NASA Logo in the header
Data Science
Operational

Alaska’s boreal-tundra resolved with vegetation height from commercial imagery and a foundation model

We fine-tuned and validated an open source Earth-observing foundation model to detect small but significant vegetation structure changes in Alaska’s boreal-tundra transition zone.

Launch Date
--
Class
--
Website
Key Staffs

Warming temperatures are rapidly altering northern landscapes through increased shrub expansion, accelerated tree growth, and widespread permafrost thaw. Accurately measuring these changes is challenging because low-statured plant types often are difficult to resolve with satellite observations. To address this challenge, researchers fine-tuned an Earth observation AI model to predict detailed vegetation height, applied it to high resolution satellite imagery, and validated it using local airborne lidar and field data. The updated, high-resolution model mapped vegetation heights across 40% of Alaska’s boreal-tundra transition zone - where tree and shrub growth acceleration over decades is altering ecological patterns and processes in place for millennia. Data revealed that approximately 82% of the mapped landscape consists of vegetation shorter than 1.5 meters, highlighting the dominance of low shrubs, and the validation indicated the model still tends to underestimate the height of the tallest portions of tree and shrub canopies. The modeling approach will benefit future analyses of detailed yet subtle tree and shrub changes across the vast northern landscapes. This work was collaboration with scientists JJ Frost and Matt Macander from Alaska Biological Research, Inc.