Bangladesh Geospatial Analysis Pipeline

40-year satellite-based water body detection, river migration, flood mapping, and wetland change analysis for Bangladesh using Google Earth Engine.

Study Area

Configurable scope via --scope:

Scope	Coverage
national (default)	All 8 divisions of Bangladesh
sylhet	Sylhet Division haor wetlands
<division>	Any specific division (e.g., dhaka, chittagong)

Key features tracked:

• 16 rivers: Padma, Jamuna, Meghna, Brahmaputra, Ganges, Teesta, Surma, Kushiyara, Kalni, Kangsha, Gorai, Arial Khan, Dharla, Karnaphuli, Sangu, Matamuhuri
• 19 wetlands/haors: Tanguar Haor (Ramsar), Hakaluki Haor, Sundarbans, Chalan Beel, and more
• Extreme flood years: 1987, 1988, 1998, 2004, 2007, 2017, 2020, 2022

Data Sources

Dataset	Period	Resolution
Landsat 5/7/8/9	1985–2025	30m
Sentinel-2	2015–2025	10m
JRC Global Surface Water	1984–2021	30m
SRTM DEM	—	30m

Pipeline Modules

Module	Description
config.py	Study area bounds, river/haor definitions, band mappings, thresholds
data_acquisition.py	GEE auth, Landsat/Sentinel harmonization, cloud masking, compositing
water_classification.py	NDWI, MNDWI, AWEI indices + Otsu thresholding + majority voting
river_analysis.py	River centerline extraction, channel migration, erosion rates
flood_analysis.py	Monsoon/dry mapping, seasonal inundation, extreme flood analysis
water_change.py	Water occurrence frequency, permanent/seasonal/rare, decade-wise change
haor_analysis.py	Haor delineation, area tracking, seasonal cycles, period comparison
visualization.py	Interactive maps (geemap/folium), matplotlib figures
export_utils.py	GeoTIFF, Shapefile, CSV export helpers
generate_report_maps.py	Static publication-quality maps (PNG, 300 DPI)
generate_pdf_report.py	Automated PDF findings report
tiling.py	Tile-based processing for large areas
run_pipeline.py	Main orchestrator with CLI modes

Quick Start

pip install -r requirements.txt
python run_pipeline.py --test                       # Verify GEE auth + 2020 dry/monsoon test
python run_pipeline.py --full                       # Full 40-year analysis (national)
python run_pipeline.py --scope sylhet --full        # Sylhet-only analysis

CLI Options

--scope SCOPE   Analysis scope: national | sylhet | <division name>
--test          Quick test (2020 dry vs monsoon)
--full          Full 40-year pipeline
--rivers        River erosion analysis only
--floods        Flood extent mapping only
--changes       Water change detection only
--haors         Haor/wetland analysis only

Outputs

All outputs are saved to outputs/:

Directory	Contents
outputs/rivers/	Per-river erosion rate CSVs, trend charts (PNG), interactive migration maps (HTML)
outputs/floods/	Flood time series CSV and chart
outputs/haors/	Per-haor area timeseries, trend charts, boundary maps, period comparisons
outputs/changes/	Water persistence stats, change maps, before/after comparisons
outputs/report_maps/	8 publication-quality static figures
outputs/	Master map, JRC overlay, multiyear comparison, findings report (HTML/MD/PDF)

Water Detection Method

1. Compute NDWI, MNDWI, AWEI for each composite
2. Apply Otsu auto-thresholding per scene per index
3. Majority voting: pixel = water if 2+ of 3 indices agree