1 James D. Doyle 1, Hao Jin 2, Clark Amerault 1, and Carolyn Reynolds 1 1 Naval Research Laboratory, Monterey, CA 2 SAIC, Monterey, CA James D. Doyle 1,

1

James D. Doyle1, Hao Jin2, Clark Amerault1, and Carolyn Reynolds1

1Naval Research Laboratory, Monterey, CA2SAIC, Monterey, CA

James D. Doyle1, Hao Jin2, Clark Amerault1, and Carolyn Reynolds1

1Naval Research Laboratory, Monterey, CA2SAIC, Monterey, CA

NRL Mesoscale Model Adaptive Observing during T-PARC/TCS08NRL Mesoscale Model Adaptive

Observing during T-PARC/TCS08

•Global-model adjoint tools (mostly dry) have been applied to TC track•Adjoints have not been applied to TC genesis systematically

-Forecasts of genesis are often poor (e.g., T-PARC/TCS08)• Is targeting for TC genesis feasible (T-PARC/TCS08)?

•Global-model adjoint tools (mostly dry) have been applied to TC track•Adjoints have not been applied to TC genesis systematically

-Forecasts of genesis are often poor (e.g., T-PARC/TCS08)• Is targeting for TC genesis feasible (T-PARC/TCS08)?

2

COAMPS Adjoint SensitivityCOAMPS Adjoint Sensitivity

J: response function xi: state vector at initial time ti xt: state vector at final time tf

MT: adjoint of non-linear model, M

Sensitivity of the response function at final time tf

to the initial (or model forecast) state at time ti

T

i f

J JM

x x

Adjoint-Based “Optimal” Perturbations

jj j

s Jx

w x

wj: weighting parameter s: scaling parameter (umax~1 m s-1)

Adjoint Allows for the Mathematically Rigorous Calculation of Forecast Sensitivity

COAMPS® Nonlinear, Adjoint, Tangent Linear Model Setup• x=40 km, 18 h (18 h lead time) (real time 24 h with 24, 36, 48-h lead times)• x=10 km, 9 h (15 h lead time) (not shown here)• PBL, Surface Flux, Microphysics (no ice)• J: Kinetic Energy in a Box (lowest 1 km)

Adjoint Sensitivity for 14 West Pacific Genesis Events (5 Super Typhoons)

• Fitow, Man-Yi, Krosa, Sepat, Trami, Wipha, Yutu (Pre- T-PARC/TCS08)• Nuri, Sinlaku, Hagupit, Jangmi, Higos, TCS025, 16W (T-PARC/TCS08)

3

COAMPS Forecasts during T-PARC / TCS-08COAMPS Forecasts during T-PARC / TCS-08

• Real-Time COAMPS Adjoint for Targeting Guidance.

• 40 km resolution for 24-h, 36-h, 48-h lead times.

• Adaptive response function box.

• Real-Time COAMPS-TC 72-h Forecasts with Moving Meshes.

• 45/15/5 km horizontal resolution.

AdjointSensitivity

C130 Flight Track

COAMPS radar reflectivity every 30 minutes on 5 km moving grid

COAMPS Track (red) & Official Warning Positions (black) (12 h)

24h959mb

48h/916mb

72h/913mb

24h/925mb

48h/916mb

72h/985mb

0h0h

24h959mb

48h/916mb

72h/913mb

24h/925mb

48h/916mb

72h/985mb

0h0h

Jangmi

4

QScat

ssmi.com

QScat

ssmi.com

Nonlinear Model EvolutionTropical Cyclone Fitow (Pre- T-PARC)

Nonlinear Model EvolutionTropical Cyclone Fitow (Pre- T-PARC)

18-h 10-m Winds & SLP (0-h Adjoint)

1000 mb1000 mb1004 mb1004 mb

ResponseFunction Area

ResponseFunction Area

• Nonlinear model captures monsoon gyre and asymmetric structure• Adjoint sensitivity computed during early development phase (18-36 h)

36-h 10-m Winds & SLP (18-h Adjoint)

• Asymmetric structure

• 18-h adjoint sensitivity (06Z 29 - 00Z 30 Aug ‘07)

> 35 kts

Best Track

5

Sensitivity FieldsMoisture & Potential Temperature

Sensitivity FieldsMoisture & Potential Temperature

36-h KE sensitivity to 18-h state

Banded Sensitivity StructuresKE/(4 km)KE/qv 500 m

KE/qv KE/

Low-Level Moisture

Sensitivity

Mid-Level Sensitivity

e Perturbation

W E W E W E

Low-Level e Maximum

(Destabilization)

• Preferred regions of large sensitivity to low-level moisture and .• Low-level e optimal perturbations: destabilize & saturate core.

6

2

0

27TLM Optimal U’ at 500 m

18h (re-scaled at 3h)10-m Wind Speed Pert.

23 m s-1

Umax=39 m s-1•Rapid Growth •Eye Structure

Tangent Linearu’ at 500 m

24 m s-1

Evolved Adjoint-Based PerturbationsPerturbation Winds

Evolved Adjoint-Based PerturbationsPerturbation Winds

correlation ~ 0.5

Evolved 18 h

Nonlinearu’ at 500 m

24 m s-1

• Evolved perturbations in TLM and NLM are well correlated.• Rapid growth over 18 h: Usfc=23 m s-1, psfc=-8 hPa,

7

T-PARC/TCS08 Real-Time Adjoint ForecastsTC Sinlaku

T-PARC/TCS08 Real-Time Adjoint ForecastsTC Sinlaku

54-h KE sensitivity to 36-h state

• Vorticity sensitivity bands that are anticyclonically curved.• Strongest sensitivity to low- and mid-level and qv.• C130 sampled key portions of the sensitivity.

24-h adjoint sensitivity 36-h lead time

Valid at 12Z 10 Sep 2008

2-km vorticity sensitivity Total energy sensitivity

C130 Flight Track

C130 Flight Track

DropsondesDropsondes

Best Track

8

T-PARC/TCS08 Real-Time Adjoint ForecastsTC Jangmi

T-PARC/TCS08 Real-Time Adjoint ForecastsTC Jangmi

54-h KE sensitivity to 36-h state

• Vorticity sensitivity shows a wave packet pattern.• The and qv sensitivity have multiple maxima over a broad area.• C130 sampled only a small portion of the sensitivity.

24-h adjoint sensitivity 36-h lead time

Valid at 00Z 25 Sep 2008

2-km vorticity sensitivity Total energy sensitivity

Best Track

9

T-PARC/TCS08 Real-Time Adjoint ForecastsTC Jangmi

T-PARC/TCS08 Real-Time Adjoint ForecastsTC Jangmi

54-h KE sensitivity to 36-h state

• SST sensitivities were computed in real time.• The SST sensitivity often showed complex patterns.• C130 deployed many AXBTs during T-PARC/TCS08.

SST Valid at 00Z 25 Sep 2008 Surface temperature (SST) sensitivity

LL

10

Genesis Sensitivity FieldsTC Nuri (T-PARC/TCS08)

Genesis Sensitivity FieldsTC Nuri (T-PARC/TCS08)

36-h KE sensitivity to 18-h state

KE/ (0.5 km) & SLP (18 h)

Vortex Structure

Evolved Optimal U’ (18-36 h)

12 m s-1

•Rapid Growth •Eye Structure

18-h adjoint sensitivity (06Z 15 Aug - 00Z 16 Aug 2008)

Low-Level e Max. Tilted

Against Shear

KE/e (18 h)

W E• Large sensitivity to radial vorticity gradient.• The qv and sensitivities are 3-4x greater than u sensitivity.• Optimal e perturbations destabilize & saturate core.

11

Perturbation Energy CharacteristicsSummary of 16 Cases (13 Events)

Perturbation Energy CharacteristicsSummary of 16 Cases (13 Events)

Initial PerturbationDomain Average of Optimal

Perturbations Evolved in TLM

• Initial total energy maximum in low-levels (bottom up).• Deep perturbation growth throughout troposphere.• Non-developers show weakest growth.

Final Perturbation (18 h)

Upper-Level Max. Consistent With

Background

Deep Response Function Box is

Consistent

12

ConclusionsConclusions

•Adjoint sensitivity characteristics for TC genesis & intensification.-Low-level maxima suggestive of bottom-up progression.-Preferred regions of moistening, destabilization, & vorticity bands.-Broad SST sensitivity regions linked to convection.-TL approximation is accurate for TCs considered (with moisture).

• Predictability of tropical cyclogenesis.-Low predictability (500-m winds: 50% cases show > 10x growth 18 h-1).-Sensitivity structures (particularly qv and ) may be observable.

• Future Directions-Higher resolution sensitivity calculations.-Data denial and targeting experiments based on T-PARC / TCS08.

13

Genesis Sensitivity FieldsTC025 (Non-Developer)

Genesis Sensitivity FieldsTC025 (Non-Developer)

36-h KE sensitivity to 18-h state

• Non-developers show slow growth, weaker sensitivity to qv and .• Greater sensitivity to vertical shear.

KE/ (0.5 km) & SLP (18 h)

No Vortex Structure

Evolved Optimal U’ (18-36 h)

3 m s-1

•Slow Growth •4x Slower than Nuri

4x Weaker than Nuri

KE/e (18 h)

W E W E

Sensitive to Shear

KE/u (18 h)

14

Dry and Moist Sensitivity ComparisonGradients and Evolved Perturbations

Dry and Moist Sensitivity ComparisonGradients and Evolved Perturbations

KE/u5 km

6.9x10-3 m s-1

Moist

KE/u5 km

-3.1x10-4 m s-1

Dry

• Moist shows faster growth (>10x) and smaller scale than dry adjoint.• Dry processes organize large-scale, moist processes dominate growth.

Evolved p’sfc

-8 hPa-5 hPa SLP Min.

Moist

-0.3 hPa

Evolved p’sfc

Dry

15

SST Sensitivity FieldsTropical Cyclones Fitow and Wipha

SST Sensitivity FieldsTropical Cyclones Fitow and Wipha

36-h KE sensitivity to 18-h state

Fitow Wipha

KE/SST KE/SST

Sensitivity of 36-h KE to 18-h SST

• Sensitivity to SST maximum near TC core & along track (asymmetry).• SST perturbations over 18 h (1°C max.) suggest pre-WISHE stage.

-Fitow: -0.75 hPa and +1.0 m s-1.-Wipha: -0.50 hPa and +1.2 m s-1.

16

Sensitivity of Adjoint Results to ResolutionTropical Cyclone Fitow (x=10 km)

Sensitivity of Adjoint Results to ResolutionTropical Cyclone Fitow (x=10 km)

qv perturbation (0-3 km) and SLP (15 h) KE/ (8 km) & SLP (15 h)

• Increase in horizontal resolution from 40 km to 10 km yields similar sensitivity structures and relationships.

• Sensitivity to moistening and warming in low-levels.• Banded vorticity sensitivity structures. • Evolved perturbations: winds~42 m s-1 (9 h), SLP ~ -8 hPa (9 h)

Banded Sensitivity Structures

Correlated and qv

Perturbations

perturbation (0-3 km)

17

NOGAPSCOAMPS

(Dry dx=100 km)

NOGAPSCOAMPS

48 h

Total Energy / SLP

Total Energy / SLP

SV: Initial

SV: Final Evolved Perturbations

Initial Sensitivity

Doyle, Amerault, Reynolds

NOGAPS SVs and COAMPS Adjoint ComparisonTropical Cyclone Fitow

NOGAPS SVs and COAMPS Adjoint ComparisonTropical Cyclone Fitow