Statistical Modeling of Extreme Precipitation with TRMM Data

Area of Science:

• Climatology
• Hydrology
• Statistical Modeling

Background:

• Existing extreme precipitation monitoring systems often suffer from noisy data due to short records.
• The Tropical Rainfall Measuring Mission (TRMM) daily product (3B42) is a key data source for precipitation analysis.

Purpose of the Study:

• To improve extreme precipitation monitoring by developing a new regional statistical modeling approach.
• To generate more accurate and intuitive average recurrence interval (ARI) maps for disaster management.

Main Methods:

• Utilized a regional modeling approach by clustering TRMM data into 28,000 non-overlapping groups using k-means.
• Adopted the peak-over-threshold method and the Point Process framework for modeling extreme precipitation events.
• Replaced the block-maxima approach with a more robust statistical method for parameter estimation.

Main Results:

• The new methodology significantly reduces noise compared to existing models, yielding more reliable ARI maps.
• The enhanced system produces ARI maps that align better with NOAA's long-term ground-based observations.
• Demonstrated the effectiveness of pooling data from similar locations to improve model parameter estimates.

Conclusions:

• The proposed regional modeling approach offers a substantial improvement for extreme precipitation analysis.
• The methodology provides valuable data for policymakers in disaster monitoring and prevention efforts.
• The developed statistical framework is adaptable for analyzing other extreme climate variables.