VENTILADORFLOR/imerg_idf_shv: Added bootstrap CI and Kirpich scripts

This repository contains the complete Python codebase accompanying the paper "Benchmark-Referenced Bias Assessment of GPM IMERG Rainfall for Drainage Infrastructure Evaluation in Data-Scarce Coastal Cambodia". All analysis steps are fully reproducible from publicly available data sources.

Study area: Sihanoukville, Cambodia (10.540139°N, 103.656389°E) Satellite data: NASA GPM IMERG V07 Final Run, half-hourly, 2000–2024 DEM: JAXA ALOS AW3D30 V4.1, 30 m resolution Official IDF benchmark: Department of Hydrology, Cambodia, 1999–2019

The code covers five steps: (1) extraction of IMERG precipitation time series and ALOS DEM from Google Earth Engine; (2) computation of rolling annual maxima for five durations (30 min, 1 h, 2 h, 4 h, 24 h) and bootstrap 95% confidence intervals for correction factors; (3) Gumbel distribution fitting using L-moments and IDF curve derivation for the 2000–2019 period; (4) automated watershed delineation using pysheds, Kirpich time-of-concentration estimation, and production of the catchment figure; (5) generation of all paper tables from processed outputs.

Key result: Raw IMERG underestimates short-duration design rainfall by 63–74% relative to the official IDF benchmark. Duration-specific correction factors (αd) with bootstrap 95% confidence intervals are: 3.65 (95% CI: 3.02–4.34) for 30 min, 2.90 (95% CI: 2.61–3.32) for 1 h, 2.08 (95% CI: 1.74–2.52) for 2 h, 1.69 (95% CI: 1.36–2.05) for 4 h, and 1.24 (95% CI: 0.98–1.47) for 24 h. After correction, the 10-year 1-hour design intensity increases from 37.8 mm/h to 109.8 mm/h, raising the estimated peak discharge from 19.8 m³/s to 57.6 m³/s.

Dependencies: earthengine-api, pandas, numpy, scipy, geopandas, rasterio, pysheds, matplotlib, openpyxl. A full requirements.txt is included.