Data and code for CHIRPS rainfall-landslide case-crossover analysis in Colombia
Description
Reproducibility materials for the article “Assessing CHIRPS rainfall indicators for landslide monitoring in Colombia: a seasonally matched case-crossover study.” The archive contains the strict pre-event analysis used in the manuscript, implemented in Python with rainfall restricted to days preceding the reported landslide date, together with the MATLAB index-date-ending workflow retained for provenance and date-alignment sensitivity. It includes derived 1:5 matched case-crossover datasets, Python and MATLAB code, result tables, grouped-transfer and temporal stress-test outputs, publication figures, variable definitions, provenance notes, and SHA-256 manifests. Rainfall-extraction units are acquisition groupings rather than native CHIRPS pixels, gauges, or independent validation sites. Raw provider-controlled CHIRPS and COOLR source archives are not redistributed; public source locations are documented.
Files
Steps to reproduce
1. Download and extract A1_RAINFALL_LANDSLIDES_REPRODUCIBILITY.zip. 2. For the manuscript analysis, install Python 3.10 or later. From the extracted repository root, install the required packages with: python -m pip install -r PYTHON_STRICT_PRE_EVENT/requirements.txt 3. Run: python START_HERE_STRICT_PRE_EVENT.py 4. The script reproduces the strict pre-event conditional logistic models, 500 support-level bootstrap replicates, 100 repeated five-fold grouped-transfer tests, leave-one-support-out checks, relative-threshold analyses, temporal stress tests, trigger-label sensitivity, and 999 matched randomizations. A successful run prints: A1 STRICT PRE-EVENT ANALYSIS: PASS and writes reproduced tables to results/STRICT_PRE_EVENT_RESULTS_REPRODUCED. 5. To reproduce the date-alignment sensitivity retained for provenance, open MATLAB, set the extracted repository root as the Current Folder, and run: START_HERE_A1_REPRODUCE.m This MATLAB route uses the earlier convention in which rainfall windows end on the reported index date t; it is not the primary temporal convention used in the manuscript.
Institutions
- Universidad Nacional de ColombiaBogota D.C., Bogotá