LHill_Geog592_ClassProjectFinalPaperDS_Final
- 1. Characterizing Pre-Fire Vegetation As Fire Fuel in the 2002 Williams Fire, San Dimas Experimental Forest, California Geography 592, SDSU ABSTRACT Analysis of whether certain types of vegetation classes are associated with higher intensity thermal features as frontline fire fuel, and to compare temperature gradients between the front and rear fire fuel areas. Lisa Hill May 10, 2016
- 2. Hill, 1 Objectives The San Dimas Experimental Forest (SDEF) is managed by the USDA Forest Service (USFS) as a field laboratory, containing chaparral and other Mediterranean-type plant communities. To measure the thermal radiance of wildland fires, the USFS utilizes the Firemapper ® system that accurately samples emitted thermal-infrared heat. Firemapper wildland fire images were orthorectified and assembled into georeferenced mosaics. Image spatial resolution ranged from 5 to 9 meters, and aerial repeat-pass images were taken on September 23, 2002 between 14:00 to 15:15 Pacific Time Standard (PST). The Firemapper images were overlaid on high altitude color infrared subset images to identify the vegetation fire front, in addition to thermal features within and behind the fire front. The objective of this study is to identify whether certain types of vegetation classes are associated with higher intensity thermal features as frontline fire fuel, and to compare temperature gradients between the front and rear fire fuel areas. Within ERDAS Imagine 2013, I utilized an unsupervised classification process to map vegetation classes within fronting spreading zones of the 2002 Williams Fire in the SDEF, and assessed the intensity values in thermal infrared images of selected fire spread events which occurred on September 23, 2002. These were compared and analyzed within ArcMap 10.3 using the Zonal Statistics as Table tool. Imagery A three band color infrared aerial photographic image of 1 meter nominal spatial resolution was obtained from the United States Geologic Survey (glovis.usgs.gov), and then orthorectified using ground control points obtained through visual assessment of ArcMap base map imagery. The aerial thermal infrared (ATIR) images had been orthorectified by the Center for Earth Systems Analysis Research (CESAR) at San Diego State University, and provided a reasonable level of alignment with the color infrared image. Software Used ERDAS Imagine 2013 ArcMap 10.3.1 Methods Two subsets from the USGS aerial color infrared image frame were selected to identify the vegetation fuel prior to the 2002 Williams fire area within the SDEF (Images 6b1.img and 8b1.img). Two sites were selected in the study area based on their fire spread, topography contour and variety of vegetation in
- 3. Hill, 2 the area. Using ERDAS Imagine, each orthorectified CIR image represented below was subset to provide a smaller image that matched to the specific fire paths identified by the selected USFS fire image mosaics (created by Pete Coulter, SDSU Geography Technical Staff). Site 1 - Ortho_subset_1.img 8b1.img Site 2 - Ortho_subset_2.img of 6b1.img Covering Big Dalton Reservoir, North of Big Dalton Reservoir, SDEF San Dimas Experimental Forest (SDEF) The two CIR subset images were classified to identify vegetation and other land cover types within the 2002 Williams fire path. The classes included subshrub, shrub, tree, herbaceous, soil, and built materials. An unsupervised classification process was utilized. I analyzed a 2000 Google Earth image within ERDAS Imagine in order to identify vegetation as subshrub, shrub, tree, herbaceous, and soil with paved streets and concrete areas. Subshrub and herbaceous were combined into one class due to the limited image spatial resolution which prevented identification and separation of these vegetation types. The following unsupervised classification steps were completed for each subset image: (1) Using ERDAS Imagine, selected the ISODATA algorithm choosing 10 iterations and 36 clusters for each subset image’s spectral-radiometric feature, and selected minimum spectral distance to assign clusters or each candidate pixel. Number of ISODATA iterations and clusters selected were based on reviewing prior Firemapping studies, class labs and recommendation from SDSU Geography remote sensing staff. (2) Within the spectral-radiometric feature space of the CIR image chose minimum pixel selection (20) threshold level and minimum cluster size (>.05%).
- 4. Hill, 3 (3) Using ArcMap 10.3, added the CIR subset images and pixels clusters were classified according to 7 cluster classes (zero classification (0)), soil (1), subshrub/herbaceous (2), shrub (3), tree (4), water (5), built (6), and reclassify (7). Vegetation cluster class labeling was based on visual interpretation of the original CIR imagery in correspondence with the location of pixels for each cluster class. (4) Assigned appropriate information class names or labels to the cluster classes. The assignment of cluster names (cluster labeling) required familiarity with the area of interest. (5) Once classes were assigned values in ERDAS Imagine, unidentified (“reclassified”) complex clusters were reclassified into 15 different clusters using an approach called “cluster-busting” (using a multiple phase approach) by running the unclassified pixels through ISODATA again. This resulted in a vegetation class map of each study site. (6) Within ArcMap, identified and examined the reliability of the vegetation classification using “heads up” visible interpretation based on the original CIR subset image and overlaid the vegetation class map on the ArcMap World Map layer. (7) Overlaid selected USFS fire mosaic images to identify fire intensity based on the vegetation land classifications identified. These fire mosaics demonstrate the fire path occurred from 2:16 to 2:50 on September 23, 2002. (8) Applied color density slice to fire mosaics to identify heat intensity based on DN’s and vegetation classes that assisted in identifying fire frontline and backline fuel composition. Roads, built surfaces and water were also classified to identify unburned areas. (9) Within ArcMap 10.3 using the Zonal Statistics as Table tool, calculated the pixel count and median Celsius temperature for fire front line and back line polygon sampling areas to estimate Sites 1 and 2 specific vegetation class temperatures and their sampling DN density. Results A conjecture was made that certain types of vegetation might burn more intensely than other vegetation, based on the higher DN’s that would also show up in the density slice layer. Red colors represented the highest heat within front fire line, with light yellow to white colors representing the coolest temperatures in the rear fire line. In Image 1, the fire is moving right to left toward the blue area (water within the Big Dalton Reservoir). Site 1 (Ortho_subset_1.img 8b1.img) covers the Big Dalton Reservoir and Site 2 (Ortho_subset_2.img of 6b1.img) lies directly to the northeast of the reservoir.
- 5. Hill, 4 Classes within the vegetation map layer were color identified as: 1 - Soil 2 – Sub-shrub (herbaceous and sub-shrubbery) 3 – Shrub 4 – Trees 5 – Water 6 – Built (roads, concrete) Site 1 - Overlay fire image at 2:16 PM Site 2 - Overlay fire image at 2:30 PM Big Dalton Reservoir Area, SDEF Northeast of Big Dalton Reservoir Fire mosaic layer with density slice (white to red color range) overlaid on classified vegetation map. Multicolored rectangle polygons within fire mosaic layer identifies sampling areas to measure fire temperature in Celsius.
- 6. Hill, 5 Site 1 - Overlay fire image at 2:43 PM Big Dalton Reservoir, SDEF Site 1 - Overlay fire image at 2:50 PM Site 2 – Overlay fire image at 2:50 PM Big Dalton Reservoir, SDEF Northeast of Big Dalton Reservoir
- 7. Hill, 6 Flickering the fire mosaics over the vegetation image layer indicated subshrub provided the greatest concentration of fuel source, followed by shrub and trees. To identify the vegetation type that yielded the highest heat response (in DNs), I used the Raster Statistics Table tool in ArcMap on selected hot spots in both images. Median DN values within the vegetation polygons were measured in degrees Celsius Statistical Analysis Within the fire mosaics, certain areas were selected toward the front and rear of the fire line as polygon samples (see attached Frontline and Backline Fire Mosaic Table). Although the front line highest temperature registered for shrub and trees indicated the highest median temperatures, the majority of fire fuel available came from sub-shrubbery. The back fire line polygons indicated the soil retained hotter median values. These results may indicate smoldering coals and debris with left over from vegetation fuel still burning (subshrub, shrub and trees).
- 8. Hill, 7
- 9. Hill, 8 Conclusion From the Unsupervised Classification vegetation map and the statistical tables, it appears trees and shrubs provide fire fuel that can generate the highest heat. However, based on the DN pixel concentration shown in the prior table and by vegetation identification from Google imagery, the greatest concentration of fire fuel provided appears to be from subshrub vegetation. Although subshrub provides a greater availability of fuel source, subshrub does not appear to generate as hot of a fire compared to measured trees and shrub temperatures. Reviewing the table data, frontline areas indicated front line fire temperatures ranging from 157° to 307° Celsius, and back fire line temperatures between 61° to 167° Celsius. Site 1 Clip 16 does not seem to coincide with the other site area measurements, which could be due to site temperature selection error. Results and interpretations of the fire intensity patterns within different vegetation types could also be affected by individual growth form map accuracy, as well as geometric misalignment between both image sets. Remote sensing images at higher resolution of the fire line area before the 2002 Williams fire would improve identification of key fuel fire species. Also, consideration and analysis of the wind, concentration of exposed soil and topography influences must be taken into account and were not analyzed in this project. This analysis aids in identifying how different types of vegetation fueled the 2002 San Dimas fire, however further analysis is required of topography, wind, soil moisture, and daily temperature within the study area.
- 10. Hill, 9 References ArcMap 10.3 World Map USFS Williams Fire JPEG map - http://www.fireimaging.com/homepage.html Stow, D. A., Riggan, P. J., Storey, E. J., & Coulter, L. L. (2014). Measuring fire spread rates from repeat pass airborne thermal infrared imagery. Remote Sensing Letters, 5(9), 803-812.