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 over 90% of natural disaster losses in India remain uninsured. 

Traditional underwriting relies on historical averages and broad regional data. 

That approach cannot keep pace with a climate that behaves differently each season and varies across districts within the same state.

Why Old Data Can No Longer Price New Risk

A district that received normal annual rainfall five years in a row may still face a 70% deficit in a single kharif season. Certain regions in Uttar Pradesh recorded a rainfall deficit during the 2025 southwest monsoon, even as the national monsoon ended 8% above normal. 

This is the core problem for climate risk assessment. Without hyperlocal climate risk data, district-level historical averages conceal extreme local variation. An underwriter relying on that average may misprice risk by either charging too much where risk is lower or too little where risk is significantly higher.

Traditional underwriting also misses the speed at which conditions are changing. The 2025 pre-monsoon season brought unseasonal hailstorms to wheat-growing areas in three of the past four years. 

This reflects a structural shift in climate patterns, not a one-off outlier that historical models can absorb.

For financial institutions, the consequence is direct. The RBI Governor stated in March 2025 that all major financial risk categories (credit, market, and operational) are influenced by climate change through physical and transition risk channels. 

Mispricing that risk means mispricing entire loan and insurance portfolios.

What Does AI-Powered Climate Risk Modeling Do

AI climate risk assessment, the use of machine learning models trained on weather, satellite, and financial data to quantify climate exposure, addresses these gaps in three specific ways.

First, it replaces historical averages with forward-looking models. AI systems trained on decades of atmospheric data can generate district-level forecasts at a fraction of the cost of traditional modelling. Google's AI weather model GenCast, for example, produces 15-day global ensemble forecasts in around eight minutes. 

That speed allows real-time recalibration of parametric triggers as conditions change.

Second, it enables hyperlocal climate risk data at the field or block level, not just the district or state level. Between 2022 and 2025, 405 Earth observation satellites entered orbit, tripling global monitoring capacity. Plutas Insure uses hyperlocal climate risk data from IMD and ERA5 satellites to improve climate underwriting accuracy.

Synthetic Aperture Radar satellites can monitor flood extent through cloud cover, day or night. Multispectral sensors detect crop stress weeks before visible damage appears.

Third, a climate risk API for financial services reduces the cost of underwriting by automating data assembly. A climate risk API for financial services (a data feed delivering hazard scores, flood heights, temperature thresholds, and drought indicators at the asset level) makes it possible to screen thousands of agricultural borrowers or insured properties without manual field surveys.

Key Data Sources Used in AI Climate Risk Modeling for Parametric Insurance

Accurate AI climate risk modeling for parametric insurance sources the following types of data:

Data Type	Source	What It Measures
Rainfall estimates	IMD weather stations + CHIRPS satellite	District-level rainfall volume and deficit
Crop health index (NDVI)	Sentinel / Landsat satellites	Vegetation stress and crop growth stage
Flood extent	SAR satellites (Sentinel-1)	Flood boundary and duration
Temperature and humidity	IMD + ERA5 reanalysis data	Heatwave onset, intensity, and duration
Soil moisture	MODIS / satellite remote sensing	Drought conditions are affecting the yield

What This Means for Parametric Insurance in India

Parametric insurance, which pays out automatically when a pre-agreed trigger is met, depends entirely on the quality and resolution of the underlying data. A poorly chosen trigger creates basis risk, where the policyholder suffers a loss, but the trigger is not met, resulting in no payout.

Better AI-powered climate risk modelling directly reduces basis risk in parametric insurance. When a rainfall trigger is set using high-resolution satellite data at the block level rather than a district-level IMD average, the trigger more closely matches what actually happened on the ground.

For Indian financial institutions, the RBI's launch of the Climate Risk Information System (RBI-CRIS) on 29 May 2025 is a significant step. RBI-CRIS provides banks and NBFCs with standardised hazard, vulnerability, and transition risk datasets. This is precisely the data infrastructure that supports both climate credit risk assessment and parametric trigger design.

In practice, a bank or NBFC underwriting agricultural loans in Bihar or an insurer pricing a crop policy in Vidarbha can now access standardised, processed climate data from a central RBI-approved repository, rather than assembling datasets independently.

Where the Data Still Has Gaps

AI climate risk modelling is improving fast, but it has limits that financial institutions must understand before building products on it.

Historical data used to train models may not fully capture how climate risk is evolving. A model trained on 30 years of monsoon data may still underestimate a once-in-20-year event that is now occurring every five years. 

A 2025 peer-reviewed study in Frontiers in Climate found that index insurance products relying on historical data models increasingly struggle to price risk accurately as the baseline shifts under climate change. 

Satellite coverage, while growing, still has gaps in data resolution for very small farm parcels. A 10-30 metre satellite resolution is now achievable for crop health monitoring, but applying it to the fragmented plots of India's smallholders requires significant processing.

Wrapping Up: Accurate Underwriting through AI Climate Risk Assessment

The quality of a parametric insurance trigger is only as good as the data behind it. AI climate risk assessment and satellite earth observation are steadily raising that quality, reducing basis risk, enabling hyperlocal pricing, and making previously uninsurable exposures tractable. 

The RBI's launch of RBI-CRIS in May 2025 formalised India's commitment to building the data infrastructure that financial institutions need. As that infrastructure develops, the gap between what climate data can tell us and what underwriting currently uses is one of the more important problems in Indian financial services today.

Ready to Explore AI-Driven Parametric Insurance Products? 

If you work in insurance, lending, or agricultural finance, it is worth looking into how parametric products are now structured using satellite and AI data and what trigger designs are available for your sector and geography.

Frequently Asked Questions

Why is data accuracy so important in parametric insurance?

Parametric insurance relies entirely on objective data to determine payouts. Accurate and reliable climate data ensures triggers reflect actual weather conditions, reduces disputes, improves customer confidence, and helps insurers price products more fairly across different regions and risk profiles.

Can AI help detect climate risks before they become severe?

Yes. AI can identify emerging weather patterns, vegetation stress, and changing environmental conditions before significant damage occurs. These early insights allow insurers, lenders, and farmers to take preventive measures and make more informed financial and operational decisions.

How often is climate data updated for underwriting purposes?

The update frequency depends on the data source. Weather stations may provide hourly or daily observations, while satellites capture images at regular intervals ranging from daily to several days, enabling insurers to monitor changing conditions throughout a policy period.

Can AI-powered climate models be used outside agriculture?

Yes. AI-driven climate risk models also support underwriting for property, infrastructure, renewable energy, logistics, and supply chains. Any sector exposed to floods, storms, heatwaves, droughts, or other climate hazards can benefit from more accurate risk assessment.