Catastrophe Scenario Modelling Framework for Indian Corporates: PML, MFL, and Multi-Peril Aggregation

Catastrophe modelling has transitioned from a tool used only by reinsurers and large multinationals to a necessity for Indian corporates selecting locations, designing insurance programmes, and managing reinsurance costs. When an Indian manufacturer chooses between a facility in Hyderabad versus Chennai, or when a logistics company decides whether to consolidate inventory in Mumbai or distribute across three cities, the catastrophe loss profile is a material financial variable. The insurance cost difference between these scenarios can range from INR 30 lakh to INR 3 crore annually, depending on the facility size and peril exposure.

Catastrophe modelling is the discipline of quantifying the probability and magnitude of losses from natural disasters and other low-frequency, high-impact events. The output is a probability distribution of potential losses, which allows corporate risk managers to answer questions like: What is the 1-in-100-year loss from earthquake at this facility? What is the aggregate loss if both my Mumbai and Pune facilities are damaged simultaneously? How much reinsurance capacity do I need, and what should I expect to pay for it?

Indian insurers and brokers increasingly expect corporates to have conducted catastrophe modelling before placing large industrial risks. The underwriter's first question on a INR 500-crore property account may be: Have you modelled the earthquake loss? This reflects the reinsurer's expectation that the insured has quantified their exposure and can defend the insurance limits and programme structure with data rather than assumptions.

The challenge for Indian corporates is that catastrophe modelling requires sophisticated software and specialist expertise. Commercial vendors (RMS, Verisk AIR, CoreLogic) provide the models, but the cost of licensing and the learning curve are substantial. This guide explains how to interpret catastrophe models and what outputs matter most for insurance programme design, allowing corporates to engage with brokers and reinsurers on equal footing.

PML (Probable Maximum Loss) and MFL (Maximum Foreseeable Loss) are the two primary outputs of catastrophe modelling, and they serve different purposes in insurance programme design. Understanding the distinction is critical for avoiding misinterpretation.

PML is the loss that is expected to be exceeded in only 1 out of 250 years on average. In technical language, it is the 99.6th percentile of the loss distribution, or the loss at a 400-year return period. PML is the metric most commonly used by reinsurers and insurance regulators because it represents an 'acceptable' tail risk. An insurer that holds capital equal to its aggregate PML across all risks can sustain a 1-in-250-year loss without becoming insolvent. Reinsurance treaties are typically structured with PML as the threshold: the insurer retains losses up to its PML and cedes excess losses to reinsurers.

For a manufacturing facility in Seismic Zone III (such as facilities in Ahmedabad, Delhi, or Kolkata), the PML from earthquake might be 65% of the building and contents value. This means a loss equivalent to 65% of the insured value is expected to occur once every 250 years on average. For a facility in Zone IV (such as Pune or western Gujarat), the PML might be 75-85% due to higher seismic intensity.

MFL (Maximum Foreseeable Loss) is more extreme. MFL is the largest loss that could credibly occur given known hazard parameters, without requiring assumptions about unlikely compound scenarios. For earthquake, MFL is typically the loss that corresponds to the maximum credible earthquake (MCE) for that region, which is the largest earthquake that scientists believe is possible in that region based on geology. For facilities in Seismic Zone IV, the MCE corresponds to approximately a magnitude 7.5 earthquake. The MFL from a Zone IV earthquake might be 95% of insured value for modern reinforced concrete buildings, or 100% (total loss) for older unreinforced masonry construction.

The practical difference: PML is used for insurance programme sizing and reinsurance purchasing. MFL is used for catastrophe bonds, for risk assessment against the insurance company's financial limits, and for assessing uninsurable risk (the loss that cannot be transferred to insurers regardless of cost). A corporate risk manager should understand both metrics but should use PML as the primary guide for insurance limit selection. Insurance limits are generally set at or above the PML so that the company retains only a small probability of an uninsured loss. Setting limits at the MFL level is excessively conservative and typically not economically justified.

Many Indian corporates operate multiple facilities across different locations. The insurance programme must address not just the loss at a single site but the accumulated loss if multiple sites are affected by the same disaster. Multi-site aggregation analysis identifies which disasters pose the greatest cumulative exposure.

For geographic perils like earthquake and flood, multi-site aggregation depends on the geographic footprint. Two facilities 200 kilometres apart in different seismic zones will have uncorrelated earthquake losses: if one facility suffers a magnitude 7.0 earthquake, the probability that the other facility (200 km away in a different zone) will simultaneously suffer the same magnitude earthquake is negligible. Conversely, two facilities 20 kilometres apart in the same industrial area are highly correlated: a single earthquake will likely damage both.

A practical example: a logistics company with distribution centres in Mumbai, Pune, and Nagpur faces earthquake risk. Mumbai and Pune are both in Seismic Zone III, separated by 150 kilometres. Nagpur is in Zone II. A magnitude 6.5 earthquake centred on Mumbai would severely damage the Mumbai DC and moderately damage the Pune DC (which is 150 km away, outside the zone of maximum intensity). The Nagpur DC would experience negligible damage. The aggregate loss would be substantial but not catastrophic. However, if the company had two DCs in Mumbai 10 kilometres apart, a magnitude 6.5 earthquake would damage both severely, creating an aggregate loss two times the single-site PML.

Flood aggregation operates on the same principle but follows watershed boundaries. Two facilities in the Mumbai metropolitan area, both in the Mithi River or Oshiwara River floodplains, have correlated flood risk. A 100-year flood event would inundate both facilities simultaneously. Facilities in different river basins (for example, one in the Mithi basin, another in the Ulhas River basin 50 km away) have lower correlation.

Cyclone aggregation depends on storm track and intensity decay with distance. Two facilities in Chennai and Bangalore, 350 kilometres apart, have low cyclone correlation because a cyclone strong enough to cause significant damage in Chennai would weaken substantially by the time it reaches Bangalore. Two facilities in Chennai and Puducherry, 150 kilometres apart and both on the East Coast, have much higher correlation because the same cyclone would affect both at destructive intensities.

For multi-site aggregation, the catastrophe model produces an accumulation schedule showing the PML and MFL for different aggregate scenarios: loss at the most exposed site alone, loss at the top two sites from a single event, loss at the top three sites, and total exposure. This schedule shows which sites drive the aggregate loss and helps the corporate prioritise diversification or risk management investments. A company discovering that 60% of the aggregate earthquake PML comes from a single cluster of facilities in Mumbai may decide to relocate some capacity to a different seismic zone, reducing the overall accumulation exposure.

Indian natural peril exposure varies by peril and region. Flood risk is the most widespread peril; RMS, Verisk AIR, and CoreLogic provide models incorporating CWC, IMD, and BIS data. Return-period analysis uses 100-year and 500-year flood levels; depth-damage curves are specific to Indian construction standards.

Cyclone risk is concentrated on the East Coast (Andhra Pradesh, Odisha, Tamil Nadu). IMD cyclone track data feeds modelling of wind speeds at each location. The 100-year cyclone intensity and potential building damage define loss profiles. For facilities within 100 km of the coast, cyclone risk is material.

Earthquake risk is highest in Himalayan foothills and along geological faults. BIS classifies India into seismic zones II-V. Zone IV (Delhi, Kutch, parts of Himachal Pradesh) faces potential magnitude 7.5-8.0 earthquakes; Zone III (Pune, Ahmedabad) faces magnitude 6.5. Zone IV PML is typically 70-80% of insured value; Zone II PML is 5-15%.

Fire conflagration risk is concentrated in older industrial areas (Surat, Tiruppur). Models estimate probability of fire spread between adjacent facilities. High-density areas may face 500-year conflagration loss of 150-200% of a single facility's value.

Three primary vendors provide catastrophe models used by Indian insurers: RMS (Risk Management Solutions), Verisk AIR, and CoreLogic.

RMS is the largest vendor with dedicated models for Indian flood, earthquake, cyclone, and wildfire. RMS incorporates CWC river discharge data, seismic hazard data from the Building Research Institute, and historical earthquake damage data. Outputs are typically accessed through brokers or insurers who license the models.

Verisk AIR is the second-largest vendor with models for flood, earthquake, cyclone, and wind. AIR's cyclone model (TC 2015+) is respected in the Indian reinsurance market. Verisk provides engineering-based models that incorporate building design standards, useful for earthquake risk where construction quality varies.

CoreLogic provides models for flood and secondary perils (hail, wildfire). Less commonly used than RMS and Verisk for Indian primary perils.

When evaluating results, ask: Which vendor? How recent is the model version? What are the building vulnerability and construction standards assumptions? Model uncertainty is significant; earthquake PML estimates can vary 20-30% between vendors depending on seismic hazard curves. Large risks are modelled by multiple vendors with results averaged or conservatively selected. Model results are best estimates with inherent uncertainty.

Insurance limits should be set at or above the PML. If earthquake PML is 60% of insured value, the limit should be at least 60%. Setting limits below the PML accepts uninsured catastrophe risk; setting significantly above is inefficient due to premium pricing for low-probability tail losses.

For a facility with INR 200 crore replacement value and 65% earthquake PML, set limit at INR 130 crore+. For aggregated sites, reflect aggregate PML.

Reinsurance is driven by insurer capacity and corporate concentration. Single facilities: insurer underwrites full limit and purchases reinsurance treaties. Multiple concentrated exposures: layered structure with primary layer (domestic insurers, INR 50-100 crore), excess layer (reinsurers, INR 100-200 crore), catastrophe layer (specialist reinsurers/cat bonds, above INR 200 crore).

Cat load is additional premium for catastrophe risk beyond routine perils. Zone IV facility (70% PML) pays more cat load than Zone II (10% PML); difference is 2-5% of total premium. IRDAI-standardised loads: earthquake 5-15% of base fire rate, flood 3-10%, cyclone 2-8% depending on location and coast proximity. These loads apply to domestic and reinsurance pricing.

Catastrophe modelling outputs feed into scenario testing and risk reporting to the board. Rather than simply stating that the company is 'well insured,' scenario testing uses specific disaster scenarios to show how the company's financial position would be affected, and how insurance mitigates that impact.

A typical scenario test identifies the top five natural disaster risks for the company and models the impact of a 1-in-250-year event for each risk. For a diversified Indian manufacturer, this might include: (1) Seismic Zone IV earthquake at the Pune facility, (2) Flood event in the Mumbai industrial area, (3) Cyclone on the East Coast affecting the Visakhapatnam facility, (4) Conflagration in the Chennai facility, and (5) Aggregate earthquake (1-in-500-year) affecting multiple facilities.

For each scenario, the financial impact is calculated as: physical loss (in INR), business interruption loss if relevant, and uninsured residual loss after insurance. The uninsured loss should be zero or negligible if the insurance programme is properly sized. If the scenario test reveals uninsured losses, the corporate has identified a gap in coverage that should be addressed before the next renewal.

Board reporting should articulate the catastrophe risks in plain language and show the insurance response. For example: 'The company's earthquake PML across all facilities is INR 180 crore. Current insurance coverage is INR 200 crore, covering 111% of the PML. The probability of uninsured earthquake loss is less than 0.4% annually (less than 1 in 250 years). This coverage is maintained through a primary policy of INR 100 crore and reinsurance excess of loss coverage of INR 100 crore.'

Regular scenario testing, conducted annually or biannually, ensures that as the company's facilities change (acquisitions, closures, relocations), the insurance programme is recalibrated and gaps are identified early. This discipline transforms catastrophe modelling from an abstract exercise into a concrete tool for financial risk management.