Journal of Forestry Research, 17(4): 263-268 (2006) 263 Evaluation of deriving fire cycle of forested landscape based on time-since-fire distribution ZHANG Quan-fa 1 , Kurt S. Pregitzer 2 , JIANG Ming-xi 1 , CHEN Wen-jun 3 1. Wuhan Botanical Garden, Chinese Academy of Sciences, Wuhan 430074, P. R. China 2. School of Forest Resources and Environmental Science, Michigan Technological University, Houghton, MI 49931, USA 3. Applications Division, Canada Centre for Remote Sensing, 588 Booth Street, Ottawa, Ontario K1A 0Y7, Canada Abstract: Estimation of fire cycle has been conducted by using the negative exponential function as an approximation of time-since-fire dis- tribution of a landscape assumed to be homogeneous with respect to fire spread processes. The authors imposed predefined fire cycles on a virtual landscape of 100 cell ×100 cell, and obtained a mosaic composing of patches with different stand ages (i.e. time since fire). Graphical and statistical methods (Van Wagner 1978; Reed et al. 1998) were employed to derive fire cycle from the virtual landscape. By comparing the predefined and the derived fire cycles, the two methods and tested the effects of sample size and hazard of burning (i.e., stand’s suscepti- bility to fire in relation to its stand age) were evaluated on fire cycle deviation. The simulation results indicated a minimum sample size of 10 times of the annual burnt area would be required for partitioning time-since-fire distribution into homogeneous epochs indicating temporal change in fire cycle. Statistically, there was significant difference among the imposed and the derived fire cycle, regardless of sample sizes with or without consideration of hazard of burning. Both methods underestimated the more recent fire cycle without significant difference between them. The results imply that deviation of fire cycle based on time-since-fire distribution warrants cautious interpretation, especially when a landscape is spatially partitioned into small units and temporal changes in fire cycle are involved. Keywords: Fire cycle; Simulation; Time-since-fire distribution; Evaluation CLC number: S762.1 Document code: A Article ID: 1007-662X(2006)04-0263-06 Introduction Fire has been the dominant disturbance and responsible for much of the structure and function of boreal ecosystems (Wein and MacLean 1983; Johnson 1992). Among the components defining fire regime (i.e., a combination of four components, namely intensity, frequency, seasonality and size), fire cycle (van Wanger 1978) or fire rotation (Heinselman 1973), is an indicator of fire frequency measuring the time (years) required to burn an area equal in size to the area under consideration. Fire cycle has been an important component in maintaining many ecosystems on earth in particular boreal ecosystems for the past 10 000 years since the last glaciation (Payette et al. 1985). Along the migration of species from south and other refugees of glacier, ecosystems adapted to the various fire regimes in the boreal region but they are not necessarily in equilibrium with the climate (Overpeck et al. 1990; Campbell and McAndrews 1993). Studies revealed a complex interaction among landscape, fire regimes, and vegetation types (Bergeron et al. 1993; Suffling 1995; Hely et al. 2001; Li et al. 2005). Changes in fire cycles would have significant consequences on landscape pattern by affecting regenerating pathways of forest after fires. Interactions of fire cycle and species' reproductive characteristics could de- termine vegetation distribution pattern of a landscape (Suffling et al. 1988; Bergeron and Dansereau 1993; Johnson at al. 1995; Suffling 1995; Gauthier et al. 1996). Biography: ZHANG Quan-fa (1965-), male, Ph.D. in Wuhan Botanical Gar- den, Chinese Academy of Sciences, Wuhan 430074, P. R. China. Email: qzhang@wbgcas.cn Received date: 2006-06-11 Accepted date: 2006-10-20 Responsible editor: Zhu Hong Practically, time-since-fire distribution has often been used to derive fire cycle (Van Wagner 1978). The derived fire cycle, commonly through a fitness of the negative exponential function, has been utilized to assess the influence of human activities and climate change on fire regime (Johnson 1979, 1992; Johnson et al. 1985; Baker, 1989; Johnson et al. 1990; Masters 1990; John- son et al. 1991). Such an approach has provided valuable in- sights in understanding fire regime, but there are outstanding issues over this approach. First, the negative exponential function is an approximation of time-since-fire distribution of a landscape subject to random, periodic fires and uniform flammability with stand age (Van Wagner 1978; Johnson and Van Wagner 1985; Johnson et al. 1994). Thus, the landscape is assumed to be homogeneous with respect to fire initiation and spread processes. Rarely is this the case in reality. For instance, biophysical factors such as fuel buildup in a stand may affect fire initiation and spread (Agee et al. 1987; Bergeron 1991; Johnson 1992). Second, fire cycle is a relative term that depends on burnt area, time period, and total area of interest (Li 2002). Studies have derived fire cycles from areas ranging from a few thousand hectares to thousands of square kilometers (Baker 1989; Johnson et al. 1991; Masters 1990). If a sample area is too small, one fire incident could burn over the entire area of interest, and derived fire cycle could be misleading. Thus, it would be of great interest in understanding how different sample sizes and landscape heterogeneity affect fire cycle deviation. Recently, Reed et al. (1998) developed an improved statistical methodology to derive fire cycle in particular concerning tempo- ral changes in fire cycle using time-since-fire distribution. The method took into consideration the fact that surviving stands originating in earlier epoch have been subjected to the more re- cent fire cycle. They argued that the previous method (Van Wagner 1978; Johnson et al. 1994) overestimated fire frequency