Investing in Maine’s future forest with high-value, low-cost geoinformatics
PI: Simons-Legaard, Erin
Abstract: The forest products industry contributes nearly $8.5 billion annually to Maine’s economy, and by some estimates this contribution could more than double with value-added processing, biodiversity offsets, forest carbon trading, and other ecosystem service credits. Realization of this potential will require adaptation of forest management strategies. Forest managers in Maine have identified a lack of spatial information on both timber and non-timber forest resources as a barrier to the planning and prioritization of management actions. Satellite remote sensing data are capable of providing near-real time mapping of forest attributes that are key to management decisions. The utility of available commercial products is limited, however, due to cost of production and reliability shortcomings. We have developed machine learning algorithms for application in remote sensing and geoinformatics that are highly adaptive and uniquely capable of addressing characteristic shortcomings of other methods. With computationally efficient software implementations that are currently under development, we plan to produce better data at lower cost than is currently available through commercial vendors. Our machine learning approach can produce a variety of products of high relevance to forest management problems, including tree species composition; intensity and time since last harvest/disturbance; estimates of volume, biomass, and carbon; and additional ecosystem services like wildlife habitat suitability. These products would provide an array of options for annual sales, and a number of forest products companies have already expressed interest in their purchase.