This is a collaborative Earth Lab Certificate project by Eric Gottlieb and Heidi Yoon studying post-wildfire recovery.

• Wildland fire is a multifaceted natural phenomenon of increasing importance to both human and ecological communities. In this project, we explore the post-wildfire recovery for the 2016 Chimney Tops 2 Fire by spatially quantifying the vegetation recovery using hyperspectral reflectance data and by categorizing the burned areas by expected fire intensity using weather, topography and fuels data.
• This project highlights how high spatial resolution (1-meter) remote sensing measurements, such as LiDAR topography and hyperspectral reflectance data, can be used to study fire recovery on the order of the spatial variation on the ground and how lower resolution (30-meter) data, such as weather and fuels, can be integrated to inform and guide a study over a large spatial area.
• In this repository, we include example notebooks that process and analyze LiDAR topography, hyperspectral reflectance, weather data, biophysical and vegetation models to assess post-wildfire recovery.

For more background about our project, please see our blog post and its figures in the Reports folder and the Graphics folder (blog-figure1-3.png and ig-raws.png, ig-raws-rose.png, ig-raws-rose-gust.png), respectively.

To install and run the python environment for this project, please use the instructions below.

1. Download the file neon-environment.yml from this repository, which contains instructions on how to install the environment, into this directory.
2. Create the environment by running: conda env create -f neon-environment.yml.
3. Once the environment is installed, activate it by running: conda activate earth-analytics-neon.

• os
• matplotlib
• numpy
• requests
• h5py
• geopandas
• earthpy
• folium
• rasterio (Workflow B only)
• rioxarray (Workflow B only)
• gdal (Workflow B only)
• richdem (Workflow B only)

Raster data

1. NEON LiDAR Ecosystem Structure

• Reference: National Ecological Observatory Network. 2022. Data Product DP3.30015.001, Ecosystem structure. Provisional data downloaded from https://data.neonscience.org on March 31, 2022. Battelle, Boulder, CO, USA NEON. 2022.

2. NEON Spectrometer Reflectance

• Reference: National Ecological Observatory Network. Spectrometer orthorectified surface directional reflectance - mosaic (DP3.30006.001), RELEASE-2022. https://doi.org/10.48443/5er3-8n49. Dataset accessed from https://data.neonscience.org on April 15, 2022

3. NEON LiDAR Elevation Digital Terrain Model

• Reference: National Ecological Observatory Network. Elevation - LiDAR (DP3.30024.001), RELEASE-2022. https://doi.org/10.48443/ymmp-fr93. Dataset accessed from https://data.neonscience.org on April 18, 2022

4. LandFire Biophysical Setting and Vegetation Departure Grids.

• Reference: LANDFIRE: LANDFIRE Remap 2016 Biophysical Settings (BPS) CONUS layer.(2020, October - last update). U.S. Department of Interior, Geological Survey, and U.S. Department of Agriculture. Available: https://landfire.gov/bps.php. Data accessed April 3, 2022.
• Reference: LANDFIRE: LANDFIRE Remap 2016 Vegetation Departure grid (VDEP) CONUS layer.(2020, October - last update). U.S. Department of Interior, Geological Survey, and U.S. Department of Agriculture. Available: https://landfire.gov/vdep.php. Data accessed April 3, 2022.

Vector data 5. Chimney Tops 2 Fire Perimeter

• Reference: MTBS Data Access: Fire Level Geospatial Data. (2022, February - last revised). MTBS Project (USDA Forest Service/U.S. Geological Survey). Available: https://mtbs.gov/direct-download. Data accessed April 3, 2022.
• Available for download in this repository as Release v1.0.0 fire-boundary

6. Great Smoky Mountains National Park Perimeter

• Reference: National Park Service- Land Resources Division. Great Smoky Mountains National Park Boundary. (December 30, 2019 - last revised). Available: https://grsm-nps.opendata.arcgis.com. Data accessed March 28, 2022.
• Available for download in this repository as Release v1.0.1 grsm-boundary

Time-series data 7. Remote Automated Weather Station (RAWS) data for Indian Grave RAWS Site.

• Reference: Indian Grave RAWS hourly data fof November 2016. U.S. Forest Service Fire and Aviation Management Information Technology Portal. 2022. FW13 hourly station data (station ID:407603). Downloaded from https://famit.nwcg.gov/applications/FAMWeb on April 16, 2022.

Tabular data 8. NEON Plant Presence and Percent Cover

• Reference: NEON (National Ecological Observatory Network). Plant presence and percent cover (DP1.10058.001), RELEASE-2022. https://doi.org/10.48443/pr5e-1q60. Dataset accessed from https://data.neonscience.org on April 18, 2022.

The overall workflow of the project is shown below.

The project workflow consists of two stand-alone workflows, Workflow A for vegetation recovery and Workflow B for fire intensity, that can each be used separately and together to relate vegetation recovery to fire intensity in specific regions of the burned area.

Workflow A is a vegetation recovery analysis in which the vegetation recovery of an 1000-m by 1000-m area within the burn perimeter can be evaluated first by calculating vegetation indices (NBR: normalized burn ratio, NDVI: normalized difference vegetation index, MSAVI: modified soil adjusted vegetation index). Then the index that can best discriminate between unburned and burned areas is determined by the random forests algorithm. Finally, multiple endmember spectral band analysis is used to classify the vegetation at a sub-pixel level into green growth [green], standing dead or dormant vegetation [deadwood], or burned areas [char].

• To use Workflow A, run the notebook vegetation_indices.ipynb to calculate the vegetation indices using a downloaded reflectance file and plot the results using matplotlib and earthpy.

Workflow B quantifies key environmental variables in the fire environment using (LiDAR Digital Elevation Model 1-m spatial resolution data for topography, (LANDFIRE grids) 30-m spatial resolution data for fuels, and time series data (RAWS wind direction) for weather. Understanding the burned area antecedent physiographic and biophysical conditions for the Chimney Tops 2 Fire allows one to relate variability in fire effects to these environmental variables that theoretically affected the fire’s intensity.The fire spread vector is quantified by determining the average wind direction from the Indian Grave Remote Automated Weather Station (RAWS) site on November 28th (with the assumption that wind direction is a proxy for spread direction). For the Chimney Tops 2 wildfire and other applications, the workflow assumes a consistent wind vector during the period in which the fire burned through the area being analyzed in order to ascertain the spread direction. The spread direction is used in conjunction with 1-m NEON LiDAR Elevation data (NEON Elevation- LiDAR: DP3.30024.001) to classify slopes as a function of their alignment with the wind direction, which is an important factor in fire intensity. Theoretically the most intense fire behavior based on wind and slope alone would have occurred in places where the wind was blowing directly upslope (i.e., azimuth of wind and aspect are subequal) because convective heating from , which we classify as head fire environments. Conversely, in locations where the wind was blowing downslope (i.e., ~180 degrees between wind azimuth and aspect), fire behavior is theoretically expected to be least intense (because the site should be sheltered from wind and convective preheating is minimal), which we classify as backing fire environments. The intermediate condition between head and backing fire environments would have occurred on slopes where the aspect was orthogonal to the wind direction, which we classify as flanking fire. The raster reclassification approach we use for this step is as follows:

Head fire environment: Aspect - Wind Azimuth >-45 and <45 Flanking fire environment: Aspect - Wind Azimuth (>45 and <135) or (<-45 and >-135) Backing fire environment: Aspect - Wind Azimuth >135 or <-135

A secondary classification using LiDAR data can be made based on steepness of slope. Head fire environments with the steepest slopes theoretically represent the areas of highest intensity. Based on established research, we will classify slopes >25 degrees in head fire environments as representing the maximum intensity environment as one end member, whereas slopes between 10-20 degrees in backing fire environments are thought to represent minimal burning conditions and are an opposite end member. Having these two end members allows us to incorporate a third variable into our analysis- that being fuels (i.e, vegetation type and condition)- to test how post-fire recovery correlates to theoretical burn intensity. Mapping the reference conditions biophysical setting (LANDFIRE BPS) of the Chimney Tops fire area and how the vegetation has departed from its historical conditions (LANDFIRE VDEP)) allows us to focus on forested areas that have experienced the most (or least) departure from their natural range of variability. The key hypothesis that will be tested is whether post-fire recovery in head fire environments on steep slopes show significant differences from backing fire environments on moderate slopes in areas with similar biophysical and vegetation departure characteristics. Our analysis will ultimately involve plotting results of the three different vegetation indices measured in Workflow A versus our various environmental variable end members.

• To use Workflow B, run the notebook wind_terrain_fuels.ipynb to reclassify the burned area of interest using a downloaded LiDAR DEM, time series wind data, and LANDFIRE fuels raster data and plot the areas of head, flanking and backing fire in the various fuel types using matplotlib and earthpy.

Workflow A

• The vegetation indices of any NEON Airborne Observation Platform Hyperspectral Reflectance hdf5 file can be calculated by Workflow A. This could include the NEON spectrometer orthorectified surface directional reflectance - mosaic data product used in this workflow and the NEON spectrometer orthorectified surface directional reflectance - flightline data product.
• In addition, any boundary shapefile for the Reflectance hdf5 file could also be plotted with the reflectance in Matplotlib using the notebook vegetation_indices.ipynb.

Workflow B

• A given area analyzed by Workflow A (typically 1 km^2 tile), or selected independently, can be analyzed in Workflow B to reclassify the area at 1-m spatial resolution by slope, aspect and at 30-m by fuels conditions into areas of expected theoretical fire intensity.

The post-wildfire-recovery project is under the MIT License.