where are we in seasonal tropical cyclone prediction?

Where Are We in Where Are We in Seasonal Tropical Seasonal Tropical Cyclone Cyclone Prediction?Prediction?Johnny ChanJohnny ChanGuy Carpenter Asia-Pacific Climate Impact CentreGuy Carpenter Asia-Pacific Climate Impact CentreSchool of Energy and EnvironmentSchool of Energy and EnvironmentCity University of Hong KongCity University of Hong Kong

2

Guy Carpenter Asia-Pacific Climate Impact Centre, School of energy and Environment, City University of Hong Kong

OutlineOutline

Statistical methodsStatistical methods

Statistical dynamical methodStatistical dynamical method

Dynamical methodsDynamical methods

SummarySummary

3


Statistical MethodStatistical Method

4


Statistical methodStatistical method

Identify a list of variables relating to Identify a list of variables relating to the atmospheric and oceanographic the atmospheric and oceanographic conditions prior to the season that conditions prior to the season that significantly correlate with seasonal significantly correlate with seasonal tropical cyclone activitytropical cyclone activity

Perform regressions to derive Perform regressions to derive prediction equationsprediction equations

5


Examples of Predictors used in the CityU ForecastsExamples of Predictors used in the CityU Forecasts

Index of the westward extent of the subtropical high over the western Index of the westward extent of the subtropical high over the western North PacificNorth Pacific

Index of the strength of the India-Burma trough (15-20Index of the strength of the India-Burma trough (15-20ooN, 80-120N, 80-120ooE)E)

West Pacific indexWest Pacific index

Sea surface temperature (SST) anomalies in the NINO3.4 region (5Sea surface temperature (SST) anomalies in the NINO3.4 region (5ooS-S-55ooN,170-120N,170-120ooW)W)

Sea surface temperature (SST) anomalies in the NINO4 region (5Sea surface temperature (SST) anomalies in the NINO4 region (5ooS-S-55ooN, 160N, 160ooE-150E-150ooW)W)

Equatorial Southern Oscillation Index (Equatorial SOI)Equatorial Southern Oscillation Index (Equatorial SOI)

Equatorial Eastern Pacific SLP - Indonesia SLP (standardized Equatorial Eastern Pacific SLP - Indonesia SLP (standardized anomalies)anomalies)

large-scale atmospheric conditions

ENSO conditions

6


Forecasts of Annual Forecasts of Annual Tropical Cyclone Tropical Cyclone Activity over the Activity over the

western North Pacific western North Pacific (Deviations from (Deviations from

Observations)Observations)

All tropical cyclonesAll tropical cyclones

Tropical storms and typhoonsTropical storms and typhoons

TyphoonsTyphoons

7/117/11

8/118/11

7/117/11

7



western North Pacific western North Pacific (Deviations from (Deviations from

Observations)Observations)



TyphoonsTyphoons

8



western North Pacific western North Pacific vs. Observationsvs. Observations



TyphoonsTyphoons

9


Annual No. of NWP TS & TYAnnual No. of NWP TS & TY

10


Statistical Dynamical Statistical Dynamical MethodMethod

11


Statistical vs. Statistical-dynamical MethodsStatistical vs. Statistical-dynamical Methods Problem with the statistical method

• Relate the past events and future conditions by statistics• Inherent problem

assumes the future would behave the same as the past, which may not be correct

Statistical-dynamical method partly solves the inherent problem by• relating dynamical model predictions with future conditions

Time

# TCsObservations statistical prediction

several months

Dynamical atmospheric

model

Predicted future conditionsIntegrate over time

statistical prediction

12


Dynamical model data – DEMETERDynamical model data – DEMETER

Development of a European multimodel ensemble system for seasonal to interannual prediction (from European Union)– 7 models (CERFACS, ECMWF, INGV, LODYC,

Météo-France, MPI and UKMO)– 9 ensemble members each– 6 months forecasts available– Base time @ 1 Feb, May, Aug, Nov– 1980-2001 (22 years hindcast)– 2.5 x 2.5 degree resolution

13


Parameter Physics

Geopotential (200-, 500-, 850-hPa)

subtropical high

Wind fields (200-, 500-, 850-hPa)

steering flow

SST TC genesis

Sea-level pressure (SLP)

subtropical high, low for TC genesis

Dynamical model data – DEMETERDynamical model data – DEMETER

14


Tracks of EC landfalling TCs 1980 – 2001, Aug – Sept Tracks of EC landfalling TCs 1980 – 2001, Aug – Sept

Subtropical High

15


Subtropical High

Subtropical High

GC

FL

Tracks of FL/GC Tracks of FL/GC landfalling TCs landfalling TCs 1980 – 2001,1980 – 2001,Aug – SeptAug – Sept

16


MethodologyMethodology Compute the 9-member ensemble mean of

each model-predicted atmospheric fields (Aug-Sept)

geopotential, zonal and meridional winds (850, 500 and 200 hPa)

SST, SLP

Extract the first 4 EOF modes of each predictor fields

11 fields x 4 modes = 44 potential predictors from each DEMETER model

Test the statistical significance of the relationship between the coefficient of each mode and the number of landfalling TCs

17


Fit a forecast equation for the number of landfalling TCs in each region

Poisson regression Cross-validation (Jackknife method)

7 forecast equations, each from an individual model

Multimodel equation derived from the 7 equations

Simple average Agreement coefficient weighted-average

MethodologyMethodology

18


RegressionRegression

Linear regression is used in most previous studies

Normality assumption of predictors and predictand Fails in # landfalling TCs (Discrete non-negative

integers)

Poisson regression Discrete probability distribution Zero probability for negative numbers

Stepwise regression

19


Factors affecting EC landfalling TCsFactors affecting EC landfalling TCsModel CERFACS

Level Parameter EOF mode

200 hPa zonal wind 1

zonal wind 3

geopotential 1

500 hPa zonal wind 1

geopotential 1

geopotential 4

850 hPa meridional wind 1

surface SST 1

MSLP 1

20


Observed vs. Observed vs. PredictedPredictedEast CoastEast Coast

Single model: CERFACS

Multimodel

21


Single model: LODYC

Multimodel

Observed vs. Observed vs. PredictedPredictedGulf CoastGulf Coast

22


Single model: LODYC

Multimodel

Observed vs. Observed vs. PredictedPredictedFloridaFlorida

23


Dynamical MethodDynamical Method

24


Dynamical method (1)Dynamical method (1)

Run a global circulation model (GCM) Run a global circulation model (GCM)

Identify and count the number of vortices Identify and count the number of vortices from the model integrationsfrom the model integrations

25


IRI forecastsIRI forecasts

26


Dynamical method (2)Dynamical method (2)

Run a global circulation model (GCM) with Run a global circulation model (GCM) with a relatively coarse resolutiona relatively coarse resolution

Solutions from the GCM are used as Solutions from the GCM are used as boundary conditions for a regional model boundary conditions for a regional model with a higher resolution that can “resolve” with a higher resolution that can “resolve” a tropical cyclonea tropical cyclone

Integrate the regional model to predict Integrate the regional model to predict seasonal activity.seasonal activity.

27


28


Example of 850-hPa flow and relative vorticityExample of 850-hPa flow and relative vorticity

29


Example of simulation of a 3-month forecastExample of simulation of a 3-month forecast

red – simulated

blue - observed

30


The modelThe model

Modified version of Regional Climate Model Modified version of Regional Climate Model Version 3 (RegCM3) developed at ITCP Version 3 (RegCM3) developed at ITCP

Horizontal resolution: 60 kmHorizontal resolution: 60 km

Emanuel cumulus schemeEmanuel cumulus scheme

Domain: 94Domain: 94ooE-172E-172ooW, 14W, 14ooS-41S-41ooNN

Initial conditions: 8 ensemble members from Initial conditions: 8 ensemble members from 00UTC on 1 May and every 6 hr after00UTC on 1 May and every 6 hr after

Integration till the end of OctoberIntegration till the end of October

31


The modelThe model

Initial and boundary conditions: ERA40 Initial and boundary conditions: ERA40

SST: weekly OISST from NOAASST: weekly OISST from NOAA

32


Detection of a tropical cycloneDetection of a tropical cyclone

Local maximum ζ850hPa ≥ ζT (450 x 10-6 s-1)

TT300hPa300hPa at centre – T at centre – Tenvironmentenvironment ≥ 1 ≥ 1ooCC

lifetime ≥ 2 dayslifetime ≥ 2 days

Genesis over the oceanGenesis over the ocean

33


Example of a tropical cyclone in RegCM3Example of a tropical cyclone in RegCM3

34


Example of a tropical cyclone in the Regional ModelExample of a tropical cyclone in the Regional Model

35


Example of a tropical cyclone in the Regional ModelExample of a tropical cyclone in the Regional Model

36


Model Climatology (1982-2001, May to Oct)Model Climatology (1982-2001, May to Oct)

37


Model vs. Observed (850-hPa relative vorticity)Model vs. Observed (850-hPa relative vorticity)

ERA40 May RegCM3 May

ERA40 Jul RegCM3 Jul

38



Monthly Distribution

Latitude Distribution

Longitude Distribution

39



40


Model vs. Observed (1997, May to Oct)Model vs. Observed (1997, May to Oct)

41


Model vs. Observed (1998, May to Oct)Model vs. Observed (1998, May to Oct)

42


Model vs. ObservedModel vs. Observed

JTWC warm

JTWC cold

RegCM3 warm

RegCM3 cold

43


Model vs. Observed (500-hPa geopotential heights)Model vs. Observed (500-hPa geopotential heights)

ERA40 May

ERA40 Jul

RegCM3 May

RegCM3 Jul

44


SummarySummary

Even with a 60-km resolution, RegCM3 is able to Even with a 60-km resolution, RegCM3 is able to generate vortices with structures that resemble generate vortices with structures that resemble those of real tropical cyclones.those of real tropical cyclones.

The model is capable of reproducing the basic The model is capable of reproducing the basic climatology, the interannual variability of tropical climatology, the interannual variability of tropical cyclones and tracks of individual tropical cyclones cyclones and tracks of individual tropical cyclones in the western North Pacific.in the western North Pacific.

It is therefore possible to use RegCM3 with global It is therefore possible to use RegCM3 with global model predictions as initial and boundary model predictions as initial and boundary conditions to produce seasonal forecasts of tropical conditions to produce seasonal forecasts of tropical cyclone activity.cyclone activity.

45


RegCM3 simulation using ECHAM5 20C runRegCM3 simulation using ECHAM5 20C run

46



47



48


SummarySummary

Statistical methods can provide some clues on Statistical methods can provide some clues on tropical cyclone activity but suffers from an tropical cyclone activity but suffers from an inherent problem of predicting future events based inherent problem of predicting future events based only on past conditionsonly on past conditions

Statistical-dynamical methods can provide Statistical-dynamical methods can provide predictive information and therefore should give predictive information and therefore should give better results, but still suffers from the statistical better results, but still suffers from the statistical nature of the method.nature of the method.

Dynamical model forecasts could be the way Dynamical model forecasts could be the way forward to predict tropical cyclone risks although forward to predict tropical cyclone risks although more research is still necessary on (a) fine-tuning more research is still necessary on (a) fine-tuning the regional model, and (b) having a better GCM.the regional model, and (b) having a better GCM.

where are we in seasonal tropical cyclone prediction?

Documents

western north pacificindex

future conditionstime

oceanographic conditions

prediction equations

statistical methodrelate

past events

nino4 region

nwp ts ty