indian Institute of Technology Bhubaneswar Outline of Presentation

• Introduction • Mesoscale modeling of TCs with MM5, ARW, NMM and HWRF systems • Conclusions and Future Directions Natural disasters Hydrometeorologi- Geophysical cal Disasters: Disasters: Earthquakes Cyclones Avalanches Flood Land slides Drought Volcanic eruption Tornadoes Dust storms Heat waves Cold waves

Warmest 12 years: 1998,2005,2003,2002,2004,2006, 2001,1997,1995,1999,1990,2000

Global warming

Period Rate 25 0.1770.052 50 0.1280.026 100 0.0740.018 150 0.0450.012 Years /decade

IPCC Introduction

• Climate models are becoming most important tools for its increasing efficiency and reliability to capture past climate more realistically with time and capability to provide future climate projections.

• Observations of land based weather stations in global network confirm that Earth surface air temperature has risen more than 0.7 ºC since the late 1800s to till date. This warming of average temperature around the globe has been especially sharp since 1970s.

• The IPCC predicted that probable range of increasing temperature between 1.4 - 5.8 ºC over 1990 levels by the year 2100.

Contd…… • The warming in the past century is mainly due to the increase of green house gases and most of the climate scientists have agreed with IPCC report that the Earth will warm along with increasing green house gases.

• In warming environment, weather extremes such as heavy rainfall (flood), deficit rainfall (drought), heat/cold wave, storm etc will occur more frequent with higher intensity.

• Climate change is now most important issue for the scientists and politicians worldwide.

• Proper disaster management can reduce the loss of lives. Global Impact of natural disaster

77.27% Hydromet 12.33% Geogological 10.40% Biological

Pre monsoon 1891-1949 70 58 58 1950-2008 D – 0% 60 50 43 39 40 31 CS – 10% 30 22 20

10 SCS – 41% No. of systems in No. years of 59 systems 0 D CS SCS Category of cyclonic disturbances Post monsoon 1891-1949 250 218 1950-2008 198 D – 10% 200 150 122 127 CS – 4% 100 81 47

50 SCS – 72% No. of systems in No. years of 59 systems 0 D CS SCS Source: Mohanty et al 2011, Natural Category of cyclonic disturbances Hazards Genesis of the cyclones over the different parts of the Bay of Bengal for the different epochs.

80 Solid bar:1901-1950 Dash bar: 1951-2007 70 Blue: Storms 60 Red: Severe Storms 50 40 30 20 10 0 NB CB SB

Mohanty et al..2011 Annual Frequency of Natural Hazards during past 60 years

16 Flood Storm 14 Earthquake/tsunami, volcanic eruption Others (Heat wave, cold wave, forest fire) 12

10

8

6

4

2

0 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 TYPES OF POTENTIAL DAMAGE DUE TO TROPICAL CYCLONES TROPICAL CYCLONES

Low Pressure, Large Pressure Gradient and Strong, Low Level Convergence of Mass, Heat and Moisture

Strong Winds Heavy Rainfall

• Damage due to Structures • Flooding of Coastal • Loss of Life • Loss of Power and Areas • Destruction of Vegetation Communications • Erosion of Beaches Crop and Livestock • Loss of Life and Injuries • Loss of Soil Fertility from • Contamination of Water • Generation of Devastating Saline Intrusions Supply Storm Surges • Loss of Life • Land Subsidence • Destruction of Vegetation, • Damage of Structures • Flooding of Land Area Crops and Livestock Real time forecast of Tropical Cyclones over Indian seas Meso-scale Modelling systems using to predict TC activities over NIO at IIT D/ IIT BBS

1. NRL ( From 1982 to 1997)

2. Meso-scale Model Version-5 (From 1995 to 2005)

3. Weather Research and Forecast Model [ARW and NMM] ( From 2005 onwards)

4. Hurricane WRF (From 2007 onwards) Recent Activities WRF Model Configuration Model Dynamics Non-hydrostatic Horizontal resolution 9 km

Forecast Length 72 – 96 hrs (depends on TCs life) Time step 30 s TCsMap over projection Arabian Sea Mercator Horizontal grid system Arakawa C-gridTCs over Bay of Bengal

Vertical co-ordinate Terrain following hydrostatic pressure co-ordinate LongitudeRadiation : 48 E – 78 E Dudhia’s long & short wave LatitudeSurface layer : 5 N – 28 N Thermal diffusionLongitude: scheme 77 E – 102 E ResolutionInitial/Lateral : 27 boundary km FNL Latitude : 3 N – 28 N IC & BC : GFS analysis & Resolution : 27 km Cumulusforecast (0.5 x 0.5) Kain Fritsch PBL scheme YSU

Microphysics WSM-3 scheme 14 TCs during 2007 – 2013 (Total 162 cases) Basin Name (Intensity) Simulations period in 12-hr interval Observed Landfall No. of forecasts

Gonu (SuCS) 00 UTC 2 – 12 UTC 5 June 2007 03UTC 6 June (over Oman) 8 Arabian Sea Yemyin (CS) 00 and 12 UTC 25 June 2007 03 UTC 26 June 2 Cyclones Phyan (CS) 12 UTC 9 – 00 UTC 11 Nov 2009 Between 10-11 UTC 11 Nov 4 (5 TCs) 17 cyclones Phet (VSCS) 12 UTC 31 May – 00 UTC 6 June 2010 12 UTC 6 June (LF-2) 12 31 cases Murjan (CS) 00 UTC 23 – 25 October 2012 18 UTC 25 October 2012 5 Akash (CS) 12 UTC 13 – 12 UTC of 14 May 2007 00 UTC 15 May 3 Sidr (VSCS) 12 UTC 11 – 00 UTC 15 Nov 2007 15 UTC 15 Nov 8 Nargis (VSCS) 12 UTC 27 April – 00 UTC 2 May 2008 12 UTC 2 May 10 Rashmi (CS) 00 UTC 25 – 12 UTC 26 Oct 2008 00 UTC 27 Oct 4 KhaiMuk (CS) 12 UTC 13 – 12 UTC 15 Nov 2008 00 UTC 16 Nov 5 Nisha (CS) 12 UTC 25 – 26 Nov 2008 00 UTC 27 Nov 3 Bijli (CS) 12 UTC 14 – 00 UTC 17 Apr 2009 15 UTC 17 April 6

Bay of Aila (SCS) 12 UTC 23 – 00 UTC 25 May 2009 9 UTC 25 May 4 Bengal Ward (CS) 12 UTC 10 – 12 UTC 13 Dec 2009 9 UTC 14 Dec 7 cyclones (14 TCs) Laila (VSCS) 12 UTC 17 – 19 May 2010 12 UTC 20 May 5

80 cases Giri (VSCS) 12 UTC 20 – 00 UTC of 22 Oct 2010 14 UTC 22 Oct 4 Jal (VSCS) 00 UTC of 4 – 7 Nov 2010 16 UTC 7 Nov 7 Thane (VSCS) 00 UTC 26 – 12 UTC 29 Dec 2011 00 UTC 30 Dec 8 Nilam (CS) 00 UTC 28 – 12 UTC 31 Oct 2012 15 UTC 31 Oct 2012 6 Mahasen (CS) 00 UTC 10 – 12 UTC 16 May 2013 9UTC 16 May 2013 13 PHAILIN (VSCS) 7-12 Oct 2013 17 UTC 12 Oct 2013 9 (SCS) Helen 19 – 22 Nov 2013 9 UTC 22 Nov 2013 6 Lehar (VSCS) 25 – 28 Nov 2013 9 UTC 28 Nov 2013 6 Madi(VSCS) 8 – 12 Dec 2013 Dissipated over BoB 8 15 Total Number of cyclones during 2007 – 13 162 Cyclone Aila (00 UTC of 23-26 May 2009) (00 UTC of 3 – 7 Nov 2010)

Cyclone Thane Cyclone Phailin (00 UTC of 26-30 December, 2011) ( 12 UTC 7 – 12 UTC 12 Oct 2013)

16 Mean track errors for NIO cyclones during 2007 - 2011 (under operational setup)

These error statistics are based on 100 TC cases

Mean Errors for NIO TCs with different resolutions Mean Intensity Errors (10m winds m/s) 500

400 Mean Errors for NIO TCs with different resolutions

42

42 42

40 40

39 39 38

27 km 18 km 9 km 38 45 36 10 36 300 35

5 23 25

0 15

-8 -7 -8 -8 -8 -7 -7 -6 -9 -6 -8 -5 -6 -4 -6 -4

-11 -12 -13 -13 -12 -11 -10 200 -10 -5 5 -5 -10 -15 Mean (km)DPE 100 -15 6 12 18 24 30 36 42 48 54 60 66 72 -25

Mean DPE (km) DPE Mean -35

113 106 104 140 129 115 186 169 141 248 222 204 316 291 274 375 359 329 -20 -45 0 18 km 9 km % of improvement -25 -55 12 24 36 48 60 72 Forecast length (hour) Forecast length (hour) Recent (20-22 Oct 2010)

Observed TC Location

Initial cyclone vortex position error is about 60 km 17 Model TC Location Osuri et al. 2013, JAMC Errors for recurving TCs 700 Mean errors for 600 recurving TCs

500 27 km 18 km 9 km 400 Improvement is 300 significant with high resolution for recurving 200

Mean (km)DPE TCs.

100

69.03 65 131 128 109 164 149 128 231 220 193 308 284 245 382 347 330 440 413 390 0 74 0 12 24 36 48 60 72 Forecast length (hour)

700 Mean errors w.r.t Intensity at initialization at 27 km resolution Mean track errors w.r.t 600 intensity at initialization 500 400 DD CS SCS Stronger cyclones can be 300 tracked with minimum 200

errors compared to Mean DPEs (km) 100

marginal cyclones or

81 60 51 131 118 92 176 146 129 211 201 165 300 285 233 334 329 264 421 412 330 depressions. 0 0 12 24 36 48 60 7218 Forecast length (hour) HWRF Prediction of TC Roanu (00 UTC 18- 22 May 2016) Real time prediction of movement, intensity of very severe cyclonic storm Hud-Hud over Bay of Bengal using High resolution dynamical model

Nadimpalli et al., 2016 Genesis prediction of Hud- Hud (at 03UTC 7 Oct 2014)

39 hour forecast 27 hour forecast 15 hour forecast Knots Movement and Intensity of Hud- Hud

Model predicted tracks

Mean Track error (km) Track errors (km) errors Track

Forecast length (hours) Model predicted 10 m maximum winds (knots)

Mean Intensity Error (knots)

10m wind Speed (knots) Speed wind 10m Mean 10m wind errors (knots) errors wind10m Mean Time (ddhh) of October 2014 Forecast Length (hr) Grey lines are different forecast Thick black line is mean value Thick line with circle symbol is IMD OBS Rainfall prediction during Landfall day of TC Hud- Hud

TRMM IMD-NCMRWF

cm 96 h fcst 72 h fcst 48 h fcst

Numbers are IMD station rainfall OBS 24-hr accumulated rainfall (cm) during landfall day for HudHud (Verification at 103 stations) Maximum Rainfall: 38 cm RMSE: 8 cm

Grey lines are model-predicted rainfall (cm) initialized at different initial times Composite Reflectivity of HudHud

Vishakhapatnam DWR station (OBS) Model predicted Real-time forecast of Phailin TC Phailin (96 hour) Forecast based on 12 UTC of 8 October 2013

White track is observed track

Time error 5 hrs ahead hrs 5 error Time Landfall point error is 29 km 29 is error point Landfall

TC Phailin (72 hour) Forecast based on 12

UTC of 9 October 2013

Time error 2 hrs ahead hrs 2 error Time Landfall point error is 16 km 16 is error point Landfall TC PHAILIN Track prediction from 00UTC 8 – 12 Oct 2013

Intensity prediction 70 Intensity (10m winds ms-1) 60

) OBS 800 812 -1 50 900 912 1000 40 1012 1100 1112 30

20 10-mwind(ms 10 0 800 812 900 912 1000 1012 1100 1112 1200 1212 1300 Mean Track errors (km) Time (date hour)

Forecast length (hour) 24-hr accumulated rainfall (cm) during landfall day for TC PHAILIN (Verification at 108 stations)

40 RMSE=7 cm 35 OBS IMD 30 Model 25 20 15 10 5

0

Puri

Korei Naraj

Tiring

Telkoi

Jaipur Tigiria

Hindol

Daitari

Bijepur

Banpur Barmul

Rajghat Salepur

Airakhol Mohana

Papunki Sukinda

Rampur

Athmalik Balikuda Nimpara

Phiringia

Daspalla

Banarpal

Odagaon

Anandpur Chaibasa Nayagarh

Madanpur

Jhumpura

Chandikhol Nuagaon K

GUdayagiri

Keonjhagarh Raghunathpur 24-hr accumulated rainfall (cm) during landfall day for TC PHAILIN (Number represents IMD OBSERVED RAINFALL at 108 stations) cm

72 hour forecast 24 hour forecast Real time Forecast of Madi (08 – 12 December )

TC Madi (96 hour) Forecast based on 12 UTC of 08 December 2013

White track is observed track

TC Madi (72 hour) Forecast based on 12 UTC of 09 December 2013

30 Track prediction from different initial conditions for TC MADI

31 Impact of DWR reflectivity and radial wind on the rainfall prediction of landfalling Tropical Cyclones over Bay of Bengal Impact of Doppler Weather Radar Reflectivity and Radial wind (model resolution = 9 km)

Radar Cyclone Name Simulations Coverage Kolkata DWR

SIDR (Very 00 UTC 13 – 12 UTC 14 Nov 2007 Kolkata severe cyclone) (4 cases) DWR

Chennai DWR Aila (Severe 00 UTC 23 – 12 UTC 24 May 2009 Kolkata cyclone) (4 cases) DWR

Laila (Severe 12 UTC 17 – 12 UTC 19 April 2010 cyclone) (5 cases) DWR

Jal (Severe 00 UTC 6 – 00 UTC 7 Nov 2010 Chennai cyclone) (3 cases) DWR

Three numerical experiments are carried out GTS includes : SYNOP, AWS, SHIP, TEMP, PILOT, CNTL - With out Data Assimilation BUOYS, SATOB, SATEM, AIREP etc. GTS - With Assimilation of GTS data DWR includes : Reflectivity and Radial velocity 33 DWR - Assimilation of GTS + DWR data Impact of TC environment data of DWR on

Track and Intensity of SIDR cyclone

GTS data Assimilation data GTS

DWR+GTS Assimilation DWR+GTS CNTL (No Assimilation (No CNTL

TC-SIDR TC-SIDR MSLP (hPa) 10m wind speed (m/s) 1010 60 OBS CNTL 1000 55 GTS DWR 50 990 45 980 40 35 970

30 MSLP (hPa)MSLP

960 (m/s) wind 10-m 25 OBS CNTL 20 950 GTS DWR 15 940 10 1300 1312 1400 1412 1500 1512 1600 1612 1300 1312 1400 1412 1500 1512 1600 1612 Forecast hour (hrs) Time (date hour) TimeForecast (date hour (hrs)hour) 34 Mean VectorMean GTSDWRand errors displacement of CNTL,

Mean VDEs (km) 100 200 300 400 500 600 CNTL Vs DWR CNTL Vs GTS GTS GTS Vs DWR 0 12 GTS vs DWR vs GTS GTS vs CNTL DWR vs CNTL DWR GTS CNTL Forecast length (hour) 24 Initial time 38 % 12 % 46% Mean improvement (%) Mean improvement 36 48 24 25 % 13 % 35 % - hr 60 48 24 % 20 % 40 % - hr 72 72 65 % 27 % 74 % - 0 10 20 30 40 50 60 70 80 hr

Skill of 3DVAR experiments (%) Prediction of TCs with HWRF Modeling system

36 HWRF predicted Life cycle of Phailin Cyclone

Mohanty et al., 2015

Glimpses of Success in Weather Science A Great Escape from the Bay of Bengal 'Super Sapphire-Phailin' 24-hr accumulated rainfall (cm) during landfall day for TC PHAILIN (stations with ≥ 15 cm are shown) HWRF model Product

38 Gain in intensity prediction with HWRF (ex.TC PHAILIN)

ARW Model predictions

m max.winds (m/s)max.winds m

- 10

Time (dd hh) of October 2013 HWRF Model predictions

HWRF system is superior for

intensity prediction over NIO (m/s)max.winds m -

basin 10

Time (dd hh) of October 2013 Very Severe Cyclonic Storm Hudhud: HWRF prediction (3 km) (12 UTC 08 Oct 2014) HWRF predicted Track HWRF predicted 10m wind (kts)

Red Track: HWRF predicted 120

) HWRF Observed

Black Track: Observed kts 100 80 60

40 10mwind ( Speed 20

Forecast Length (Hrs)

HWRF model simulated vortex at initial time OBS Initial time (12 UTC 08 Oct 2014)

HWRF model simulated vortex at 10m wind speed at the stage of VSCS OBS 66h fcst (06 UTC 11 Oct 2014)

Multiplatform Satellite Surface Wind Analysis (CIRA, NOAA satellites) Cyclone centric HWRF domain for NIO

D 01 Prediction of intensity changes of very severe cyclonic storm Phailin over Bay of Bengal using HWRF modelling system D 02

D 03

Osuri et. al., 2016 Track and intensity prediction IC: 12 UTC 09 October 2013

(a) Track Prediction

(c) 10 m maximum sustained winds (knots)

IC: 12UTC 09 Oct 2013 Oct 09 12UTC IC: 10m 10m speed max. (knots) wind Time (ddhh)

(b) Mean Track Error (km) (d) Forecast Length (hours)

Mean track track Meanerror (km)

Mean errors Mean Meanintensity error (knots)

Forecast Length (hours) 42 Radius of maximum winds (km)

IC: 12 UTC 09 Oct 2013 IC: 12 UTC 09 Oct 2013 Radius of max. winds (km) windsmax. of Radius

Mean errorError of Rmax 70 60 H2D_IMD H3D_IMD 50 40 30 20 10 0

Radius of max. winds (km) windsmax. of Radius 6 12 18 24 30 36 42 48 54 60 66 72 78 84 Forecast length (hours) 43 Probability of RI (%) of VSCS Phailin

IC: 12 UTC 09 October 2013 Run

(a) H2D (b) H3D Probability of RI (%) RI of Probability

Time (ddhh) of October 2013 Bars for Probability of RI (%). RI = ≥ 30 knots in 24 h period Line graph: Next 24h intensity change (knots) Blue rectangular Box: Observed RI phase

There is almost 12 h delay in H2D system. H3D version predicted RI at right time. 44 Height- radial warm core (temperature anomaly, °C) Structure of Phailin valid at 03 UTC 10 October 2013

IC: 2013-10-09_12 UTC RI onset (valid at 03 UTC 10 October 2013)

(a) Satellite derived (b) H2D (c) H3D Height (km) Height

Peak Intensity (03 UTC 12 Oct 2013) Height (km) Height

45 Mean Track and intensity error statistics: 2010 Vs 2013 versions of HWRF system

250 Mean Track Errors (km) 200 27 cases corresponding 150 to 4 TCs of

100 2013

50 CNTL 2Domains 3Domains [Phailin, Oct 0 Helen, Nov 6 12 18 24 30 36 42 48 54 60 66 72 78 84 Lehar, Nov

Madi, Dec] Mean Intensity errors (knots) Mean track errors (km) errors track Mean (knots) errors Intensity Mean There are two contrast cyclones during post monsoon season of 2013

Phailin – showed [CS – very severe cyclonic storm in 18 hours]; undergone rapid intensification during 10-11 October of 65 knots in 24 hours]

Lehar – showed rapid weakening [very severe cyclonic storm to depression in 18 hrs] Rapid Intensification of TC PHAILIN Weakening of TC LEHAR (IC: 00UTC 10 Oct 2013) (IC: 00UTC 26 Nov 2013)

Black: Observed Red: HWRF

Black: Observed Red: HWRF

SCS CS DD Landfall Contrasting Dissipating characteristics of TCs Lehar and Phailin

Phailin - HWRF simulated 10 m winds every 24 hours

Lehar - HWRF simulated 10 m winds every 24 hours Weakening of TC LEHAR : SST Impact

27.2 °C 27.2 °C 27.2 °C

27.8 °C 27.5°C

Experiments conducted

1. Removing the SST gradient by modifying the entire basin SST to 27.8 C 2. Modifying the gradients specifically at the location where the weakening began 3. Updating SST every six hours using INCOIS (Indian National Center for Oceanic Information System) and GODAS SST data

Conclusion : These experiments revealed no significant impact of SST on Lehar’s weakening. Ventilation of mid-level low entropy (⍬e ) air above the boundary layer of Lehar

Mid level low ⍬e suppresses the convective instability and triggers the spin down of vortex

References : Smith and Montgomery (2015), Tang and Emanuel (2012), Riemer et al. (2009) , Nolan and McGauley (2012)

Mid-low level vertical shear creating thermodynamic pathways

for the low ⍬e air to intrude in to the storm environment

Key notes : Evidence of an interplay of mid-low (850 – 400 mb) level vertical shear and low entropy air towards Lehar’s weakening. - What is the driver - Shear or dry air? Can we isolate the two? - Results from idealized experiments indicate that dry air can only impede the storm. Shear is essential to weaken the vortex. - Could the cause of shear be the winds carrying low entropy air?

51 Evolution of equivalent potential temperature ⍬e and wind vectors overlaid

Phailin Lehar

Key Points :

- ⍬e gradients showcase the spiral from which the low entropy air is funneled in to the storm in the case of Lehar. - The air is also seen to enter from the rear quadrants

-On the other hand, in case of Phailin, the storm ⍬e does not drop until the storm makes landfall -Is it just the absence of cold/dry air over the land during Phailin or is the shear in Phailin’s storm environment such that thermodynamic pathways are52 not created for dry air to intrude? Comparison of Vortex structures (vertical)

Phailin Lehar

Before intensification or weakening

Key Note: Decoupling of upper and lower levels in Lehar (left) seen in the form of Vortex tilting, a clear sign of shear interaction with the vortex. No such tilt is observed in Phailin After intensification or weakening

53 Impact of upper level anomalous warming/cooling – A thermodynamic outlook

Phailin Lehar

- Temperature anomaly plots (Time-Height) showing upper level cooling in the case of Lehar and upper level as well as mid (above the boundary layer) warming in case of Phailin. - Upper level warming promotes coupling between the upper and lower levels and is correlated to the drop in pressure.

54 Role of land surface – Potential modulators/sources of low entropy air/ shear?

Phailin Lehar

Lagrangian Back Trajectory analysis of low level influx of dry air ( < 8 g/kg Specific Humidity) for TCs Phailin and Lehar

Intrusion of dry air originating from the land (Northern/Eastern India) in the case of Lehar as the storm approaches land. No intrusion of dry air observed with Phailin until it makes landfall.

55 Real-time prediction of Storm Surge

56 1-way coupled Storm Surge Prediction for TC PHAILIN

PEAK SURGE ELEVATIONS AT THE BOUNDARY (IC:2013100912) 3.5

3

2.5

2

1.5

1

0.5

0

Vishakhapatnam Srikakulum §Landfall Gopalpur Puri Paradip Dhamara Chandipur

Left_LF 3.18 3 Landfall 2.93 Gopalpur 2.55 Puri 2.24 2 Max. peak surge = 3.2 m (Predicted before 84 hours of Landfall) 1

0

Sea surface elevationSea (m) 60 70 80 90

-1 Time (hrs) Storm Surge Prediction for TC Hud-Hud (a) IC: 12 UTC 08 Oct 2014 (91 h Fcst) (c) IC: 00 UTC 11 Oct 2014 (31 h Fcst)

Max. value of peak surge=1.4 m Max. value of peak surge=1.7 m

(b) IC: 12 UTC 08 Oct 2014 (91 h Fcst) (d) IC: 00 UTC 11 Oct 2014 (31 h Fcst) Peak surge envelop along the coast (m) coastthe alongenvelop surge Peak

(e)

Skill (%) Skill Peak surge (m) Peak

Different initial conditions (ddhh) CONCLUSIONS

• ARW model provides better track guidance, while, intensity error is more.

•HWRF provides better track and intensity forecast. Intensity improvement is mainly due to vortex initialization and relocation.

•The model performance for the prediction of tropical cyclones has improved with the data assimilation.

•The assimilation of Satellite derived winds have significantly improved the prediction of tropical cyclones in terms of track and structure. However, improvement in intensity is comparatively less. The landfall prediction is also improved considerably.

•DWR assimilation, improved the intensity and the associated rainfall patterns both spatially and temporarily. What we are doing to the nature is but a mirror reflection of what we are doing to ourselves and to one another.

60