Uncertainty analysis for computer model projections of hurricane losses

Area of Science:

• Meteorology and atmospheric science
• Risk management and actuarial science
• Structural engineering and wind engineering

Background:

• Hurricane Charley (2004) demonstrated significant uncertainty in real-time hurricane loss projections, with estimates fluctuating dramatically.
• The Florida Commission on Hurricane Loss Projection Methodology (FCHLPM) mandates rigorous auditing of loss projection models for insurance rate filings.
• Audits include crucial uncertainty and sensitivity analyses performed by multidisciplinary expert teams.

Purpose of the Study:

• To perform an uncertainty and sensitivity analysis on loss projections derived from a Holland B parameter hurricane wind field model.
• To quantify the impact of input variables on wind speed and loss uncertainty in hurricane modeling.
• To inform and improve future uncertainty and sensitivity analyses for hurricane loss projection models.

Main Methods:

• Utilized a Holland B parameter hurricane wind field model for loss projection.
• Conducted uncertainty analysis to quantify the reduction in uncertainty attributable to each input variable.
• Performed sensitivity analysis to assess the influence of input variables on model outputs.

Main Results:

• The analysis quantified the expected percentage reduction in uncertainty for both wind speed and financial loss.
• Identified key input variables that significantly influence the uncertainty of hurricane loss projections.
• Provided a framework for understanding and reducing uncertainty in hurricane damage assessments.

Conclusions:

• Uncertainty and sensitivity analyses are critical for validating hurricane loss projection models.
• The Holland B parameter model's uncertainty can be systematically analyzed to improve forecasting.
• This research contributes to more reliable insurance rate filings and better hurricane risk management.