Since the beginning of 2011, the state of North Carolina has experienced dry conditions and high winds, which have increased the fuel load on the ground. This extreme weather led to several severe wildfire events resulting in nearly 100,000 burned acres. These wildfires caused considerable damage to the Piedmont and Coastal Plains regions’ ecosystems and greatly affected the livelihoods of many North Carolinians.
The two largest wildfires from the 2011 wildfire season occurred in the wetlands of North Carolina, which are primarily underlain with peat soil. Due to its high organic content, peat burns similarly to coal, as a smoldering fire rather than the more iconic flaming fire, and is highly susceptible to ignition during drought conditions. The risk of severe wildfires in the wetlands of North Carolina has increased as a result of growing land-use practices that have altered the hydrology of the ecosystem. Practices such as the construction of roads, ditches and canals to facilitate logging activities drain the peat soils which exacerbates fire hazard. Once ignited, peat fires are extremely difficult to contain and extinguish as they can travel undetected underground and cross fire control lines.
The economic impact of wildfires is not limited to the immediate cost of containing the fire. These challenging peat fires create a strain on resources for firefighters, equipment and water supplies, which in turn creates an imminent hazard for the surrounding habitats and homes. The persistent drought conditions in North Carolina since June 2010, especially in the coastal plains, contributed to the spread and size of the wildfires. Together, these factors caused the Pains Bay Fire (in Dare County) and Juniper Road Fire (in Pender County) to be among North Carolina’s most expensive and damaging wildfires in recent years, destroying over 75,000 acres of wildlife reserves and protected game land.
Recognizing this environmental danger, this past summer a team of eight DEVELOP student interns on the North Carolina Disasters and Ecological Forecasting team at NASA Langley Research Center set out to analyze these two devastating wildfires. The team used remote sensing data to provide its project partners with a tool to enhance their current management and decision making process for wildfires.
Utilizing NASA’s Earth Observing Systems (EOS), the DEVELOP team conducted several analyses on these two extreme fires using remote sensing data and tools in GIS software.
First, the team assessed drought conditions before each fire using a suite of sensors. Three indices were used in order to provide project partners with the methodology to thoroughly analyze drought conditions using NASA EOS. Two of these indices were derived from Landsat 5 Thematic Mapper (TM), while the third was from Aqua and Terra Moderate-resolution Imaging Spectroradiometer (MODIS). The team was able to use the Landsat 5 TM images to perform a long-term study of the ground conditions leading up to the fire because Landsat 5 TM images are taken every 16 days. The MODIS data allowed a near real-time study immediately before the fires because the images are taken daily. Combined, the data products provided a comprehensive analysis of the condition of the vegetation before the fires started and which areas may have been vulnerable to wildfire.
The Normalized Difference Vegetation Index (NDVI) and Normalized Difference Moisture Index (NDMI) were derived from atmospherically corrected Landsat 5 TM images to show changes in vegetation cover and moisture for up to a year before each fire was ignited. The NDVI is used to measure the amount of greenness in vegetation by comparing the Landsat 5 TM red and near-infrared bands. The NDVI can be used as an indicator of plant health because plant greenness portrays how well the plant is carrying out photosynthesis. Similarly, the NDMI is used to measure moisture in the leaves of the plant. This index compares the Landsat 5 TM near-infrared and shortwave-infrared radiation bands to indicate whether the leaves are water-filled or wilted. Low water content in the leaves creates an indicator that the plant is stressed and this can be attributed to drought conditions.
The final index used to assess drought conditions, the Normalized Multi-band Drought Index (NMDI), is a relatively new index that uses MODIS Daily Surface Reflectance product (Aqua and Terra). The DEVELOP team used this index to not only estimate daily real-time drought severity of vegetation health, but also soil moisture. The NMDI utilizes three near-infrared bands that are highly sensitive to soil and vegetation moisture. However, it is important to be familiar with the study area because this is a versatile index that inverts depending on whether vegetation or bare soil is exposed. Due to the high vegetation within both studied areas, the team was able to analyze the derived images more clearly.
Daily MODIS imagery was run through a model made in ArcGIS 10 for Dare County (Pains Bay Fire) and Pender County (Juniper Road Fire) in order to evaluate the vegetation and soil moisture stress two weeks before the fires started and one week after the fires ended for both study areas. This step is meant to be a potential near real-time method for end users to use as an additional measure to monitor an important factor that leads to where fires may spread.
Together, these three indices were used as indicators of stressed vegetation and low soil moisture for both study areas to further show areas vulnerable to wildfire.
The next step of the project included mapping burn severity of the land from the Pains Bay and Juniper Road fires to show the impact of fires on the ecosystem and the magnitude of ecological change they caused. To create these maps, the team used Landsat 5 TM to calculate the Relative difference Normalized Burn Ratio (RdNBR) (Figure 1) from pre- and post-fire Normalized Burn Ratios (NBRs) using various equations in GIS software. The difference Normalized Burn Ratio (dNBR) was also created, but the RdNBR was found to be more accurate for the study area because of the dense vegetation present before the fires. The RdNBR allows for better estimates of burn severity by removing the correlation of dNBR and pre-fire biomass.
Landsat 5 TM was also utilized to create burn scar maps to show the burn extent of each fire. These burn scar maps were composed by stacking three Landsat 5 TM bands (the green band, the near-infrared band and the shortwave-infrared (SWIR) band). This band combination effectively shows the scarring of the land. The green band is highly reflective of leaf pigments, as well as the near-infrared band, that is highly reflective of healthy chlorophyll walls. The SWIR band can penetrate clouds and emphasize the change in thermal properties of damaged-vegetation surfaces. Therefore, this band combination shows scarring of the land due to wildfires by highlighting unhealthy vegetation and damaged areas.
Understanding how severely the land was burned is important for project partners because it indicates which areas received the most damage and where they may first expect areas of vegetation regrowth. NASA EOS is able to capture these results without the need for costly and untimely field work.
Finally, fire-risk assessment maps (Figure 2) were created using Landsat 5 TM data in GIS software to provide project partners with a reliable and accurate tool to aid in fire protection, prevention and soil analysis. A multi-criteria evaluation (MCE) method was used to incorporate multiple land cover components to determine areas of high fire risk. The MCE utilized the fuzzy logic (a new tool to ArcGIS 10) to describe geographic components that are essentially continuous in nature. This method also tries to correct the inherent error added when crisp lines are formed to describe seamless geographic phenomenon, allowing for risk values to be continuous rather than categorical.
There are five variables that were chosen for inclusion in the fire risk map: proximity to the wildland-urban interface (WUI), proximity to roads, slopes, fuel type, and soil type. These factors were chosen because they contribute to fire ignition, intensity or movement. Each of these factors was normalized before it was given a fuzzy membership. This normalization included reprojecting each layer into the same coordinate system, normalizing the spatial extent to include both study areas, and verifying that each spatial resolution was 30 meters. After various fuzzy memberships were applied to each component of the fire-risk map, the “Fuzzy Or” method, which takes the maximum cell value for every pixel, was used. This scenario is the worst-case scenario, and exemplifies peak drought condition and risk. The final map output effectively displays areas at high risk for future fire ignition based on the five previously mentioned parameters.
The North Carolina Disasters and Ecological Forecasting team partnered with Tim Howell, Fire Environment Branch Head at the North Carolina Forest Service (NCFS), and Kelley Van Druten, WUI Specialist at the Alligator River National Wildlife Refuge (ARNWR). Both organizations are interested in this project because it represents an opportunity to use NASA EOS to aid in their wildfire analysis methods and decision making processes. Currently, the NCFS and ARNWR rely on outside agencies to assist with their fire-risk predictions and burn severity analysis. However, through their partnership with NASA DEVELOP, the NCFS and ARNWR have been given the tools to create these products on their own and use them in near real-time.
The methodologies developed during this project will serve to complement existing methodologies and assist each partner organization in creating a possible future product of higher accuracy, reliability and efficiency.
In addition to a technical paper and data products, a user-tutorial was provided to the partners to assist in familiarization with NASA remote sensing capabilities and aid in replicating and adding the project’s methods to their current fire assessment strategies. Furthermore, a team member was able to share project results at the North Carolina Prescribed Fire Council's Annual Meeting this fall.
Figure 1. Post-fire Relative difference Normalized Burn Ratio (RdNBR) images created using Landsat 5 TM data in ArcGIS 10 to qualitatively display areas of high burn severity for the two study areas (click for larger image).
Figure 2. The fire-risk assessment map product created by the NC Disasters and Ecological Forecasting team that will assist project partners with future fire-risk planning and mitigation for the two selected counties containing the wildfire study areas (click for larger image).