The Data behind the Risk: How AI and Satellite Tech Are Rewriting Climate Underwriting

India ranked sixth in the 2025 Climate Risk Index among the top ten countries affected by extreme weather. Yet more than 90% of natural disaster losses in India remain uninsured.

The challenge is not simply a lack of insurance products.

It is a lack of sufficiently precise data to understand and price the risk.

Traditional underwriting often relies on historical averages, broad geographic classifications, and limited ground-level observations. But climate risk is becoming increasingly local, dynamic, and difficult to predict using historical patterns alone.

A district can record normal rainfall over several years and still experience severe drought in a single season. Two farms only a few kilometres apart can experience very different levels of flood or heat exposure.

This is where AI climate risk assessment, satellite earth observation, and hyperlocal climate risk data are changing the way insurers and financial institutions approach underwriting.

Why Old Data Can No Longer Price New Risk

Traditional underwriting works best when historical patterns are relatively stable.

Climate change challenges that assumption.

A district that received normal annual rainfall for five consecutive years can still experience a severe rainfall deficit during a single Kharif season. During the 2025 southwest monsoon, several regions of Uttar Pradesh recorded significant rainfall deficits even though India's overall monsoon rainfall ended above normal.

The national average therefore tells only part of the story.

This is the fundamental challenge of climate risk assessment.

Averages can hide exposure.

For an underwriter, relying solely on district or state-level averages can create two problems:

• Overpricing: Low-risk areas are charged premiums that do not reflect their actual exposure.
• Underpricing: High-risk areas are priced too cheaply because localised hazards are hidden inside broader averages.

Neither outcome is desirable.

Underpricing creates portfolio losses when extreme events occur.

Overpricing reduces insurance adoption among customers who may already struggle to afford coverage.

Climate risk also changes rapidly.

Extreme heat, rainfall variability, drought patterns, and flood behaviour are evolving. A model based entirely on historical averages may therefore struggle to represent emerging risks.

For financial institutions, the implications extend beyond insurance.

The RBI Governor noted in March 2025 that major financial risks—including credit, market, and operational risk—are influenced by climate change through physical and transition-risk channels.

If climate risk is mispriced, financial institutions can end up mispricing entire portfolios.

What Does AI-Powered Climate Risk Modeling Do?

AI climate risk assessment uses machine learning and advanced statistical models to combine large volumes of weather, satellite, geospatial, and other relevant data to estimate climate exposure.

For insurers and lenders, this creates three major advantages.

1. Moving From Historical Averages to Forward-Looking Risk

Traditional models often ask:

"What usually happens here?"

AI-driven climate models can instead ask:

"Given current conditions and historical patterns, what is likely to happen next?"

This distinction matters.

AI systems can process decades of atmospheric observations alongside current weather conditions and forecasts to generate more dynamic risk assessments.

Google's GenCast, for example, demonstrated the ability to generate global ensemble weather forecasts for up to 15 days in a matter of minutes.

For parametric insurance, faster forecasting creates the possibility of continuously monitoring conditions against predefined triggers.

The result is a more dynamic underwriting process rather than one based solely on a static historical risk score.

2. Creating Hyperlocal Climate Risk Data

The second advantage is spatial resolution.

A state-level climate score is useful for understanding broad exposure.

A district-level score is better.

A block-level or asset-level score can be significantly more actionable.

Hyperlocal climate risk data allows insurers and lenders to assess differences in rainfall, temperature, vegetation health, flood exposure, and drought conditions across much smaller geographic areas.

Satellite earth observation is central to this capability.

Synthetic Aperture Radar (SAR) satellites can detect flood extent even through cloud cover and during night-time conditions.

Multispectral satellite imagery can identify changes in vegetation and crop health before physical damage becomes obvious on the ground.

This creates a new underwriting layer:

Location → Climate exposure → Asset vulnerability → Financial risk

3. Automating Underwriting Through Climate Risk APIs

The third shift is automation.

A climate risk API for financial services can deliver processed climate-risk information directly into an insurer's or lender's existing technology infrastructure.

Instead of manually collecting weather reports, satellite observations, hazard maps, and historical datasets for every customer, an API can provide structured risk indicators at the required geographic level.

For example, a financial institution could receive:

• Rainfall deficit scores
• Temperature thresholds
• Heatwave indicators
• Flood exposure
• Drought probability
• Soil-moisture conditions
• Vegetation stress
• Geographic hazard scores

This makes it possible to screen thousands of agricultural borrowers, properties, or insured assets without conducting manual assessments for every location.

The result is potentially faster underwriting, lower operational costs, and more consistent risk assessment.

The Data Behind AI Climate Risk Modeling

AI models are only as useful as the data feeding them.

Modern climate underwriting therefore combines multiple data sources rather than relying on a single weather station or dataset.

Data Type	Example Sources	What It Measures
Rainfall	IMD, CHIRPS	Rainfall volume, deviation, and accumulation
Crop health	Sentinel, Landsat	Vegetation stress and crop development
Flood extent	Sentinel-1 SAR	Flood boundaries and duration
Temperature & humidity	IMD, ERA5	Heat intensity, duration, and extreme-temperature events
Soil moisture	Satellite remote sensing, reanalysis datasets	Drought and agricultural water stress
Historical climate	IMD, ERA5 and other long-term datasets	Long-term climate patterns and variability
Forecast data	Weather forecasting systems	Near-term climate conditions and potential trigger events

Table 1: Key data sources used in AI-enabled climate risk assessment.

The power comes from combining these datasets.

For example, rainfall alone may indicate that a region is dry.

Rainfall + soil moisture + vegetation stress provides a much stronger signal of agricultural drought.

Similarly, temperature alone indicates heat.

Temperature + humidity + duration can provide a more meaningful picture of heat stress.

This data fusion is one of the key advantages of modern climate underwriting.

What This Means for Parametric Insurance in India

Parametric insurance depends fundamentally on the quality of its trigger.

A policy may specify that a payout occurs when rainfall falls below a particular threshold or temperature exceeds a defined level for a specified number of days.

If the trigger is poorly designed, the policy can create significant basis risk.

Basis risk occurs when the policyholder experiences a real loss but the predefined trigger is not breached.

This makes data quality critical.

Better AI-powered climate risk modeling can help insurers design triggers using more granular information.

Instead of relying only on a single district-wide rainfall average, an insurer can incorporate high-resolution satellite and weather data to understand the spatial distribution of the hazard.

The objective is not to eliminate basis risk completely.

It is to reduce the gap between the measured trigger and the actual experience of the policyholder.

For example, consider a rainfall-based product for farmers.

A traditional approach might use:

District rainfall → Single threshold → Fixed payout

A more sophisticated approach could incorporate:

Historical rainfall + current observations + satellite indicators + soil moisture + forecast data → Localised risk model → Better-calibrated trigger

This creates the potential for products that are more closely aligned with actual climate exposure.

RBI-CRIS and the Growth of India's Climate Data Infrastructure

India's climate-risk data infrastructure is also evolving.

The RBI launched the Reserve Bank Climate Risk Information System (RB-CRIS) in May 2025 to provide regulated entities with standardised information relating to climate hazards, vulnerability, and other climate-risk parameters.

For banks and NBFCs, this is important because climate-risk assessment has historically required financial institutions to combine data from multiple sources themselves.

A centralised climate-risk data infrastructure can help reduce this fragmentation.

For an NBFC underwriting agricultural loans in Bihar, for example, physical climate-risk information can become another input alongside income, repayment history, collateral, and other credit variables.

For an insurer pricing agricultural exposure in Vidarbha, the same type of data can support hazard assessment and parametric trigger design.

The broader direction is clear:

Climate data is becoming financial data.

Where the Data Still Has Gaps

AI climate risk assessment is improving rapidly, but it is not a perfect solution.

Financial institutions need to understand several limitations.

Historical Data May Underestimate Emerging Risks

AI models still rely heavily on historical observations for training and calibration.

But climate change is shifting the baseline.

A weather event that historically occurred once every 20 years may become substantially more frequent.

A model trained primarily on historical relationships can therefore underestimate the probability or severity of emerging risks.

A 2025 peer-reviewed study in Frontiers in Climate highlighted challenges for index-insurance models that rely heavily on historical climate relationships as those underlying conditions change.

This means underwriting models need continuous recalibration.

Satellite Resolution Is Not the Same as Asset Resolution

Satellite technology has advanced significantly.

Resolutions of 10–30 metres are increasingly available for many earth-observation applications.

But India's agricultural landscape is highly fragmented.

A single satellite pixel can contain multiple small farms, crops, roads, trees, or buildings.

Turning satellite observations into reliable individual-asset risk assessments therefore requires significant processing and contextual information.

Data Quality Still Determines Model Quality

AI does not automatically solve poor data.

Missing observations, inconsistent weather stations, geographic inaccuracies, cloud contamination, and gaps in historical records can all affect model performance.

This makes data validation, quality control, and uncertainty measurement essential components of climate underwriting.

From Climate Data to Climate Underwriting

The most important change is not simply that insurers now have access to more data.

It is that climate data can increasingly become part of the underwriting workflow itself.

A modern climate underwriting process can look like:

1. Identify the asset

Locate the farm, property, borrower, business, or insured exposure.

2. Assess the hazard

Measure exposure to rainfall deficit, flood, heat, drought, cyclone, or other relevant hazards.

3. Analyse current conditions

Combine historical climate data with real-time observations and forecasts.

4. Apply AI models

Estimate probability, severity, and potential financial impact.

5. Design the insurance trigger

Translate the climate signal into a measurable parametric threshold.

6. Monitor continuously

Track conditions throughout the policy period.

7. Trigger automatically

When the agreed threshold is crossed, initiate the predetermined payout mechanism.

This creates a fundamentally different underwriting architecture.

The underwriter is no longer working with a static climate profile.

The risk can be monitored throughout the life of the policy.

What AI and Satellite Technology Change for Financial Institutions

For insurers, the benefits can extend across the insurance value chain:

• Underwriting: More granular risk differentiation.
• Pricing: Better alignment between premiums and geographic exposure.
• Product design: More precise parametric triggers.
• Portfolio management: Better identification of accumulated climate exposure.
• Claims and payouts: Automated trigger verification.
• Distribution: Digital products that can be priced and issued at scale.

For banks and NBFCs, the same data infrastructure can support:

• Climate-adjusted credit scoring
• Agricultural portfolio monitoring
• Physical-risk assessment
• Climate stress testing
• Collateral-risk assessment
• Resilience-linked lending

The result is a convergence between climate intelligence and financial risk management.

Wrapping Up: The Future of Underwriting Is Data-Driven

The quality of a climate insurance product is ultimately constrained by the quality of the information behind it.

AI climate risk assessment, satellite earth observation, and hyperlocal climate risk data are steadily changing that equation.

They make it possible to move from broad geographic averages toward more granular and dynamic assessments of climate exposure.

For parametric insurance, that matters enormously.

Better data can produce better triggers.

Better triggers can reduce basis risk.

And lower basis risk can make climate protection more useful and scalable.

India's development of RB-CRIS adds another important layer by strengthening the country's climate-risk data infrastructure for financial institutions.

The next phase of climate underwriting will therefore not be defined simply by having more insurance products.

It will be defined by whether insurers and financial institutions can measure climate risk precisely enough to price it, monitor it, and respond to it.

Ready to Explore AI-Driven Climate Underwriting?

Climate risk is becoming more local, dynamic, and financially material.

For insurers, lenders, and agricultural-finance providers, combining satellite intelligence, AI models, and real-time climate data can create a more responsive approach to underwriting.

Explore how AI-powered climate risk assessment and parametric insurance can help financial institutions build products designed around the climate risks their customers actually face.