4.03.00 - Uncertainty analysis, computational ecology, and decision support
2nd North American Mensurationists Conference
Portland, Oregon, USA; 11-14 December 2022
IUFRO Units involved: 4.01.00, 4.03.00
The North American biometricians are geographically separated into three regions: southern (SOMENS), northeastern (NEMO), and western (Western Mens). Each region holds individual meetings each year, sometimes in conjunction with other organizations. In the late 1980s, the first continental event was held in Saint Louis, MO. After more than three decades, the three regions decided that a new continental event should be held.
The 2022 North American event will present traditional biometrics and forest measurements techniques as well as novel modeling methods, such as machine learning, or forest inventory based on phodar or UAV (to list just a few). The event captured the interest of two research groups from the IUFRO Division 4 (Forest Assessment, Modelling, and Management), namely the 4.01 (Forest mensuration and modeling) and the 4.03 (Uncertainty analysis, computational ecology, and decision support), which co-sponsor the conference.
Keith Reynolds, United States
Paul Hessburg, United States
Bogdan Strimbu, United States
Chonggang Xu, United States
IUFRO Research Group 4.03 includes three Working Parties that cover:
- Characterization of uncertainty in forest modelling systems (4.03.01).
- Machine learning and computational ecology, especially as these methods apply to quantitative landscape ecology and forest modelling (4.03.02).
- Decision support systems and related information technologies for sustainable management of forest systems (4.03.03).
Each Working Party spells out its scope of work and the state of knowledge on its respective web page. For the 4.03 Research Group as a whole, though, our over-arching objective is to promote collaboration among the disciplines of the Research Group in ways that are not only mutually supportive but build upon the strengths of each discipline to produce synergistic advances in the science behind, and the management supporting, sustainable forestry.
We encourage visitors who visit the 4.03 web pages to join the Research Group's email list, https://www.iufro.org/science/divisions/division-4/40000/40300/mailing-list/, to stay informed on unfolding plans for the group and related meetings hosted or co-hosted by 4.03.
State of Knowledge
DSS frameworks are systems for building DSS applications, as opposed to the conventional notion of a DSS as a one-off application designed to support solutions for very a specific problem type. Frameworks are a higher order abstraction of the DSS concept, with the advantage that they are suitable for developing DSS solutions for a very broad range of problems.
Workflow engines and interfaces extend the capabilities of DSS frameworks with graphical programming tools that support data analysis, data processing (especially complex geoprocessing tasks), data transformation, scenario modelling, and system extensibility (by calling analytical engines external to the framework when needed).
Ontologies (or knowledge graphs) similarly are a way to enhance the power of DSS frameworks, in this case by invoking an ontology engine to query an ontology for relevant entities and relations formalized within it. For example, what are some alternative analytical sequences that be used to solve decision problems of type X? A bit out on the horizon, combining the capabilities of ontology engines and workflow engines in DSS frameworks opens up intriguing possibilities for querying for solution methods and then implementing them in workflows.
Parsimonious multiscale nonlinear models are difficult to develop given the computational challenges and hierarchical structure innate to forest ecosystems and landscapes. However, it is paramount to develop accurate description of the processes characterizing the status and dynamic of the forest.
Spatially explicit models for the development of tree components became the one of the central part of the current forest modelling, as the focuses shifted historically from stands to trees and now to branches and leaves.
Integration of remote sensing data into landscape modelling. Data acquired with remote sensing technologies became the source of almost unlimited amount of information that must be analysed. However, remote sensing brought a new challenge in modelling, namely wall to wall data that contains errors. Therefore, the models developed from remote sensing data must include now an assessment of error, or error propagation is now a part of the forest modelling process.
Machine learning modelling for the exploration and prediction of landscape disturbance patterns. The adaptability of machine learning to particular situation while considering the general trend in data made them suitable for solving a variety of forest problems, from landslides to growth and yield and inventory.