The objective of this study is to establish a system for predicting the tropical cycle (TC) activity in the South Pacific zones during week-long periods. There are three key requirements to accomplish this objective. First, an investigation of physically meaningful and predictable weekly TC variability is required. Second, based onphysical considerations, potential predictors associated with weekly TC variability need to be identified. Third, a forecast model based on selected predictors needs to be developed and model performance in predicting sub-seasonal TC activity evaluated. This study is designed as a three-year project. Year 1 (2016): identification of potential predictors for sub-seasonal tropical cyclone genesis. Year 2 (2017): development and verification of a model for predicting sub-seasonal tropical cyclone genesis. Year 3 (2018): improve the prediction model and disseminate the results, including outreach. In this report, we will focus on the identification of possible sources (i.e., potential predictors) associated with weekly TC variability based on physical considerations.

Based on composite analysis, the MJO phase has a strong relationship with sub-seasonal TC genesis in the South Pacific Ocean. TIGGE forecasts are capable of predicting the MJO itself; therefore, it is reasonable to use MJO forecasts from TIGGE as one of the primary predictors of weekly TC genesis prediction. Various SPCZ characteristics, strength, area, centroid latitude/longitude, are dominated by either seasonal or semi-annual variability. However, the maximum relative vorticity axis always lies 6°-10°S to the south of the SPCZ position, independent of its wide year-to-year excursions. Given the collocation between the large cyclonic vorticity and TC genesis areas, it is reasonable to use relative vorticity averaged in each zone instead of directly using various SPCZ characteristics. Based on the relationship between the seasonal climate cycle of TC activity and local relative vorticity, the area-averaged relative vorticity field is regarded as a good predictor for weekly TC genesis prediction, especially for regions west of the dateline. Considering the relationship between ENSO and weekly TC genesis probabilities for each zone, we decided to use the Nino3.4 index calculated from TIGGE forecasts as one predictor of weekly TC genesis. Based on the relationship between the seasonal cycle of TC activity and ENSO phases, TC genesis increased east of the dateline in the warm phase ENSO years, and increased in the western part in the cold phase ENSO years.

There is a room for further improvement of potential predictors. For instance, the terms used to generate the Genesis Potential (GP) index will be further considered. The four different terms in GP index are absolute vorticity at 850 hPa, relative humidity at 700 hPa, vertical wind shear between 850 and 200 hPa, and maximum potential intensity. As the GP index empirically quantifies the effects of large-scale environment on tropical cyclogenesis, and considers both thermodynamic and dynamic factors for cyclogenesis, the consideration of terms in GP index would be good predictors for sub-seasonal TC genesis prediction. This work will be done in the next year.

Even though all predictors can explain part of the sub-seasonal TC activity, the inclusion of as many potential predictors as possible in the multivariate regression does not necessarily guarantee better predictability. In particular, for the case of mutually correlated potential predictors with stepwise procedures for determining the predictor, screening is required in the logistic regression. We will use the stepwise regression approach to select a best set of predictor variables from pools of potential predictors for each zone and each forecast lead-time. This work will also be done in the next year.

1. INTRODUCTION

1.1 Background and motivation

1.2 Research objectives

2. DATA

2.1 TC best track dataset

2.2 TIGGE forecast

2.3 Reanalysis data

3. CLIMATOLOGY OF TC GENESIS IN THE SPO AND THE ASSOCIATED BASIC STATE

4. TARGET ZONE OF INTERESTS

5. SEASONAL CYCLE OF WEEKLY TC PROBABILITIES

6. IDENTIFICATION OF POTENTIAL PREDICTORS

6.1 MJO

6.1.1 Relationship between MJO phase and TC genesis

6.1.2 MJO predictability with TIGGE

6.2 SPCZ

6.2.1 SPCZ variability and relationship with TC genesis

6.3 ENSO

6.3.1 Relationship between ENSO and TC genesis

7. SUMMARY

8. FUTURE PLANS

8.1 Improvement of predictors

8.2 TC statistical forecast model

8.3 Statistical modeling issues

8.4 Stepwise predictor screening

REFERENCE