82° ?! Something Feels Off!

Growing up in the Caribbean, you develop a relationship with weather that most people don't have. Not in a dramatic, storm-chaser way. More in a quiet, constant way. You feel the shift in the trade winds before you check anything. You know what rain smells like twenty minutes before it arrives. Someone's grandmother already told you what's coming.

That's part of why I never paid much attention to weather apps. I'd just glance at them the same way you glance at a clock you already know is wrong. The number says 82°F. Fine. But 82 in Antigua doesn't feel like 82 in New York, and it definitely doesn't feel like 82 in London. The humidity here sits and hits differently. The wind off the Atlantic carries something that a "Real Feel" number on AccuWeather doesn't always quite translate.

Sometime last year I started building a weather site. For which I've written about the origins before: how it started with my father, with the calls that used to come in during hurricane season, asking whether this storm was going to hit, how strong, and when. After he died, those calls came to me. That part of the story hasn't changed.

What has changed is what I started to notice once I got deeper into the data.

When you build a weather platform from scratch, you have to make choices about how to present information. What words to use. What thresholds matter. What "windy" means. What "heavy rain" means. These seem like small decisions until you realize that every major weather service has already made those decisions for you, and almost none of them had your region in mind when they did.

Take wind descriptions. On most global platforms, 15 mph wind speeds are often described as "moderate." In our EC islands, that's a normal Tuesday afternoon!

The trades blow like that most of the year. If a weather site is telling someone in Dominica that the wind is moderate every single day, the word loses all meaning. It stops being information you can feel and just becomes noise.

Or take "feels like" temperatures. The algorithms behind those numbers were largely developed for temperate climates. They account for wind chill very well, because that matters in New York in January.

But the way humidity and solar radiation interact in a tropical maritime climate, where you're at sea level, surrounded by warm water, with persistent moisture in the air. Well, it doesn't really map as cleanly onto models built for continental weather.

I'm not saying the data is wrong. The sensors are fine. But the interpretation, the layer between raw data and what a person actually experiences when they step outside, that layer is where things get lost.

Researchers have reached the same conclusion from the data side. A 2022 study in the International Journal of Climatology looked at 40 years of heat stress data across the Caribbean and found that standard indices don't capture how conditions are actually experienced here. The interaction between humidity, radiation, and wind speed matters more than temperature alone, and the Lesser Antilles showed the steepest increase in heat stress trends. The science confirms what any one of us standing outside already knew: the number on the screen doesn't match what we feel!

This is what I've been focused on with Dewedda.com.

Not adding more data. The opposite, really. Making the existing information land the way it should for someone standing on one of our islands.

Wind descriptions that reflect what trade winds actually feel like here, not in San Francisco. Condition summaries are written the way we'd describe weather to each other. "Boy, today breezy eh?!" Yeah, like that, not the way a meteorological text would phrase it.

Thresholds adjusted so that "heavy rain" actually means heavy rain for our tropical islands.

Some of this is technical. Adjusting the breakpoints in the code where one description transitions to another. Rewriting the logic that decides when conditions are "comfortable" versus "oppressive." Pulling in sea-level pressure and dew point alongside temperature so the "feels like" calculation accounts for what it's actually like to be here, at this latitude, at this humidity, with this kind of wind.

But most of it is not technical at all. Most of it is just paying attention. Reading what the site says, then stepping outside and asking, does this match what I'm feeling right now? And if it doesn't, figuring out why.

Choosing the right data source was part of getting this right. I tested several weather APIs and settled on AccuWeather for the hourly data.

After comparing readings against what I was actually feeling outside, their data consistently matched conditions here better than the alternatives. When every description, threshold, and summary on the site depends on that foundation, accuracy is the thing you don't compromise on.

The big weather platforms will never do this for the Caribbean or specifically the Eastern Caribbean islands. It's not that they can't. It's that there's no reason for them to. We are a few million people spread across many small islands. Not exactly a priority market.

And that's fine. That's not a complaint. It's just the long-standing reality that pushed me to build something different.

What surprised me is how much there is to get right. I assumed the hard part would be hurricane tracking, storm proximity calculations, satellite imagery, the technically complex stuff. And last year, developing that part was hard. But the thing that keeps pulling me back in, the thing I keep refining, is the simple stuff. The temps, wind, rain and the words and descriptions.

Because for us, weather isn't dramatic most of the time. It's not always hurricanes and tropical storms. It's knowing whether to hang laundry outside. It's knowing whether the sea will be rough for the ferry. It's knowing whether that cloud buildup to the east is going to drift over or push past.

Getting that right turns out to be harder, and more satisfying, than I expected!