Counting Trees on Our Adirondack Property Using LiDAR, Aerial Imagery, and AI

It started with a pretty simple question: How many trees are actually on our Adirondack property?

Our parcel is relatively small, but like much of the Adirondacks, it is heavily wooded. Walking the property and manually counting trees would be possible in theory, but it would also be time-consuming and surprisingly difficult. Instead, we decided to see what we could learn using publicly available geospatial data and artificial intelligence.

The combination ended up being pretty powerful:

Claude + NYSGIS LiDAR data + aerial photographs

Using these tools, we were able to develop an estimated tree count for the property and, even more interestingly, estimate the heights of many of those trees.

What is LiDAR?

LiDAR stands for Light Detection and Ranging. An aircraft sends laser pulses toward the ground and measures how long it takes for those signals to return.

Because some laser pulses hit the tops of trees while others make it through openings in the canopy and reach the ground, LiDAR can create a detailed three-dimensional representation of a forest.

For our property, the available NYSGIS LiDAR data was collected in 2018. That means the analysis is essentially giving us a snapshot of what the forest looked like at that time.

By comparing the elevation of the ground to the elevation of the vegetation above it, it is possible to estimate tree height.

If the ground elevation at a particular location is 500 meters and the top of the vegetation is 520 meters, for example, the approximate tree height would be 20 meters.

Finding Individual Trees

Knowing the height of the forest canopy is one thing. Figuring out how many individual trees are within that canopy is another.

Trees growing close together create overlapping crowns, particularly in a mature forest. The analysis therefore involved looking for high points within the canopy that could represent the tops of individual trees.

Those potential treetops can then be separated into individual tree crowns and counted.

It is important to call this an estimate, not a perfect census of every tree on the property. Small understory trees may be hidden beneath larger trees, and closely spaced trees can sometimes appear as a single crown.

Still, it provides a remarkably interesting look at the structure of the forest without having to physically measure every tree.

Adding Aerial Photography

Aerial imagery provided another useful layer.

LiDAR tells us about height and three-dimensional structure, while aerial photographs let us visually examine the landscape.

Together, they help distinguish forest canopy from clearings, roads, buildings, and other features on or near the parcel.

That combination makes it much easier to check whether the results actually make sense.

Where Claude Came In

The AI component did not magically look at a picture and count the trees.

Instead, Claude served as a tool for working through the geospatial analysis. It helped develop and troubleshoot the code needed to process the LiDAR data, create canopy-height information, identify potential treetops, separate tree crowns, calculate heights, and summarize the results.

That is one of the parts of AI that is becoming increasingly interesting to explore.

AI can make specialized datasets and analytical techniques much more approachable. A project that might once have required extensive programming or GIS experience can now become something that a curious person can experiment with using real scientific data.

More Than Just a Tree Count

The original goal was simply to answer:

How many trees are on our property?

But once the data was processed, the possibilities became much more interesting.

The same dataset can potentially tell us about:

• approximate tree heights
• canopy density
• gaps in the forest
• forest structure
• particularly large trees
• changes in vegetation when compared with newer imagery or future LiDAR datasets

Because our LiDAR data dates to 2018, it also creates something of a historical baseline. Trees have grown, some may have fallen, and others may have become established since the data was collected.

That makes the analysis a snapshot of our little piece of the Adirondacks at a particular moment in time.

AI Meets Environmental Science

This experiment ended up being a great example of what happens when environmental science, GIS, remote sensing, open data, and artificial intelligence come together.

We started with a question about our own property and ended up using the same basic types of data and analytical approaches that can be used in forestry, ecology, conservation, and land management.

And there is something especially fun about looking across a wooded Adirondack hillside and realizing that a dataset collected from an aircraft years ago can help tell the story of the trees growing there.

What started as “I wonder how many trees we have?” became a much bigger exploration of what modern geospatial data and AI can help us discover about a forest.