How AI Is Strengthening Resilient Infrastructure

Contents

Executive Summary

This panel discussion explores how artificial intelligence, combined with satellite imagery, geospatial modeling, and digital infrastructure, can enhance infrastructure resilience against disaster risks. The panelists emphasize that while AI offers transformative potential, scaling from experimental pilots to governmentwide implementation requires bridging critical gaps in data infrastructure, institutional capacity, policy frameworks, and financial investment — particularly in low- and middle-income countries.

Key Takeaways

AI is transitioning from experimental to strategic enabler — but only if institutional capacity-building, policy frameworks, and data standardization keep pace with technological innovation. Technology alone is insufficient.
The "bridge problem" is as important as the technical problem — breaking silos between datasets (cyclone, heat, hydrology, livelihood), between technology creators and policymakers, and between national experts and ground-level government engineers is essential for scalability.
Localized, low-compute models may outperform centralized AI — "pocket-sized" AI models that run on local devices with minimal data can serve rural and remote communities better than cloud-dependent systems, reducing digital inequities.
Infrastructure resilience requires integrated risk assessment — treating infrastructure as isolated assets (roads, water, power) rather than interconnected networks that serve communities is why past mitigation projects fail to prevent repeat erosion and flooding in adjacent locations.
Equity must be built into AI design from the start — AI applications for disaster resilience must explicitly protect marginalized groups (tribal women, coastal communities, farmers) and address livelihood disruption, not just asset damage.

Panel Discussion at AI Summit

Key Topics Covered

AI and Disaster Risk Reduction: Role of AI in infrastructure planning, siting, operation, maintenance, and recovery
Satellite Imagery & Geospatial Data: ISRO's use of remote sensing and physics-based models combined with AI for flood, cyclone, and climate monitoring
Flood Forecasting & Early Warning Systems: Integration of hydrodynamic modeling, machine learning, and SAR data for rapid flood inundation mapping
Decision-Ready Intelligence: Translating complex climate and hazard datasets into actionable insights for administrators and policymakers
Mobile Network Resilience: Using AI to predict disasters and protect telecommunications infrastructure in low- and middle-income countries
Data Integration & Silos: Breaking down departmental data silos to create unified risk assessment frameworks
Capacity Building: Training government engineers and local authorities to operationalize AI-based tools
Urban Heat Mapping: Applying AI-driven satellite analysis to identify urban heat islands and vulnerable populations
Infrastructure Investment at Scale: Budget implications of building resilient infrastructure (India's ₹12 lakh crore capital expenditure plan)
Equity & Ethics in AI: Ensuring AI applications reduce inequalities and protect marginalized communities

Key Points & Insights

Scale of Infrastructure Risk is Massive: Global average annual losses from infrastructure damage exceed $700 billion (physical assets only); accounting for service disruption losses increases this to 7.4 times higher. India alone has documented ₹19,000 crore in direct infrastructure losses across 11 states in two years.
Over 50% of Global Infrastructure Needed by 2050 Remains Unbuilt: This creates a unique opportunity to integrate resilience and AI into infrastructure design from inception, rather than retrofitting existing systems.
AI Accelerates Post-Disaster Assessment: Traditional flood mapping via satellite imagery relies on manual threshold-setting and before/after image comparison. AI-trained models can automate flood inundation mapping in under one minute by integrating multiple datasets (elevation, slope, historical flood events), enabling rapid PDNA (Post-Disaster Needs Assessment) completion.
Data-Scarce Environments Can Benefit from AI: Even with incomplete datasets, AI can generate proxy datasets and localized models informed by citizen input, enabling villagers and local administrators to make informed infrastructure decisions without waiting for complete data collection.
Generative AI Simplifies Complex Technical Models: Flood velocity curves and hydrodynamic models are incomprehensible to village administrators. Generative AI converts these into plain-language advisories and visual reports, making risk analytics actionable at grassroots level.
Critical Gap Between Technology and Implementation: Policymakers face a "digital divide" and skepticism about AI. Solutions exist but aren't reaching government engineers, state disaster management authorities, or infrastructure operators — requiring policy bridges and institutional change management.
Satellite Data & AI Enable Targeted Early Warnings: High-resolution satellite imagery combined with AI can predict cloud bursts, cyclone paths, storm surge, and rainfall with precision sufficient for hyperlocalized alerts to vulnerable populations (farmers, coastal residents) rather than generic broadcast warnings.
Mobile Networks Are Both Critical Infrastructure and Vulnerability: AI can optimize mobile tower resilience by predicting extreme heat or cyclones and triggering cooling mechanisms or antenna removal. Conversely, network outages during disasters block communication; AI-driven "disaster connectivity maps" track infrastructure status in real time to support humanitarian response.
Upfront Investment & Data Costs Are Barriers: State governments' first questions are: "What will this cost?" and "What data do we already have?" Solutions require standardized cost-benefit analyses, pre-packaged datasets, and modular learning modules for capacity transfer to engineers across multiple departments (CWC, PWD, RD).
Agentic AI Represents Next Frontier: Systems that answer questions like "Should I build a school here?" and auto-calculate budget requirements (₹10,000 vs. ₹80 lakh difference) can empower local administrators to self-assess and submit evidence-based budget proposals to state disaster management authorities, streamlining preparedness funding.

Notable Quotes or Statements

Rita Misal (NDMA): "We seem to patch up one piece of land and the next piece of land, after several years we see that other piece of land along the same coastline or river is broken down. We don't know whether the impact of it or what is it happening... This is a list of failures we have spent crores of rupees and we have not been successful yet to address the main problem." (On repeated infrastructure failures in coastal and river erosion mitigation)
Rita Misal: "How do we then move from pilots or experiments to scalable solutions? There has to be a strategy for this governance of giving this tech, taking this technology, AI technology right from the national level from experts like you all, from institutions like ISRO, to the government people who need to understand this and apply it. That's where I think there is a huge gap."
Samita (Resilience AI): "Generative AI simplifies the information to land with the person who is actually taking the decision... A village pradhan and a patail is using a system which is Resilience 360. Now that has converted the entire conversation of just computational AI [into] simplified, actionable recommendations."
Dr. Praep Tapl (ISRO): "AI can help in how to send these alerts to targeted people. It's not just that you hear from radio or TV that there will be heavy rain or there will be flood. This information should reach targeted people." (On AI-driven alert dissemination)
Dr. Saharia (IIT Delhi): "We can train this model over three months, but the inference it only takes one minute to get a flood inundation map... You are able to utilize a lot of information that traditionally we would never utilize."
Anu (SEEDS): "We speak about cyclones... but it was very slow onset erosion based work that pushed this village out very slowly... We need to break out of these silos in which we've put our data. This is cyclone-centric. This is hardware. This is infrastructure. It needs to merge."

Speakers & Organizations Mentioned

Speaker	Affiliation	Role/Expertise
Rita Misal	National Disaster Management Authority (NDMA)	Member; disaster risk reduction, post-disaster recovery, resilience planning
Dr. Praep Tapl	ISRO (Space Application Center, Ahmedabad)	Group Director, Atmospheric & Ocean Science; satellite imagery, flood/cyclone monitoring
Samita	Resilience AI (founding CEO)	Disaster management practitioner; intersection of humanitarian systems, tech, institutional resilience
Dr. Manavindra Saharia	IIT Delhi; CDRI Fellow	Geospatial science, hydrodynamic modeling, machine learning for flood forecasting
Anu	SEEDS; CDRI associate	Disaster risk reduction, climate resilience, governance innovation; co-founder SEEDS
Kimberly Brown	GSMA (Mobile for Humanitarian Innovation)	Head of emerging tech & humanitarian innovation; mobile network resilience
Dr. Sriant K Panigrai	Indian Institute of Sustainable Development (IISD)	Civil servant, policy maker; 37 years experience (World Bank, UNEP); urban heat mapping, climate risk modeling
Ranjini (Moderator)	CDRI (Coalition for Disaster Resilient Infrastructure)	Led panel; contextualized on CDRI's Giri platform and data infrastructure work
Deepali & Co-organizers	CDRI	Session coordination

Technical Concepts & Resources

AI/ML Methodologies

Computational AI: Large-scale data processing and modeling (traditional approach)
Generative AI: Creating human-readable reports and plain-language advisories from complex models
Agentic AI: Question-answering systems that autonomously assess and recommend actions (emerging)
Machine Learning Models: Trained on historical flood/cyclone data to automate prediction and classification

Data Sources & Tools

ISRO Satellite Data:
- Cloud imagery, 3D temperature profiles
- Synthetic Aperture Radar (SAR) for flood detection
- Digital Elevation Models (DEM)
- Resource monitoring satellites (high-resolution geospatial)
External Datasets: NASA Earth Data, World Pop, OpenStreetMap
Google Flood Hub: AI-driven flood prediction tool (mentioned in Madagascar example)
Disaster Connectivity Maps: ITU-backed tool leveraging AI to track infrastructure status in real time
GIRI Platform (CDRI): Infrastructure risk assessment tool; quantifies average annual losses

Modeling Approaches

Physics-Based Models: Hydrodynamic flood modeling combined with AI for speed & accuracy
Threshold-Based Classification (traditional): Manual cutoffs for flooded/non-flooded pixels → automated via AI
Multi-Source Integration: Combining satellite imagery, elevation, slope, historical events, and hydrology into unified models
Rapid Assessment Automation: Flood inundation mapping in <1 minute vs. weeks for traditional approaches

Key Processes & Concepts

Post-Disaster Needs Assessment (PDNA): Quantifying damage to infrastructure; AI accelerates completion
Impact-Based Forecasting: Predicting not just weather events but downstream consequences (inundation areas, vulnerable populations affected)
Disaster Risk Reduction (DRR): Long-term mitigation through siting, design, and operational resilience
Slow-Onset vs. Rapid-Onset Disasters: Erosion vs. cyclones require different AI approaches
Resilience 360: Proprietary tool from Resilience AI; village-level risk assessment and budgeting interface
Heat Island Mapping: Satellite thermal analysis to identify urban heat vulnerabilities

Policy & Governance Frameworks

16th Finance Commission: References India's disaster resilience index budgeting
India's 2026 Capital Budget: ₹12 lakh crores; 7 new rail corridors, 20 national water highways
Digital Public Infrastructure (India Stack Model): Proposed framework for shared, interoperable AI/data platforms across government
Capacity Building Modules: Translating complex AI insights into modular training for government engineers across sectors (CWC, PWD, Rural Development)

Case Studies & Examples

Satbaya, Odisha: Multi-decade coastal erosion; 3–4 cyclone displacements; resettled village now facing river erosion — illustrates slow-onset, chronic risks
Bihar (Kosi River): Annual flooding; lacks accurate projections for damage/loss estimation
Delhi (March–June 2023): 34 heat-related deaths in two city locations; urban heat mapping critical
Madagascar: GSMA/ITU partnership using Google Flood Hub + mobile networks for flood alerting
80 Indian Villages: Currently using Resilience 360 system with UN and NDMA support

Disclaimer & Limitations

Transcript Quality: Portions contain minor transcription artifacts (repeated phrases, unclear terms) which have been clarified contextually; some speaker attributions inferred from role descriptions
Time Constraints: Panel was cut short; not all speakers completed remarks; some technical details were truncated
Organizational Names: One speaker's name and institution were mispronounced by moderator but corrected in discussion
Specific Data/Figures: Some statistics (₹19,000 crore, 34 deaths) are specific to examples cited; generalization not implied
Technology Maturity: Some tools (agentic AI, Resilience 360) appear still in early deployment; claims of "success" are based on pilot evidence