Multi-event flood inventory
Produced event labels and recurrence evidence for 2018, 2020, 2022 and 2024 from Sentinel-1 SAR.
Independent flood-resilience project · Lokoja, Nigeria
Lokoja Flood Risk and Resilience Planning
Where should flood-resilience work begin in Lokoja? Satellite-observed floods, mapped people and infrastructure help identify places for further investigation.
Project objective
Translate flood evidence into transparent planning priorities.
The project combined multi-event Sentinel-1 flood mapping, terrain and environmental predictors, spatially validated machine learning, population and building exposure, and infrastructure screening. Its purpose was to move beyond susceptibility mapping toward reproducible intervention planning.
17Model predictors
0.813Mean spatial PR-AUC
1,475Robust priority units
24Priority bridges and fords
My contribution
Designing the complete analytical and communication workflow.
I developed the data preparation, validation, modelling, exposure, infrastructure and WebGIS stages as a reproducible end-to-end project.
Spatial machine learning
Compared logistic regression, Random Forest and gradient boosting using spatial cross-validation and population weighting.
Independent validation
Assessed transfer to the 2024 event and later completed chronological validation across all events.
Exposure analysis
Integrated official constrained WorldPop, building footprints and susceptibility classes.
Infrastructure screening
Assessed mapped roads, critical facilities, bridges and fords using hazard context.
Intervention planning
Scored 600 m planning units and tested robustness under alternative weighting scenarios.
Spatial workflow
From satellite observations to intervention themes.
- 01Map events
Extract comparable flood labels from multi-date Sentinel-1 observations.
- 02Build predictors
Align terrain, drainage, water occurrence and land-cover predictors to a 30 m grid.
- 03Validate models
Use spatial folds, population weighting, ablation and chronological transfer tests.
- 04Assess exposure
Combine hazard with population, buildings and mapped infrastructure.
- 05Prioritise action
Rank planning units, assign intervention themes and publish the evidence through WebGIS.
Project gallery
Model evidence, validation and planning outputs.
Outputs and value
A complete open decision-support workflow.
The project produced analysis-ready rasters, validation tables, intervention units, publication graphics, documentation and an interactive dashboard.
- Continuous probability and five-class susceptibility rasters
- Spatial and chronological model-validation evidence
- Population, building and infrastructure exposure summaries
- Robust Critical and High intervention units
- Recommended intervention themes and screening shortlist
- Open interactive WebGIS dashboard and reproducible project documentation
Project attribution: Independent work by Samuel Bahago Gabriel using open geospatial data and open-source analytical components alongside ArcGIS Pro.
Continue exploring