Saved in:
Bibliographic Details
Main Author: He, Kunlong
Format: Recurso digital
Language:
Published: Zenodo 2025
Online Access:https://doi.org/10.5281/zenodo.16263506
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • <p><br>README: Typhoon Extreme Rainfall Event Analysis and Return Period Estimation</p> <p>Overview<br>--------<br>This MATLAB script processes historical typhoon-induced rainfall data to identify extreme rainfall events on a pixel-by-pixel basis across multiple years. It estimates the frequency (λ) and return period of these extreme events using a sliding window approach, statistical thresholding, and Generalized Pareto Distribution (GPD) modeling.</p> <p>Input Data<br>----------<br>- Typhoon rainfall data (.nc4): Half-hourly precipitation datasets for typhoon periods, stored under:<br>  G:\广东台风降雨\typhoonyear\{year}\*.nc4<br>  Each year has multiple NetCDF files (each representing a time step).<br>  'precipitation' is the main variable.</p> <p>- Static threshold reference:<br>  G:\文章\台风-地灾模型\数据\typoon10kmtypre.mat<br>  Contains a 3D matrix 'globalMatrix' used to compute the 95th percentile rainfall threshold across time.</p> <p>- Geo-reference raster:<br>  G:\文章\台风-地灾模型\数据\ygahazardpredictclip.tif<br>  Used to define spatial extent and coordinate reference.</p> <p>Key Steps<br>---------<br>1. Threshold Calculation:<br>   - Compute the 95th percentile of historical rainfall data ('globalMatrix') at each grid cell.<br>   - Extend the threshold grid by one row/column to match the interpolated domain.</p> <p>2. Event Detection:<br>   - Group .nc4 rainfall files within each year by typhoon event (gap ≥ 5 days separates events).<br>   - For each event, interpolate rainfall data onto a regular 0.1° grid (111:0.1:116, 25:-0.1:21).<br>   - Use a sliding window of 24 steps (12 hours) to detect extreme rainfall windows where ≥6 time steps exceed the threshold.</p> <p>3. Extreme Event Statistics:<br>   - For each pixel, count how many extreme rainfall events occurred across all years.<br>   - Store the maximum rainfall per detected extreme event for later statistical analysis.</p> <p>4. Return Period Calculation:<br>   - Compute annual frequency λ = number of events / number of years (excluding missing years).<br>   - For each pixel with ≥5 events:<br>     - Fit exceedances over the 95th percentile to a Generalized Pareto Distribution (GPD).<br>     - Calculate return period for each extreme value using:<br>       T = 1 / (λ * (1 - F(x)))<br>     where F(x) is the GPD cumulative distribution.</p> <p>5. Extract Key Metrics:<br>   - Maximum return period (MaxReturnPeriod.tif)<br>   - Mean rainfall intensity of all extreme events (MeanRainfallIntensityTC.tif)<br>   - 95th percentile rainfall of each pixel (Rainfall95Percentile.tif)<br>   - Return period corresponding to the 95th percentile rainfall (ReturnPeriod95PercentilesTC.tif)</p> <p>Output Files<br>------------<br>GeoTIFFs saved in the working directory (same spatial resolution and extent):</p> <p>| File Name                          | Description                                    |<br>|-----------------------------------|------------------------------------------------|<br>| MaxReturnPeriod.tif               | Max return period value per pixel              |<br>| MeanRainfallIntensityTC.tif       | Mean rainfall intensity of extreme events      |<br>| Rainfall95Percentile.tif          | 95th percentile of rainfall intensities        |<br>| ReturnPeriod95PercentilesTC.tif   | Return period associated with 95th percentile  |</p> <p>Notes<br>-----<br>- Years with missing data (e.g., 2004, 2019) are skipped.<br>- A detailed error log is recorded in error_log.txt.<br>- Requires MATLAB's Statistics and Machine Learning Toolbox for GPD fitting (fitdist).<br>- Can be extended to non-typhoon events or other regions with appropriate data substitution.</p>