Developing an Inner-Core SST Cooling Predictor for Use in SHIPS

(Joint Hurricane Testbed funded project 2003-2005)

Principal Investigator: Joseph Cione

NOAA/AOML/ Hurricane Research Division

Co- Investigators:

John Kaplan (HRD); Chelle Gentemann (RSS); Mark DeMaria (NESDIS/ORA)

1.  Introduction

Results from Cione and Uhlhorn (2003) illustrate a clear link between inner-core sea surface temperature (SST) cooling (relative to ambient SST conditions ahead of the storm) and subsequent TC intensity change (Figure 1).

Simply stated, hurricanes that cooled the least intensified the most. Relatively modest changes in inner-core SST (order 1.0K) were shown to alter maximum total enthalpy flux (latent plus sensible) by 45% or more. Changes in surface energy flux of this magnitude can be the difference between a system that rapidly intensifies and one that quickly decays. Complicating matters further, inner-core SST cooling patterns often go undetected since it is the most difficult region of the hurricane to accurately and routinely observe. Since operational coupled models have had difficulty simulating and validating inner-core SST cooling patterns (Bender et al 2000), the recent findings from Cione and Uhlhorn (2003) highlight the critical (and immediate) need to accurately account for inner-core SST conditions.

For several years, the Statistical Hurricane Intensity Prediction Scheme (SHIPS) (DeMaria and Kaplan 1999) has shown skill in predicting TC intensity change (DeMaria et al. 2002). The primary goal of this JHT project is to build on SHIPS existing skill by providing improved ambient SSTs as well as new and critically important SST estimates within the difficult to observe high wind inner-core TC environment. It is believed that incorporating a more accurate depiction of the ocean surface boundary condition (SST) into SHIPS will directly result in improved forecasts of TC intensity change, and as such, address the Tropical Prediction Center's (TPC) highest forecast priority (TPC A-1).

2. Significant Accomplishments

Building upon in-situ inner-core hurricane observations documented in Cione and Uhlhorn (2003), a statistically stable cooling algorithm that utilized along track SST, TC latitude and storm speed as predictors was developed.

2a. Dependent Sample Results (1989-2004)

The newly developed TC Inner-Core SST Cooling algorithm was tested in a dependent mode using 1000s of individual forecasts taken from the SHIPS 1989-2004 Atlantic basin storm database. Analyses incorporating observations from tropical depression strength or greater systems resulted in improved SHIPS intensity forecasts over all time periods between 12-120h. These results are encouraging since no mean degradation was found at any forecast interval even though many weak systems were included in the analysis. When results were stratified by initial storm intensity and observed intensity change, the positive impact of utilizing inner-core SSTs on SHIPS forecasts was found to be even more significant (see Figures 2-3).

Figure 2 Figure 3

Figures 2 and 3 illustrate that intensity forecasts involving major (Category 3 4 and 5) hurricanes and forecasts associated with rapid intensity change (top and bottom 10% of the hurricane sample) netted the largest forecast improvements (7-13.5%) when estimated inner-core SST was used in lieu of Reynolds SSTs (used operationally in SHIPS prior to 2005). These very encouraging results were recently presented in detail at the 2005 IHC held in Jacksonville Florida (March, 2005).

2b. Independent Sample Results and Testing of High Temporal and Spatial Resolution SSTs in SHIPS (2003 Atlantic hurricane season)

While the dependent sample results are very significant, the next step was to test the impact of using the inner-core SST cooling algorithm on independent storm samples (i.e. 2003 and 2004 Atlantic Hurricane season tropical cyclones). Baseline SHIPS 2003 forecasts (that use Reynolds weekly SSTS) were compared with re-run SHIPS forecasts using inner-core SST estimates derive from the algorithm. The summary results from these SHIPS model re-runs for the 2003 season are shown in Table 1.

Table 1. SHIPS Forecast Sensitivity to SST (2003 North Atlantic Hurricane Season)

SHIPS Average Intensity Forecast Skill (relative to SHIFOR) Using Alternate SSTs (%)
	Forecast Interval (hr)
Model	12	24	36	48	72	96	120
SHIPS w/Reynolds (SST: weekly-100km-IR-operational)	16.3	30.2	31.1	35.3	31.7	13.9	-10.0
SHIPS w/inner-core SSTs (SST: predicted-inner-core)	15.3	28.9	32.5	38.2	36.4	22.4	-1.3
SHIPS w/Inner-core SST minus SHIPS w/Reynolds	-1.0	-1.3	1.4	2.9	4.7	8.5	8.7
No. Cases	318	289	260	232	183	146	121

Table 1. SHIPS average forecast sensitivity using two different SST estimates for the 2003 North Atlantic hurricane season. SSTs tested include Reynolds (operationally used in SHIPS) and inner-core SST algorithm-derived SSTs. The skill shown is calculated as the fractional increase/decrease in forecast error from the baseline SHIFOR intensity forecast. Positive (negative) values denote forecast improvement (degradation) over SHIFOR. Shown in blue (red), SHIPS w/ inner-core SSTs minus SHIPS w/Reynolds depicts the forecast skill improvement (degradation) found when SHIPS forecasts using inner-core estimates of SST are compared with SHIPS forecasts using Reynolds SSTs.

From Table 1, we see that significant average skill improvements were found (over SHIPS w/Reynolds) between 72-120h when Inner-core SST estimates were used. Over the 72-120h forecast interval, average forecast average skill improvement (relative to SHIPS w/Reynolds) was found to be 7% (450 forecasts) and 8.6% between 96-120h (267 forecasts).

2c. Additional Independent Sample Results (2004 Atlantic hurricane season)

Table 2. SHIPS Forecast Sensitivity to SST (2004 North Atlantic Hurricane Season)

Average Intensity Error (knots) and % Skill Improvement/Degradation to SHIPS
	Forecast Interval (hr)
Model	12	24	36	48	72	96	120
SHIPS w/Reynolds (SST: weekly-100km-IR-operational)	7.8	10.7	12.7	13.6	16.9	22.5	26.2
SHIPS w/inner-core SSTs (SST: predicted-inner-core)	7.7	10.6	12.6	13.5	16.5	21.6	25.1
% Improvement/Degradation (when Inner-core SSTs were used)	0.9	0.7	0.3	0.7	2.4	4.3	3.8

Table 2. SHIPS average intensity error (kts) and % change in skill for the 2004 North Atlantic hurricane season. SSTs tested include Reynolds (operationally used in SHIPS) and inner-core derived SSTs. The improvement in skill shown (in blue) is calculated as the fractional decrease in forecast error from the baseline SHIPS intensity forecast.

From Table 2, we see that average SHIPS forecasts for the 2004 Atlantic hurricane season improved at every forecast interval between 12-120h. Over the 72-120h forecast interval, average forecast average skill improvement (relative to SHIPS w/Reynolds) was found to be 3.5%. It should also be noted that for both the 2003 and 2004 samples the average percent improvement to SHIPS forecasts were much larger for both the 'Hurricane Only' and Rapid Intensifier' (as defined by J. Kaplan) sub samples.

-Exact statistics/graphics unavailable at time of writing -

3.  Final Comments/Summary of Work

Prior to the onset of the 2005 Atlantic hurricane season it was decided that the inner-core SST cooling algorithm would be used operationally within SHIPS due to the strong dependent sample (1982-2004) and independent sample (2003 and 2004) results already shown. The PIs are very pleased with these findings and decision to use the algorithm operationally in 2005 and beyond. Future planned work in this area includes the testing of a next generation inner-core SST algorithm that will be basin independent (E and W Pacific as well as Atlantic). This version of the algorithm would also incorporate new data (2000-2005) and most likely utilize information on storm intensity and subsurface ocean structure. This work, while preliminary, shows great promise.

Another important effort will be to see if a surface flux predictor can be developed for use in SHIPS. The inner-core SST predictor is only "one half of the puzzle" with regard to estimating surfaces fluxes in within the high wind TC environment. Successfully estimating inner-core surface air temperature and moisture, will in turn enable testing of 'bulk' heat flux parameters for possible future use in SHIPS.

In closing, the PIs feel fortunate that they were given the opportunity to assist the operational forecast community. We are confident that we can build upon the progress we have already made by significantly improving the inner-core SST algorithm as well as potentially deriving new inner-core air-sea predictors for future use in SHIPS.

Key Reference

Cione, J.J. and E. Uhlhorn 2003: Sea surface temperature variability in hurricanes: Implications with respect to intensity change. Mon. Wea. Rev., 131: 1783-1796.