Assessing the Potential for Wildfire Using ERS-2 SAR Imagery

Laura L. Bourgeau-Chavez, Gordie Garwood, Kevin Riordan, Brad Cella, Sharon Alden, Mary Kwart and Karen Murphy, Michigan Research and Development Center, General Dynamics AIS, National Park Service, U.S. Fish & Wildlife Service

Wildfire is a common occurrence in boreal Alaska and natural resource agencies devote considerable resources to fire management and suppression. Currently, these agencies rely on the Canadian Forest Fire Danger Rating System’s Fire Weather Index (FWI) for the assessment of the potential for wildfire. FWI is based solely on point-source weather data collected daily in a sparse network across the state of Alaska. Although currently invaluable, there are issues with the FWI system and the agencies recognize a need for improvement. Recent research (NASA grant NAS5-03113) has been conducted to enhance the prediction of wildfire potential in Alaska using satellite C-band (5.3 cm wavelength) synthetic aperture radar (SAR). SAR is sensitive to the moisture content of the features being imaged including vegetation and soils. The relationships between in situ soil moisture, C-band backscatter, and fire danger codes have been under investigation for several years at a variety of burned and unburned sites in interior Alaska. Focus has been on recently burned (0 - 7 years) boreal forests because they allow moisture in the organic ground surface layer to be measured directly from a satellite sensor without interference of the forest canopy and because they are a common feature across the Alaskan landscape. Comparisons of unburned forests to adjacent burned forests have revealed similarities in the temporal patterns of in situ moisture monitored throughout a fire season. Thus, soil moisture monitored from burned forests may be used as surrogates for unburned adjacent forests.

In our research to improve fuel moisture monitoring for fire danger prediction, we have focused on three approaches: 1) to use C-band SAR to initialize and calibrate the existing FWI codes; 2) to map soil fuel moisture directly across a burned landscape using SAR algorithms developed for different burn severity types; and 3) to map fuel moisture across a landscape using time-series analysis of SAR data. The first two of these approaches have been demonstrated and the third continues to be investigated.

Canadian and Alaskan resource managers have noted issues with the determination of the spring start-up values of one of the FWI codes, drought code (DC). DC is an index of moisture in the lower duff layer and it has a 52 day lag period, thus it is most affected by inaccurate start-up values. Another problem with the DC in Alaska is that of mid-summer variations in measured moisture values within permafrost regions that are not accounted for in the FWI system. We developed an algorithm relating ERS-2 SAR backscatter from recently burned forests to DC. By measuring the ERS-2 backscatter from a burn scar in spring, after snowmelt, we can predict the DC and use this value to initialize the weather-based DC. Figure 1 shows an example of this process for the Fort Greely weather station using ERS-2 backscatter from the neighboring 1999 Donnelly Flats burn. The DC based on the default spring initialization value of 15 is shown as the solid black line of Figure 1. By initializing the code in the spring with SAR from a 2 May 2004 image (dashed line), the DC jumps from 55.6 to 271, indicating much drier (higher DC indicates drier status) conditions than what weather alone was predicting. SAR-derived DC values are plotted throughout the summer on Figure 1 showing that the new DC curve is more in line with what was observed from the SAR sensor until late July when the SAR sensor shows the site as getting wetter, while DC continues to climb. This dramatic increase in moisture observed in the SAR may be due to melting of frozen ground layers.

The second approach that we investigated was to map spatially varying soil moisture across a burned landscape. To do this, a combination of Landsat and C-band SAR was used. Maps shown in Figure 2 show soil moisture within a burned boreal forest and how it changes over the 2002 summer. The maps were created from ERS-2, C-band imagery using algorithms developed that relate surface soil moisture to ERS-2 backscatter. The technique involves subdividing the burn scar by burn severity class and then applying SAR algorithms developed for each of three burn severity classes (light, moderate, and severe) to convert backscatter to soil moisture. To categorize sites by burn severity to the ground, Landsat data and field collections were used. Dividing the sites in this way reduces errors due to variation in surface roughness, revegetation, and soil type, all of which influence SAR backscatter and all of which are affected by burn severity.

The maps were validated with independent in situ data which resulted in an overall 13.1% rms error. This soil moisture retrieval technique represents a first step in landscape scale monitoring of surface soil moisture and research is ongoing to expand to unburned areas. The retrieval of fuel moisture information from SAR imagery represents an innovative technique which would allow fire managers to directly assess the potential for wildfire over large regions at high spatial resolution. Such moisture monitoring has applications not only for fire danger assessment but also for modeling carbon gas exchange, net primary productivity, and assessing possible effects of a changing climate on hydrologic regimes.

Click here to download a copy of the newsletter featuring this article