Using NASA Earth Observations to Monitor Post-Fire Vegetation Recovery in the Colorado Front Range

I led a team in conducting this analysis for NASA DEVELOP in partnership with the US Forest Service Rocky Mountain Research Station. We used Landsat TM, ETM+, and OLI data, Sentinel-1 C-SAR data, ALOS-2 PALSAR-2 data, and ancillary datasets to conduct time-series analysis of post-fire forest recovery, model post-fire tree regeneration, and model conifer and deciduous tree cover.

  • Project lead Eric Jensen
  • Collaborators Kristen O'Shea, Anthony Vorster, Charles Rhodes, Zachary Werner, Lauren Kremer, and Audrey Colley
  • Website Project Webpage on NASA DEVELOP site
  • Tools Google Earth Engine, R, ArcGIS Pro
  • Completed August 2020

ABOUT THE PROJECT


Project summary

Forest composition and structure in the Colorado Front Range has been altered by changing wildfire regimes. In particular, increased moderate- and high-severity fire significantly reduces forest cover following fire and often results in reduced seedling regeneration. Reduced tree canopy regrowth has chronic effects on upland ecological function and downstream water quality. This project partnered with the US Forest Service to estimate long-term vegetation recovery following four Colorado Front Range fires between 1996 and 2002—the Bobcat, Buffalo Creek, Hayman, and High Meadows fires—using Landsat 5 Thematic Mapper (TM), Landsat 7 Enhanced Thematic Mapper (ETM+), and Landsat 8 Operational Land Imager (OLI). The random forest algorithm was applied to produce maps of percent forest canopy cover for coniferous trees, deciduous trees, and all trees using time-series variables for pre- and post-fire as inputs. Similarly, maps of post-fire seedling regeneration were produced using random forest for coniferous trees, deciduous trees, and all trees using ecological drivers (soil, climate, fire, and topography) and pre-fire remote sensing predictors. Relationships between ecological drivers of post-fire vegetation trajectories were also evaluated. Additional analyses were conducted to (1) assess whether seedlings could be detected by Landsat or synthetic aperture radar (SAR) time-series analysis (2) assess pre-fire and post-fire Landsat variables against pre-fire and post-fire tree cover estimates to evaluate whether magnitude of forest change can be detected. Understanding variables that influence vegetative recovery, vegetation type conversion, and watershed characteristics will aid forest restoration efforts and water quality management.

Study area

Study area map depicting the four fires analyzed in the study, Bobcat, Buffalo Creek, Hayman, and High Meadows. The regional map on the right shows the location of each fire in the Front Range of Colorado, overlaid upon a layer depicting tree cover from the National Land Cover Database. The fire maps on the right depict classified fire severity using dNBR and RdNBR from the Monitoring Trends in Burn Severity dataset for each fire analyzed here.


Key figure

Our analysis had two key components, 1) modeling current tree cover percent of deciduous and conifer trees (left) and 2) modeling suitability of deciduous and conifer tree regeneration (right). The figure on the left depicts conifer tree cover in the Hayman Fire, which is important for identifying seed sources in ponderosa pine forest. The figure on the right depicts suitability for conifer tree regeneration, which can help the USFS crews to identify areas where tree regeneration is likely to occur and areas where reestablishing tree cover will only come through management intervention.