Cultivating Sustainable Forests with Satellite Imagery Analysis
Remotely sensed imagery is enhanced by the richness of data captured by people on the ground. So the USDA’s Forest Inventory and Analysis (FIA) team created the Big Data Mapping and Analytics Platform (BIGMAP), a cloud-based, national-scale modeling, mapping, and analysis environment for US forests. BIGMAP is optimized and tuned to leverage the parallelization and mass storage required for raster processing at scale. The FIA team fused thousands of Landsat scenes with hundreds of thousands of plots, ultimately processing tens of trillions of pixels in the cloud, all in a matter of days.
For more than 100 years, the United States Department of Agriculture (USDA) Forest Service has worked to sustain the health, diversity, and productivity of the nation’s forests and grasslands for current and future generations. The USDA’s Forest Inventory and Analysis (FIA) program was created to improve the use and integration of advanced remote sensing technologies to aid in that mission.
In this video, officials from the USDA Forest Service discuss the FIA program and demonstrate the science behind their mapping project. To tackle complex, multidimensional raster analysis problems, the USDA Forest Service leverages the versatility of ArcGIS and Python to build custom machine learning algorithms that model the relationship between satellite imagery and the forest characteristics they measure. By tracking changes over time for each pixel in a satellite image, they can monitor seasonality and vegetation, predict droughts and environmental impacts, and better understand forest species. The data being developed is then used to inform policy and management decisions regarding our nation’s forests; it’s also available to the public to facilitate collaboration on creating a sustainable future.
Play the video to see demonstrations of the tools used for the FIA program or continue reading for key takeaways.
This content is available publicly, published in the US Forest Service ArcGIS Online organization, ArcGIS Living Atlas of the World, and open data portals. Maps like these give the ability to populate spatially explicit tools that may be integrated into conservation planning to support carbon management, wildlife stewardship, watershed restoration, and other environmental services.
Here are two examples of how you might consume these services in your own analyses using geoprocessing tools, raster functions, and more.
- Droughts—Droughts are having huge impacts on forests and woodlands, and agency scientists are already using BIGMAP results to make drought model projections available widely. Significant droughts were shown from an examination of exposure from recent years around the Central Valley in California. Projecting forward to 2040, drought exposure shifts from Central California to the Southern Rocky Mountains.
- Carbon sequestration—Adding data to a suitability model helps identify areas in the Pacific Northwest where planting opportunities exist without significant threats. These could be important areas to realize shared stewardship to implement climate mitigation or restoration plans.
Additional sample mapping applications include the following:
- Mapping major forest carbon pools
- Monitoring seasonality in vegetation and overall forest structure
- Showing losses in forest carbon due to disturbances such as wildfires and tornadoes
- Modeling suitability for planting opportunities without significant threats
Here are five modeling techniques being used:
By tracking changes over time, we can monitor seasonality, vegetation cycles, and senescence across large geographic areas to identify different types of forest, composition of tree species, and overall forest structure.
Used to analyze the time series captured in vegetation phenology, this technique allows us to characterize not only the average condition of vegetation but also how conditions change over the course of a year.
These coefficients describing seasonal changes in vegetation, along with other auxiliary data like climate and topography, can be combined with response data collected on forest inventory plots to order tree species along environmental gradients.
k-Nearest Neighbors imputation
The location of plots in the feature space of environmental gradients can be used with the k-Nearest Neighbors (kNN) algorithm, which works by assigning a “bucket of plots” to each pixel based on their proximity as measured in the feature space.
Prediction and mapping
Each bucket represents a group of records, stored in FIA’s database, from which one can make pixel-level predictions, quantify uncertainty, and map a variety of forest attributes.
Need help finding the right solution?
Schedule a conversation with one of our experienced sales consultants. Tell us how you're using imagery and remote sensing data now, and we'll show you where a comprehensive geospatial system can take your work next.