- NOAA forecast a below-normal 2026 Atlantic season, but the seasonal count is the wrong number to plan against. Your business faces the one storm that reaches it, not the basin average.
- The right tool is expected annual loss: add up, across every severity of storm, its annual probability times the damage it would cause. We work it in full below.
- For our hypothetical Montego Bay hotel with US$5 million at risk, the expected annual loss comes to US$160,000, and the rare severe storms drive most of it.
- A return period is just one over the annual probability. A 1-in-50-year storm still has about an 18 percent chance of striking within a 10-year window.
- Anchor your continuity budget, your insurance and your parametric cover to that expected-loss figure, not to how the last season felt.
NOAA issued its 2026 Atlantic outlook in May, forecasting a below-normal season: 8 to 14 named storms, 3 to 6 hurricanes and 1 to 3 major hurricanes, against a 30-year average of 14, 7 and 3. The odds it gave were 55 percent below-normal, 35 percent near-normal, 10 percent above. A business owner reads that and relaxes a little. That instinct is the expensive one, because the seasonal count describes the whole Atlantic, and no business trades against the whole Atlantic. It trades against the single storm that happens to arrive on its coast, and the season that number belongs to tells you almost nothing about that.
The proof is one year old. The 2025 season produced a normal number of storms and one of them was Hurricane Melissa, a Category 5 that made landfall on 28 October and that the World Bank and the Inter-American Development Bank later priced at US$8.8 billion in damage to Jamaica, equal to 41 percent of the country's 2024 GDP and the costliest storm in its recorded history. An ordinary season, a catastrophic outcome. If you want to plan a business around storm risk, you have to stop reading the season and start doing the arithmetic on your own exposure. Here is exactly how.
Step One: Expected Annual Loss
The number that actually governs a risk decision is the expected annual loss, sometimes called average annual loss. It answers a single question: if you could run this year thousands of times, what would the storm damage cost you on average per year? You build it from severity scenarios. For each one, take the annual probability it happens and multiply by the loss it would cause you. Then add the scenarios together. That is the whole method.
Take a hotel in Montego Bay with US$5 million of insurable value in building and contents. The four scenarios below, their probabilities and their damage ratios are illustrative inputs you would replace with your own location's hazard data and your own loss estimates. The method does not change when the numbers do.
| Scenario | Annual probability | Damage | Loss | Probability × Loss |
|---|---|---|---|---|
| Tropical storm / minor | 20% | 2% of value | $100,000 | $20,000 |
| Category 1–2 hurricane | 6% | 15% of value | $750,000 | $45,000 |
| Major hurricane (Cat 3+) | 2% | 50% of value | $2,500,000 | $50,000 |
| Catastrophic (Melissa-scale) | 1% | 90% of value | $4,500,000 | $45,000 |
| Expected annual loss | $160,000 | |||
The expected annual loss, in one line
EAL = (0.20 × $100,000) + (0.06 × $750,000) + (0.02 × $2,500,000) + (0.01 × $4,500,000) = $20,000 + $45,000 + $50,000 + $45,000 = $160,000 per yearUS$160,000 is the fair annual price of this hotel's storm risk. It is the anchor for every decision that follows: how much to spend on mitigation, how much cover to buy, and what a fair premium looks like.
The Rare Storms Own The Risk
Look again at the last column. The minor event has a 20 percent annual chance, ten times more likely than the major hurricane, yet it contributes only US$20,000 to the expected loss. The major and catastrophic scenarios together are responsible for US$95,000, nearly 60 percent of the total, despite a combined annual probability of just 3 percent. This is the pattern that trips up intuition. People plan for the events they see often and underprice the ones they rarely see, and in storm risk the rare events carry most of the money.
Where The Expected Loss Comes From
Each scenario's share of the US$160,000 expected annual loss. The two rarest events drive most of it.
Illustrative model. Shares of a US$160,000 expected annual loss for a US$5m-value property.
Step Two: Return Periods And The 10-Year Window
A return period is nothing more than one divided by an annual probability. The major hurricane at 2 percent a year is a 1-in-50-year event. The catastrophic one at 1 percent is 1-in-100. The language misleads people into hearing "we are safe for decades," which is not what it means at all. Each year is an independent roll of the dice. What a business should actually ask is: over the years I plan to be open, how likely is at least one of these?
The formula is one minus the chance of it never happening. For an event with annual probability p over n years, the probability of at least one occurrence is 1 minus (1 minus p) to the power n. Run it for the major-hurricane band over ten years.
Chance of at least one major-or-worse storm in 10 years
Combined annual probability of a major or catastrophic event: 2% + 1% = 3%.
P(at least one in 10 years) = 1 − (1 − 0.03)^10 = 1 − 0.97^10 = 1 − 0.737 = 0.263, about 26%A "3 percent a year" risk is a roughly one-in-four risk across a decade. The single major hurricane on its own (2% a year, the 1-in-50 event) works out at about 18 percent over ten years. Rare per year is common per business lifetime.
The Exceedance-Probability Curve
Insurers and risk modellers rarely stop at a single expected-loss number. They draw the whole shape of the risk as an exceedance-probability curve: for every possible loss size, the annual probability of a loss at least that large. It is the clearest single picture of an exposure, because it shows both the everyday small losses and the tail that can end a business. For our hotel, adding the scenario probabilities from the top down gives the curve below.
Exceedance-Probability Curve, Montego Bay Hotel
Annual probability of a storm loss at least as large as the value on the horizontal axis.
Illustrative, built from the four-scenario model above. The long low tail on the right is the storm that closes a business.
Read the curve from left to right. Small losses are common: nearly a 29 percent chance in any year of at least US$100,000 of damage. Large losses are rare but never zero: a 1 percent annual chance of a US$4.5 million loss that would take out most of the building. The flat tail on the right is the part that matters most and gets managed least, because it is the loss that does not just hurt the year, it can end the business.
Jamaican coastline. Photo via Unsplash.
What The Number Tells You To Do
Once you have US$160,000 as an expected annual loss and a curve with a US$4.5 million tail, the decisions get concrete. Spending on mitigation and insurance that together cost meaningfully less than US$160,000 a year while removing most of that tail is a good trade. Paying far more than the expected loss for cover that only trims the common small losses is not. The maths converts "how worried should we be" into a budget line a board can approve or reject on its merits.
It also reframes insurance. A premium always sits above the pure expected loss, because the insurer has to fund the capital it holds against bad years, its reinsurance, and its costs. After Melissa, that gap widened sharply: Caribbean property premiums rose 15 to 25 percent in storm-exposed areas, and Jamaica's general insurance industry profit fell from J$2.6 billion in 2024 to J$30 million in 2025. Knowing your own expected loss is what lets you look at a renewal quote and judge whether it is fair, or whether hardening the roof would cut the risk more cheaply than paying the higher premium every year.
The one exposure the property model above leaves out is the one that often does the most damage: business interruption. A closed hotel loses revenue every day the doors are shut, and that loss is not in the US$5 million of building value at all. This is where parametric cover has become the sharp tool in the region. It pays automatically when a measured trigger is hit, wind speed or rainfall at a defined point, rather than waiting on a loss adjuster. Jamaica has moved hard in this direction: the Office of Utilities Regulation approved a US$106.6 million parametric package for Jamaica Public Service ahead of the 2026 season, and GK General Insurance's GK Weather Protect now covers roughly 5,000 farmers with automatic weather-triggered payouts. Cash in days, when continuity depends on it, is worth more than a larger cheque that arrives after you have already missed payroll.
Model Your Own Storm Exposure
Give us your location, your insured values and your revenue-per-day-closed, and we will build the expected-loss and exceedance-probability model for your business, then use it to size continuity, insurance and parametric options. It is the same maths above, run on your numbers.
Explore Our Data Science Service ↗Frequently Asked Questions
What is expected annual loss and how do you calculate it?
Expected annual loss, also called average annual loss, is the sum across every damage scenario of its annual probability multiplied by the loss it would cause. Multiply each scenario's probability by its loss and add them up. It is the fair annual price of the risk you carry, and it tells you roughly what to be willing to spend each year on mitigation and insurance combined.
What is a return period and what does a 1-in-50-year storm actually mean?
A return period is one divided by the annual probability of an event. A 1-in-50-year storm has a 2 percent chance in any single year. It does not mean the last one was 50 years ago or that you are safe until then. Over 10 years the chance of at least one is 1 minus 0.98 to the power 10, about 18 percent.
Why is a below-normal hurricane season still dangerous?
A seasonal forecast counts storms across the whole Atlantic basin and says nothing about which storm makes landfall on your parish. NOAA forecast a below-normal 2026 season, and the 2025 season produced a normal count yet included Hurricane Melissa, a Category 5 the World Bank and IDB priced at US$8.8 billion of damage to Jamaica. Your business experiences the one storm that reaches it, not the average.
How much should a business spend on hurricane protection?
Anchor it to your expected annual loss. If that figure is around US$160,000, spending less than it on mitigation and insurance that removes most of the tail risk is rational, and spending far more for marginal protection is not. The maths turns an emotional question into a defensible budget line.
Why is an insurance premium higher than the expected loss?
An insurer covers its expected payouts plus the cost of capital held against bad years, reinsurance and administration, so the premium sits above the pure expected loss. After Melissa, Caribbean property premiums rose 15 to 25 percent in exposed areas. Knowing your own expected loss lets you judge whether a quote is fair or whether mitigation is cheaper.
What is parametric insurance and why does it matter for storm risk?
Parametric insurance pays automatically when a measured trigger is hit, such as a wind speed at a defined location, rather than after a slow claims process. Jamaica approved a US$106.6 million parametric package for Jamaica Public Service ahead of the 2026 season, and GK Weather Protect covers roughly 5,000 farmers. For a business, the value is cash arriving in days, when continuity depends on it.
What data does a business need to model its own hurricane risk?
You need the insurable value of property and stock, the loss each severity would cause as a percentage of that value, the annual probability of each severity for your location, and your interruption exposure in revenue lost per day closed. StarApple Analytics builds this from parish-level hazard data and your own financials.
About StarApple Analytics
StarApple Analytics is Jamaica's leading data science, business intelligence and market research company, a subsidiary of StarApple AI, the first AI company in the Caribbean, established by Adrian Dunkley in Kingston in 2023. We build expected-loss and climate-risk models for Caribbean businesses through our data science service, and run training with certificates for teams that want the skill in-house. Contact us at insights@starapple.ai.
Related reading across the Caribbean AI network