Detailed land cover map for the entire United States

Land cover type plays a major role in determining how much surface runoff occurs during a rainfall event. Artificial surfaces generate more runoff, while natural areas reduce runoff by promoting infiltration. As a result, detailed land cover maps are essential for accurate flood mitigation, stormwater management, and broader surface water planning.

The culmination of years of development

Scalgo has developed an AI model that can process trillions of high-resolution aerial imagery pixels to classify land cover across the United States. The new U.S. land cover map builds on earlier work in Denmark, Sweden, Norway, Finland, Germany, France, Poland and the United Kingdom. With each country mapped, the AI model has been refined through detailed training data and repeated evaluation, steadily improving its ability to interpret aerial imagery with high precision.

“Our AI method is very powerful because it keeps learning and getting better every time we apply it,” says Jonas Tranberg, Head of Machine Learning at Scalgo. “In the U.S., an existing model trained on European data already performed well in most temperate areas. Deserts were new to the model, so we had to create new training data for these areas. We also added a global average temperature dataset as an input to help the model distinguish between the otherwise similar classifications of snow-ice and bare land.”

Rigorous model methodology

The Scalgo Land Cover map distinguishes between 12 classes at 25 cm resolution. The nationwide dataset is produced based on sources from the National Agriculture Imagery Program (NAIP), which is managed by the U.S. Department of Agriculture’s Farm Service Agency (USDA FSA). NAIP imagery is collected during the growing season across the USA every few years, and includes natural color and near-infrared bands critical for land cover analysis.

The machine learning model first produces preliminary classes:

• Impervious

• Bare land

• Bare rock

• Snow/ice

• Shallow Vegetation

• Dense Vegetation

Auxiliary vector data from Open Street Map is then utilized to supplement the imagery, enabling a more detailed classification of the natural and artificial areas:

• Paved Road

• Unpaved road

• Railroad

• Building

• Other Paved

• Freshwater and Sea

As Jonas explains, “We create the AI output by alternating between training and evaluation phases. We start with a model trained on orthophotos outside of the USA to identify areas where performance is insufficient. Next, we add new training data from these areas and re-train. We repeat the process until the model is sufficiently accurate, which means at least 95% accuracy on randomly sampled locations.”

Start using the land cover map

Everyone now has instant access to the U.S. land cover map. The land cover layer will live in the familiar left side menu. As with all data in Scalgo Live, the land cover map is fully interactive.

Detailed breakdowns of land cover by watershed are now automatically available anywhere in the United States. Download data in a few clicks for use in GIS or other flood models.

Land cover in your projects

In a Workspace, you can now adjust land cover just as easily as you edit elevation data. Draw polygons to change land cover manually and add your own types, such as green roofs or permeable parking areas. You can also update land cover in your project by importing shapefiles.

Apply infiltration parameters and assign specific runoff functions to all land cover classes in the Flash Flood Map to more accurately depict flooding. Review upstream infiltration and runoff volumes anywhere along a flow path with a single click.

Check out this Getting Started video to learn more about how to work with land cover editing and infiltration in Scalgo Live.

Discover Scalgo Live

Please book a meeting with our U.S. team to learn more, or request a free 30-day trial.