1. Michigan Tech Research Institute (MTRI)  Michigan Technological University 3600 Green Court, Suite 100  Ann Arbor, MI 48105 (734) 913-6840 – Phone  (734) 913-6880 – Fax  www.mtri.org Moderately Dry Date (11 July 2009) Co-Author Contact Info: Brigitte Leblon bleblon@unb.cal 506 453-4924 Author Contact Info: Laura Bourgeau-Chavez lchavez@mtu.edu 734 913-6873 Developing Improved Predictions of Fuel Moisture with Radar in Alaskan Boreal Ecosystems Fire danger monitoring can be improved with the inclusion of remote sensing data. Specifically, satellite synthetic aperture radar (SAR) data can be used to provide a more direct measure of moisture conditions to improve Fire Weather Index (FWI) predictions. Satellite Imaging Radar backcatter has been found to be highly correlated with the FWI Drought Code (DC) as the long wavelengths penetrate the vegetation and ground to detect moisture in the deeper organic soil layers Two Methods Developed to Assess Fuel Moisture using C-band Radar 1.Application of SAR for FWI Drought Code Initialization in Spring and recalibration through the fire season using single channel C-band SAR data. Satellite sensors available since 1992. 2.Advanced Polarimetric SAR retrieval of moisture for mapping capability. Satellites available since mid-2000s. Burned Boreal Forests 27 July 1992 Low Fire Danger 12 August 1992 High Fire Danger 22 June 1992 Moderate Fire Danger 26 August 1994 Extreme Fire Danger ©ESA 1992-5 Burned Forest DC 270 DC 231 DC 250 DC 616 Large area averaging averages out spatial variation so that a relationship can be developed Predictive Equation based on multiple sites DC = -45.592 * (ERS-2 backscatter dB)-114.68 1. Application of SAR for Drought Code Initialization in Spring Fort Greely Weather Station, Delta Jct, AK Spring image (2 May 2003) used to initialize DC for the Ft. Greely weather station -- Donnelly interrogation area backscatter = -8.60 dB -- SAR predicted DC = 281.7 Weather based DC (default of 15 used mid-April) = 79.4 on 2 May  weather underestimated drought conditions Donnelly Flats 1999 burn Interrogation Area 2 May 2003 ERS Image Weather Station Predictive Equation DC = -45.592 * (ERS-2 backscatter dB)-114.68 ERS Radar Drought Code Initialization vs. Default Delta Junction example 2003 fire season Large Area C-band SAR backscatter Relationship to DC in low biomass sites a.Corrects spring initialization values for existing weather stations b.Correction of DC values in mid summer due to frozen soil thaw c.Allows additional point locations to increase the sampling area d.Moisture in burned areas demonstrated as strongly correlated to moisture patterns in unburned adjacent areas e.Alaskan Fire managers indicated that such a system to improve the current weather-based approach would be of high utility % Volumetric Moisture Application of the polarimetric algorithm for “all sites” to wet, moderate and dry Radarsat-2 image dates resulting in accuracy of 6.7% volumetric moisture content (RMSE). Mature forest areas (> 1.7 kg/m 2 biomass) are removed from maps. All Sites Burned Sites Unburned Sites Predicted 12-15 cm % Volumetric Soil Moisture Actual 12-15 cm % Volumetric Soil Moisture 2. Advanced Polarimetric SAR Retrieval of Moisture Beneath Vegetation Predicted vs. Actual Moisture Content Using polarimetric information (e.g. D max ) with the SAR backscatter in algorithms improves predictive capability Previous studies using single channel C-band SAR were limited by confounding factors of surface roughness and biomass. Polarimetric variable D max Single Channel C-HH D max – maximum degree of polarization does not vary much by date (moisture condition) and appears to increase with increasing structural complexity C-HH backscatter is strongly affected by moisture status Goal: to further develop radar remote sensing methods to monitor and map spatially explicit organic layer fuel moisture for fire danger prediction. Caveat: SAR sensors are sensitive to moisture of image elements, but also vegetation structure and surface roughness. Solution: Polarimetric SAR can help account for variability in these confounding factors and is promising for future application. Combining SAR variables that appear strongly correlated to vegetation structural complexity with variables strongly related to organic layer soil moisture improved empirical algorithms by 27-33% over traditional single channel SAR and four band SAR backscatter capability. Next Steps To extend this research, additional sites should be evaluated that have greater biomass as well as a range of structure L-band (~24 cm wavelength) data should be evaluated across a range of forest sites (including > 3 kg/m 2 ) C-band (~5.7 cm) data should be evaluated across additional low biomass sites (< 3 kg/m 2 ) Empirical and physically based models should be studied in tandem to further theoretically understand the polarimetric scattering for expanded application Dry Date (23 August 2010) Wet Date (09 August 2008) Soil Moisture Maps from Polarimetric Radarsat-2 Predictive Algorithm for All Sites: %VMC = 319.31 D max -612.50 Unpol max + 5190.4 C-VH -266.37 R 2 = 0.77 This work was funded by NASA grants NAS5-03113, NAS-98-129, NAG-51-0097, NNX09AM15G and NNG04GR24G; Canadian Space Agency SOAR grant #445; and an NSERC Discovery grant.