Printer Friendly Version
Back to Intensity Change projects | Back to Main Projects Page

Probabilistic Rapid Intensification (RI) Forecasts for the Atlantic and E. Pacific Basins

Principal Investigator: John Kaplan
Collaborating scientist:
Mark DeMaria (NOAA/NESDIS/ORA)
John Knaff (NOAA/NESDIS/ORA)
John Dostalek (CIRA)
Jason Dunion (CIMAS/HRD)
Joe Cione (NOAA/HRD)
Jun Zhang (CIMAS/HRD)
Jeffrey Hawkins (NRL)
Thomas Lee (NRL)
J. Solbrig (NRL)
Evan Kalina (CU)
Paul Leighton (HRD)


Methodology:

Although some modest improvements in tropical cyclone intensity guidance models has been made over the past couple of decades, predicting changes in tropical cyclone intensity remains a very challenging problem, especially for cases of rapid intensification (RI). In recent years, a statistically based rapid intensity index (RII) that utilizes large-scale predictors derived from the SHIPS model as well as a few GOES inner-core predictors (Table 1) has been developed to estimate the probability of RI over the succeeding 24-h for systems in both the Atlantic and eastern North Pacific basins (Kaplan et al. 2010). (Table1). Although the original version of the RII only provided estimates of the probability of RI for the 30-kt RI threshold, the version that was adopted for operational use by the NHC prior to the 2008 Hurricane Season also provided probability of RI estimates for the 25-kt and 35-kt RI thresholds. Figure 1 shows a verification of the operational RII forecasts for the 2008-2010 Hurricane Seasons for all three of the aforementioned RI thresholds. It can be seen that the SHIPS RII was generally skillful (save for the 35 kt threshold in the Atlantic) for this time period with the level of skill being higher in the eastern North Pacific than the Atlantic basin.

Results

Table 1
Predictors used in the current operational
Atlantic and eastern North Pacific RII
Previous 12-h intensity change
850-200 mb vertical shear from 0-500 km radius (24-h mean)
200 mb divergence from 0-1000 km radius (24-h mean)
850-700 mb relative humidity from 200-800 km radius (24-h mean)
% area from 50-200 km covered by -30°C GOES-IR brightness temperatures at T=0 h
Std. dev of 50-200 km GOES-IR brightness temperatures at T= 0 h
Potential intensity (Current intensity - maximum potential intensity)
Oceanic heat content (24-h mean)

Figure 1. The skill of the 2008-2010 operational RII forecasts. The number of RI cases for each RI threshold is also provided for both the Atlantic (blue) and E. Pacific (red) basins along the x-axis.

Although the above results indicate that the SHIPS RII exhibited some skill for the period of study, the skill was on the low side underscoring the need for additional research in this area. Thus, a recently completed JHT project was conducted to develop an improved version of the operational RII using data from three new data sources. The first of these three sources was the time evolution of inner-core structure as deduced from GOES IR imagery. Although some basic parameters from GOES IR imagery such as counts of cold cloud pixels are already included in the RII, the time evolution of the inner-core structure is not and recent research has shown that such information is correlated with RI. Thus predictors derived from complex principle principal analysis of the GOES IR imagery were employed to determine if such information could be used to improve the existing RII. The second source was microwave-derived total precipitable water. Research involving the SHIPS model has shown that total precipitable water is statistically correlated with intensity change. Thus, total precipitable water was employed to try to improve the RII. Finally, predictors that provided a measure of how conducive the overall thermodynamic environment surrounding the inner-core was for tropical cyclone intensification were also examined for their ability to improve the RII.

Table 2 shows the predictors that were ultimately selected for inclusion in the new experimental Atlantic and eastern North Pacific version of the RII. A comparison of Table 1 and Table 2 shows that the new experimental version of the RII includes six of the same predictors that were used in the operational version as well as two additional predictors and two replacement predictors. A more detailed description of the new experimental RII can be found in Kaplan et al. (2011).

Figure 2 shows the improvement in skill of the new experimental versions of the RII over the current operational versions for the 1995-2009 Atlantic and eastern North Pacific basin developmental samples. The improvements in skill are shown for the 25-kt, 30-kt, and 35-kt RI thresholds as well as an additional RI threshold of 40-kt for which a new version of the RII was recently developed. Inspection of the results in Fig. 2 indicate that the experimental version exhibited anywhere from about 2-5% more absolute skill (10-30% relative improvement) in the Atlantic and 0.2-3% greater absolute skill (1-12% relative improvement) in the E. Pacific.

Although the results in Fig. 2 are encouraging, they were obtained using observed fields. Thus all of the forecast cases from 2008-2010 were re-run using the real-time GFS forecast fields and NHC forecast tracks that were archived for these cases. Prior to performing the re-runs for a given year, both the operational and experimental versions of the RII were first re-derived by excluding all cases from that particularly year. Consequently, the results discussed below should closely mimic those that would have otherwise been obtained had these forecasts been made in real-time for this three-year period.

Table 2
Predictors used in the new experimental RII
Previous 12-h intensity change
850-200 mb vertical shear from 0-500 km radius (24-h mean)
200 mb divergence from 0-1000 km radius (24-h mean)
Percent area with TPW < 45 mm within 500 km radius 90° up-shear at T=0 h
Second principle component of GOES-IR imagery within 440 km radius at T= 0 h
Std. dev of 50-200 km GOES-IR brightness temperatures at T= 0 h
Potential intensity (Current intensity - maximum potential intensity)
Oceanic heat content (24-h mean)
Inner-core dry air predictor (24-h mean)
T=0 h maximum sustained wind

Figure 2. Skill improvements of the new Atlantic and eastern North Pacific Experimental RII over the current operational version for the 1995-2009 dependent sample (N=2524). Improvements are expressed as the increase in absolute skill for each of the RI thresholds. The numbers of RI cases for each threshold are shown along the x-axis with blue numbers for the Atlantic and red for the E. Pacific basin.

Figure 3. Skill of the experimental and operational versions of the Atlantic (top) and eastern North Pacific (bottom) versions of the RII for the independent 2008-2010 re-run forecasts. The Skill of each model is assessed relative to climatology as discussed in Kaplan et al. (2010). The number of RI cases for each threshold are given in orange along the x-axis.

Figure 3 shows a comparison of the skill of the experimental and operational versions of the RII for the 2008-2010 Atlantic and eastern North Pacific basin samples. It can be seen that the experimental RII was generally more skillful than both climatology and the operational version (save for the 40-kt RI) in both basins. However, these improvements where observed to be substantially larger for the lower RI thresholds perhaps due to the relatively small RI sample sizes available for the highest RI thresholds. It is interesting to note that both versions of the eastern North Pacific RII were found to exhibit increasing skill with increasing RI threshold magnitude while the opposite trend was observed for the Atlantic basin version. These opposing trends in skill as a function of RI threshold magnitude were also found for the 2008-2010 operational RII forecasts (Fig. 1).

Figure 4 shows an example of the performance of the new experimental version of the RII as well as the operational version for Hurricane Adrian (2011). The figure indicates that although both versions showed high RI probabilities at the time when RI first commenced, the experimental version correctly showed continuously high probabilities of RI during the entire period during which RI was observed while the operational RII probabilities decreased too quickly. Inspection of the individual RI forecasts indicates that the rapid decrease in RII probabilities of the operational version of the RII was due to the values of the relative humidity RI predictor falling outside the range of values for which RI had been previously been observed.

Figure 4. The probability of RI for the operational (solid blue) and new experimental (dashed blue) versions of the RII for eastern North Pacific Hurricane Adrian (2011). The NHC real-time maximum sustained wind estimates (solid red line) and the initial (T=0 h) times at which RI commenced (red triangles) are also provided.

In the future, additional research will be conducted to incorporate satellite- derived microwave imagery into the RII and to extend the RII out to 48-h as part of an ongoing NOAA/JHT project. In addition, an ensemble-based RII that employs both the operational version of the RII as well as additional versions that are generated using different statistical methods will be developed. Finally, a new intensity consensus model that makes use of probabilistic RII forecasts to generate deterministic intensity forecasts will be developed.

FY10-11 Achievements

  • A new experimental version of the current operational SHIPS RII was developed for use in the Atlantic and eastern North Pacific basins as part of the NOAA Joint Hurricane Testbed (JHT).
  • Presentations describing the new experimental version of the RII were made at both the 65th Interdepartmental Hurricane Conference and the 29th Conference on Hurricanes and Tropical Meteorology.
  • The new experimental RII was run and output made available to forecasters at the NHC in real-time commencing on Aug 1, 2011.
FY12-13 Milestones

  • Develop versions of the RII for the additional lead times of 12,36 and 48 h.
  • Develop improved RII that includes predictors derived from microwave imagery.
  • Run the new microwave-based and new lead-time versions of the RII in real-time and provide output to NHC forecasters.
  • Evaluate the skill of the new versions of the RII for independent Atlantic and eastern North Pacific samples.

Key references:

Kaplan, J., J. Cione, M. DeMaria, J. Dostalek, J. Dunion, J. Knaff, J. Zhang, T. Lee, J. Hawkins, E. Kalina, J. Solbrig and P. A. Leighton, 2011: Improvement in the rapid intensity index by incorporation of inner-core information. JHT Year 2 Final report.

Kaplan, J., M. DeMaria, and J.A. Knaff, 2010: A revised tropical cyclone rapid intensification index for the Atlantic and eastern North Pacific basins. Wea. Forecasting, 25, 220-241.

Kaplan, J. and M. DeMaria, 2006: Estimating the likelihood of rapid intensification in the Atlantic and E. Pacific basins using SHIPS model data, Preprints 27th Conference on Hurricanes and Tropical Meteorology , Monterey, CA, Amer. Meteor. Soc..

Kaplan, J., and M. DeMaria, 2003: Large-scale characteristics of rapidly intensifying tropical cyclones in the North Atlantic basin. Wea. Forecasting, 18, 1093-1108.

DeMaria, M., M. Manelli, L.K. Shay, J.A. Knaff, and J. Kaplan, 2005: Further improvements to the Statistical Hurricane Intensity Prediction Scheme (SHIPS), Wea. Forecasting, 20, 531-543.

Wilks, D.S., 1995: Statistical Methods in the Atmospheric Sciences, Academic Press, 467pp.


Back to Intensity Change projects | Back to Main Projects Page


Last Modified : August 16, 2011

Stay Connected