Every quarter, the School of Environmental and Forest Sciences holds weekly SEFS Seminars with faculty and experts. These seminars provide a space for presentation and discussion of a variety of topics relevant to the school and its students. Each seminar is held at 3:30 p.m. on Wednesdays in the Forest Club Room of Anderson Hall. After each presentation, a reception and discussion time is held in the room. You can watch previous Seminars on the SEFS YouTube channel.
The SEFS Seminar Series is made possible with support from the Corkery Family Environmental and Forest Sciences Director’s Endowed Chair fund.
Winter 2020 Seminars:
Feb. 19 – Dave Thau, World Wildlife Federation
Title: Forest Centaurs: The Past, Present, and Future of Human and Machine Collaborative Learning
Abstract: Humans and computers have been learning about forests together for decades. In recent years, as machine learning and big data processing have advanced, this collaborative learning has accelerated considerably. This talk will cover the past, and present of using machine learning and other artificial intelligence techniques for forest monitoring and management with an eye specifically on applications relevant for the United Nations Sustainable Development Goals. It will then move to describe very recent and planned work around forest monitoring and management aimed at achieving the World Wildlife Fund’s goal of helping build a future in which people live in harmony with nature.
Feb. 26 – Tony Chang, Ph.D., CSP Inc.
Title: “Chimera: A deep-learning approach for fusing multi-sensor data for forest classification and structural estimation”
Abstract: In recent years, the increased availability and amount of high-resolution (<30-m) imagery and advancements in machine learning algorithms have opened up a new opportunity to fuse multiple datasets of varying spatial, spectral, and temporal resolutions. In this talk, I describe a new approach to simultaneously classify forest land cover type and estimate continuous forest structure metrics using a deep learning model ensemble. This approach applies an ensemble of multi-task convolutional neural network (CNN) model we call Chimera. The Chimera ensemble integrates varying resolution, freely available aerial and satellite imagery (e.g., NAIP, Landsat), as well as relevant environmental factors (e.g., climate, terrain) to classify five forest cover types (conifer, deciduous, mixed, dead, none (non-forest)) and to estimate four continuous forest structure metrics (biomass, quadratic mean diameter, basal area, canopy cover). I demonstrate this approach by training the Chimera ensemble on georeferenced Forest Inventory and Analysis (FIA) field plots from the USDA Forest Service within California and Nevada, and highlight validation metrics. This modeling approach can estimate forested conifer type locations and structural attributes with high accuracy in a repeatable and cost efficient manner. Applications include inputs to fire behavior modeling and conservation area monitoring. Future implementations of the Chimera ensemble on a distributed computing platform could provide annual estimates of forest structure (biomass) and measurements of land-use change in the western U.S. over time.
March 4 – Simone des Roches, Ph.D., post-doctoral researcher, UW Department of Urban Design and Planning
Title: “The interacting effects of climate change and urbanization on Threespine Stickleback evolution”
Both climate change and urbanization can shape variation within species across space and time. Still, few studies have investigated how potential interactions between these two drivers affect contemporary adaptive trait change. We resurveyed populations of Threespine Stickleback that had been sampled from Californian estuaries in the last 40-100 years. Historical samples demonstrated that stickleback lateral plate number, a defensive, streamlining trait with substantial genetic and environmental associations, declined with decreasing latitude and precipitation. Plate number is linked to a single Mendelian-inherited gene (“EDA”). The derived “low plated” phenotype, associated with two copies of the “low” EDA allele, often evolves rapidly in lentic (“lake-like”) habitats where plates hinder maneuverability in densely-vegetated, slow-moving water. We show that the low EDA allele may be increasingly selected for in estuaries that have become more lentic with decreased precipitation and streamflow. This habitat transformation is likely responsible for increases in low-plated stickleback both over time and with decreasing latitude. Stickleback from estuaries surrounded by extensive urbanization, however, highlight a notable exception: plate number has increased in estuaries that have undergone significant hydromodification – including the channelization and dredging that tend to increase streamflow and reduce biological habitat complexity. Variation in stickleback plate phenotype and genotype, therefore, may reflect divergent selection in estuary habitats that are transforming under urbanization versus climate change.
March 11 – Paul Hessburg, USDA Forest Service