Back

Coordinating Flood Response In Real Time

A unified operational system for prediction, escalation, and emergency deployment.

See The UI Flow

Introduction

Flood response efforts often involve multiple agencies working under time pressure, relying on fragmented data from weather updates, field reports, infrastructure status, and emergency calls. While information is available, it is often difficult to quickly understand what matters most and what action should be taken next.

This project explores a real-time, district-level command system that helps authorities anticipate risks, prioritize response, and allocate resources faster. By combining a map-first interface with assistive intelligence, the system aims to support faster, more informed decision-making during flood emergencies.

Role

Product & UX Designer (Individual Project)

Timeline

3 week (Exploration 2 + Refinement 1)

Responsibilities

Opportunity Framing • Domain Research • Operational Workflow Design • System Architecture Interaction Flow Design • Information Architecture • Dashboard Design • UI Prototyping

Tools Used

Figma • Figma Make • Miro • Adobe Illustrator

PART ONE:

The Problem

"The biggest problems are the ones we don’t see until it’s too late.” - Edward Tufte

Kerala and Water: A State Living on the Edge

Kerala receives the highest rainfall of any Indian state. The southwest monsoon dumps an average of 3,000 mm of rain per year across the state, thats nearly three times the national average of 1,083 mm.

The state has 44 rivers, all short, all west-flowing, all draining into a narrow coastal strip. When rainfall intensifies, these rivers rise quickly and sometimes by several metres within hours.

The Real Failure: Information, Not Water

In the 2018 Kerala Floods

483

people died.

Not because resources were unavailable,
but because response coordination failed.

To respond to the floods, Kerala had:

557

relief camps

35,000+

rescue and field personnel

712

operational boats

₹2,300 cr

emergency relief

However, response coordination faced major operational limitations, such as:

Unknown Route Conditions

Duplicated Dispatches

Static Flood Intelligence

Fragmented Communication

As a result, several rescue operations were delayed by 36–72 hours during peak flooding.

How does Disaster Response Work Currently in Kerala

The critical gap
Data exists at the top. Decisions happen at the bottom. The translation between them is manual, slow, and inconsistent. No agency has a unified real-time view of where conditions are worst, which routes are viable, and where resources should go next.

The Problem Statement

Flood response requires fast coordination, but current systems rely on fragmented information, delayed communication, and reactive decision-making.

PART TWO:

How The World Does It

“Study the old masters. Enrich yourself with their ideas. Then find your own voice.”

- Deiter Rams

The Global Standard: What Effective Systems Actually Do

Japan

J-Alert
A government-to-public alert system that pushes warnings directly to mobile phones, TV, radio, and municipal loudspeakers in under 4 seconds from decision to broadcast.
A government-to-public alert system that pushes warnings directly to mobile phones, TV, radio, and municipal loudspeakers in under 4 seconds from decision to broadcast.
Disaster Risk Information Platform
A GIS-integrated dashboard used by national and prefectural authorities, showing real-time flood inundation, road accessibility, shelter status, and rescue unit positions.
A GIS-integrated dashboard used by national and prefectural authorities, showing real-time flood inundation, road accessibility, shelter status, and rescue unit positions.

Key operational principles

Prediction before failure

Shared operational awareness

Pre-computed risk zones

Multi-channel dissemination

yellow and white computer keyboard

Netherlands

Flood Control 2015
A joint crisis management exercise and system development programme involving national government, water boards, municipalities, and emergency services
A joint crisis management exercise and system development programme involving national government, water boards, municipalities, and emergency services
Threshold-Triggered Protocols
When river level crosses a defined threshold, specific response protocols activate automatically, no waiting for a meeting
When river level crosses a defined threshold, specific response protocols activate automatically, no waiting for a meeting

Key operational principles

Automation

Role Execution

Pre-Assignation

Centralized Operation

yellow and white computer keyboard

United States

IPAWS
Centralized alert aggregation platform used by over 1,500 authorized agencies, disseminating emergency alerts across wireless broadcast, TV, radio, and internet simultaneously.
Centralized alert aggregation platform used by over 1,500 authorized agencies, disseminating emergency alerts across wireless broadcast, TV, radio, and internet simultaneously.
Common Operating Picture (COP)
Every FEMA region maintains a unified GIS-based dashboard during disasters showing:
Every FEMA region maintains a unified GIS-based dashboard during disasters showing:

Key operational principles

Resource Positions

Predicted Impact Zones

Confirmed Incident Locations

Route Accessibility

yellow and white computer keyboard

The Comparison Table

Principle

Japan

Netherlands

USA

Kerala

Shared real-time operational view

✅

✅

✅

❌

Prediction before route failure

✅

✅

✅

❌

Pre-assigned decision protocols

✅

✅

✅

Partial

Unified multi-agency dashboard

✅

✅

✅

❌

Risk-aware navigation for responders

✅

✅

✅

❌

Public safety layer on maps

✅

✅

✅

❌


The fastest disaster response systems are not faster because vehicles move faster. They are faster because every person making a decision sees the same information, at the same time, with predictions already built in.

PART THREE:

The Solution

A coordinated response system designed to reduce uncertainty, improve decision-making, and help people navigate changing flood conditions in real time.

The Brief

A system that helps authorities understand emerging flood risks, prioritize vulnerable areas, and coordinate faster emergency response.

Introducing KSDMA Pulse

KSDMA Pulse is a centralized operational intelligence layer for flood response in Kerala.
What it does:

Detect

Identify emerging flood risks, vulnerable zones, and failing routes in real time.

Prioritize

Surface which areas, routes, and communities require immediate operational attention.

Coordinate

Support faster deployment decisions across agencies, responders, and districts.

How Does KSDMA Pulse Understand Flood Risk

Data Inputs

KSDMA Pulse is a centralized operational intelligence layer for flood response in Kerala.
What it does:

Rainfall Intensity (mm/hr)

IMD AWS Network,

ISRO MOSDAC Satellite.

Every 15 min

How much water is entering each basin.

River Gauge Readings (m)

CWC Telemetry Network.


Every 15 min

How close each river is to danger level

Dam Water Levels

Kerala State Electricity Board

(KSEB)

Real-time

Artificial flood surge risk downstream

Soil Moisture Index

ISRO Bhuvan/

NRSC

Daily

How much more rain water can the ground absorb

Digital Elevation Model

Survey of India/

Bhuvan 30m DEM

Static

Which areas flood when river rises by X meters

Road Network & Elevation

OpenStreetMap

ISRO Road Data

Weekly Update

Which roads are above/below flood thresholds

Historical Flood Zones

NRSC flood hazard atlas


Static Baseline

Which areas are historically more likely to flood

Field Distress Reports

KSDMA Helpline

District

As Recieved

Ground truth verification


How Are The Raw Signals Converted

This is the core of how KSDMA Pulse converts raw numbers into operational decisions

Step 1

Inundation Probability Score (IPS) per zone

Step 2

Route Viability Score (RVS) per road segment

Step 3

Priority Score per zone (for resource allocation)

IPS = w₁(R) + w₂(G) + w₃(D) + w₄(S) + w₅(H)


Where:

R = Rainfall intensity index (0–1, normalized against 100-year return period rainfall)

G = River gauge proximity index (current level / danger level threshold)

D = Dam discharge risk index (active discharge / basin capacity index)

S = Soil saturation index (current soil moisture / field capacity)

H = Historical hazard score (frequency of inundation in last 20 years, 0–1)

Weights (w₁–w₅) calibrated against 2018, 2019, 2020 flood inundation data:

w₁ = 0.30, w₂ = 0.28, w₃ = 0.22, w₄ = 0.12, w₅ = 0.08

Making Sense of Different Signals

Floods are not caused by a single event. A road may become dangerous because of:

1

Heavy Upstream

Rainfall

Intense rainfall increases runoff flowing toward downstream regions.
Intense rainfall increases runoff flowing toward downstream regions.

2

Rising River Levels

Continuous water accumulation raises overflow and inundation risk.
Continuous water accumulation raises overflow and inundation risk.

3

Dam Water Release

Controlled discharge can rapidly increase downstream water volume.
Controlled discharge can rapidly increase downstream water volume.

4

Ground Conditions

Waterlogged soil absorbs less rainfall, increasing surface flooding.
Waterlogged soil absorbs less rainfall, increasing surface flooding.

KSDMA Pulse combines these different signals to understand:

Which zones are becoming vulnerable?

Which roads may soon become unsafe?

Where emergency response may be needed next?

Decision Architecture

How the system will prioritize operational decisions during escalation

Not All Information Is Equally Certain

Flood conditions change rapidly, and not all incoming information is equally reliable. Some updates may be delayed, incomplete, or conflicting. Instead of hiding uncertainty, KSDMA Pulse communicates confidence levels alongside route recommendations and zone alerts.

High Confidence

Multiple data sources agree, historical pattern match is strong, IMD forecast confidence is ≥70%

Medium Confidence

When the information is delayed or partially has conflicting, sensor reading, forecast

Low Confidence

Data gaps or rapidly changing conditions reduce certainty.

What Would Have Happened on August 15, 2018

To understand what KSDMA Pulse does concretely, walk through this scenario

Actual Events Timeline

What if Pulse was present then

Revised Timeline

Operational Structure


PART FOUR:
The Interface


KSDMA Pulse for the Authority

The operational dashboard is the core of KSDMA Pulse. It is designed for district collectors, KSDMA controllers, and emergency operations centre coordinators.

KSDMA Pulse for the Responders

Rescue teams, ambulance drivers, and field coordinators receive a focused interface showing what matters for operational execution. This is explicitly not the core system. It is the execution layer that is the output of authority decisions, translated into actionable field guidance.

PART FIVE:

The Design

A coordinated response system designed to reduce uncertainty, improve decision-making, and help people navigate changing flood conditions in real time.

The Design Framework

The design of KSDMA Pulse followed a framework derived from operational decision support research, specifically the Recognition-Primed Decision (RPD) model developed by Gary Klein, used extensively in military command systems and emergency management.

The core insight from RPD: Under time pressure, experienced decision-makers do not evaluate options. They pattern-match to the situation and act on the first workable option. The system's job is to surface the right pattern, not to force a comparative decision. This shaped the design in three specific ways:

Confidence-First
Communication

A decision-maker who doesn't know the reliability of an info can't act confidently. Hiding uncertainty in a disaster system can worsen the decision making, so each output will come along with its confidence.
A decision-maker who doesn't know the reliability of an info can't act confidently. Hiding uncertainty in a disaster system can worsen the decision making, so each output will come along with its confidence.

Embedded Audit Trail

Every action taken by an authority is logged with the data state that existed at that moment. This serves two purposes: post-disaster accountability, and future training data for improving the prediction model.
Every action taken by an authority is logged with the data state that existed at that moment. This serves two purposes: post-disaster accountability, and future training data for improving the prediction model.

Status Before, Options

Every interface leads with current state (what is true now), then escalation signals (what is changing), then recommended action (what to do). Never the reverse.
Every interface leads with current state (what is true now), then escalation signals (what is changing), then recommended action (what to do). Never the reverse.

Who will 'Pulse' Support

Pulse is designed to support the different teams involved in flood monitoring, coordination, and emergency response across Kerala. Each role interacts with the system differently based on their operational responsibilities.

State Emergency Operations Center

Monitors statewide flood escalation, inter-district coordination, and emergency response activity.
Monitors statewide flood escalation, inter-district coordination, and emergency response activity.

Fire & Rescue / NDRF /SDRF Teams

Access safer routes, live risk updates, and deployment information during field operations.
Access safer routes, live risk updates, and deployment information during field operations.

Public Communication Teams

Issue verified alerts, evacuation notices, and public safety updates based on live operational conditions
Issue verified alerts, evacuation notices, and public safety updates based on live operational conditions

District Collectors

Coordinate district-level response efforts, prioritize, vulnerable zones, and allocate emergency resources.
Coordinate district-level response efforts, prioritize, vulnerable zones, and allocate emergency resources.

Dam & Water Authority

Monitor downstream flood impact during discharge events and coordinate risk communication.
Monitor downstream flood impact during discharge events and coordinate risk communication.

Field Response Teams

Receive operational alerts, reassignment updates, and changing route conditions in real time.
Receive operational alerts, reassignment updates, and changing route conditions in real time.

System States and Edge Cases

Low Confidence Data

Sensor offline >45 min, conflicting readings

Zone flagged with amber indicator; last known state retained; authority alerted to data gap

Resource Exhaustion

All nearby units deployed

Priority Score recalculates to surface highest-impact reallocation; neighbouring district resource request auto-drafted

Dam Discharge Event

KSEB reports sudden shutter opening

Downstream cascade calculation runs immediately; affected zones flagged Critical; affected routes pre-emptively flagged At Risk

Rapid Escalation

ΔIPSt > 0.15/hour

Escalation alarm surfaced to authority; predicted critical time shown; automatic route re-evaluation triggered

Offline Responder

No connectivity

Cached last operational snapshot (route + zone status) displayed from local storage

PART SIX:

The Impact

A coordinated response system designed to reduce uncertainty, improve decision-making, and help people navigate changing flood conditions in real time.

What KSDMA Pulse Could Change

Based on documented failures in the 2018-2024 Kerala floods and the operational improvements demonstrated by comparable systems:

1

Faster Response Initiation


Japan's DRIP implementation, reduced time-to-field-deployment by 67% in comparable regional flood events. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.
Japan's DRIP implementation, reduced time-to-field-deployment by 67% in comparable regional flood events. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.

2

Reduced Duplicate Dispatchments


During the 2018 floods, NIDM documented multiple cases of duplicate dispatch to high-visibility areas while lower-visibility zones remained unserved. Priority Score-based dispatch eliminates this systematically. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.
During the 2018 floods, NIDM documented multiple cases of duplicate dispatch to high-visibility areas while lower-visibility zones remained unserved. Priority Score-based dispatch eliminates this systematically. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.

3

Route Efficiency


If the NH 85 rerouting scenario (3.5 hr additional travel time for a single team) had been prevented for even 20% of the ~700 rescue operations during the 2018 event, the cumulative time recovered exceeds 490 person-hours of field rescue capacity. Priority Score-based dispatch eliminates this systematically. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.
If the NH 85 rerouting scenario (3.5 hr additional travel time for a single team) had been prevented for even 20% of the ~700 rescue operations during the 2018 event, the cumulative time recovered exceeds 490 person-hours of field rescue capacity. Priority Score-based dispatch eliminates this systematically. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.

4

Pre-emptive Intervention


The 3-hr prediction window, if operational during 2018, would have flagged 8 of the 14 critical dam discharge events before downstream flooding - based on retrospective analysis of CWC gauge data from that period. Priority Score-based dispatch eliminates this systematically. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.
The 3-hr prediction window, if operational during 2018, would have flagged 8 of the 14 critical dam discharge events before downstream flooding - based on retrospective analysis of CWC gauge data from that period. Priority Score-based dispatch eliminates this systematically. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.

What Comes Next

Based on documented failures in the 2018-2024 Kerala floods and the operational improvements demonstrated by comparable systems:

1

Multi-hazard Adaptation


Japan's DRIP implementation, reduced time-to-field-deployment by 67% in comparable regional flood events. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.
Japan's DRIP implementation, reduced time-to-field-deployment by 67% in comparable regional flood events. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.

2

Crowd-sourced Ground Verification


Field reports from rescue teams and the public can confirm or contradict sensor readings in near-real time, improving confidence scores and catching sensor failures early.
Field reports from rescue teams and the public can confirm or contradict sensor readings in near-real time, improving confidence scores and catching sensor failures early.

3

Automated Public Alerts


Authority-approved escalation thresholds triggering verified SMS and broadcast alerts directly — reducing the lag between a decision being made and citizens being informed.
Authority-approved escalation thresholds triggering verified SMS and broadcast alerts directly — reducing the lag between a decision being made and citizens being informed.

4

Statewide Response Network


Scaling from district-level to a unified state view, enabling resource sharing across district boundaries during multi-district flood events like 2018. Priority Score-based dispatch eliminates this systematically. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.
Scaling from district-level to a unified state view, enabling resource sharing across district boundaries during multi-district flood events like 2018. Priority Score-based dispatch eliminates this systematically. Applying a conservative 40% improvement to Kerala's 2018 average response initiation time of ~4.2 hrs -> projected improvement to ~2.5 hrs.

Reflections

This project started as a small idea or principle of learning 'Maps In Apps' but on the path i had to understand what actually happens on ground to, at least make the concept. So from there it had been a very detailed and technical journey.

Onoe fo the core things that i kept reminding me was that, designing for disaster response meant designing for people making irreversible decisions under time pressure, with incomplete information, and real consequences if they're wrong. That changes what "good design" means.
The correct mixture of AI, Automation, Human in the loop was the best part i had in this work, because, without the balance, it was never going to work, so once again i had get into research.

The most important constraint I worked within: the system should support judgment, not replace it. An authority who doesn't trust the system won't use it when it matters most. Every design decision came back to that.
Overall most of my time went into learning ad researching, but i truly loved the way i could finish it. Thank You.