Introduction
A Sri Lanka-focused research prototype connecting flood impact mapping, explainable landslide prediction, GN-level human impact assessment, and hazard-aware emergency response.
Back to Projects
Final Year Research · Disaster Risk · Case Study
Integrated multi-hazard risk assessment and emergency response
Role
Team Research Project
Core Stack
Python · GIS · Machine Learning
Research Components
Flood · Landslide · Response
Status
Active Research
Research Journey
From research question to response system.
A progress timeline for Project J26-IT-334. The project began in July 2025; the source summary records milestone status, but does not assign individual dates.
Project started
July 2025
2
Completed
milestones
5
In progress
milestones
3
Planned
milestones
Project milestones
Ordered by progress · undated stages remain undated
Research project begins
Project startKick-off for the integrated flood and landslide risk assessment and emergency response research.
Date not specified
Proposal & literature review
CompletedDefine the research problem, review related systems, and establish objectives and research questions.
Date not specified
Data acquisition
Partly completed · continuingGather and prepare NBRO, DMC, Census, OpenStreetMap, and historical disaster data.
Date not specified
Landslide data preprocessing
CompletedPrepare an event and control dataset with rainfall, soil, elevation, and slope attributes.
Date not specified
Model development & comparison
In progressCompare logistic regression, XGBoost, and LightGBM for landslide prediction.
Date not specified
Explainability layers
In progressDevelop SHAP and LIME explanations for landslide and impact decision support.
Date not specified
Inter-component interface contracts
In progressDefine versioned interfaces and structured payloads exchanged between the four components.
Date not specified
Prototype dashboards
Static prototype availableDemonstrate the Risk Atlas and integrated user experience with illustrative data.
Date not specified
Integration with live component outputs
PlannedConnect the prototype experience to outputs produced by the research components.
Date not specified
Evaluation, backtesting & usability
PlannedValidate against historical events and conduct planned usability evaluation.
Date not specified
Final report & viva
PlannedConsolidate findings and present the completed research.
Integrated Research System
Four components. One response chain.
The research connects hazard assessment to local impact priorities and response decisions through structured component interfaces.
Four-component architecture
Hazard input · 01
Flood impact assessment
Kalu Ganga · Kelani Ganga
Hazard input · 02
Landslide prediction
Nuwara Eliya District
Decision layer · 03
Human impact & decisions
Combines hazard severity with exposure, sensitivity, and capacity to prioritize local alerts.
GN-division prioritiesResponse layer · 04
Emergency response
Supports resource allocation and route planning that accounts for hazard-blocked roads.
Coordinated actionIntended information flow only. Live integration between components is still in development.
Component breakdown
Flood impact assessment
Kalu Ganga · Kelani Ganga
Flood severity assessment using rainfall, river-level, terrain, and historical flood information.
Rainfall & river levels
Hydrometeorological signals
Terrain context
Catchment and elevation layers
Flood history
Past events for context
Intended contribution
A severity view and map layer intended to inform downstream impact analysis.
Landslide prediction
Nuwara Eliya District
Compare Logistic Regression, XGBoost, and LightGBM using rainfall, soil, elevation, and slope attributes.
Model comparison
Logistic Regression · XGBoost · LightGBM
Environmental data
Rainfall · soil · elevation · slope
Interpretability
SHAP & LIME under development
Intended contribution
A landslide likelihood estimate with SHAP/LIME explanations under development.
Human impact & decisions
GN-division level
Bring hazard severity together with exposure, sensitivity, and capacity at GN-division level.
Hazard severity
Flood and landslide context
Community exposure
People and assets at risk
Vulnerability factors
Sensitivity · response capacity
Intended contribution
Prioritized local impact information for decision support.
Emergency response
Multi-district pilot
Consider response resources and routes while accounting for roads affected by hazards.
Resource coordination
Plan allocation under constraints
Route planning
Consider hazard-affected roads
Response support
Proposed coordinated actions
Intended contribution
A proposed response plan to support coordinated action.
These cards describe the research design; they do not represent live, operational hazard services.
Illustrative scenario
How the response chain is intended to work.
Inputs · illustrative only
Stage 01 of 04Illustrative rainfall and river-level information enters the flood assessment; terrain and rainfall attributes support landslide analysis.
RainfallRiver levelsTerrain attributes
This walkthrough describes a proposed workflow only. It does not show real-time weather, validated risk scores, or operational instructions.
Research Methods
Compare carefully. Explain clearly.
The landslide component combines structured data preparation, model comparison, and explainability. Performance results are not shown here because verified final results were not included in the project summary.
Three approaches · no scores published
Landslide model comparison
Logistic Regression, XGBoost, and LightGBM are being compared using the prepared landslide event/control data.
Logistic RegressionXGBoostLightGBM
Interpretability layer · in progress
Explainability
SHAP and LIME are being developed to help explain how environmental and human-impact factors influence model outputs.
SHAPLIME
Landslide component
Data preparation
The event/control dataset includes rainfall, soil, elevation, and slope attributes for model development.
RainfallSoilElevationSlope
Research Prototype
Explore the research prototypes.
Open the prototypes to see how the ideas are presented. Both demonstrate intended interfaces and data flow; they are not operational hazard systems.
Prototype interfaces use illustrative/mock data. They do not provide live hazard predictions, validated risk scores, or emergency instructions; live integration and evaluation remain planned research work.
Up Next in Portfolio · Campus Management
Clever Campus→
A modern campus management system that streamlines student, staff, schedule, and administrative operations.
JavaSpring BootMySQLReact