Project Title: Climate-Related Disaster Coordination - Digital

Health Resilience During Urban Floods

ITBM: B – Group: 6 – Topic: 2

Team Members:

1) Chemudupati Venkata Surya Sasidhar, 24030241064, [Link]@[Link]

2) Kavala Nirmal Swaroop, 24030241075, [Link]@[Link]

3) Khushboo Jangam, 24030241076, [Link]@[Link]

4) Keira Ruth Lobo, 24030241078, [Link]@[Link]

5) Meda Sesha Sai Varun, 24030241080, [Link]@[Link]

Domain (Healthcare / Sustainability / Supply Chain / E-commerce & Quick Commerce &
Food Delivery): Healthcare & Sustainability

Audience Focus (B2B / B2C / Both): Both (B2B - Government, NGOs, large hospitals; B2C -
citizens, volunteers)

News Article / Report : Sphere India (2023), Cyclone Michaung - Joint Rapid Needs
Assessment (JRNA) Report, Tamil Nadu, available at: [Link]
default/files/202312/TN_Cyclone%20Michaung_JRNA%[Link]
Abstract

Climate-induced disasters increasingly threaten urban resilience, with healthcare access

emerging as a critical point of failure. The 2023 Cyclone Michaung and subsequent floods in
Chennai highlighted how high-tech cities can face paralysis when essential medical
connectivity collapses. According to the Sphere India Joint Rapid Needs Assessment (2023),
over three-quarters of surveyed households in Chennai reported difficulties in reaching health
facilities. This gap between patient needs and provider capacity underscores the urgency of
coordinated, technology-enabled interventions.

This study proposes a Digital Disaster–Health Coordination Platform that integrates real-
time flood intelligence, hospital availability, fleet telemetry, and citizen-facing support
channels. The system design incorporates features such as flood-aware ambulance routing,
pre-triage of patients through symptom reporting, and a multi-stakeholder dashboard for
agencies and NGOs. Evaluation metrics—including dispatch time, hospital load balancing,
and citizen satisfaction—were defined to benchmark effectiveness in simulated and pilot
environments.

Findings highlight that resilience in disaster healthcare depends as much on information

logistics as on physical infrastructure. The platform offers actionable implications for
policymakers and practitioners: municipal authorities can embed it into disaster management
frameworks, NGOs can leverage it to extend outreach, and hospitals can use it to manage
patient inflows. Beyond Chennai, the approach demonstrates a scalable model for other
flood-prone metros in India, aligning with SDGs 3 (Health), 11 (Sustainable Cities), and 13
(Climate Action).

Keywords

Urban flooding; healthcare access; disaster coordination; AI-enabled logistics; Cyclone

Michaung; digital resilience; SDGs

1. Introduction

Urban flooding has become an increasingly frequent and destructive hazard in Indian
metropolitan regions, exposing critical dependencies between transport, communications, and
healthcare delivery. The December 2023 floods in Chennai, driven by Cyclone Michaung,
demonstrated how hyperlocal inundation can sever patient–provider connectivity even in
cities with comparatively advanced hospital capacity: residents in submerged suburbs were
cut off from essential care, ambulance routes became impassable, and ad-hoc triage and relief
camps struggled to meet latent demand. (Sphere India, 2023)

Several structural drivers make such outcomes likely to recur. First, climate change is
increasing the intensity and unpredictability of extreme rainfall events and coastal storms,
raising the probability of urban inundation and cascading failures across infrastructure
networks. The World Bank and allied analyses emphasize that urban populations -
particularly those in low-lying coastal metros - will face rising flood risk without significant
resilience investments. (World Bank, 2025). Second, urban development patterns and ageing
drainage/transport systems concentrate vulnerability in peri-urban and informal
neighbourhoods, where healthcare access is already marginal and evacuation options limited.
(Singh et al., 2023)

From a health systems perspective, disasters translate rapidly into unmet primary-care needs,
interrupted chronic-disease management, broken referral pathways, and overwhelmed
emergency departments. Post-event assessments of Cyclone Michaung report high
proportions of affected populations who could not reach nearby health facilities and relied on
temporary health camps - illustrating that hospital capacity alone is insufficient when
logistics and information flows falter. (Sphere India, 2023). The World Health Organization’s
Health Emergency and Disaster Risk Management (Health EDRM) framework foregrounds
this linkage: resilient health systems must integrate risk information, local surge capacity, and
community-level access solutions to reduce avoidable morbidity and mortality during
disasters. (World Health Organization, 2019)

This project positions the primary problem as a breakdown in hyperlocal patient–provider

connectivity and medical logistics during urban floods, with Chennai (Cyclone Michaung,
2023) used as the focal case. The problem is not merely flooded roads, but the intersection of
(a) absence of real-time situational information for citizens and first responders, (b)
disconnected ambulance and fleet coordination, and (c) lack of pre-triage and demand-
management tools that would allow hospitals to optimize limited capacity under constrained
access. A targeted technological and operational intervention could therefore mitigate
mortality and morbidity by restoring information flows, coordinating transport alternatives
(boat/raft, high-clearance ambulances), and enabling remote triage/telemedicine for non-
transportable patients. (Lokmic-Tomkins et al., 2023)
Key stakeholders include city health authorities (municipal health departments), hospital
networks (public and private), emergency services and ambulance providers, municipal
disaster management agencies, community organisations and resident welfare associations,
telemedicine providers, and technology platform operators (mapping, weather/flood sensors,
fleet telematics). Each stakeholder has distinct priorities: hospitals seek demand forecasts and
pre-triage; ambulances need real-time routing and flood-aware navigation; residents need
credible, localized guidance on accessible care; and civic agencies require situational
awareness to deploy rescue assets effectively. (District Disaster Management Authority,
Tiruvallur, 2024)

Scope and limitations: this report focuses on an integrated socio-technical platform that fuses
real-time hydrometeorological/flood-inundation data, hospital service-availability feeds, and
ambulance/alternative-vehicle fleet coordination, augmented by an LLM-driven citizen agent
for routing and pre-triage. The geography of interest is Chennai metropolitan and adjacent
peri-urban districts (Tiruvallur, Chengalpattu), assessed through the Cyclone Michaung
events and generalizable to other Indian coastal metros. The project deliberately excludes
large-scale physical infrastructure investment planning (e.g., drainage retrofits) but will
describe how digital coordination complements such investments. It also recognizes data-
availability constraints (real-time flood models, hospital bed APIs) and ethical/privacy trade-
offs when designing patient-facing agents. (Sphere India, 2023)

Roadmap - what follows in the report:

• Section 2 (Objectives & Research Questions): measurable objectives and 1–3 focused
research/solution questions that guide design and evaluation.

• Section 3 (Background & Related Work): literature on urban flood resilience, disaster
health systems, telemedicine in emergencies, and location-allocation models. (Singh et al.,
2023)

• Section 4 (Stakeholders & Use Cases): concrete personas and use-case tables mapping
pain points to desired outcomes.

• Section 5 (Solution Overview): high-level description of the integrated platform and how it
addresses the Chennai case pain points.

• Section 6 (System Design): architecture, key features, data and integration patterns, user
journeys, and risk mitigation.
• Section 7 (Evaluation Plan & Results): proposed pilot design, metrics (access time, first-
contact resolution, mortality reduction proxies), baseline and targets.

• Section 8 (Business Model & Financials): unit economics, cost/revenue assumptions, and
a 12–24-month projection for a pilot rollout.

• Sections 9–11: sustainability impacts (SDG mapping), limitations/future work, and a

concise conclusion with policy and practitioner recommendations.

In sum, the Chennai 2023 floods provide a stark example of resilient-systems failure at the
intersection of climate, urban form, and health service delivery. This study interrogates that
failure and proposes a pragmatic, testable platform design intended to restore hyperlocal
healthcare connectivity during urban inundation events - with explicit attention to operational
feasibility, privacy, and equity.

2. Objectives & Research Questions

The 2023 Chennai floods, exacerbated by Cyclone Michaung, demonstrated how climate-
induced urban inundation can trigger healthcare access collapse. Despite the presence of
advanced hospitals and emergency services, patients in submerged suburbs were unable to
reach facilities, and ambulance networks were paralyzed (Sphere India, 2023). Addressing
this challenge requires a coordinated, socio-technical solution that integrates flood-risk
intelligence with medical logistics and citizen-facing support.

Objectives

1. To design and propose an integrated disaster-health coordination platform that

fuses real-time flood data, hospital capacity information, and ambulance fleet
management.
2. To assess the feasibility of LLM-driven agents in providing citizens with safe
routes, remote triage support, and access to alternative medical logistics during flood
emergencies.
3. To evaluate the societal, sustainability, and ethical implications of deploying such
a platform in urban Indian contexts, particularly regarding inclusivity, privacy, and
resilience.
Research / Solution Questions

 RQ1: How can urban flood data, hospital service availability, and transport fleet status
be integrated into a unified system to preserve healthcare access during inundation
events?
 RQ2: What role can large language models (LLMs) play in enabling real-time
communication, safe routing, and pre-triage for patients and caregivers during
disasters?
 RQ3: What evaluation metrics and sustainability criteria should be applied to ensure
that the proposed solution improves resilience while safeguarding data privacy,
fairness, and inclusivity?

3. Background & Related Work

3.1 Urban Floods and Healthcare Disruptions

Urban flooding has emerged as a recurrent hazard in Indian metropolitan areas, exposing
systemic weaknesses in infrastructure and public health systems. A 2023 World Bank
analysis projects that climate-driven flooding will impact more than 700 million urban
residents globally by 2050, with South Asia among the most vulnerable regions due to rapid,
unplanned urbanization (World Bank, 2025). In Chennai, Cyclone Michaung inundated
suburbs and cut off access to essential healthcare, underlining how urban floods trigger
cascading failures: blocked transport routes, electricity outages, and disrupted supply chains
all converge to create healthcare access collapse (Sphere India, 2023) . Similar outcomes
were observed in the 2015 Chennai floods and the 2020 Hyderabad urban floods, which left
diabetic, dialysis, and cardiac patients stranded without timely intervention (Ravindran et al.,
2021).

The literature consistently shows that mortality during floods is not solely caused by
drowning or direct trauma but also by secondary health effects: interruption of chronic
disease management, maternal health services, and emergency referrals (Patel et al., 2020).
These effects disproportionately impact marginalized groups, including the elderly, informal
workers, and residents of peri-urban settlements where healthcare infrastructure is sparse.
3.2 Disaster Health Logistics and Coordination Gaps

Health systems during disasters depend not only on hospital capacity but on the logistics of
connecting patients to providers. A recent WHO Health Emergency and Disaster Risk
Management (Health EDRM) report emphasizes that surge capacity must be
complemented by real-time situational awareness and access management (WHO, 2019).
In practice, this means that even when hospitals are physically functional, patients cannot
reach them without reliable transport and coordinated routing.

Studies on disaster logistics highlight three critical bottlenecks:

1. Transport paralysis - conventional ambulance fleets are unable to navigate

submerged roads (Sphere India, 2023).
2. Information gaps - citizens lack visibility into which hospitals remain accessible and
which services are functional.
3. Decentralization failures - fragmented coordination between municipal authorities,
hospitals, and fleet operators delays resource deployment (Gupta and Dey, 2022).

Global cases reinforce these findings. During Hurricane Harvey (Houston, 2017),
coordination failures left dialysis and critical-care patients stranded; during the 2022 Pakistan
floods, tens of thousands were unable to access routine immunizations and maternal health
services due to logistics paralysis (UNICEF, 2022). These parallels show that Chennai’s
crisis is part of a wider, repeating pattern of “healthcare isolation during floods.”

3.3 Digital Health and Telemedicine in Disaster Contexts

Digital health tools have increasingly been positioned as solutions for resilience. During
COVID-19, telemedicine platforms in India expanded rapidly, with usage increasing fivefold
in metropolitan cities (EY - IPA, 2021). This demonstrated that remote triage and digital
consultations can maintain care continuity when mobility is constrained. In disaster
contexts, telemedicine has been applied to cyclone shelters in Odisha and flood-prone areas
of Assam, providing remote prescriptions and initial triage (Lokmic-Tomkins et al., 2023).

However, literature also cautions that telemedicine alone cannot solve disaster health gaps:
connectivity disruptions, lack of devices, and low digital literacy limit uptake in vulnerable
communities (Khatri et al., 2022). Moreover, without integration into logistics - e.g., linking
triage with ambulance dispatch or medication delivery - telemedicine risks being a parallel,
under-utilized system rather than a core disaster-health mechanism.
3.4 AI, Data Integration, and Disaster Platforms

Emerging work on AI and disaster management emphasizes the potential of large language
models (LLMs) and agentic AI to improve information access, situational awareness, and
coordination. A 2024 MIT–IBM report notes that LLMs can act as “cognitive orchestrators” -
interpreting citizen queries, integrating sensor data, and triggering automated workflows
(MIT–IBM, 2024). Applied to urban floods, such agents could route citizens to safe medical
facilities, trigger alternative transport requests (e.g., boats, high-clearance vehicles), and filter
misinformation.

Research on location–allocation models for disaster health demonstrates that integrating

flood maps with hospital networks optimizes patient allocation and reduces preventable
delays (Gulzari, 2021). Similarly, AI-enabled flood-inundation forecasting systems
deployed in Jakarta and Bangkok show the feasibility of real-time data integration into urban
decision-making (WMO, 2022).

Yet, the ethical literature warns of risks: data privacy breaches (especially if medical and
geolocation data are combined), algorithmic bias in triage prioritization, and the digital divide
limiting equitable access (Carter et al., 2023). Thus, any solution must incorporate privacy-
by-design and fairness checks, as well as offline fallback mechanisms.

3.5 Gaps Identified

Synthesizing prior work, four gaps emerge:

1. Lack of integration: Disaster logistics studies and telemedicine pilots often operate
in silos, without an integrated “platform” linking flood intelligence, hospital status,
and patient communication.
2. Hyperlocal focus missing: Existing models emphasize macro-level disaster relief
(e.g., state/national coordination) but fail at neighborhood-level patient–provider
connectivity.
3. Limited AI deployment: While AI has been studied for flood forecasting and
demand allocation, few real-world implementations demonstrate LLM-driven,
citizen-facing disaster health coordination agents.
4. Ethics and inclusivity risks: Literature highlights digital divide and privacy
concerns, but frameworks for mitigating these in Indian disaster contexts remain
underdeveloped.
3.6 Relevance to the Chennai Case

The Chennai floods represent a convergence of these gaps. Hospitals remained operational,
but connectivity failures created an artificial healthcare collapse. Existing relief camps and
telemedicine pilots were reactive and fragmented, lacking integration with transport logistics.
By proposing a platform that fuses real-time flood data, hospital feeds, and ambulance
routing, with an LLM-powered citizen interface, this project directly addresses the
identified research gaps and situates Chennai as both a case study and a testbed for scalable
solutions in other flood-prone metros.

4. Stakeholders & Use Cases

Effective disaster-health coordination requires identifying the full ecosystem of actors

impacted by urban floods and their specific needs. In the Chennai Cyclone Michaung floods
(2023), both institutional and community stakeholders experienced barriers that compounded
healthcare access failures. Stakeholders can be broadly divided into business/government
entities (B2B) and citizens/communities (B2C), with the platform designed to serve both
simultaneously.

4.1 Target Stakeholders

 Government agencies: Municipal disaster management authorities, state health

departments, and local civic bodies responsible for emergency planning and resource
deployment.
 Healthcare providers: Public hospitals, private clinics, and specialty centres
(dialysis, maternal care, cardiac care) that faced sudden surges and lacked visibility on
patient inflow.
 Emergency transport services: Ambulance operators (government, private, NGOs),
as well as boat operators and high-clearance vehicles deployed ad hoc.
 Community organisations: Resident welfare associations (RWAs), local NGOs,
volunteer networks supporting evacuation, relief, and telemedicine access.
 Citizens/consumers: Patients, caregivers, and vulnerable populations (elderly,
chronically ill, pregnant women) stranded without safe routes or real-time
information.
 Technology/data providers: Mapping services, flood-sensor networks, telecom
operators, and AI solution developers enabling integration.
4.2 Stakeholder Needs Table

Stakeholder Current Pain Point Desired Outcome Priority

Government Disaster Lack of real-time Integrated dashboard High
Agencies situational awareness; combining flood data,
fragmented coordination hospital capacity, and fleet
with hospitals/fleets. status for rapid decisions.
Public & Private Unpredictable patient Real-time demand High
Hospitals inflow; inability to triage forecasts, pre-triage via
remotely; beds patient symptom reporting,
under/overutilized. and optimized referrals.
Ambulance/Fleet Blocked roads; no Flood-aware routing with High
Operators alternative routing live traffic/flood overlays;
information; resource dispatch prioritisation for
wastage on failed critical cases.
attempts
Community Manual coordination; Access to platform for Medium
Organisations/NGOs limited reach; delayed alerts, coordination with
rescue/medical support civic agencies, and on-
for vulnerable groups. demand telemedicine
support.
Citizens/Patients Lack of reliable Personalized guidance: Very
information; stranded nearest functional clinic, High
without access to safe safe route, or alternative
transport or healthcare. (boat/telemedicine).
Technology/Data Data fragmentation; no Clear integration standards Medium
Providers structured mechanism for (APIs) and responsible data
deployment in crises. governance framework.

4.3 Concrete Use Cases

1. Patient in a flooded suburb (B2C): A diabetic patient in North Chennai requires

insulin but cannot leave her area due to waist-high flooding. Using the citizen app
interface, she reports symptoms; the system identifies a nearby accessible primary-
 care centre, arranges delivery via a boat-ambulance, and provides interim
teleconsultation.
2. Ambulance routing (B2B/B2C): An ambulance operator in central Chennai receives
a cardiac emergency request. The platform integrates real-time flood maps and
suggests an alternate elevated road, preventing a failed route and reducing response
time by 40%.
3. Hospital demand management (B2B): A tertiary hospital in southern Chennai sees
sudden inflows during the flood. The dashboard shows nearby clinics that still have
available capacity; non-critical cases are redirected, freeing emergency beds for acute
patients.

5. Solution Overview

The Chennai floods highlighted that healthcare collapse during disasters is not only due to
damaged infrastructure but primarily due to coordination gaps between patients, hospitals,
transport services, and civic agencies. The proposed solution is a Digital Disaster–Health
Coordination Platform that integrates real-time flood intelligence, hospital availability,
ambulance/alternative fleet routing, and citizen-facing support. At its core, the platform
is designed to restore hyperlocal patient–provider connectivity during floods while respecting
privacy, inclusivity, and operational feasibility.

The solution avoids technology/vendor lock-in by focusing on capabilities and workflows:

real-time data fusion, intelligent triage, and decentralized routing. The approach is modular,
allowing integration into existing municipal systems or NGO-operated platforms, and is
scalable to other flood-prone metros.

Pain Point (from Capability Workflow/Automation Expected Outcome

news) Required
Residents cut Flood-aware Integrate flood-inundation Patients reach
off from patient maps with fleet data; citizen nearest functional
hospitals due to navigation and queries trigger routing for clinic/hospital
submerged alternative ambulances/boats/high- safely, reducing
suburbs (Sphere transport clearance vehicles. preventable
India, 2023). coordination. morbidity and
mortality.
Emergency Fleet management Real-time rerouting based Ambulances and
 services unable with adaptive on flood sensors and relief vehicles avoid
to navigate routing. crowdsourced updates; failed routes,
inundated areas. dispatch prioritization based reducing wasted
on patient severity. time and increasing
response success.
Hospitals Pre-triage and Citizen app collects Balanced patient
overwhelmed by demand symptoms; AI agent load across
unpredictable management. provides initial triage and facilities;
inflows and lack recommends facility emergency capacity
of triage. allocation; hospitals see reserved for acute
inflow forecasts. cases.
Citizens lacking LLM-driven Citizen asks for nearest safe Reduced panic,
trustworthy, citizen facility; system integrates misinformation, and
real-time health engagement and data and responds with delays; inclusivity
information multi-channel clear, accessible instructions via low-bandwidth
during floods. communication access (voice/SMS)
(SMS, app, IVR).

Fragmented Centralized API-based integration of Improved

coordination disaster-health hospital capacity, flood situational
between dashboard with sensors, fleet telemetry; awareness, faster
government decentralized data dashboards for civic inter-agency
agencies, NGOs, inputs. agencies and NGOs. decisions,
and hospitals. coordinated relief
deployment.

5.2 Key Characteristics of the Solution

 Integrated: Combines flood, hospital, and fleet data in one ecosystem.

 Citizen-facing: Provides actionable, hyperlocal guidance across multiple channels.
 Modular & Scalable: Can be deployed at city level and replicated across coastal
metros.
 Inclusive: Offers low-bandwidth modes (SMS/IVR) for digitally excluded
populations.
 Ethically aligned: Incorporates privacy-by-design, consent for symptom reporting,
and auditability for AI decisions.
6. System Design
6.1 Architecture Overview:

System Architecture

6.2 Key Features

1. Flood-Aware Routing & Safe-Corridor Generation

o User story: As a stranded resident, I want to receive a safe route to a
functional clinic so I can reach care without taking flooded roads.
o Acceptance criteria: route shown within 60s; alternate transport suggested if
route impassable; route validated against latest flood layer and 2 recent
telemetry updates.
o Risk / Mitigation: stale flood data → require timestamped layers and
confidence metric; fallback to “nearest accessible assembly point” if uncertain.
 2. Real-time Hospital Capacity & Pre-triage
o User story: As hospital staff, I want pre-triage reports and inflow forecasts so I
can allocate resources before patients arrive.
o Acceptance criteria: symptom report accepted → triage level assigned
(urgent/non-urgent) with 80% agreement vs clinical nurse triage in pilot;
inflow forecast error <20% in 6-hour horizon.
o Risk / Mitigation: incorrect triage → require clear “non-diagnostic”
messaging, human-in-loop escalation, and conservative dispatch rules for
critical cases.
3. Fleet Orchestration & Dispatch
o User story: As fleet operator, I want prioritised dispatch instructions and
flood-aware routes so I avoid failed attempts.
o Acceptance criteria: dispatch suggestions provide ETA and obstacle
probability; aborted runs reduce over time after learning.
o Risk / Mitigation: operator non-compliance → integrate with incentives and
SOPs; provide offline maps for GPS loss.
4. Multi-channel Citizen Agent (LLM constrained)
o User story: As a low-bandwidth user, I want to call or SMS and get simple,
actionable guidance (route / teleconsult) in my language.
o Acceptance criteria: responses under 2 messages for SMS; calls provide IVR
menu and live-agent escalation; language support for Tamil + English.
o Risk / Mitigation: hallucination risk → use retrieval-augmented architecture
with hard constraints, canned responses for safety, and "source needed" on
uncertain claims.
5. Coordination Dashboard & Alerts for Authorities
o User story: As municipal control room officer, I want a consolidated
situational view to deploy assets effectively.
o Acceptance criteria: show heatmap of requests, available vehicles, hospital
capacity; enable assignment within 3 clicks; alerts for clusters above
threshold.
o Risk / Mitigation: information overload → provide filters and role-based
views.

6.3 Data & Integrations (sources, APIs, and responsible integration)

Primary data sources & integration approach

 Flood models & sensors: ingest via NOAA/IMD feeds, local sensor APIs, and
crowdsourcing endpoint. Use timestamped GeoJSON tiles and WMS/WFS layers.
Validate with redundancy (satellite vs gauge vs crowdsourced reports).

 Hospital systems: integrate with hospital management systems or provide a

lightweight CSV/API portal for small clinics to update status. Where available, align
with NDHM / Ayushman Bharat (ABDM) standards for health data exchange and
health ID mapping. Consent required for PHI access. (National Digital Health
Mission, 2020)
 Fleet telematics: ingest GPS + status via secure REST/WebSocket APIs or SMS
gateway for low-tech operators; normalize to unified telemetry schema.
 Telemedicine / clinician workflows: integrate with telemedicine platforms per India
Telemedicine Practice Guidelines (2020) for clinician registration, identification, and
prescription rules. (Dinakaran et al., 2021)

Responsible integration practices

 Minimise PII/PHI: only collect minimum data needed (location, triage level), avoid
storing full health records unless explicit informed consent.
 Consent & purpose limitation: implement explicit consent flows; log consent; map
retention to DPDP Act and NDHM guidance. (The Digital Personal Data Protection
Act, 2023) (NDHM,2020)
 APIs & authentication: OAuth2 / mTLS for partner APIs; role-based access control
(RBAC) for dashboards; strict audit logging.
 Data sovereignty & storage: store sensitive data in approved locations per
government policy; policy to delete or de-identify after retention period.

6.4 User Journeys (primary flows)

1. Onboarding (citizen; low friction)

o Open app / send SMS keyword → system requests minimal location consent
and language preference.
 o User selects “I need medical help” → quick symptom checklist (3–6 Qs) →
system responds with pre-triage and recommended action
(route/teleconsult/dispatch).
o If dispatch needed, user confirms; consent recorded; dispatch triggered with
ETA SMS.
2. Emergency Task Execution (fleet + hospital coordination)
o Citizen request arrives → decision engine scores urgency → nearest available
vehicle receives dispatch with flood-aware route → hospital alerted with pre-
triage packet and ETA.
o If vehicle cannot reach, system suggests alternate pickup point or boat
operator; citizen receives instructions (map + voice).
3. Exception Handling (connectivity/failsafe)
o If user connectivity lost mid-flow → system sends SMS with meeting point
coordinates and simple directions; nearby volunteers / NGOs alerted.
o If triage uncertain, escalate to human clinician (teleconsult) within SLA (e.g.,
15 minutes) or route to nearest open facility as conservative default.

6.5 Risk, Safety & Compliance

Privacy & Data Protection

 Comply with India’s Digital Personal Data Protection Act (DPDP) 2023: lawful
basis for processing, purpose limitation, user consent, data minimisation, and breach
notification. Maintain consent logs and data-subject access processes. (DPDPA, 2023)

Health Data Standards

 Align to NDHM / ABDM interoperability and consent frameworks for health records
where PHI/health IDs are used; restrict linkage to health IDs to explicit use cases with
patient consent. (NDHM, 2020)

Clinical & Telemedicine Governance

 Follow Telemedicine Practice Guidelines (2020) for clinician identification,

permissible prescriptions, and documentation. Maintain human-in-loop policy for all
critical triage decisions. (Dinakaran et al., 2021)

Explainability & LLM Safety

  LLM outputs limited to structured intents and templated responses; require
provenance tags (“data timestamp, source”) and a fallback “human clinician”
escalation. Maintain model decision logs and expose simple rationales (e.g., “route
recommended because water depth < X and vehicle high-clearance available”) for
audit.

Fairness & Inclusion

 Provide multi-channel access (SMS/IVR) and language support (Tamil + English).

Monitor usage and outcomes by geography and demographic proxies to detect
disparities. Implement offline community-agent workflows (NGO volunteers) where
digital access is constrained.

Security

 TLS for all transport; field-level encryption for sensitive fields; regular penetration
tests; hardened key management and least privileged RBAC. Implement incident
response and breach notification timelines aligned with DPDP.

Regulatory & Operational Alignment

 Integrate SOPs for ambulance services and emergency contact numbers; map dispatch
rules to local emergency response frameworks. Keep chain-of-custody logs for
decisions and routes to support post-incident review and liability mitigation. (National
Health Mission Odisha, 2020)

Ethical considerations

 Avoid automated life-and-death autonomous decisions; ensure transparency about

limitations; design for opt-out and human oversight. Provide explicit consent
language for location and health data and offer data deletion mechanisms.

7. Evaluation Plan & Results

Evaluating the proposed disaster–health coordination platform requires a multi-layered

approach, combining simulations, pilot deployments, and user testing. The goal is to
establish whether the solution improves healthcare access and coordination during urban
flood events without introducing unacceptable risks to privacy, safety, or equity.
7.1 Evaluation Methods

1. Simulations and Digital Twins

o Flood scenarios will be simulated using historical data from Chennai (Cyclone
Michaung, 2023; 2015 floods) and integrated with synthetic hospital/fleet
datasets.
o Key metrics such as patient travel time, successful dispatch rates, and load
balancing across hospitals will be observed under controlled conditions.
2. Pilot Deployment
o A small-scale pilot can be run in two flood-prone wards of Chennai, involving
2–3 hospitals, 10 ambulances, and ~1,000 citizen users.
o Data collected will include real-time dispatch logs, patient feedback, and
hospital utilisation.
3. User Testing (Citizen Interface)
o Usability tests with ~50–100 community participants (elderly, chronic
patients, volunteers).
o Evaluation criteria: comprehension of instructions, trust in the system,
inclusivity (SMS/IVR effectiveness).
4. Expert Validation (Clinicians & Authorities)
o Pre-triage algorithm outputs compared with human nurse/doctor triage across
200 sample cases.
o Dashboard usability reviewed with municipal disaster officers and hospital
administrators.

7.2 Metrics & Targets

Metric Definition Baseline Target Result

(if tested)
Turnaround Time from citizen Current floods: ≤20 mins in 80% TBV
Time request → 45–60 mins avg of cases. (pilot
(Dispatch) ambulance/boat (Sphere India, required).
dispatch confirmed. 2023).
First- % of citizen requests Current: <50% ≥85% success. TBV
Contact resolved without (delays, failed
Resolution repeat attempt routes).
(routing failure / info
 gap)
Routing % of routes suggested Current: ~60% ≥90% validated TBV
Accuracy that are usable (not (manual routing against
blocked). by operators). telemetry/flood
data.
Hospital % of hospitals Current: 2–3 ≥75% balanced TBV
Load operating within 80– tertiary utilisation.
Balancing 90% of surge capacity hospitals
(avoiding overload or overloaded;
underuse). others
underused.
Pre-triage Agreement rate Baseline: none ≥80% agreement. TBV
Agreement between AI triage (manual only).
output and nurse
triage (for urgency
categorisation).
Citizen User survey score Current: ad hoc ≥4.0/5 across TBV
Satisfaction (trust, clarity, systems, low diverse
usefulness) on 5-point trust (~2.5/5). demographics.
Likert scale.
Carbon Reduction in wasted Current: ~20– ≤10% wasted trips. TBV
Efficiency ambulance kilometres 30% trips
(lower emissions, wasted due to
resource savings). failed routing.
Compliance % of data flows with Current: 100% adherence to TBV
Adherence logged consent, fragmented, DPDP/NDHM
encryption, and audit non-digitized Telemedicine
trail. processes. guidelines.

7.3 Expected Outcomes

If the evaluation targets are met, the platform will demonstrate:

 Operational impact: Reduced patient isolation, faster dispatches, balanced hospital

utilisation.
  Societal value: Lower mortality/morbidity during floods, improved trust in public
health systems, better inclusion of vulnerable populations.
 Sustainability impact: Reduced fuel wastage and carbon emissions from failed
dispatches, aligning with SDG 3 (health), SDG 11 (resilient cities), and SDG 13
(climate action).
 Ethical robustness: Verified compliance with privacy, consent, and fairness
standards in disaster contexts.

8. Business Model & Financials

8.1 Revenue Model

The platform follows a hybrid B2B + B2G model, with sustainability embedded:

1. Government contracts (B2G): Municipal/state disaster management agencies pay

annual subscription/license fees to deploy the platform for urban flood resilience.
2. Healthcare partnerships (B2B): Public and private hospitals subscribe for access to
dashboards, demand forecasting, and triage integration.
3. NGO & CSR-funded access (B2B/G): NGOs and corporate CSR programmes
sponsor citizen access (SMS/IVR costs, awareness campaigns) in vulnerable areas.
4. Ancillary services: Value-added integrations such as telemedicine consultations, fleet
telematics, and analytics offered at premium pricing.

This model balances public-good funding (government/CSR) with enterprise-grade

partnerships (hospitals, telemedicine providers), ensuring affordability for citizens (no
direct charges).

8.2 Unit Economics (Assumptions)

 Technology costs: Cloud hosting + API usage ≈ ₹8–10 lakhs/month at pilot scale.
 Team costs: 8–10 FTEs (data engineers, AI/ML, compliance, operations) ≈ ₹12
lakhs/month.
 Integration costs: Initial setup with hospitals/NGOs ≈ ₹25 lakhs (one-time).
 Revenue potential:
o Govt/state contract: ₹1.5–2 crores annually per city.
o Hospital partnerships: ₹5–10 lakhs annually per hospital.
o NGO/CSR contributions: ₹30–50 lakhs annually (regional sponsorship).
Estimated break-even point: within 18–24 months if 1 city + 10 hospitals + 2 NGO partners
onboard.

Cost/Revenue Assumptions Monthly Annual Notes

Item (In Rs. (In Rs.
Lakhs) Crores)
Cloud & Infra Hosting, APIs, storage, 8 0.96 Scales with
Costs scaling at 10,000 users user growth.
Staff & Operations 8–10 FTEs (tech + ops + 12 1.44 Stable; may
compliance) expand post 18
months.
Integration/Support Hospital onboarding, 5 0.6 Higher in Year
NGO training, call-center 1, drops after
fallback integration
phase.
Awareness & Citizen engagement 2 0.24 May be
Outreach campaigns, IVR/SMS NGO/CSR-
access subsidized.
Total Monthly — 27 3.24
Costs
Govt/City 1 city govt contract (₹2 16.7 2.00 Anchor revenue
Subscription crore/year) stream
Hospital 10 hospitals × ₹7 5.8 0.70 Expands as
Partnerships lakhs/year hospital
network scales.
NGO/CSR Regional CSR + NGO 3.3 0.40 Ensures
Contributions grants (₹40 lakhs/year) inclusivity for
vulnerable
groups.
Ancillary Services Premium 2.1 0.025 Optional but
telemedicine/analytics scalable.
(~₹25 lakhs/year)
Total Monthly - 27.9 3.35 Surpasses cost
Revenue by Year 2.
Projection Summary:

 Year 1 (Pilot): Costs > revenue; CSR/NGO funds needed to cover gap.
 Year 2 (Expansion): Costs stable (~₹3.2 cr); revenues rise to ~₹3.3 cr; breakeven
achieved.
 Long-term: Each additional city contract or 5+ hospitals generate margin (~ ₹50–80
lakhs/year surplus).

8.4 Financial Sustainability & Risk Considerations

 Risks: Govt budget delays, hospital reluctance, CSR variability.

 Mitigations: Start with pilot partnerships, modular pricing (basic dashboard vs full
integration), and explore donor/UN funding for first 18 months.
 Sustainability Alignment: Revenue reinvested into low-bandwidth citizen channels
(SMS/IVR) and inclusive services, ensuring no direct citizen charges.

9. Sustainability & Societal Impact

The proposed disaster–health coordination platform is designed not only as a technical
solution but as a sustainability-aligned intervention that contributes to multiple global
development goals. By ensuring continuity of healthcare access during climate-induced urban
floods, the system advances public health, resilient infrastructure, climate action, and
inclusive communities.

9.1 SDG Mapping

SDG Alignment / Contribution

SDG 3: Good Health Ensures uninterrupted access to emergency and chronic care
and Well-Being during floods; enables pre-triage, telemedicine, and reduced
preventable mortality.

SDG 9: Industry, Builds resilient digital infrastructure for disaster response;

Innovation, and integrates AI, geospatial analytics, and health system APIs into
Infrastructure a scalable innovation ecosystem.

SDG 11: Sustainable Strengthens urban resilience by connecting citizens, hospitals,

Cities and Communities and authorities; fosters safer, more inclusive disaster
management systems in vulnerable communities.

SDG 13: Climate Action Enhances adaptive capacity to climate-driven extreme events;
reduces emissions via optimized fleet routing and fewer failed
dispatches.

SDG 17: Partnerships Encourages multi-stakeholder collaboration between

for the Goals governments, hospitals, NGOs, and technology providers.

9.2 Environmental Impact

 Positive contributions:
o Optimized routing reduces fuel wastage and emissions, with fewer failed
ambulance trips (potentially lowering carbon footprint by 15–20% compared
to status quo).
o Encourages sustainable use of alternative transport modes (boats,
community vehicles) during floods.
 Risks/mitigations:
o Increased cloud hosting has associated carbon costs; mitigated by using green
cloud providers and carbon-offset programmes.

9.3 Ethical Impact

 Privacy and Consent: Patient data and location sharing raise ethical risks. The
platform adheres to India’s DPDP Act (2023) and NDHM guidelines, with strict
consent flows, audit logs, and deletion policies.
 Explainability: LLM outputs are constrained to rule-based responses with human
oversight for critical triage; logs ensure transparency.
 Fairness: The system avoids bias by ensuring multi-channel access (app, SMS,
IVR), multilingual support (Tamil + English), and inclusive design for digitally
marginalized groups.
 Accountability: All automated decisions maintain human-in-loop validation
pathways, reducing ethical concerns around autonomous decision-making in life-and-
death contexts.

9.4 Inclusion & Social Equity

  Equity focus:
o Prioritizes vulnerable groups disproportionately affected during floods
(elderly, chronically ill, pregnant women).
o NGOs and CSR programmes subsidize SMS/IVR channels, ensuring zero cost
for citizens.
 Accessibility:
o Designed for low-literacy and low-bandwidth environments through IVR
voice prompts and SMS-based interactions.
 Community empowerment:
o Resident welfare associations (RWAs) and volunteer groups receive real-time
dashboards, enhancing local self-reliance.

9.5 Anticipated Societal Benefits

 Health resilience: Reduced mortality and morbidity during floods through faster
access to care.
 Trust building: Improved transparency and efficiency of civic response increases
trust in public health institutions.
 Scalability: Lessons from Chennai can be replicated across India’s other flood-prone
metros (Mumbai, Kolkata, Hyderabad), creating nationwide impact.

10. Limitations & Future Work

Despite its promise, the proposed disaster–health coordination platform has several
limitations that must be acknowledged. These constraints stem from technical maturity,
operational dependencies, regulatory contexts, and adoption dynamics. Recognizing
them helps ensure realistic expectations and guides the roadmap for future improvements.

10.1 Current Limitations

1. Data Availability and Accuracy

o Real-time flood data (depth, spread, velocity) in Indian metros is often
fragmented or delayed.
o Hospital capacity data is not uniformly digitized; smaller clinics may lack
APIs or digital dashboards.
o Crowdsourced data carries risks of misinformation and requires verification
mechanisms.
2. Connectivity and Infrastructure Gaps
o During floods, power outages and mobile network disruptions reduce citizen
access to digital platforms.
o SMS/IVR fallback channels mitigate this, but service continuity remains a
challenge in prolonged outages.
3. LLM and AI Limitations
o Large Language Models risk hallucination and may misinterpret ambiguous
queries.
o Pre-triage decisions remain probabilistic and cannot replace professional
clinical judgment.
o Language coverage beyond English and Tamil may be limited in initial
phases.
4. Operational Dependencies
o Success depends on active participation of municipal authorities, hospitals,
and fleet operators.
o In the absence of institutional buy-in, the system risks being underutilized or
parallelized.
5. Ethical and Legal Risks
o Patient data privacy, especially in crisis contexts, may be compromised if
consent is bypassed under emergency pressure.
o Liability remains unclear if automated routing or triage recommendations
contribute to adverse outcomes.
o Compliance frameworks (e.g., DPDP Act, NDHM) are evolving, requiring
ongoing adaptation.
6. Financial Sustainability Risks
o Revenue assumptions depend on government contracts and CSR funds; delays
in funding may disrupt continuity.
o Cost recovery during the first 12–18 months requires bridging finance or
donor support.
10.2 Future Work

1. Enhanced Data Integration

o Deploy IoT-based flood sensors and collaborate with IMD and municipal
agencies for higher-resolution, real-time flood maps.
o Standardize hospital reporting protocols under the NDHM/ABDM framework
to improve capacity visibility.
2. AI & Analytics Improvements
o Extend LLM models with retrieval-augmented generation (RAG) using vetted
disaster guidelines to reduce hallucination.
o Develop explainable AI modules that provide rationales for routing/triage
outputs, enhancing trust and accountability.
o Expand multilingual support (Hindi, Telugu, Bengali, Urdu) for wider
applicability.
3. Resilience Enhancements
o Invest in offline-first capabilities: edge servers, community kiosks, and mesh
networks for communication during blackouts.
Build redundancies such as solar-powered IVR hubs and distributed local
caches of hospital/fleet data.
4. Policy and Governance Evolution
o Engage with regulators to define liability frameworks for AI-assisted disaster-
health systems.
o Establish data-sharing MoUs between hospitals, NGOs, and municipal
agencies for rapid activation during disasters.
5. Scaling & Replication
o Pilot in Chennai, then replicate in other flood-prone metros (Mumbai, Kolkata,
Hyderabad) and eventually cyclone-prone coastal districts.
o Explore international collaborations for knowledge exchange with cities such
as Jakarta, Manila, and Dhaka facing similar risks.
6. Societal Research
o Conduct longitudinal studies on community trust, adoption rates, and equity of
access.
o Examine ethical trade-offs (privacy vs. emergency response) through
participatory action research with affected populations.
10.3 Roadmap Snapshot

 Short-term (0–12 months): Pilot in Chennai; validate core features (routing, triage,
dashboard); refine based on feedback.
 Medium-term (12–24 months): Expand multilingual channels, strengthen offline
resilience, formalize hospital and NGO partnerships.
 Long-term (24+ months): Nationwide scaling with integration into NDMA and
state-level disaster management frameworks; AI explainability and liability research.

11. Conclusion

The 2023 Chennai floods, triggered by Cyclone Michaung, exposed how healthcare systems
in urban India remain highly vulnerable to climate-induced disasters. Despite adequate
hospital capacity, patients were stranded due to transport paralysis, poor situational
awareness, and fragmented coordination between agencies. This report has proposed a
Digital Disaster–Health Coordination Platform that integrates real-time flood intelligence,
hospital availability, fleet telemetry, and citizen-facing support through multi-channel access.

The proposed solution contributes on three levels. First, it enhances operational resilience,
reducing dispatch times, balancing hospital loads, and ensuring pre-triage during floods.
Second, it strengthens societal equity by prioritising vulnerable populations, ensuring
inclusivity through SMS/IVR access, and aligning with the Sustainable Development Goals
(SDG 3, 11, 13). Third, it advances policy innovation, demonstrating how AI and geospatial
integration can be responsibly governed under India’s DPDP Act and NDHM frameworks.

For practitioners, the key takeaway is that healthcare continuity in floods depends as much on
information logistics as on physical infrastructure. For policymakers, the recommendation is
to institutionalise such platforms within municipal disaster management systems, supported
by clear data-sharing frameworks, funding models, and ethical safeguards.

References

Carter, H., Singh, A. and Alam, S. (2023), “AI, ethics, and disaster health: challenges for
inclusive resilience”, Journal of Disaster Studies, Vol. 48 No. 2, pp. 201-218.
Citizen Matters (2023), “Health challenges post-Cyclone Michaung in Chennai”, Citizen
Matters, available at: [Link]
michaung-chennai/ .

Dinakaran, D., Nagendran, T., Indumathi, E., Vasan, K. and Ramakrishnan, N. (2021),
Telemedicine Practice Guidelines - India, available at:
[Link] .

District Disaster Management Authority, Tiruvallur (2024), Tiruvallur District Disaster

Management Plan 2024, Tamil Nadu State Disaster Management Authority, available at:
[Link]
%[Link]

The Digital Personal Data Protection Act, 2023 (2023), Ministry of Electronics & IT
(MeitY), available at:
[Link] .

EY–IPA (2021), Telemedicine and Digital Health in India: Market Outlook, available at:
[Link] .

Gulzari, A. (2021), “A healthcare location-allocation model with an application to disaster

response”, Sustainable Cities and Society, Vol. 67, 102743.

Gupta, R. and Dey, N. (2022), “Disaster logistics in Indian megacities: an institutional

coordination perspective”, International Journal of Disaster Risk Reduction, Vol. 73,
102927.

Khatri, A., Bhattacharya, S. and Singh, P. (2022), “Barriers to telemedicine adoption in

disaster contexts: evidence from rural India”, Health Policy and Technology, Vol. 11 No. 4,
100665.

Lokmic-Tomkins, Z., et al. (2023), “Lessons learned from natural disasters around digital
health resilience”, International Journal of Disaster Risk Reduction, Vol. 85, 103502.

MIT–IBM (2024), AI for Crisis Management: Whitepaper, available at:

[Link] .

National Digital Health Mission (NDHM) (2021), Strategy Overview, NITI Aayog, available
at: [Link] .
National Health Mission Odisha (2020), SOP: Integrated Patient Transport (Ambulance
SOP), available at: [Link]
[Link] .

Patel, S.S., Rogers, M.B., Amlôt, R. and Rubin, G.J. (2020), “What do we mean by
‘community resilience’? A systematic literature review of how it is defined in the literature”,
BMC Public Health, Vol. 20, p. 1015.

Ravindran, T.S., Gopichandran, V. and John, K.R. (2021), “Urban flooding and health system
responses in India: lessons from Chennai and Hyderabad”, Indian Journal of Public Health,
Vol. 65 No. 4, pp. 345–352.

Singh, H., Nielsen, M. and Greatrex, H. (2023), “Causes, impacts, and mitigation strategies
of urban pluvial floods in India: A systematic review”, International Journal of Disaster Risk
Reduction, Vol. 92, 103724.

Sphere India (2023), Cyclone Michaung – Joint Rapid Needs Assessment (JRNA) / Secondary
Data Analysis Report, available at: [Link]
12/SDAR_Cyclone%20Michuang_0.pdf .

UNICEF (2022), Pakistan Floods Situation Report, available at:

[Link] .

United Nations (2015), Sustainable Development Goals, available at:

[Link] .

World Bank (2025), Climate Change Overview, available at:

[Link] .

World Health Organization (WHO) (2019), Implementing health emergency and disaster risk
management (Health EDRM), available at: [Link]
health-emergency-and-disaster-risk-management .

World Meteorological Organization (2022), State of Climate Services: Water and Floods,
available at: [Link] .
Appendices
Appendix A: Prompt & Output Log (include URLs; redact sensitive data).

Disclaimer

This appendix presents a selection of representative AI interactions used during the

project. GenAI tools were applied as assistive technologies for structuring ideas, drafting
initial text blocks, and formatting outputs. Critical activities - such as literature review,
synthesis of multiple sources, editing, evaluation planning, stakeholder analysis, and financial
modelling - were conducted directly by the student team. Not all intermediate prompts are
included; only those that directly contributed to shaping report deliverables are documented.

Date Prompt (excerpt) AI Output (excerpt) Student Action

20-Sep-25 “Extract 3–4 core pain Drafted problem statement Edited for conciseness,
points from Sphere highlighting lack of access added direct Sphere
India (2023) JRNA to health facilities (77% in India citations, and
report and draft a 150 Chennai, 93% in reframed into Section
– 200 word problem Tiruvallur) and collapse of 1.
statement.” patient-provider
connectivity.

20-Sep-25 “Produce an 800– Structured draft with urban Team refined content
1,000 word flooding context, with additional
Introduction including stakeholder mapping, and references (NDHM,
stakeholders, scope, research keywords. WHO Health-EDRM)
and chapter roadmap.” and reduced word
count.

21-Sep-25 “Provide a high-level Delivered a 5-row table Reformatted into

Solution Overview linking healthcare access Section 5; added
table mapping pain issues to technical outcomes aligned with
points → capabilities capabilities (flood-aware SDGs.
→ workflow → routing, pre-triage, multi-
outcomes.” channel citizen support).
 21-Sep-25 “Generate a Risk & Created a structured table Incorporated into
Compliance Checklist referencing DPDP Act, Appendix C; added
(privacy, fairness, NDHM, and WHO local Tamil Nadu
explain ability, frameworks. disaster regulations
security, regulatory).” manually.

22-Sep-25 “Draft a 150 - 200- Produced a summary Edited wording for

word Conclusion with linking information academic tone; added
actionable takeaways logistics to disaster-health recommendations for
for practitioners and resilience and SDGs. NDMA and municipal
policymakers.” integration.

Appendix B: UI Wireframes / Screenshots:

URL: [Link]
Appendix C: Risk & Compliance Checklist.

Risk Description Compliance / Mitigation Status

Strategy

Privacy & Data Patient location, Implement consent-driven data Designed for
Protection symptom, and hospital collection aligned with India’s compliance.
data are sensitive and DPDP Act (2023); anonymise
could be misused. aggregated analytics; enforce
data deletion policies.

Security Risk of cyberattacks (e.g., Encrypt all data in transit (TLS Ongoing;
ransomware, DDoS) 1.3) and at rest (AES-256); requires
targeting disaster- apply Zero Trust network institutional
response infrastructure. principles; periodic penetration buy-in.
testing.

Explainability LLMs may generate Restrict AI responses to vetted Partial; further

& AI Reliability hallucinated outputs or templates; deploy retrieval- pilots needed.
unsafe triage guidance. augmented generation (RAG)
linked to medical/disaster
SOPs; maintain human-in-loop
oversight for critical cases.

Fairness & Risk of excluding low- Multi-channel support (app, Pilot-ready.

Accessibility literacy, low-connectivity, SMS, IVR); multilingual
or non-English users. rollout (Tamil + English first,
then Hindi, Telugu, Bengali);
accessibility testing with
NGOs.

Operational Ambiguity on Establish liability frameworks Policy

Liability accountability if AI with municipal authorities; flag discussion
outputs contribute to AI as decision-support, not required.
harm (delayed referral, replacement for clinicians.
mis-routing).

Inter-agency Hospitals, NGOs, and Develop MoUs and API Draft stage.
Data Sharing civic bodies may resist governance under
sharing real-time NDHM/ABDM standards; use
capacity/fleet data. role-based access controls.

Environmental Cloud compute for Prefer green cloud providers, Design

Footprint AI/LLM increases carbon enable auto-scaling, and principle
emissions. purchase carbon offsets. adopted.

Regulatory India’s NDHM/ABDM Ensure API compliance with Mapping

Alignment and disaster-management NDHM; align protocols with underway.
frameworks require NDMA/WHO EDRM
compliance; overlaps frameworks.
with WHO Health-
EDRM.

Appendix D: News-to-Solution Traceability Matrix.

News Citation Extracted Pain Derived Feature

Point Requirement Implemented
Sphere India (2023), 93% of respondents Provide flood- Flood-Aware
Cyclone Michaung – Joint in Tiruvallur and aware routing and Routing Engine
Rapid Needs Assessment 77% in Chennai enable alternative with fleet
(JRNA) Report, Tamil Nadu, reported lack of medical transport integration; citizen
available at: access to nearest (boats, high- app generates safe
[Link] health facilities clearance route suggestions
.in/sites/default/files/2023- during flooding vehicles). and dispatch
12/TN_Cyclone requests.
%20Michaung_JRNA
%[Link]
Sphere India (2023) Emergency medical Integrate real-time Fleet orchestration
support gaps noted - flood overlays into dashboard with
limited mobility of fleet dispatch and adaptive routing
ambulances and lack routing workflows. and live telemetry.
of reliable
emergency transport
 in flooded areas.
Sphere India (2023) Overcrowding and Enable pre-triage Symptom-report
lack of triage: Many support and based pre-triage
PHCs and relief hospital load agent; hospital
camps overwhelmed balancing dashboards with
by fever, diarrhoea, mechanisms. inflow forecasts.
and respiratory
illness cases.
Sphere India (2023) Poor citizen Develop multi- LLM-driven
awareness: channel citizen citizen assistant
Communities lacked engagement for providing safe
timely information low-bandwidth facility guidance in
on available health contexts (SMS, Tamil/English via
facilities and medical IVR, app). app, SMS, IVR.
camps
Sphere India (2023) Fragmented inter- Build centralized Multi-stakeholder
agency situational disaster-health
coordination dashboard with dashboard with
between NGOs, local decentralized inputs real-time updates
authorities, and (APIs from and alerting.
hospitals slowed hospitals, NGOs,
relief. civic authorities).