Comparison of Local, Regional, and Scaling Models for Rainfall Intensity-Duration-Frequency Analysis

Area of Science:

• Hydrology
• Extreme weather analysis
• Urban planning

Background:

• Intensity-Duration-Frequency (IDF) analyses are crucial for urban flood management.
• Accuracy of IDF models relies on historical rainfall data, often sparse in developing countries.
• High-resolution rainfall data is essential but frequently unavailable.

Purpose of the Study:

• To quantify the performance of different IDF models based on the number of available high-resolution and daily rain gauges.
• To assess model accuracy and uncertainty under varying data availability scenarios.
• To provide insights for conducting IDF analyses in data-scarce regions.

Main Methods:

• Cross-validation framework using Monte Carlo bootstrapping experiments.
• Application of five IDF models (local, regional, scaling) with the generalized extreme value (GEV) distribution.
• Analysis of annual rainfall maxima from 30-min to 24-h durations using 223 high-resolution gauges in Arizona.

Main Results:

• All tested models showed similar performance for return periods up to 30 years.
• Local and regional models performed best when more than 10 high-resolution gauges were available.
• Incorporating daily rainfall data (≥10 gauges) significantly reduced uncertainty and increased accuracy, particularly with scaling models.

Conclusions:

• Bias correction of the GEV shape parameter is recommended for estimating quantiles with large return periods.
• Simple scaling models effectively infer subdaily rainfall statistics from daily data when high-resolution data is scarce.
• Model performance is influenced by the ability to account for elevation effects on parameters.