Mong-Ming Lu and Ru-Jun May Research and Development Center Central Weather Bureau

Exploring the Possibility of Using Tropical Cyclone Numbers

to Project Taiwan Summer Precipitation Patterns

Mong-Ming Lu and Ru-Jun May

Research and Development Center

Central Weather Bureau

Motivation

Taiwan summer rainfall prediction beyond one month has no skill.

Tropical cyclone numbers show promising predictability.

Can we use tropical cyclone numbers to project Taiwan rainfall patterns?

Available Information

NW Pacific tropical cyclone activity FORCASTS

• The IRI Typhoon Activity Forecast

(http://iri.columbia.edu/forecast/typhoon/)

• Korean Meteorological Agency’s Forecast

• Johnny Chan’s Forecast

(http://aposf02.cityu.edu.hk)

• Prof. Joe Kwon’s Forecast

NW Pacific tropical cyclone activity MONITORING

* Total TC number in January-May

Analysis Procedure Predictors

TC numbers in January - May

January - December

July - October

June – August

Regional Difference

Group the stations into 5 groups –

North (7 stations), Central (4), South (3), East (6) and All (20)

Contingency Table Method• Categorize Data

TC numbers (More, Normal, Less)

as close to tercile as possible,

the members in “Normal” must be more

than in “More” and “Less”

Precipitation (More, Normal, Less) - Tercile

• Significance Test - Mason and Goddard (2001)

Identify which months of precipitation with statistically significant signals at 85%, 90% levels of confidence

Limitations:

* Significance criterion is sensitive to the category member

numbers.

* Satisfactory test is difficult to do when stations are too few.

Identification Procedure and Criteria:

• Data length: 1959-2001

• The contingency tables are made for each station.

• Signal Identification :

The 20 stations are grouped into 5 groups. The signal magnitude is represented by the percentage number of stations showing the signal of satisfying the significance criterion of any category. The uncertainty is represented by the percentage number of stations showing no signal. If the signal magnitude is less than the uncertainty, we decide that this group has no signal. The relative signal magnitude is defined as the difference between the signal magnitude and uncertainty. Significant signal is identified only when the relative signal magnitude is larger than 33%.

**The signal in south Taiwan should be treated with extreme caution because there are only 3 stations in that group. The signal can be identified if one statio

n satisfies the signal magnitude and two stations are uncertain.

http://e-service.cwb.gov.tw/docs/V3.0/knowledge/typhone/ch2/005.htm

Typhoon Invading Paths (1960-1999)

DATA

• Tropical Cyclone -

JTWC best track data final v., 1959-2002

• Precipitation -

CWB 20 stations

RESULTSPredictor: Annual Total Number of TCs

More Normal LessAnnual TC Number Above Normal Annual TC Number Normal Annual TC Number Below NormalW: Wet, D: Dry, N: Normal, W: Wet, D: Dry, N: Normal, W: Wet, D: Dry, N: Normal,no *: 85% , with *: 90% significance level no *: 85% , with *: 90% significance level no *: 85% , with *: 90% significance level

N. TWNC. TWNS. TWNE. TWNAll TWN N. TWNC. TWNS. TWNE. TWNAll TWN N. TWNC. TWNS. TWNE. TWNAll TWNJun D D D D Jun N JunJul N* N* Jul W* JulAug W Aug AugSep Sep SepOct Oct D* Oct W W W

JJ JJ W* JJ D*

JA N* JA JA D

AS AS AS D

SO SO SO W W W W

JJA JJA W* JJA D

JAS JAS N JAS D* D

ASO ASO ASO

JFMAM TC Number Above Normal JFMAM TC Number Normal JFMAM TC Number Below NormalW: Wet, D: Dry, N: Normal, W: Wet, D: Dry, N: Normal, W: Wet, D: Dry, N: Normal, no *: 85% , with *: 90% significance level no *: 85% , with *: 90% significance level no *: 85% , with *: 90% significance level

N. TWNC. TWNS. TWNE. TWNAll TWN N. TWNC. TWNS. TWNE. TWNAll TWN N. TWNC. TWNS. TWNE. TWNAll TWNJun D* D* D* Jun W* Jun W N*

Jul N* Jul W* Jul W

Aug Aug W* W* Aug N*

Sep Sep W SepOct N* N* Oct Oct W

JJ D* JJ JJJA D* N* JA W* JA W*

AS D* N AS N* ASSO D* SO N* SOJJA D* D* JJA W* JJA W

JAS D* JAS W* JASASO D* D* D ASO N* ASO

Predictor: January-May Total Number of TCs

More Normal Less

Predictor: July-October Total Number of TCs

JASO TC Number Above Normal JASO TC Number Normal JASO TC Number Below NormalW: Wet, D: Dry, N: Normal, W: Wet, D: Dry, N: Normal, W: Wet, D: Dry, N: Normal,no *: 85% , with *: 90% significance level no *: 85% , with *: 90% significance level no *: 85% , with *: 90% significance level

N. TWNC. TWNS. TWNE. TWNAll TWN N. TWNC. TWNS. TWNE. TWNAll TWN N. TWNC. TWNS. TWNE. TWNAll TWNJun D* D* D* Jun W Jun W N*

Jul Jul W W W W Jul D* D* N D*

Aug Aug W Aug N* N*

Sep Sep Sep N

Oct N Oct Oct N

JJ JJ W* W W* W JJ N D*

JA N* JA W* JA D*

AS AS W AS N D*

SO N N SO SOJJA N D* D* D* JJA W W* W JJAJAS JAS W W JAS D*

ASO N ASO W W ASO

More Normal Less

RAINFALL PROJECTIONS

What did we know in June 2003 ? 1. Jan-May total number (6) - above normal. What can we project ? High possibility for North Taiwan to have a DRY summer (June-October).

2. Dr. Johnny Chan’s Forecast: Annual total TC number will be normal / slightly below normal (263). What can we project ? None.

3. IRI’s Forecast: July-October total (named) TC number will be normal / slightly above normal. (reference Normal: 17-20)

Dr. Joe-H. Kwon’s Forecast: JJA total number will be below normal. (6.3±1) KMA: JJA total number below normal (?) What can we project ? normal / slightly above normal ? (Normal Wet; Above Normal Dry)

below normal High possibility for East Taiwan to be DRY.

June 2003 IRI Typhoon Activity Forecast

There is a slightly enhanced probability (approximately 40%) that the number of named tropical cyclones in the western North Pacific during the 2003 peak season (July to October) will be in the normal category, and also a 40% that the number of cyclones will be in the above normal category. The normal category is defined as between 17 and 20 named tropical cyclones. These probabilities are slightly greater than the long-term average probability of 33%. The accumulated cyclone energy (ACE*) index during these months also has an enhanced (approximately 45%) probability of being in the above normal range. Furthermore, a slight shift in the average longitude (westward) and latitude (nortward) of tropical cyclone tracks is predicted. This forecast is consistent with the possible onset of La Niña sea surface temperature anomalies that could develop, with a region of above normal SST in the central and western North Pacific. Normally, La Niña conditions would result in below normal SSTs in both the eastern and central tropical Pacific. Our SST forecast, however limits the below normal SST to east of the date line.

TC NUMBER VERIFICATION

Jan-May total number (6) of the Western North Pacific TCs this year is above normal.

Annual total TC number will be normal / slightly below normal (263).

up-to-now 19 named

x IRI’s Forecast: July-October total TC number will be normal / slightly above normal.

actually 13 named Dr. Joe-H. Kwon’s Forecast: JJA total number will be below normal

(6.3±1).

actually 8 named

Taiwan Precipitation Percentile Rank

Jan-May TC

MORE

SLP Ano. Composite - June

2003 June

SLP Anomalies

JASO TC

LESS

SLP Ano. Composite –

July

2003 July

SLP Anomalies

SUMMARY• TC numbers may be useful for projecting Taiwan summer pr

ecipitation patterns.• January-May total number of TCs shows most consistent sig

nals in the precipitation of northern Taiwan. More Dry Normal Wet Less signals are weak• Peak season (July-October) total number of TCs also show c

onsistent signals in the precipitation of eastern Taiwan. More, Less Dry Normal Wet• No Taiwan precipitation signals can be extracted based on

the annual total number of TCs.

Questions

• Where does the predictability of TC numbers come from?

• How have the TC number predictions been used? (Who are your customers?)