What Is NatCat? A Broker's Guide to Catastrophe Modeling

TL;DR

NatCat (natural catastrophe) modeling estimates insured losses from perils like hurricanes, earthquakes, floods, and wildfires by combining hazard, vulnerability, exposure, and policy terms to produce metrics such as average annual loss (AAL) and probable maximum loss (PML), which carriers use to set premiums, limits, and capacity. Because models fill missing or vague property data with conservative defaults, brokers who submit precise geocodes, complete COPE details, current replacement costs, and populated hazard fields can secure more accurate loss estimates and better terms for their clients.

Your client's property schedule hits an underwriter's desk. A few days later, a model has priced their hurricane, flood, and earthquake exposure. That number drives their premium, their limits, and sometimes whether they get coverage at all. Ever seen two similar buildings get wildly different quotes? The model is usually why.

So, what is NatCat, and why should you care about how it works? This guide explains it in plain English. You'll see how natural catastrophe modeling combines hazard, vulnerability, and exposure data to produce the loss estimates carriers price from. You'll learn what AAL and PML actually measure and who relies on them. Most importantly, you'll see how the quality of the data you submit shapes the results your clients get back and what you can fix before you hit “Send.”

What Is NatCat?

The abbreviation “NatCat” is short for “natural catastrophe.” It describes a large, sudden event caused by nature that damages property across a wide area and produces a wave of insured claims at the same time. A hailstorm hitting one roof is a claim. The same storm tearing across three counties and hitting thousands of roofs is a NatCat event.

Most carriers group the following perils under the NatCat label:

• Hurricanes: Wind, storm surge, and heavy rain that damage coastal properties and often reach well inland.
• Earthquakes: Ground shaking plus follow-on losses, such as fire following the quake.
• Floods: Overflowing rivers, surface water after intense rainfall, or coastal inundation.
• Wildfires: Fires that spread into suburban and commercial areas bordering wildland (see our guide to wildfire risk modeling for how this risk is assessed).
• Severe convective storms: Tornadoes, hail, and straight-line winds.
• Winter storms: Ice, snow load, and freeze damage to roofs, pipes, and equipment.

You'll see the term written as “nat cat,” “NatCat,” or just “cat.” It appears in carrier submission requirements, reinsurance treaty wording, rating agency reports, and quarterly market updates. When an underwriter asks about a client's “cat exposure,” they almost always mean exposure to these natural perils.

Scope matters here: NatCat covers natural hazards only. Man-made catastrophes, including terrorism, large cyber events, and major casualty losses, sit in separate categories with their own models, pricing, and capacity. If you want a broader view of how these categories relate, our overview of the types of risk in insurance lays them out side by side.

NatCat Modeling vs. Catastrophe (Cat) Modeling

Catastrophe modeling is the broad practice of estimating what a catastrophic event could cost. Natural catastrophe modeling is the branch that focuses specifically on hazards caused by nature.

In day-to-day conversations, most people treat the two terms as interchangeable because natural perils drive the bulk of property cat analysis. Technically, though, cat modeling can also cover man-made scenarios, while NatCat modeling stays limited to natural events. Knowing the difference helps when you read treaty language or a carrier's modeling assumptions, since the scope of the model shapes what the numbers include.

cta-inline-card

AI Agent for Insurance Brokers

• 

  Ingests and repairs your data automatically

  SOVs, Loss Runs, Revenue, Payrolls, Vehicle lists, and more

• 

  Remediates issues before they hit modeling

  Fix problems, explain impact, and track progress

• 

  Provides action-oriented recommendations

  Prioritize open items and resolve gaps faster

Watch the Demo

AI Agent for Insurance Brokers

• 

  Ingests and repairs your data automatically

  SOVs, Loss Runs, Revenue, Payrolls, Vehicle lists, and more

• 

  Remediates issues before they hit modeling

  Fix problems, explain impact, and track progress

• 

  Provides action-oriented recommendations

  Prioritize open items and resolve gaps faster

Watch the Demo

How Natural Catastrophe Modeling Works

The Core Components of a NatCat Model

Most vendor models use the same basic structure. Think of it as an assembly line where each station passes its result to the next. The four stations are:

1. Hazard: This module sets out where events can happen, how often, and how strong they get. Examples include peak wind speeds along a coastline or ground shaking near a fault line.
2. Vulnerability: This step estimates how much damage a building will take at a given intensity. A 1960s unreinforced masonry warehouse and a new steel-frame office end up with very different damage curves.
3. Exposure: The model maps your client's assets with their values attached, so it knows exactly what's at risk.
4. Insurance conditions: Deductibles, limits, sublimits, and exclusions turn the ground-up loss (total physical damage) into the insured loss a carrier would actually pay.

Where do the events come from? Reliable hurricane records go back about a century, and most regions have seen only a few major earthquakes. That history is far too thin to price a 1-in-250-year storm. Natural catastrophe modeling fills the gap with probabilistic event catalogs: tens of thousands of simulated years of plausible events, built from scientific research and past observations. These catalogs include storms that haven't happened yet but realistically could.

Model Inputs and Outputs

Every model run starts with exposure data, and this is where brokers have the most influence. The model needs a geocoded location (ideally at the rooftop or parcel level), physical details such as construction type, occupancy, year built, and number of stories, and the financial terms of coverage. Any blank field gets filled with a default assumption, and those defaults are usually conservative. Clean, complete submissions give the model less to guess at, which is why many teams now use insurance broker software to gather and validate property data before it goes to market.

Models then run in one of two modes:

• Deterministic modeling asks what happens if a specific event strikes, for example, a Category 4 hurricane making landfall in Tampa.
• Probabilistic modeling runs the full catalog and returns a range of outcomes, each with its likelihood.

Carriers usually price from probabilistic results and use deterministic scenarios to stress-test accumulations.

The table below lists the output metrics you'll see most often in quotes and model reports, along with what each one means for your client.

AAL tells you what an average year costs. PML and TVaR tell you how bad a very bad year could get. Carriers weigh both, so a small change in submitted data, such as a corrected roof age or a more precise geocode, can shift a quote noticeably. That makes data quality one of the most practical levers you have when advocating for a client.

Who Uses NatCat Modeling and Why

Model results travel further than most people expect. The loss estimate that shapes your client's quote also feeds into reinsurance purchasing, capital planning, and credit ratings. Knowing who reads these figures explains why carriers push so hard for detailed, accurate submissions.

Key Users Across the Insurance Value Chain

Insurers and reinsurers are the heaviest users. They run NatCat models to decide which risks to write, what to charge, and how much reinsurance to buy so one bad season doesn't drain their balance sheet.

Underwriting is where brokers feel things the most. The modeled average annual loss (AAL) anchors the technical premium, while probable maximum loss (PML) figures shape the limits, sublimits, and deductibles a carrier is willing to offer. Two carriers can model the same schedule of values and reach different conclusions, which is one reason marketing a property account widely often pays off. The quality of the underlying exposure data plays a big part in how far apart those results land.

Brokers and risk managers use model output to gauge a client's overall cat risk, compare program structures, and support their position in negotiations. Rating agencies review modeled cat losses when assessing a carrier's financial strength. Regulators pay attention as well. For example, Europe's Solvency II framework requires insurers to hold capital against catastrophe risk, and US regulators factor cat exposure into their own capital requirements.

Here's a straightforward process many brokers follow to turn natural catastrophe modeling output into leverage:

1. Request the data: Ask each carrier or your firm's analytics team for modeled AAL and PML by peril, and by location where possible.
2. Compare year over year: Line up the results against last year's figures to spot large jumps that need an explanation.
3. Find the loss drivers: Identify the handful of locations responsible for most of the modeled loss, since a few coastal or quake-zone sites often dominate.
4. Translate for your client: Explain in plain language what the numbers mean for premium, limits, and deductibles.
5. Test alternatives: Use what you've learned to model options, such as higher wind deductibles or separate quake limits, before going back to market.

Working through these steps turns a model report into a negotiating tool, so you can walk into carrier conversations with evidence instead of guesswork.

Why NatCat Modeling Matters Now

Catastrophe losses keep rising, and the pressure shows up in nearly every property renewal. Swiss Re's research tracks both growing insured losses and a wide protection gap (the share of economic losses from natural disasters that goes uninsured). It put global insured natural catastrophe losses at $107 billion in 2025, below the roughly $140 billion implied by the long-term trend. It attributes the below-trend year to favorable variability rather than any easing of underlying risk, and if 2026 losses follow the long-term average, they would total $148 billion.

So what's driving the long-term rise? Climate change is shifting the frequency and severity of some perils, especially severe convective storms, floods, and wildfires. Meanwhile, more people and buildings continue to move into coastal areas, floodplains, and the wildland-urban interface. Property values and rebuilding costs have climbed too, so each event hits a larger pool of insured assets.

That mix makes NatCat modeling more important with every renewal cycle. When carriers feel uncertain about a risk, they add a pricing cushion or cut back capacity. A clear, well-supported model result is often what keeps your client's program placed on reasonable terms.

cta-inline-card

Why Data Quality Determines NatCat Model Accuracy

Every metric in the table above rests on one input you control directly: the property data in the submission.

Garbage In, Garbage Out

A NatCat model never sees your client's building; it only knows what the statement of values tells it. When a field is missing or vague, the model plugs in a default, and those defaults lean conservative. Unknown construction is often treated as weaker construction, and a missing year built may be read as older building stock. The outcome is a loss estimate that looks worse than the actual risk.

These are the gaps that cause the most trouble on submissions:

• Inaccurate geocoding: A building placed at a ZIP code centroid instead of its rooftop can end up miles from its true distance to the coast or a fault line.
• Incomplete construction and occupancy details: The model falls back on a generic damage curve that doesn't reflect how the building will actually perform in a storm or earthquake.
• Stale replacement-cost values: Outdated figures understate or overstate what's at risk, which skews both AAL and PML.
• Blank hazard fields: Missing flood zone or distance-to-coast data leaves the model guessing about the exact exposure it's supposed to measure.

Modelers run into this pattern constantly. When one property's results look off, the cause usually traces back to the schedule, and the science inside the model is rarely at fault.

How Better Property Data Improves Modeling Outcomes

Model-ready data shares three traits: accurate geocoding, complete COPE attributes (construction, occupancy, protection, and exposure), and populated hazard fields. With those in place, natural catastrophe modeling has far fewer blanks to fill with worst-case assumptions. The table below shows how an incomplete submission compares with a model-ready one across each data element.

For brokers, the payoff shows up in client and carrier conversations. Risk assessments match the real portfolio, fewer defaults inflate loss estimates, and underwriters spend less time sending questions back and forth.

That's why data preparation, validation, and enrichment belong upstream, before anything reaches a model. Many broker teams now rely on tools built specifically for this step. Archipelago's Agent, for example, reads SOVs alongside supporting files like roof inspections, seismic reports, and property condition assessments. It then fills gaps using geocoding, hazard data, construction codes, and structural engineering rules. It flags issues, shows how each fix changes the picture, and gives your whole team a single shared dataset to work from, with accounts typically processed in under 24 hours. See how Archipelago's Agent can prepare your model-ready property data.

Conclusion: Better NatCat Results Start with Better Data

If you've ever typed “what is NatCat” into a search bar after a rough renewal, you already know that the modeling side can seem like a black box. The mechanics are simpler than they look, though. A model throws thousands of simulated events at the buildings it has been told about, layers in the policy terms, and spits out the numbers carriers use to set price. Cat losses keep going up, so underwriters lean on those outputs harder every year, and they pad the margin anywhere the data leaves them guessing. Of everything in that chain, the submission is the piece you actually control.

Ahead of your next property renewal, pull your client's highest-value locations and go through them one by one. Look for exact coordinates, an accurate construction class, up-to-date replacement costs, and hazard fields that are filled in. Clean up those gaps before anything else. After that, ask your carriers how their natural catastrophe modeling results moved compared to last year. That's the conversation where a tidier schedule starts to show up for your client through better premiums, limits, and terms.

FAQs

How often do insurers update their NatCat models?

Major vendors typically release updated model versions every few years to reflect new science, claims data, and building code changes. Carriers may also adjust their internal view of risk annually, which is one reason why modeled results can shift at renewal even when a portfolio hasn't changed.

Can a broker challenge a carrier's catastrophe model results?

Yes. Brokers can request location-level modeled losses, point out incorrect or defaulted data, and provide supporting documents such as roof inspections or seismic reports to justify a lower loss estimate.

Why do different carriers produce different NatCat loss estimates for the same property?

Carriers may license different vendor models, run different model versions, or apply their own adjustments for factors like demand surge and secondary perils. How each carrier handles missing data also varies, which can widen the gap between quotes.

What property details have the biggest impact on catastrophe modeling results?

Precise location, construction class, year built, and roof characteristics tend to move modeled losses the most, especially for wind and earthquake. Secondary modifiers like roof covering, roof anchorage, and opening protection can further reduce estimates when documented.

Does a lower modeled loss guarantee a lower property insurance premium?

Not always, since premiums also reflect market capacity, reinsurance costs, loss history, and each carrier's appetite. However, a well-supported NatCat result gives underwriters less reason to add pricing loads for uncertainty.