The 2026 Atlantic hurricane season starts June 1, and while early outlooks suggest that a developing El Niño might result in a tamer season than in the past few years, with below-average hurricane activity, all it takes is one big storm hitting a populated area to make it a bad hurricane season.

Every year, Americans rely on accurate forecasts when hurricanes might be developing to know when to stock up on supplies, prepare for power outages or evacuate.

Those forecasts have improved dramatically in recent decades, but the improvements can’t be taken for granted. Over the past year, federal funding cuts and job losses in the very programs that are helping make Americans safer from extreme weather threaten to stall progress and stretch forecasting resources to the breaking point.

I am an atmospheric scientist whose research focuses on hurricanes, including how and why they intensify or weaken. I also work with scientists at the National Oceanic and Atmospheric Administration, NOAA, to analyze observations collected by reconnaissance aircraft and evaluate computer model forecasts of hurricanes.

Here’s what forecasters rely on during hurricane season and why investing in science, forecasting technologies and the people who run them matters.

Flying through hurricanes

To have the best chance of an accurate hurricane forecast, computer models and meteorologists need to know about the location, intensity and structure of a hurricane, along with the environment that surrounds it. Satellites are crucial for tracking storms from above, but many details can be collected only inside the storm, where satellites can’t see.

That’s why NOAA relies on “hurricane hunters” – a group of skilled pilots and scientists who fly through storms all season long to collect storm data, which is quickly transmitted to forecasters and computer models.

When storms are developing, the U.S. Air Force Reserve and NOAA conduct several hurricane hunter flights per day to provide the most up-to-date storm information. During these missions, the crews often fly directly into the storm, through screaming winds and heavy rain, to release instrument packages called dropsondes.

The dropsonde is a feat of science and engineering, able to accurately measure the temperature, humidity, wind and pressure in hostile conditions. This data is radioed back to the aircraft. From there, it is processed and transmitted to NOAA, where forecasters analyze it and computer models use it to initialize forecasts.

I and many hurricane scientists have used dropsonde data collected over the years to build a better understanding of how hurricanes behave. A recent study showed that computer model forecasts of hurricane tracks were up to 24% more accurate when they included dropsonde data than those that didn’t.

Simulating hurricanes

A big reason hurricane forecasts have gotten better has been federal investments in computer models that can simulate these storms.

In 2008 the U.S. government funded the NOAA Hurricane Forecast Improvement Project, leading to substantial advancements in computer modeling and forecast accuracy. Computer models got better at incorporating the observations gathered by aircraft, showing air movements and rain bands in greater detail.

The flagship NOAA hurricane model is now the Hurricane Analysis and Forecast System, which does a better job of predicting rapid intensification, among other things, than its predecessors.

When storms rapidly intensify, as several have done in recent years, they can pose an acute risk to coastal communities. More accurate forecasts give people and communities better information to decide how to prepare and when they need to evacuate. Improvements since 2007 have resulted in an estimated US$2 billion in savings per hurricane landfall and many lives saved.

That’s a huge return on investment. In 2024, NOAA’s entire budget was $6.7 billion.

Keeping an eye on the storms ahead

There are some exciting developments ahead in hurricane observations and modeling.

NOAA in 2024 ordered two new aircraft, expected to be delivered by 2030, to begin replacing its aging hurricane hunter fleet so fights and their data collection can continue.

Private companies working with NOAA have deployed and tested autonomous drones – both in the air and sail drones on the ocean surface – that can collect data in areas where quality observations are hard to get.

Additionally, artificial intelligence weather models have emerged, such as Google DeepMind, which made a big splash as the most accurate forecast model of the 2025 hurricane season.

Some lingering dark clouds

Despite these promising developments, a different storm is eroding the bedrock upon which the national weather forecast enterprise sits.

Cuts in funding and staffing have stressed NOAA’s ability to collect critical observations. Last year, retired NOAA scientists volunteered to staff hurricane hunter reconnaissance flights so the missions could still be flown.

The Trump administration proposed cutting NOAA’s budget by more than a quarter, including dismantling its Office of Oceanic and Atmospheric Research. Congress rejected many of the administration’s proposed budget cuts, ultimately approving a $6.1 billion budget in March 2026, still down from the previous budget.

The National Center for Atmospheric Research, which led the development of computer models and dropsonde technology, has also been targeted by the Trump administration to be dismantled. The American Meteorological Society warns this decision “will harm meteorological research and innovation in the United States with severe consequences to current and future efforts of the weather enterprise to protect life, property, and the nation’s economy.”

I worry about the funding and staff cuts stressing systems that keep scientific progress marching forward and warn Americans about hazardous weather. Losing staff and support raises the risk of critical failures, such as delayed severe weather warnings and broken equipment causing new blind spots when storms threaten. In the long run, failing to invest risks stagnation or even reversing the hard-fought progress the U.S. has made in advancing weather prediction.

With coastal populations and development expanding over the past few decades, and storms becoming stronger, the vulnerability of the U.S. to costly, damaging hurricanes has increased dramatically. It is more important than ever that public investment in hurricane science and forecasting continue.

This article, originally published May 18, 2026, has been updated with NOAA’s 2026 Atlantic Hurricane Season outlook.

Brian Tang receives funding from the National Science Foundation, the National Aeronautics and Space Administration, and the Center for Western Weather and Water Extremes. He has research collaborations with the National Oceanic and Atmospheric Administration's Hurricane Research Division. He is a member of the American Meteorological Society.

Students who are Black, Hispanic, female, from low-income families or multilingual learners are less likely to be identified with autism in U.S. elementary schools than their white, male, higher-income or English-speaking peers. This finding comes from our new research, published in April 2026 in the academic journal Autism.

These disparities appear even among students who have similar levels of academic achievement and who are attending the same schools.

Our research shows there are big and recurring gaps in whether students are identified with having autism while they attend U.S. elementary schools. In both 2003 and 2019, for example, fourth grade female students were about 80% less likely to be identified with autism, as compared to similarly situated boys.

We found that for every 10 boys identified with autism, only about two girls in a comparable situation – including those displaying similar levels of reading achievement and attending the same schools – were identified.

We analyzed data repeatedly collected from 2003 to 2022, using large, nationally representative samples of about 160,000 fourth grade students participating in the National Assessment of Educational Progress.

We specifically looked at data that included student academic achievement. This approach let us consider potential bias in how a student’s disability is identified.

Why it matters

Understanding these disparities in U.S. elementary schools is important to help ensure that all students with disabilities have equal access to appropriate services and supports.

Schools are one of the most common places that provide disability services to children and adolescents. This includes students who have autism.

Some research finds that teachers are more understanding of a student’s classroom struggles when informed that the student has autism.

School-based special education services, such as speech therapy, often benefit students with disabilities, including those of color. Student will not receive these services without an identified disabilty.

For example, recent analyses of public data from Massachusetts, Indiana and Connecticut compared the achievement trajectories of the same students before and after they received special education services. The students did better in both reading and mathematics when they received special education services.

Students with disabilities are also more likely to graduate from high school and attend college if they receive special education services.

What still isn’t known

We do not know whether these disparities in autism identification are occurring in other elementary grades, at least based on the National Assessment of Educational Progress data.

In another of our recent analyses, though, we did observed racial disparities in autism identification across elementary grades.

Some other research suggests that students of color and girls experience significant delays in receiving autism diagnoses.

Our analysis is based on students who completed the National Assessment of Educational Progress reading test. Students with severe autism and higher support needs who were unable to complete these assessments, even with accommodations, were not included in our analysis.

Future studies could examine whether sociodemographic disparities in autism identification are occurring in U.S. middle and high schools as well for students with significant impairments.

What’s next

Our additional preliminary analysis indicates there are other types of disparities at play. For example, we are finding that Black and Hispanic girls, low-income Black students and multilingual learners who are white or Hispanic are especially unlikely to be identified as having autism.

We are also exploring whether some of these disparities have grown, or otherwise changed, following recent increases in autism prevalence rates, including for students of color and girls.

The Research Brief is a short take on interesting academic work.

Paul L. Morgan received funding from the U.S. Department of Education's Institute of Education Sciences to support these analyses. Opinions expressed here are those of the author and do not represent the view of the U.S. Department of Education.