Big Data, Remote Sensing and Inventory Fusion for Forest Ecosystem Assessments

Conceptual diagram of big data remote sensing and inventory processing approach illustrating various inputs (top) and outputs (bottom), as well as lidar- (left) and imagery-based (right) approaches evaluated and included within the platform.


The success of forest management projects requires detailed information on forest structure and composition, which are then included as decision-making and modeling inputs. This information is used to assess current condition, evaluate and compare treatment outcomes, and to monitor the success/failure of moving an ecosystem towards desirable conditions. Often referred to as the first steps of ecosystem assessment and forest management planning (Jensen and Bourgeron 2001), these efforts typically include precise data on the amount and distribution of desirable resources, estimates of how resources may change over time, and hazard assessments (e.g., wildfire). Assessing resources, hazards, and dynamics over a large landscapes requires costly and resource-intensive spatially-explicit data. To be useful for resource managers, this data often needs to be site-specific information on tree sizes and species, fuels condition and arrangement, wildlife habitat suitability, and operational condition.

This work is funded by a USDA NIFA McIntire-Stennis grant and seeks to develop and statistically validate an open-source, big data remote sensing and inventory data fusion software platform which will provide enhanced forest structural and compositional information in support of forest resource decision making in the Southwest. Specifically, the objectives of this research are to:

  1. Assess and statistically validate algorithms for identifying individual trees and species from remotely sensed data of Southwestern forests using new and/or existing stem-mapped, area- and tree-based sample
  2. Using a suite of methods identified in the previous objective, design and implement a platform integrating multiple data sources (data fusion) which are typically too large to analyze in traditional manners (big data) to provide detailed forest resource information at the tree- to stand- to landscape-level.
  3. Assess the accuracy and statistical properties of forest resource estimates (e.g., estimate bias, consistency, error, spatial uncertainty) and use these to provide improved information used for land management decision
  4. Apply the platform to two Southwestern landscape case studies – i.e. applications of the framework and analyses in ponderosa pine forests of northern Arizona with the US Forest Service and The Nature Conservancy to quantify existing condition, assess low-value biomass product availability, facilitate watershed treatment implementation, and monitor forest restoration treatments.