Abstract
An understanding of the relationship between the physical characteristics of the landscape and fire occurrence is necessary to determine the potential risk of fire to human development adjacent to wilderness. Boundaries of severe fires were delineated on recent aerial photography (1994) and correlated with United States Forest Service (USFS) fire records of the area. A geographic information system (GIS) was then used to determine which attributes of the physiography in the study area played leading roles in influencing the accuracy of aerial photo delineation of recent fires. This accuracy assessment was used to improve estimates of historical fire derived by delineating fires on older (1937) photos.
Introduction
Wilderness managers are mandated by the Wilderness Act of 1964 to protect natural conditions from human-caused change. Managers now recognize the dynamic nature of ecological systems, and the fundamental role of fire in causing these changes. Because fire changes every aspect of an ecosystem, from its constituent species to rates of nutrient cycling, managers need to understand when and where past fires have burned. Managers need this information for two reasons: to allow natural ignitions to play their vital role in wilderness ecosystems, and to separate natural from human-caused change in fire regimes and fire effects. However, for a variety of reasons, many if not most areas have virtually no information on how fire fuels have changed over time, and on past fire regimes and the spatial and temporal variability of these fires. One proposed solution to this lack of information is to map dominant fire fuels and the patterns of spatial occurrence in severe fires, which would leave a sufficiently visible record, from historical aerial photos.
The purpose of the research reported here is to use historical aerial photos to quantify changes over time in the cover of dominant tree species and the sizes of these trees. Dominant trees and their sizes are vital for understanding fuels and fire behavior. Cover and size-class are also directly affected by management policies. For example, fire suppression allows shade tolerant tree species to grow under the canopies of fire-resistant, shade intolerant tree species. One of the key questions in managing western landscapes is whether fire suppression is actually allowing shade tolerant trees to flourish across a landscape. Historical aerial photos may provide an opportunity to quantify how the cover and size of the dominant trees have changed over time across an entire landscape. Further, if spatially explicit, then change in cover and size-class could be related to the influence of physiographic factors of slope, aspect, and elevation, as well as past fire suppression efforts.
The other purpose of this research is to evaluate the accuracy of using historical aerial photos to identify spatial patterns in the occurrence of severe fires. We were interested in whether factors such as slope, aspect, and elevation might affect the accuracy of using historical aerial photos in this way. If there was a systematic influence of these physiographic factors, then the results of this research could be used to improve estimates of historical fire occurrences.
The Bitterroot Mountains Project Area
The project area is 11,400 ha lying on the eastern slope of the Bitterroot Mountains 10 km west of Florence, Montana and 3 km east of the Montana/Idaho border. Elevations range from 1200 m to 2700 m. The area consists of forested slopes with highly incised drainages trending to the east. There are some peaks and ridges above treeline consisting of talus and boulders and a few small lakes. Canopy tree species are dominated by subalpine fir (ABILAS), Englemann spruce(PICENG), western larch(LAROCC), and ponderosa pine (PINPON), with lesser amounts of four other species.
Methods
A wilderness resource manager with photo-interpretation skills was enlisted to delineate polygons showing evidence of severe fire on recent (1994) and historic (1937) aerial photography. This evidence was based on vegetative indicators that suggest a disturbance other than avalanche, insects, or logging. The primary indicator was even-aged reproduction, especially in lodgepole, blackened or red vegetation, snags; even-aged reproduction under a fire resistant overstory; or a mosaic of even-aged reproduction and unburned overstory. The estimated age of the fire in the resulting polygons was based on vegetation.
To determine the change in vegetation from 1937 to 1994, the following vegetative characteristics were photo-interpreted. Polygons were characterized by dominant tree species, size class, and per cent cover; per cent cover of herb, shrub, bare soil; and per cent cover of rock and water. Representative sample points within 10 different polygons were identified for each vegetation class and the per cent tree cover and size class were determined for the 1937 photos. These sample points were transferred to the 1937 Mylar overlays, then transferred to the 1994 resource photography where estimates of per cent tree cover and size class were determined from the recent photography. Sample points were scanned into ARC/INFO and registered to the project area. All tabular data resulting from the photointerpretation was entered into an Applix spreadsheet and later joined to the ARC/INFO coverages. The Digital Elevation Model (DEM) for the project area was acquired to be used for analysis of physiographic variables in relation to vegetation change and fire history.
The USFS fire history data for the project area was acquired and stored as a separate ARC/INFO coverage. A coordinated effort was necessary between two adjoining forests to insure full coverage of the project area.
Analysis
These methods yield two different data sets: per cent canopy cover and size classes for 1937 and 1994 photos for each vegetation type derived from the sampling point data, and the area and location of each vegetation type from the polygon data.
The point data was used to derive estimates of mean per cent canopy cover and size class for each vegetation type was then compared between 1937 and 1994. To examine potential factors affecting differences in canopy cover over 50 years, the sampling point locations were draped on a DEM slope, aspect, and elevation for each point. These data were used to examine the influence of topographic position on canopy cover change. In addition, if large changes in canopy cover are observed, these sampling point locations can be laid on air photos from intervening years to establish the trend in canopy cover changes.
The areal data will be used to derive estimates of the per cent of the total study area covered by each vegetation type to quantify changes between 1937 and 1994. As for canopy cover, the polygonal coverage will be draped on a DEM to examine changes in the location of each vegetation type within the watershed.
The USFS fire history data and the photointerpreted 1994 areal data were be overlaid to determine the accuracy of date and size of fire perimeters between the two methodologies.
Results
T-tests were used to determine significance of changes between the 1937 and 1994 sample periods. Overall, the mean per cent conifer cover increased significantly between 1937 and 1994. The average size class increased but not significantly. The mean per cent conifer cover for ABILAS, LAROCC, and PINPON increased significantly. Only PINPON increased in average size class between 1937 and 1994.
Testing was done to determine correlations between change in per cent cover and change in size class. Greater increases in cover were correlated with larger size classes in the 1994 sample and greater increases in size were correlated with greater per cent cover in both time periods. The smaller the size class in 1937, the greater the change in size, and the larger the size class in 1994, the greater the change in size.
Correlations were computed to determine the influences of physiographic variables on per cent cover and size class. Aspect is not correlated with per cent cover in 1937 and 1994. Aspect and elevation are related to conifer cover, but neither are related to change in per cent cover. Size classes tend to be smaller and experience less change at higher elevations. Aspect and elevation are correlated with vegetation type at each sample point, but the slope is not correlated with any other measures associated at that site.
Stepwise linear regression was performed on the change in per cent cover and change in size class to predict change between the variables. This revealed that the vegetation type in 1994 was the single best predictor of change in per cent conifer cover. The regression also showed that elevation was the single best predictor of the amount of change in size class between sample periods.
Photointerpretation of the 1994 data produced estimates of fire dates in selected polygons. None of these dates of polygons matched the dates of polyonal extent of any of the USFS fire atlas data. Only seven polygons of USFS fire data coincided with the study area and those were much larger in areal extent than the underlying photointerpreted polygons. Where estimated fire dates did overlay the USFS fire data, the span varied from 17 to 61 years.