A selection of research projects in
Using Machine Learning and Airborne Lidar to Map Broad-Scale VisibilityKnowing how much of your surroundings you can see in a wildland environment has immense value in a range of fields of study. For wildland firefighters, visibility is a key component of situational awareness, the lack of which can result in dangerous or life-threatening situations. Viewshed analysis facilitates quantifying visibility, but using viewsheds to map proportional visibility across entire landscapes is computationally prohibitive. We developed a new approach for mapping landscape-wide visibility that only uses a few sample viewsheds along with a suite of terrain and vegetation predictors and random forests. We also developed a new R package, VisiMod, to implement our workflow.
Click here to access the published paper. |
Refining Hike Time Prediction EstimatesAs described below, the relationship between slope and travel rates is a fundamental human-environment interaction, the quantification of which has many potential benefits. Knowing how long it will take a pedestrian to travel along a particular route has immense value in emergency response, defense, and archaeology, to name a few applicable disciplines. Building on the Strava-based work below, we dug deeper into an alternative crowdsourced database from AllTrails, which enabled us not only to get hike-specific data (as opposed to all pedestrian movement), but also track entire hikes, providing an estimate of total hike time. We demonstrated an impressive capacity to predict total hike time among a diverse population with a series of new slope-travel rate predictive models.
Click here to access the published paper. |
Crowdsourcing the Relationship between Terrain Slope and Pedestrian Travel RatesDo you know how long it takes you to walk a mile on flat ground? How about up a 10 degree slope? Yes, you know you move more slowly (or at least you exert more energy) while walking uphill, but how much more slowly? One of the most literal human-environment interactions is the manner with which we move through a landscape. And one of the most important drivers of that interaction is the slope of the terrain. In this study, we leveraged the immense power of a massive crowdsourced dataset of GPS tracks collected using the fitness app Strava in order to uncover unparalleled insight into the pedestrian slope-travel rate relationship.
Click here to access the published paper. |
Detecting and Quantifying Timber Harvesting Activity in Oregon Using a Landsat Time SeriesRural communities throughout working landscapes of Eastern Oregon are highly dependent on timber resources and have been over a century. However, changes in the timber economy in recent decades have altered this coupled human-natural system. These changes can be observed from space, as timber harvesting creates a clear signal of change between successive satellite images. To assess prevailing trends in timber resource extraction in this region, we used a time series of Landsat images from 1986 to 2011 to explore the best practices for detecting and quantifying harvesting activity.
Click here to access the published paper. |