How Much Does an Hour of Rain Downtime Cost Your Operation?
Rain without flooding and without a casualty already costs money every month, only the bill never arrives under that name.
A maintenance job scheduled for an outdoor area without a reliable forecast hires a contractor and mobilizes equipment for a service that rain interrupts halfway through. Installing an industrial awning stalls because the safety standard does not allow work on roofs and coverings in the rain. At a steel plant, a chemical plant or a food and beverage line, the scheduled intervention on outdoor equipment enters the same queue and gets no date to resume. The work front stops with no one knowing when it will restart, because no schedule treats forecast rain as a constraint, only as an unforeseen event after the fact. On the highway, wet pavement changes braking behavior and delays the decision to reduce speed or close a lane before the first car stops, and the paving improvement project itself stalls mid-deadline.
The market has already measured this, with numbers and sources. In the United States, common rain, not storms, increases highway crash risk by 38% on average across 16 years of Texas data, with a minimum of 32% in the worst year, and the effect is larger on high-speed highways than on urban streets, according to a study published in 2025 in Scientific Reports. Still in the United States, 77% of weather-related traffic crashes happen in rain or drizzle, not snow or fog, according to the Federal Highway Administration. In construction, light rain, less than 4 mm in 12 hours, already reduces crew productivity by up to 40% in a concrete pour operation, according to a study published in 2025 in Applied Sciences. The same holds for outdoor maintenance: the same light rain, the same exposed crew, the same drop in output.
The bill that rain leaves at the end of the month rarely carries the event's name. It arrives as a contractor hired for a service that could not be done, a concrete slab that had to be redone, fewer delivery riders on the street. Different line items, different owners, and almost never added up as what they were: predictable rain that no one turned into a decision.
Without timely warning about the exact point, it charges on two fronts:
Operations: what was scheduled for that window stops happening.
Budget: every hour of that stoppage enters the month's cost, with shift, contractor and equipment already paid for.
In this article
The danger of rain: why knowing how long it rains matters more than knowing whether it will rain.
How to turn the danger of rain into calculated risk per asset.
What the same rain impacts in each sector, starting with the table that sums it all up.
Industry: the outdoor maintenance that the safety standard does not allow to happen.
Highways: common rain already raises crash risk before the first car stops.
Construction: the predictable rain that the schedule treats as chance.
Mobility and on-demand delivery: when the whole business feels rain in real time.
The return on investment from anticipating the danger.
The method: three steps to apply starting tomorrow.
Where this becomes a daily decision.
The danger of rain: predicting how long it lasts matters more than predicting whether it will fall
Rain happens when air loaded with vapor rises, cools and can no longer hold the water it carries. What pushes that air upward varies: a cold front, heat that organizes convection, a tropical wave, a low-level current, terrain that forces the air to climb. Rain does not have one fixed mechanism per region: the same operation receives rain from different physical origins in different weeks of the year, and what decides the operation is the window, not the mechanism.
A phrase circulates in operations, with a grain of truth: "it's impossible to predict rain." To this day, numerical modeling has real difficulty predicting the exact occurrence of rain at a specific point. It is good at flagging, days in advance, that a moisture-loaded system is forming. It is poor at pinning down what time it arrives, how long it stays and how strong it gets, especially for convective rain, which forms and moves on a scale of minutes. But the phrase misses the point: the operation does not need to predict rain perfectly, it needs the best decision given the chance of prolonged hours of rain.
That is why, for the nearest horizon, from 15 minutes to 4 hours, i4sea complements the numerical model with radar nowcasting: instead of only running the simulation, the reading tracks in real time where rain is already falling and where the cell is moving, and updates the window as it moves. The numerical model gives days of warning. The radar pins down the hour.
A 20-minute rain and a 6-hour rain call for opposite decisions: one is worth waiting out, the other is worth rescheduling the whole day. To tell the two apart, the reading identifies the physical type of the system that is falling, because each type has a different typical duration and path, and that identification is what turns "it's going to rain" into a window with a start, duration and intensity.
How to turn the danger of rain into calculated risk per asset
Two operations receive the same rain. One knows how to distinguish, within the same window, the different possibilities for how that rain will behave, and reschedules the shift with that reading in hand. The other decides based on the rain probability from a free app or the forecast from most meteorologists, including on television: a grid calculation, how many squares of the model show rain divided by the total squares in the region, weighted by the model's confidence that day. The problem is that no one discloses which grid square covers your asset or what confidence was used in the calculation, so the same 60% chance of rain can mean different things at different points and on different days.
Rain is the danger. Loss is the risk. Between the two sit factors that multiply or reduce the impact, and those factors decide whether the same 5 mm fall on a rural area with nothing exposed or on the only maintenance front that was open that morning.
Public forecasting does not make that conversion, and it fails on three points at the same time.
Scale. Public forecasting operates at a resolution of about 25 km. At that scale, the model cannot see the maintenance line, the mountain stretch or the work front. It describes a state of the region.
Format. "Chance of rain in the region" contains no decision, and it is the same grid probability used in free apps and in most television forecasts, without disclosing the square or the confidence behind the calculation. What is missing is the threshold that changes the protocol, the window in which it is crossed, the matching action and the name of who carries it out.
Measurement. Almost no one measures accuracy, and those who do usually measure whether it rained, not whether the decision was the right one. The scorecard lives in the memory of whoever remembers the time the forecast failed, and memory keeps the error far sharper than the hit.
The second and third points cost the most. The second, because not knowing that the rain probability you use does not mean it will rain on your asset creates misalignment and a lack of trust in the forecast itself. The third, because it blocks learning: without a record of forecast, decision and outcome, there is no accuracy rate of your own, and without an accuracy rate, confidence turns into opinion, and opinion does not change operational protocol.
i4sea follows three layers to make that conversion.
Layer 1, the data. The foundation is a proprietary numerical model with 1 to 3 km resolution for all of Latin America, with oceanographic grids of 10 m to 500 m and a proprietary reanalysis of more than 10 years per location. Where public forecasting describes the region, i4sea already starts from your asset, your stretch of road or your work front. For the nearest horizon, radar nowcasting corrects the window in real time.
Layer 2, interaction with the terrain and the business. The same millimeters do not produce the same result at two different points. The reading crosses the danger with:
Prior rain and soil saturation: the same 40 mm that runs off on dry soil soaks a slope that already received rain the day before.
Land use, slope and drainage: the real geography of that point, and its capacity to drain what it receives.
Intensity versus duration: 40 mm in one hour and 40 mm in twelve hours hit drainage systems, slopes and schedules in completely different ways.
What depends on the point hit: the more people, assets and revenue pass through that specific point, the higher the risk, even with an identical climate danger.
Layer 3, the risk matrix. i4sea connects the threat index, the intensity and the window at the exact point to the impact on that specific business, and turns "it's going to rain in the region" into a risk reading per asset, ready for decision.
These three layers support the three gains the operation feels: lower cost, crews allocated where the risk is, and planning locked in before the event.
What the same rain impacts in each sector
The physics stays the same. What gets wet, the indicator that suffers and who signs off change completely.
Table 1 · What rain charges by sector
Industry. What rain interrupts: Outdoor maintenance, anticorrosive painting, rail welding, industrial roofing. Where the bill shows up: Postponed intervention, wasted mobilization, rework
Highways. What rain interrupts: Reduced-grip pavement, rescue on a risk route, paving work. Where the bill shows up: Crash cost, rescue mobilization, driver alerts, lost work day
Construction. What rain interrupts: Concrete pours, earthmoving, trench excavation, lifting. Where the bill shows up: Day of delay, rework, deadline extension, accident with leave
Mobility and on-demand delivery. What rain interrupts: Fewer delivery riders on the street, ride-hailing and app order flow, shared bike and scooter use. Where the bill shows up: Higher cost, demand drop, compromised deliveries, customer satisfaction
Industry: the outdoor maintenance the safety standard does not allow to happen
At a steel plant, a chemical plant or a food and beverage line, a good part of routine maintenance happens outside the building: anticorrosive painting of metal structures, roof repair, intervention on exposed equipment. Rain is more than discomfort for this maintenance. On at least two points, it is a prohibition.
The Brazilian standard NR-18 (item 18.7.8.2, clause "c") bans work on roofs or coverings "in rain, strong winds or adverse weather conditions." The Brazilian standard NR-35 (item 35.5.5.1, clause "d") requires that the Risk Analysis for any work at height consider "adverse weather conditions" before clearing the crew. Neither standard sets a millimeter threshold, but both turn rain into a legal stop criterion, not slack in the schedule.
Anticorrosive painting is the most recurring example. N-13, the Brazilian CONTEC standard used in steel, petrochemical and practically all of Brazil's heavy industry for industrial painting, explicitly bans it: "no paint application shall be made in rain, fog or mist, nor when relative humidity exceeds 85%." Exposed metal structure, tank, roofing: the maintenance that protects all of it against corrosion only happens within that window, and the window closes with rain.
Steel and chemical plants: anticorrosive painting of structures and tanks, stopped by N-13's 85% humidity rule.
Food and beverage: maintenance of outdoor storage roofing and structure, under the same NR-18 ban on roofs.
Industrial awnings: installation and maintenance of tensioned covering, the same NR-18 provision, applied to the letter.
Railways. Track maintenance also has a written rule. Rumo's operating procedure for thermite rail welding bans starting the job with rain forecast, and if rain arrives mid-weld, it requires covering the rail with a tarp and a temporary splint. A weld that gets wet is logged as a defect and must be replaced within 7 days. Each weld lost to rain means more than one lost service day: it becomes scheduled rework.
One caveat. No public source consolidates a number for how many days of industrial maintenance, Capex or Opex, Brazil loses to common rain each year. What exists is formal recognition of the problem. In the Brazilian power sector, ONS, in Submódulo 6.5, authorizes the utility to cancel scheduled maintenance, without penalty, due to adverse weather conditions. In the rest of industry, what is missing is turning that into a calculated decision, with a window and an action.
What an accurate forecast delivers here is a binary decision, made in advance: paint today or reschedule the crew for tomorrow, weld now or wait for the dry window, close the covering before the front arrives instead of after it passes. The mobilized crew enters the right window, instead of being canceled at the last minute.
Highways: common rain already raises crash risk before the first car stops
On the highway, wet pavement changes braking behavior, reduces tire grip on the asphalt and increases stopping distance, and the effect shows up in crash statistics well before any accumulation that would justify closing the road. Common rain, not storms, already raises this risk, as the numbers at the opening of this article show. In Germany, the same mechanism appears with an even larger number: summer rain on divided highways with speeds above 130 km/h increases the relative risk of a single-vehicle crash by 536%, according to a 2022 study published in the European Transport Research Review, based on data from 4.7 million crashes across 401 German districts between 2006 and 2017.
In Brazil, that cross-reference has not been published yet. The open microdata from the Federal Highway Police (PRF) already records the weather condition of each crash, but no official body, PRF, ANTT, DNIT or CNT, has published to date the national percentage of crashes caused by rain or wet pavement. It is a real market gap, and i4sea is positioned to be the first to cross that open data and publish the missing number.
Anticipating that window changes two decisions for the highway concession at the same time. The first is mobilization: knowing in advance which stretch and what time the risk rises lets the operator position tow trucks, ambulances and rescue crews closer to the point with the highest chance of occurrence, instead of dispatching after the crash has already happened. The second is the driver alert: variable message signs, the concession's app and radio can warn about wet pavement risk before the first car stops on the lane, not after.
Rain also delays the highway before it is finished. Brazilian paving and improvement projects run on their own schedule, and rain in that schedule is not crash risk, it is a lost service day. DNIT (Brazil's federal highway agency) has already rescheduled a paving project due to rain on BR-280, in Santa Catarina. The BR-381 duplication in the Serra da Fernão Dias remains stalled, and ANTT (Brazil's federal highway regulator) cites the end of the rainy season among the conditions to resume. A repaving project in Vitória da Conquista, with an original deadline of 20 days, had to be suspended until the rain let up. Public works auditing already has a methodology for this: in a case reviewed by IBRAOP (Brazil's public works auditing body), a 90-day deadline extension request was reduced to 19 days after comparing the period's rainfall to the region's climatological norm, and another request, for 30 days, was denied because rainfall came in below the historical average. The decision mechanism is the same as in industrial maintenance: reallocate the crew to the dry stretch, instead of discovering on the day that rain will disrupt the service. It also accounts, in an auditable way, for the schedule according to the chances of forecast rain.
What is missing here is not data, it is application: hyperlocal forecasting by stretch, instead of a regional bulletin, with the alert reaching the concession's control tower and the project schedule at the same time, with stretch, window and action.
Construction: the predictable rain that the schedule treats as chance
On a construction site, anticipation shows up directly in the schedule and in the contractual claim. A day of delay on a large project costs between US$ 50,000 and US$ 150,000.
Rain on a construction site is treacherous because the damage is not always the lost day.
Concrete pours: take on water after placement and come back as rework.
Earthmoving: pools water and blocks the next front.
Trench slopes: give way when the soil is saturated, with crew underneath.
Lifting: stops, and the crane sits idle while still being rented.
This is where the most evident gap among the sectors in this post shows up. The most widespread body of method on the job site, Lean Construction and the Last Planner System, organizes constraints, sequencing, reliable commitment and percent plan complete. Forecast rain rarely shows up as a named constraint in that method. It enters later, in the causes report, labeled as unforeseen, and the job site ends up treating one of the schedule's most impactful variables as chance.
The fix is the same as in the other sectors, applied to the work front. Each critical front gets its own threshold, because concrete pours, earthmoving, foundations, lifting and coastal works have different tolerances. Each threshold gets a warning window and a defined action. And each decision enters the record, with timestamp and source data, which later supports the conversation about deadline extensions, contractual penalties and insurance claims.
Mobility and on-demand delivery: when the whole business feels rain in real time
On-demand mobility enters this post because it shows the scale of the problem in a way no other sector here does: in real time, without waiting for the month to close.
Ride-hailing trips rise with rain, and rise more for operators with dynamic pricing than for those without it. In New York, Uber rides increase by about 22% during rain hours, against 19% for Lyft and only 5% for traditional taxis, according to a study published in the Journal of Economic Behavior & Organization. In Brazil, ride-hailing fares have risen by as much as 70% on heavy rain days in São Paulo, according to a report based on official price data.
Delivery: a Chinese study published in a climatology journal confirms that rain has a statistically significant influence on delivery order volume, without publishing the isolated percentage.
Bikeshare: a survey of nearly 100 million trips across 40 bikeshare systems in 16 countries points to rain as the second most influential variable on use, behind only time of day. As far as the research found, there is no public study specific to Brazil.
Shared e-scooters: in Brisbane, Australia, 31.8% of 817,000 trips analyzed happened in wet weather against 68.2% in dry weather, a ratio of nearly 2 to 1, according to a study published in the Journal of Transport Geography.
The safety of people on the street enters the same account. In Brazil, 41.3% of app delivery workers surveyed in São Paulo and Rio have already had a work accident, and motorcycles were involved in 38.6% of the more than 34,000 traffic deaths recorded in the country in 2023. Neither source isolates the effect of rain specifically, but the logic is the same as on the highway: wet pavement changes risk, and app delivery workers are on it in any condition.
If apps worth billions of dollars in market value feel the effect of common rain on their own demand, the same effect on an industrial or infrastructure operation, measured in idle assets and lost shifts, is even more direct to prevent: the business does not depend on the customer's mood, it depends on a decision someone can already make today.
The return on investment from anticipating the danger
No system prevents rain. What changes is the cost of dealing with it. A real case shows the size of that delta, with the scope clearly stated.
Ultracargo (real case, Brazil). In the windstorm of July 29 and 30, 2026, the maintenance team rescheduled the preventive maintenance that had been set for the day with the highest wind-related accident risk, avoiding the accident and optimizing internal and contractor crew resources. The originating event was wind, not rain. The case appears here as a reference for the order of magnitude of the same mechanism, not as a rain measurement.
"i4sea's analyses, directed specifically at our business, change the game in how we deal with weather conditions, allowing that information to be used in a much more strategic way in the operation."
Carlos Gasparotto, Maintenance Manager, Ultracargo
NIST, in a survey on maintenance in American industry, measured 3.3 times more downtime among plants most dependent on reactive maintenance than among those that plan the intervention. The World Meteorological Organization (WMO), in "The Triple Dividends of Early Warning Systems and Climate Services" (2024), points to a return of about 9 to 1 on early warning systems and up to 30% less damage with 24 hours of lead time.
The third is the most direct on the value of forecasting. Molina and Rudik, in NBER Working Paper 32548 (2024), measured the effect of forecast accuracy improvements on US hurricane forecasts: the accumulated improvements since 2007 reduced total costs by 23%, equivalent to about US$ 2 billion per hurricane. The same authors measured the reverse: underestimating wind speed by 10 m/s, an error equivalent to up to two categories on the Saffir-Simpson scale, raised damages by about US$ 177 million and federal recovery spending by US$ 23 million per affected county. The hurricane happened all the same. What changed was the quality of the preparation.
It is the same principle as rain, on a smaller scale and with much greater frequency.
The method: three steps to apply starting tomorrow
These steps work with a spreadsheet, an in-house system or a vendor. None of them depend on hiring i4sea. Start with just one asset.
1. Write the threshold before the next event. Choose the asset that exposes you most, a maintenance line, a mountain stretch, a concrete pour front, and define the number that changes the decision: intensity in mm/h, accumulation over 24 h and 72 h, soil saturation. Next to each number, write the action and the name of who carries it out. Until the threshold is written down, every forecast turns into a debate.
2. Measure accuracy by the correctness of the decision. Open one line per event with four fields: what was forecast, which threshold was triggered, what was decided and what happened. In one quarter you have your own accuracy rate, by asset and by threshold. That is the argument that convinces the skeptic on your team, because it is that team's own track record, not a vendor's promise.
3. Deliver the alert where the decision happens. An alert that arrives in a daily report dies in the report. Put the alert on the channel the team already opens, WhatsApp, Teams, SMS or email, with the stretch or asset, the time window, the threshold crossed and the recommended action. Keep the record: it supports the contractual claim later and calibrates the threshold itself.
Treat lead time probabilistically, because each window calls for a different owner and action: accumulated rain usually opens a window of 24 to 72 hours, severe storms 1 to 6 hours, and rain already in motion 15 minutes to 4 hours.
The gain from these three steps is the same across every sector: the structurally lower cost of preparation replaces the cost of reaction. It is the Delta Framework that i4sea uses in its benefit studies, and it fits in one line: benefit equals the sum of the difference between unplanned cost and planned cost, multiplied by the number of events.
Where this becomes a daily decision
A threshold written into a procedure only protects the operation when the alert arrives in time, with the recommended action attached.
The difference shows up in the sentence that reaches the supervisor. A generic tool answers "conditions may deteriorate between Thursday and Friday." A response calibrated to the asset says "Thursday, 2pm to 8pm, 73% chance of rain, reschedule the structure painting to before 1pm." The first generates debate. The second generates a work order.
The AI Climate Agent delivers the ready decision to the team before the event: which asset, which window, which action. It already knows your operation's thresholds, protocols and history, arrives on the channel the team already uses, and logs in every response the source of the data behind it.
If you would rather start with the diagnostic, we map the climate dangers to your assets and show what the system would see today, with real data.
Request a free climate exposure diagnostic per asset!
Ask the Agent what your team asks on Monday:
Can we pour concrete on Thursday?
Does the mountain stretch close?
Can the structure painting be done today?
The hours of warning your operation had, and what it did with them, decide the cost of the next rain.
Sources
Common rain increases highway crash risk by 38% on average across 16 years of Texas data (minimum 32%), with a larger effect on high-speed highways than urban streets. Tafazzol et al., "Relative crash risk and road safety during rainfall in Texas from 2006 to 2021", Scientific Reports, 10/21/2025. nature.com
77% of weather-related traffic crashes in the United States happen in rain or drizzle, not snow or fog. FHWA (Federal Highway Administration), Road Weather Management, NHTSA data 2019-2023. ops.fhwa.dot.gov
Light rain, less than 4 mm in 12 hours, can reduce labor productivity by up to 40% in a concrete pour operation. Larsson, R.; Rudberg, M., Journal of Construction Engineering and Management, 2019, cited and reconfirmed by Schuldt et al., Sustainability, 2021 (mdpi.com) and Šopić et al., Applied Sciences, 2025 (mdpi.com). The parallel to outdoor maintenance is i4sea's reading of the same mechanism (crew exposed to the same light rain), not an extrapolation of the percentage.
Brazil, Inmet red alert for accumulated rainfall above 100 mm in one day in Rio Grande do Sul on 07/21/2026, from a cold front coming from Argentina. Extra Classe
Chile, Dirección Meteorológica de Chile alert at the Alarma level on 07/14/2026, for a frontal system associated with an atmospheric river: 130 to 180 mm in Coquimbo, 120 to 160 mm in Valparaíso and 100 to 150 mm in the Metropolitan Region. El Mostrador
Mexico, Servicio Meteorológico Nacional forecast of 75 to 150 mm in Chihuahua, Sinaloa, Durango and Nayarit, and 50 to 75 mm in Sonora, Jalisco and Michoacán, on 07/18/2026. El Informador
Europe, stationary system over Valencia, Cuenca, Albacete and Murcia on 10/29/2024; Turís (Valencia) with 772 mm in 24 hours, national record. Aemet, official report
NR-18, item 18.7.8.2, clause "c": bans work on roofs or coverings "in rain, strong winds or adverse weather conditions" (Brazilian standard). Official text, gov.br/trabalho-e-emprego (PDF)
NR-35, item 35.5.5.1, clause "d": the Risk Analysis for any work at height must consider "adverse weather conditions" (Brazilian standard). Official text, gov.br/trabalho-e-emprego (PDF)
N-13 (CONTEC/ABRACO), item 4.9.8.3: bans paint application in rain, fog or mist, or when relative humidity is above 85% (Brazilian technical standard). N-13 REV. K (PDF)
ONS, Submódulo 6.5, item 8.7.5: authorizes cancellation of scheduled maintenance, without penalty, due to adverse weather conditions (Brazilian power sector). Submódulo 6.5, rev. 2020.03 (PDF)
Rumo (Brazilian rail operator), operating procedure for thermite rail welding: bans starting a weld with rain forecast; a weld that gets wet is logged as a defect and must be replaced within 7 days. Official procedure (PDF)
Summer rain on divided highways above 130 km/h increases the relative risk of single-vehicle crashes by 536%, based on 4.7 million crashes across 401 German districts between 2006 and 2017. Becker, N.; Rust, H. W.; Ulbrich, U., European Transport Research Review, vol. 14, article 37, 2022. link.springer.com
DNIT rescheduled a paving project due to rain on BR-280 (Santa Catarina). ndmais.com.br
The BR-381 duplication project in Serra da Fernão Dias remains stalled; ANTT cites the end of the rainy season among the conditions to resume. atibaiahoje.com.br
A repaving project in Vitória da Conquista (Bahia), originally scheduled for 20 days, was suspended until the rain let up. emurc.com.br
Brazilian public works audit methodology using rain as a criterion: a 90-day deadline extension request reduced to 19 days, and another 30-day request denied, after comparing period rainfall to the region's climatological norm. IBRAOP, 14th National Symposium on Public Works Auditing, 2011. Official PDF
Uber rides increase by about 22% during rain hours, Lyft by about 19%, traditional taxis by about 5%. Brodeur, A.; Nield, K., "An empirical analysis of taxi, Lyft and Uber rides: Evidence from weather shocks in NYC", Journal of Economic Behavior & Organization, vol. 152, 2018. ideas.repec.org
Ride-hailing fares rose by as much as 70% on heavy rain days in São Paulo (Brazilian data). SBT News
Rain has a statistically significant influence on online delivery sales, without an isolated percentage published in the abstract. Wang, Y.; Zhao, H.; Liu, L., Theoretical and Applied Climatology, vol. 153, 2023. scholar.hit.edu.cn
Rain is the second most influential variable on bikeshare use, behind only time of day; nearly 100 million trips across 40 systems in 16 countries. Bean, R.; Pojani, D.; Corcoran, J., Journal of Transport Geography, vol. 95, 2021. ideas.repec.org
In Brisbane, Australia, 31.8% of 817,650 shared e-scooter trips analyzed occurred in wet weather versus 68.2% in dry weather (ratio 1.9). Kimpton, A. et al., Journal of Transport Geography, vol. 104, 2022. Open-access author PDF
41.3% of app delivery workers surveyed in São Paulo and Rio have already had a work accident; 38.6% of Brazil's 2023 traffic deaths involved a motorcycle (Brazilian data). CUT, citing research from the NGO Ação da Cidadania and the 2025 Atlas da Violência. cut.org.br
3.3 times more downtime among plants most dependent on reactive maintenance (13.0% versus 4.0% downtime). Thomas, D. S.; Weiss, B. A., Economics of Manufacturing Machinery Maintenance, NIST AMS 100-34, June 2020. Official NIST PDF
A return of roughly 9 to 1 on early warning systems, and up to 30% less damage with 24 hours of lead time. Liu, E.; Kull, D.; Chaponda, M., The Triple Dividends of Early Warning Systems and Climate Services, WMO, 10/30/2024. wmo.int
US hurricane forecast improvements since 2007 reduced total costs by 23%, about US$ 2 billion per hurricane; underestimating wind speed by 10 m/s raised county-level damages by about US$ 177 million and federal recovery spending by US$ 23 million. Molina, R.; Rudik, I., "The Social Value of Hurricane Forecasts", NBER Working Paper 32548, 2024. nber.org/papers/w32548