An Evaluation of Environmental Effects on Forest Regeneration of the 1988 Yellowstone Fire

Feng Zhao, University of Maryland, College Park, fengzhao@umd.edu (Presenter)
Ran Meng, University of Utah, mengran07@gmail.com
Chengquan Huang, University of Maryland, cqhuang@umd.edu
Maosheng Zhao, University of Maryland, zhaoms@umd.edu

The 1988 Yellowstone Fire was the largest observed since the national park was established in 1872, affecting approximately 45% (about 4000,000ha) of the park. Understanding forest recovery following this catastrophic fire event is critical for understanding long term regional C fluxes and informing policy makers on the impacts of disturbance on carbon cycle. Remote sensing technique provides a viable tool to examine long-term effects of environmental variables on forest regeneration after such a large fire.

In this study, we examined the effects of environmental factors on the regeneration of major forest species following the 1988 fires in YNP using time series (1988-2010) forest regrowth product from Landsat and Vegetation Change Tracker (VCT). Three major steps were involved in this study. First, we validated the VCT forest regrowth product using high spatial resolution images from National Agricultural Imagery Program (NAIP) and Google Earth. Second, we used the Random Forest (RF) machine learning algorithm to evaluate the impacts of topography and climate variables on the forest regeneration status (recovered vs. non-recovered) for major forest species in YNP. Third, we focused on the recovered forests and examined how the rate of forest recovery (using years till recovery as response variable) has been affected by environmental variables. Preliminary results show that the VCT regrowth product is highly reliable, with overall accuracies of 80% for all forest species and 76% - 88% for specific forest types. The RF models revealed that whether regeneration occurred or not after/during? more than the two decades after the fire were strongly correlated to topography and climate variables with error rates ranging from 4.5% - 18.2% for major forest species in YNP. The dominant variables used by the RF models in determining forest regrowth status included aspect, elevation, soil type, and post-disturbance climate. Within the regenerated stands, environmental variables explained 48.6%-79.5% of the variance of the years it took for the regenerating forests to become spectrally detectable with Landsat. For all the species examined, terrain played the most important role in determining the recovery rate following the 1988 fires.

Presentation Type:  Poster

Session:  Theme 1: Tracking habitat change through new integrative approaches and products   (Mon 1:30 PM)

Associated Project(s): 

• Huang, Cheng: Role of Forest Disturbance and Regrowth in the US Carbon Budget ...details

Poster Location ID: 71