I am developing an automated workflow to map forest trees over large (> 100 ha) areas using drone imagery. The workflow incorporates:

• Construction of 3D stand structure models from drone imagery using photogrammetry
• Segmentation of individual trees from structure models
• Identification of trees to species using convolutional neural networks for computer vision

More details >

I am building a non-linear spatially-explicit Bayesian model for predicting seed rain (and, subsequently, regeneration) of conifers following high-severity wildfire given a known spatial arrangement of surviving reproductive trees. Data on surviving tree spatial arrangement will be derived from drone imagery. To otbain dispersal data to train the model, I am coordinating intensive field surveys that involve documenting the density of regenerating conifers in a dense grid of plots across several recent severely burned areas.

When replanting trees after high-severity wildfire, managers need to decide what trees to plant and where to plant them. These decisions can be difficult given anticipated climate change and limited resources. I am working on several projects to guide reforestation efforts:

• Identification of the topoclimatic conditions associated with poor natural regeneration but high success of planted trees. This project involves development of an empirically-trained geospatial management tool for prioritizing reforestation efforts across a given landscape.
• Empirical tests of assisted gene flow, or planting seedlings from warm- and dry-adapted populations into cooler sites in anticipation of drought and warming. This project involves observing the performance of seedlings planted 1,000 to 3,000 feet in elevation above their origins.

The extreme California drought of 2012-2015 triggered extensive tree mortality across large areas of the state. I am analyzing large geospatial and plot-based datasets to understand:

• What biophysical factors explain spatial variation in mortality patterns? Can these relationships be used to predict future mortality?
• How can management such as forest thinning reduce furture mortality risk?

Recent results:

• Dense stands in arid sites are most susceptible to drought
• Forest thinning reduces future mortlity risk, especially in pines
• Most areas that experienced extreme mortality related to the California drought show strong natural regeneration

Other current research projects include:

• Identifying drivers of California tree species range limits and inferring sensitivity to climate change
  • Includes analysis of ~800 tree cores that we collected and processed
• Evaluating the suitability of modeled topoclimatic water balance variables as drought metrics for landscape ecology applications

Find my publications on Google Scholar.

Google Scholar links to free PDFs for most of my publications, but if you need a PDF, contact me and I’ll happily send one!