FloodGuard

A Software-Based Flood Early-Warning and Reservoir Decision Support System
Focused on Short-Term Water-Level Prediction and Risk Classification

Team

Table of Contents

  1. Introduction
  2. Solution Architecture
  3. Software Designs
  4. Testing and Validation
  5. Conclusion
  6. Links

Introduction

Flooding in Sri Lanka, particularly in the Mahaweli River Basin, has resulted in significant social and economic damage. Existing reservoir management approaches primarily depend on monitoring current water levels. During extreme rainfall events, this reactive strategy often leads to delayed gate operations and downstream flooding.

FloodGuard proposes a predictive, software-based decision-support system that estimates short-term future reservoir water levels (30–60 minutes ahead) using rainfall trends and water-level rise rates. Instead of relying solely on real-time sensor readings, the system integrates:

This project is implemented entirely as a software system using simulated and historical data sources. It does not perform real-time dam control and does not interface with physical sensors. The architecture is designed to allow future integration of advanced machine learning models.

The objective is to demonstrate how predictive analytics can improve early flood warnings compared to traditional monitoring systems.

Solution Architecture

FloodGuard follows a modular and scalable system architecture consisting of four major components:

1. Simulation Engine

Generates realistic rainfall and reservoir water-level data based on:

2. Prediction Engine (Rule-Based)

Implements a deterministic trend-based extrapolation model:

This replaces machine learning in the current implementation while maintaining architectural compatibility for future ML integration.

3. Risk Classification Module

Classifies predicted water levels into:

Thresholds are configurable and reflect reservoir capacity levels.

4. Frontend Dashboard

Provides visualization of:

The system is implemented using a backend API architecture and a responsive web dashboard.

Software Designs

1. Data Simulation Model

Rainfall Simulation:

Reservoir Simulation:

2. Prediction Model

Let:

Future water level:

W(t + h) = W(t) + (ΔW × h) + (k × R_avg)

Where k is a rainfall influence coefficient.

This model ensures:

3. Backend API Design

The backend exposes endpoints for:

4. Scalability Considerations

The architecture allows:

Testing and Validation

The system is validated using simulated flood scenarios representing:

Validation Metrics:

Comparative testing is performed between:

Results demonstrate earlier warning generation compared to traditional reactive monitoring.

Conclusion

FloodGuard demonstrates a structured, scalable, and industry-aligned approach to flood early-warning decision support.

The system successfully:

Although machine learning integration is reserved for future enhancement, the current deterministic model ensures reliability, explainability, and academic validity.

The architecture is intentionally designed to allow seamless integration of advanced predictive models and real-time sensor networks in future research phases.