In many California forests, wildfires are increasingly burning at high severity over large areas, creating landscapes with low potential for natural regeneration. Post-fire tree planting is accordingly playing an important role in post-fire forest management, but the large extent of high-severity burned areas is stretching resources. Accurate prediction of natural regeneration patterns could improve the efficiency of post-fire restoration efforts, but current methods for predicting regeneration are limited by (1) lack of detailed information on the location and density of the surviving trees that serve as seed sources and (2) coarse representation of seed dispersal processes. We are working to improve predictive accuracy through a “big data” approach that explicitly maps seed-source trees and accounts for biologically realistic seed dispersal processes. Our study aims to establish a new methodology that involves inferring individual-tree dispersal patterns without needing to isolate individual trees.

First, we are creating spatially extensive (~300 ha) maps of surviving trees (potential seed sources) from recently burned landscapes by collecting drone imagery and processing it into 3D forest models and 2D stem maps using automated photogrammetry and tree detection algorithms. We are also conducting intensive plot surveys across these same landscapes to quantify the spatial variation in natural regeneration and ground-truth the drone-derived tree maps. To predict tree recruitment across these landscapes, we are applying a novel spatially explicit modeling framework to simultaneously characterize (a) individual-tree dispersal distances and (b) the joint contribution of multiple seed sources (mature trees) to seed rain in a given site. We will use our fitted model to make spatially continuous predictions of natural regeneration at high resolution (10 m) across these landscapes. We will package our resulting model into a web application for easy application to future fires.

Young, D. J., Koontz, M. J. & Weeks, J. (2022). Optimizing aerial imagery collection and processing parameters for drone-based individual tree mapping in structurally complex conifer forests. Methods in Ecology and Evolution, 13, 1447– 1463. https://doi.org/10.1111/2041-210X.13860

Metashape Github: https://github.com/ucdavis/metashape

Presentation to Ecological Society of America 2022: https://www.eventscribe.net/2022/ESA/fsPopup.asp?Mode=presInfo&PresentationID=1106070

For more information on this project please visit:

The project websites:
https://www.changingforests.com/drones-ai/
https://changingforests.com/current-research/dispersal-modeling/

Contact Information:

Derek Young, Ph.D. (PI)
djyoung@ucdavis.edu

CAL FIRE Forest Health Research Program
FHResearch@fire.ca.gov