Old Maps, Hyperspectral Imaging, and Using Technology to Visualize the Natural World

2.7K

We have a multidisciplinary episode of Coast to Canopy this month. It’s airing in advance of the Wind Water Fire Conference, which will take place at the Monticello Opera House on October 3-4. The conference has its roots in Jefferson County’s status as an archeological and paleontological hotspot. But archeology is a science that draws from multiple other disciplines, from geology to biology to climate science. And so, the conference will cover a breadth of topics about natural science in the Big Bend region of Florida.

After all, we live in a biodiversity hotspot as well as an archeological hotspot. The two are likely connected. Looking back through the millennia, the richness of flora, fauna, and water resources would have made this area an attractive place to live, even in the cooler, drier Pleistocene era.

In a way, this is three episodes in one. We talk about this area’s unique climate, a technological approach to researching longleaf one ecosystems, and archeology both hundreds and thousands (and even tens of thousands) of years old.

It’s a diverse lineup, and an opportunity to delve into multiple aspects of our natural north Florida. But, as our conversation unfolded, a theme emerged. Our guests are all using advanced technology to interpret our landscape in new ways. We can contrast the technologies used by each researcher, and contrast them also with the subjects in James Dunbar’s presentation. Five hundred years ago, Spanish explorers had to map an unexplored (to them) hemisphere. To do so, they used the most advanced science of their day.

Full schedule and presenter info for the Water-Wind-Fire Conference.

Meet our guests

• Dr. James Dunbar is the Board Chair of the Aucilla Research Institute, and a former state archeologist for the state of Florida.
• Jenny Rogers is a Remote Sensing and Geospatial Technology Analyst with Tall Timbers Research Station and Land Conservancy.
• Dr. Ryan Truchelut is the co-founder and Chief Meteorologist of WeatherTiger, a Tallahassee agricultural meteorology, expert witness, and hurricane forecasting company.

Our guests shared a lot of information in this show. We were able to add a lot of video and images to show Dr. Dunbar’s 500 year old maps or Jenny Rogers’ hyperspectral images, and we’ll share more in these posts. Because their two presentations focus on visualizing our area, I’ve kept them together for this accompanying blog post. Ryan Truchelut joins in later to talk about technology in science, but I’ve separated his section of the interview pertaining to weather and climate. You can read that here.

Coast to Canopy blog posts are curated transcripts. My notes appear in italics.

We start by having Dr. Dunbar describe the conference. The Aucilla Research Institute shortened its name to the Wind Water Fire Conference after we recoded this episode. The original name included the term First Floridians, a nod to the conference’s roots in prehistoric archeology.

The multidisciplinary nature of archaeology

James Dunbar: [The conference is] sort of archeology, but it’s also natural sciences. And it’s trying to look at the Big Bend area and how it developed through time. And just basically looking at, the environment, the vegetation, the, animals, the people and how things change through time. But it it involves a lot of geology.

It involves multispectral imaging. It involves weather forecasting and those type of things. And one of the reasons we decided to go on the theme of water, wind and fire is the last three hurricanes that have hit here. We’ve had some wildfires and thought it would be a good time to address sort of a broad spectrum of different topics that are all related in a way, but they’re not related directly to each other.

Rob Diaz de Villegas: In the discipline of archeology, you draw upon all of these different, other disciplines too. It’s all very heavily intertwined, right?

James Dunbar: Right. As much as we can bring to the context of things, the better off we are. And what I mean by that, if you find an archeological site in the middle of nowhere and you have to draw on, well, what are the diagnostic artifacts? That gives you a sense of time, but also, well, how did these people live?

If you have a place where you can core for pollen and get that kind of information… LiDAR may give you some kind of information of structures at the time. We have somebody that’s a geologist out of Canada that’s going to come and talk, Brendan Fenerty, and his interest is looking for weather conditions in ancient times. And he’s found some evidence of hurricane activity in Florida 12,500 years ago or so.

Page-Ladson and the ever-changing dates for human occupation of the Americas

Dr. Dunbar is best known for his work on an archeological site in the Aucilla River, called the Page-Ladson site, in the 1980s and 90s. We tell a detailed version of his and subsequent excavations in the WFSU documentary, Finding the First Floridians.

James Dunbar: I was working with a fella, David Webb, who was a paleontologist, and myself and some archeologists, and Lee Newsom, who’s a paleobotanist. And we really pieced together… it’s not an occupation site. It’s more like a human activity area that took place in the Aucilla River when the water tables were lower.

In a layer they determined was mastodon feces, divers found bones from a mastodon that may have been butchered by humans.

James Dunbar: We have cut marks on a tusk and some other things. And it was 14,500 years old, which caused all kinds of consternation in the community of archeology, because everybody knew nothing could be older than 13,000 years. And here we are at 14.5.

Archeologists were skeptical of their radiocarbon dates, and their work did not initially receive the recognition it later would. In 2012, James Dunbar helped a team of archeologists from Texas A&M, who found a knife in the mastodon waste layer, which they dated to 14,570 years ago. This time, the date caused the archeological community to revise estimates of human arrival in the Americas.

Pushing back the date: how long ago did people arrive in the Americas?

James Dunbar: Since that time, there have been other discoveries. White Sands National Park is the famous one that dates from, I think, 24,000 to 23,000 or 22,000 years old. Human footprints followed in the tracks of giant sloths. They were hunting the animals, and it’s fully carbon datable because they’re seeds that were in the sediment and that kind of thing. And there’s been two major studies done now that say this is real archeological site.

So, getting into my end of things, why is that important? Because [archaeologists] were saying that people couldn’t have been here longer than 13,000 years ago, because the Cordilleran and Laurentide ice sheets closed off [prior to that] and there was no path to come into the Americas.

Well, if people were here in the late glacial maximum, they were already here. And I think we’re going to find human occupation sites or activity sites somewhere in the Americas that are dating to the 30,000 range. I think there’s a high possibility now because at the late glacial maximum, maybe people weren’t coming in because the Cordilleran and Laurentide ice sheets blocked off everything, but they were already here.

The last glacial maximum ended about 20,000 years ago. Glacial ice melted for thousands of years until an opening formed in the ice sheets separating Alaska, and the Beringean land bridge, from the rest of North America. This opening is thought to have formed just over 13,000 years ago. DNA evidence has shown that indigenous Americans likely came from northeast Asia, and, when sea level was lower, the land bridge was where they likely crossed.

If people were here during the last glacial maximum (between 22,000 and 24,000 years ago), when there was no terrestrial path between Asia and much of North America, what Dr. Dunbar is saying is that people may have first come here in the previous interglacial period, tens of thousands of years earlier.

Hyperspectral imaging in longleaf pine ecosystems

Jenny Rogers: I do remote sensing. So that just means observing the Earth from above.

So we use satellites or drones, but really, I work with any data as long as it’s associated with a location… we’re using satellite imagery and we’re actually using the next generation of satellite imagery. We’re actually using something called hyperspectral.

So, these are satellites taking images of the ground. In a typical camera image, you have pixels. When you were a kid, maybe you pressed your face up against the TV screen like your parents told you not to. And you could see the individual pixels, and they were a line of red, green and blue. Eeach of those are at different intensities. Then you back up and you’re like, wow, this creates so many different colors. Each of those intensities for those three colors, is a number. And so for something like a multispectral satellite, you have, you know, somewhere like 11 to 13 different colors with their own number of their own intensity.

I’m working with a satellite that has 240 different colors. In the multispectral… you have red, green, blue, and then you get into the infrared. And in the same way that we can say that a color is red, the satellite can pick out maroon or fuchsia or crimson.

In a theoretical sense, they’re each wavelength of light.

Flavors of longleaf pine ecosystems

Jenny Rogers: This became a really valuable technology… it’s used a lot for plant health. So, plants… to our eyes they’re green because they’re absorbing all of the different colors and reflecting the green.

It does the same thing with the infrared. And so we use that as a measure of plant health. There’s different pigments that we could be picking up from a satellite. This is the idea that we’re using to detect different plant communities. We’re interested in the native pine savannas of the region: flatwoods, pine savanna, sandhills, upland pines.

But we’re also interested in separating those from the pine plantations of the region that don’t manage with fire as much. And, more mixed pine forests that have maybe previously been these pine savannas that have had fire excluded. And so the hardwoods kind of take over.

The upland pines are more in this Red Hills region. They have a lot more clay soil. Each of these ecosystems is dominated with an overstory of longleaf pine, and a lot of wiregrass understory.

This differs when you go to the Sandhills… our main site was the Ordway-Swisher Biological Station near Gainesville. And they have more of a turkey oak component, whereas [in the Red Hills] we have the southern red or southern red oak.

In the Orlando area at the Disney Wilderness Preserve, which is one of our other sites for this project, has the flatwoods. And on the coast here, we also have flatwoods, and that has more of the saw palmetto component in the understory.

Detecting the frequency of fire on the landscape

Jenny Rogers: Long history of fire in the southeast. And we still have a lot of fire on the landscape. I believe that 5 million acres are being managed with fire in the southeast annually. It’s very important for ecosystems, like I said, so that… we can maintain the understory.

For example, in our upland pine savanna ecosystem, we can have up to 52 species in a three foot by three foot area. And so and that’s probably the most in North America that you’ll find. To keep that, you need fire on the landscape. So that’s one thing we’re looking for. Because 80% of the southeast is privately owned land, we don’t really know where all of those examples of ecosystems are.

And so it’s important for us to know so we can prioritize our conservation efforts.

Jenny Rogers says we have a long history of fire in the southeast. This area has been at the forefront of using prescribed fire to mimic the way fire had naturally burned. In Finding the First Floridians, we learned how Florida landscapes burned over tens of thousands of years, and the relationship between fire and megafauna such as mastodons and mammoths.

Teaching an Artificial Intelligence to look for forest types

Jenny Rogers: In the way that we do the work, we do use machine learning. So, it tends to determine for itself what is the most important. We don’t really know, but it could be related to leaf waxiness, or even in the understory, the soil moisture, or things like the anthocyanins, the red pigments, or the keratin or the orange pigments in the plants.

We really just let the model sort of determine what’s the most important, and a lot of it tends to be even just like the amount of chlorophyl, that you see in the plants.

Rob Diaz de Villegas: I saw in your description for [your presentation], 76% accuracy when you ground truth. Talk about that teaching [the AI], and having it teach itself what these ecosystems look like.

Jenny Rogers: Thankfully, we’re working in an area that has a long history of research in the plant communities. We know where these ecosystems are for the areas that we work in. And we’re able to tell an algorithm, hey, these are known areas. Can you help us with these photographs to find other examples?

We use random forest modeling, which is a made up of decision trees, which are essentially flowcharts. So hundreds of flowcharts that help with pattern recognition, essentially.

From hyper spectral imaging to old fashioned cartography

James Dunbar: Within the last decade, anyway, archives all over the world… have started to make and digitize their historic maps, which is wonderful because you can just download them and then compare and contrast. And that really hasn’t been around that long.

Specifically, Dr. Dunbar has been looking at Spanish maps from the 1500s.

James Dunbar: So, if you’re a Spaniard, you’ve got a new world you just found and you need to map it.

How do you map a sphere and put it on a flat piece of paper? The Spanish government first employed cartographers and then, “No, we need cosmographers. The people that look at the observable universe, bring it back down to Earth, and let’s see what we can do.”

One of the most famous, I think, was Alonso de Santa Cruz. He and a fellow you may have heard of before, Sebastian Cabot. Sebastian Cabot was the pilot major for Spain. And they went on an expedition from 1526 to 1530. And they wanted to get around, into the Pacific, but the headwinds were pretty strong. So they stayed in on the coast of South America, then into the Caribbean from 1526 to 1530.

Mapping the Americas in the sixteenth century

James Dunbar: So, Alonso de Santa Cruz comes back to Spain and gets his degree in cosmology and becomes the cosmologist for the King of Spain, King Charles the Fifth, and begins mapping the New World.

And there were other people doing that at the time. But, one of the problems was nobody could figure out in the ocean, with an astrolabe (which was the instrument used), they couldn’t calculate longitude. So, they only had latitude to go by an estimated distance traveled.

When they were in the middle of the ocean and trying to figure out where they were, the Spanish call it a point of fancy because they didn’t know really where they were. They knew the latitude, they didn’t know the longitude. They had a compass heading, they stick to the compass heading.

Cruz was definitely a scientist of his time. He actually developed a map of Spain and 1543 or somewhere in there. That was a longitude latitude map. Now, wait a minute. You couldn’t do longitude back then?

Well, you could if you were on land and you paid attention to things like solar eclipses or lunar eclipses. And he did a map of Spain that looks like a modern map. It’s longitude and latitude with a little grid ticks on the side of the map that researchers since that time say it was more accurate than Mercator’s maps, which were done in the 1590s.

He did these in the 1540s.

The disastrous expedition of Pánfilo de Narváez

Dr Dunbar has found a discrepancy between different maps of the era. It concerns the placement of the Bay of Miruelo, which is where an expedition led by Pánfilo de Narváez landed. Narváez is believed to have landed near Tampa and then traveled by land to Apalachee Bay.

James Dunbar: So, [Alonso de Santa Cruz’s] map of 1542 and then 1545 places Florida in a – well, you can find the Big Bend on that map, let me put it that way. Earlier maps, he couldn’t because the Gulf of Mexico looks like a semicircle. But he’s definitely got the Big Bend and he puts an object there that says the Bay of Miruelo [Dr. Dunbar pronounces this Bay of Merlot in the podcast]. But anyway, it was Diego Miruelo who was the pilot for Narvaez, and they end up finding a bay at 29.5 degrees latitude.

And everybody that’s looked at maps [previously], they never had the ability to get all these maps together and look at them at the same time and cut them apart because they’re digital. Anyway, the Bay of Miruelo was in the Big Bend area. It was not down in Tampa, the way it gets mapped out in 1542 and 1544.

And then Sebastian Cabot comes out with a map in 1544 and is here is where the Narváez expedition landed and why that has been ignored all these years. I have no idea.

Earlier maps by Alonso de Santa Cruz have the Bay of Miruelo closer to the mouth of the Aucilla River, at the eastern edge of Apalachee Bay. Dr. Dunbar believes that Narváez may have landed near the Aucilla, and not Tampa. As the story was told, they traveled west to Apalachee Bay and lost most of their men along the way. The survivors then set sail for Mexico, where only three of their expedition landed.

Mapping Apalachee Bay with LiDAR

James Dunbar: I’m just probably insane for doing this, but I’m trying to get a map piece together that covers from Alligator Point to just south of Deadmen Bay, Steinhatchee Florida.

And it goes out 30, 35 meters below sea level, this brand new LiDAR. And then I’m trying to pull the land-based LiDAR angle, go all the way up and have a map for the conference. I don’t know if it’s going to work. I’m about ready to kill my computer or my phone.

Rob Diaz de Villegas: Talk about the LiDAR scanning. How does how does that work?

James Dunbar: It’s similar to what Jennifer’s doing, but it’s different. It’s a laser. For land based LiDAR, you use a red laser. For bathymetric LiDAR, use a green laser. And you need clear or fairly clear water to be able to do the bathymetry part of it. But it when you get it, it’s really high fidelity.

I think USGS, probably in cooperation with the Florida Geological Survey, have been collecting this data.

Now in the first stretch of LiDAR for offshore is for Apalachee Bay. And I found a lot of interesting things out there. I don’t dive anymore. I have to get some of my friends to do it.

The Tech Discussion: How quickly is technology changing the way researchers work?

Ryan Truchelut: I’d certainly say in meteorology that things are changing quite quickly. One of the things that is just now coming to the forefront are machine learning or artificial intelligence means of doing the type of weather prediction that we’re more used to full physics models doing. And we’ve gotten familiar with those over the last 20, 30 years.

Now we’re starting to have some pretty strong evidence that these almost more statistical approaches, which use huge training sets, can approximate some of the skill levels for certain applications. But this is very new stuff. And a lot of meteorology is being able to recognize, predict biases in models.

And then you can correct for [things like]… the American model tends to have areas of high pressure that are too weak off the East coast. So it tends to take hurricanes too far east. Over time you can observe cases like that and then correct for it. Well, we don’t really know what the biases are in the artificial intelligence based model. This is the first hurricane season that a number of them are available. A few of them were available last season as well, in a more kind of widespread way.

It’s something I’m looking at in evaluating as a forecaster.

The power of big data is in how you use it

Ryan Truchelut: But again, also, this should go back to that conversation we were having about bad actors acting on information. You can read that discussion in a separate post about north Florida weather and climate.

James Dunbar: Garbage in, garbage out.

Ryan Truchelut: Exactly. And you if you have far more models, you have all the physical models that we used to have before and now you have these AI models. It’s a choose your own adventure. There’s more things to select from and show.

Well in 15 days the AI model is showing a major hurricane approaching Florida. (This episode was recorded on August 15, 2025, as Hurricane Erin moved up the Atlantic)

Well, does that mean anything? No. Furthermore, we don’t really know because we don’t have the context of seeing how it’s performed. We don’t have those skill scores going back a few years, and I think that’s really key in meteorology to be able to kind of understand not only what the forecast is, but, as a meteorologist, why a forecast is what it is. That’s something that the AI approach really takes away from you.

To sum it up, the field is changing very quickly. I think it’s very promising for certain applications. But we need to evaluate those very carefully.

And of course, just with the media environments that we’re in, people are very afraid of artificial intelligence. I’m more afraid of natural stupidity.

The machine learning black box

Jenny Rogers: Yeah. Even, for me, because I do use machine learning and I am given a lot of opportunities to continue learning in my career in my position, which is really lovely. But even still there’s lots of machine learning models that I know very little about. And then then it gets into deep learning and and that’s a whole new ballgame.

Ryan Truchelut: Yeah, I think the black box approach to science is not a good way to go, because so much of science is the ability to help people understand and explain something and in a responsible way to tell a story. And I fear that as we’re automating, if we automate too much, we’ll get away from that kind of meaning that we can extract from data.

Jenny Rogers: Yeah. And I can see the pros and cons. I think there are efforts to try to make the machine learning even more, or less black box, and, you know, get into what it is actually doing. And so there are ways to, to get at what it’s picking up.

James Dunbar: I have a GIS program that has AI built into it now, but I haven’t used it yet. Where I guess you could tell it, well, this is what this kind of target looks like. Find more of them with square miles of area to look at.

Jenny Rogers: I think in some ways you do get better accuracy with machine learning. And so if you are if you’re robustly testing that accuracy, if you have examples from a large area and you know you’re not applying the model to things that’s never seen before.

So I think it is really context specific. And people definitely misuse those sorts of things.

How well can AI predict extreme weather events?

Ryan Truchelut: And that’s the big problem in meteorology, is that so much of the value in my world is driven from being able to make good probabilistic forecasts of extreme events, and by definition, extreme events are rare. So it’s difficult to establish trends. It’s difficult to have enough events that go into a training set that are relevant, that makes sense.

And that’s what’s causing the damage. AI does a great job of telling you that today is going to be 92 [degrees] with a 30% chance of thunderstorms. We just don’t know how it’s going to do how it’s going to fare forecasting a Hurricane Andrew or an Irma or a or even a Helene.

It’s something that we’re going to have to evaluate over time and we’re going to have to… tweak.

Predicting rapid intensification with artificial intelligence

Rob Diaz de Villegas: In theory, if it keeps learning, it should only become more accurate, even if you really can’t see inside of how it’s working. Is that accurate to say?

Ryan Truchelut: It’s hard to say because the physics inside the core of a major hurricane are very complex. Yeah. And the greatest gains that we’ve seen in forecasting accuracy, we do a much better job now than we did ten years ago, even than we did during, say, with Hurricane Michael in 2018. We are doing a much better job of forecasting cases where a hurricane jumps very quickly from a tropical storm to, you know, a category three, 4 or 5 hurricane.

You know, 72 hours before landfall, Michael was a tropical storm its strength into a category five hurricane in three days. National Hurricane center. Initial forecasts were, you know, calling for a category 1 or 2. Now, fast forward to Helene in 2024. The initial forecast for Helene right off the bat. Advisory one calling for a category three hurricane.

It’s anchoring those expectations to something much more impactful and much more realistic, and giving us better time. That’s not due to the that’s not due to AI. It’s not due to machine learning. It’s due to the hard work of NOAA scientists who have been conducting field experiments and building better physically based models of what’s going on in the core of a major hurricane.

And it’s given those tools to the forecasters at the National Hurricane Center to be able to just to have a better guideline for when is rapid intensification most likely. And how can we communicate that risk to the general public? We’re doing a much better job of it now, and it’s such a shame that we’re not funding those efforts fully.

Yeah, because they produce such great gains for society. The idea that we can just replace all that with artificial intelligence is just specious.

James Dunbar: You know, it should be a tool, not something that’s heavily relied on and just go away and have a Coke or something.

Ryan Truchelut: Oh yeah. I mean, it’s a great efficiency tool. Yeah. But it’s a great efficiency tool to allow human beings to do what we do best, which is actually do that critical thinking. Yeah. And also, you know, balance that with what’s going to benefits society the most too.

Using data to tell a story

Rob Diaz de Villegas: Talk about the human part of this. There’s a lot of data… shaping it into something useful, but also communicating. Talk about your job as a storyteller, as far as taking all of this data and making it meaningful to people, to the public.

Ryan Truchelut: Useful and usable is my is my motto at WeatherTiger. Yeah. And, you know, I mean, that is scientific knowledge in a vacuum that isn’t understood or applied, is not providing benefit to society. So that’s, you know, everything that I do is very much trying to think hard about how to get people to take the right things away from my forecast and the right things.

My number one goal is that people have the most lead time that they can possibly have for a high impact hurricane event or other weather event. And the flip side of that is not over-warning people, not scaring people, not having excessive false alarms. Weather really provides that fast feedback on on being able to communicate, because… they’re either taking the right message away from your forecast or they aren’t.

And unfortunately, I’ve had a lot of experience in the past ten years of having to communicate highly impactful hurricane event risks here in North Florida.

There’s never been another period in hurricane climatology going back 175 years, where we’ve had ten major hurricane landfalls in a nine year period along the Gulf Coast. We truly are living through an extraordinarily active period of those high impact events, both in North Florida, specifically, and for the Gulf as a as a general region.