The purpose of this study was to partition the components of a mixed fire frequency, and empirically relate these components to temporal and spatial differences in fire frequency. The method of reconstructing the fire frequency was to build a stand-origin map, from this estimate the mixed time-since-fire distribution, and then use a graphic technique to partition the mixed distribution into two homogeneous fire-frequency distributions. The fire frequency for the last 380 yr in the 495-km2 Kananaskis Watershed showed a temporal change in fire frequency at almost-equal-to 1730. This change in fire frequency is related to a change from a warmer and drier climate before 1730 to a cooler and moister climate since then. When the fire-frequency data were partitioned by a graphic technique, the two resulting fire frequencies fit negative exponential distributions. For the period 1730-1980 the fire cycle (the time required to burn an area equal to the area of study, or one divided by the scale parameter of the exponential distribution) was 90 yr. For the period before 1730 the fire cycle was 50 yr. The Watershed could not be subdivided into smaller, spatially homogeneous fire-frequency units. The negative exponential distribution of fire frequencies suggests a constant hazard function (mortality force). This constant hazard was corroborated by a spatial correlation (Moran's I) test which found no spatial pattern other than ones expected by chance between forest ages on either side of all fire boundaries. Thus, there was no tendency for young stands to be associated either with only younger- or only older-aged stands. Other studies on fire behavior in the Watershed corroborate the short fire cycle, constant hazard, and lack of spatial fire-frequency differences, and suggest that the regional climate control of the temporal fire frequency is related to a characteristic synoptic weather pattern and the resulting high intensity and high rate-of-spread of fires.
