Seventh Annual AMS Student Conference

P1.53

Development of a hurricane secondary eyewall formation index

Matthew Sitkowski, CIMSS/Univ. of Wisconsin, Madison, WI; and J. Kossin

Increased satellite coverage and improvement of remote sensing instrumentation has allowed for frequent monitoring of tropical cyclones (TCs) over the past few decades. Microwave imagery highlights the most intense convection within the inner core of a TC allowing for a clear depiction of how the eyewall and rainbands are arranged. The evolution of the inner core convective pattern plays a vital role in the intensity of the TC. Major (V > 100 kts) TCs are commonly observed to undergo a transformation where a second eyewall forms at some distance around the primary eyewall. This secondary eyewall formation (SEF) is generally the precursor to an eyewall replacement cycle. These cycles often cause dramatic and rapid changes in intensity and are very important to recognize in a forecasting setting, particularly when a TC is approaching land. At present, forecasters have no objective methods to help recognize SEF, and they need to rely on aircraft reconnaissance data, coastal radar imagery, or satellite microwave imagery to make a subjective determination of whether SEF is occurring or not. These data are not always available in a timely manner. Over 4,500 SSMI, SSMIS, TRMM, and AQUA microwave images in combination with aircraft, radar, and additional satellite data, were used to examine nearly 175 major TCs over a ten year period (1997-2006) to develop SEF climatology for the Northern Hemisphere. Roughly 56% of all major TCs undergo SEF at least once. The West Pacific, North Atlantic, and East Pacific basin account for 60%, 25% and 10% of the SEF cases respectively. On average, there is a 26% probability of SEF occurring within 12 hours when TC winds exceed 95 kt. NCEP/NCAR reanalysis data were used to examine the ambient environment of the TCs included in the climatology. To identify the environmental features, environmental reanalysis fields were centered about the storm center at every 6-hourly storm position fix, and then a composite analysis was performed, which was then subjected to statistical significance tests. Additionally, Principal Component Analysis (PCA) was applied to the storm-centric environmental fields and the expansion coefficients of the leading modes of environmental variability were stratified by cases of SEF and tested for statistical significance. Thus far, a number of environmental features that relate significantly to SEF have been identified. The amplitude and meridional structure of the ambient sea surface temperature (SST) anomalies, low-level and mid-level moisture, vertical wind shear, and lapse rates are a few examples of some of the environmental features that play of role in SEF. The difference of the SST composite images is displayed in figure 1. The SEF TCs are generally warmer with differences pronounced north of the TC center. This indicates a reduction of meridional SST gradient associated with SEF, which is physically reconcilable with the hypothesis that SEF is more likely in an environment that's more symmetric around the storm. Features identified in the composite analysis and PCA were applied in a Bayesian classification framework that combines climatological probabilities of SEF with “class conditioned” probability density functions for the environmental features. Two classes (binary classification) were developed. Class 1 comprises TC fixes with intensity greater than 95 kt, but no SEF occurred in the following 12 hours. Class 2 comprises the sample of cases where SEF did occur within 12 hours. The Bayesian classifier provides probabilities that a particular group of features belongs to a particular class. Significantly higher probabilities were found in class 2 and lower probabilities in class 1, as desired (Figure 2). Plans to include GOES infrared imagery and microwave imagery in the analysis will likely add substantial skill in the Bayesian classifier. Another important future step will be the inclusion of the existing features that are used in the SHIPS model.

Poster Session 1, Student Conference General Poster Session
Sunday, 20 January 2008, 5:30 PM-7:00 PM, Exhibit Hall B

Previous paper  Next paper

Browse or search entire meeting

AMS Home Page