Relationships between landscape patterns and fire occurrence within a successional gradient in sagebrush steppe–juniper woodland Aaron D. Roth A,B,C , Stephen C. Bunting A and Eva K. Strand A A Department of Rangeland Ecology and Management, College of Natural Resources, University of Idaho, PO Box 441135, Moscow, ID 83844, USA. B Present address: USDA-Natural Resource Conservation Service, John Day Service Center, 721 S Canyon Boulevard, John Day, OR 97845, USA. C Corresponding author. Email: aaron.roth@or.usda.gov Abstract. Expansion of western juniper (Juniperus occidentalis Hook. var. occidentalis) has altered vegetation composition, fire behaviour and fire potential throughout south-western Idaho and eastern Oregon. Utilising GIS-derived products and fire-simulation software, the influence of the spatial arrangement of different woodland developmental stages on simulated surface fire occurrence was evaluated. Custom fuel models and a recent vegetation map processed in FARSITE under moderate fire conditions were used to create a fire-occurrence grid in three sixth-order watersheds on the Owyhee Plateau of south-western Idaho. Landscape pattern metrics were selected to quantify the spatial arrangement of plant communities within a neighbourhood around points from each successional stage randomly placed within each watershed. Linear regression analysis of fire occurrence and each of the selected landscape metrics was compared for four successional stages of western juniper encroachment to assess the effect of landscape-scale vegetation arrangement on fire occurrence. The landscape structure had little influence on whether an early-successional area burns in a surface fire, whereas the surrounding landscape structure influenced whether a late-successional or mature woodland area burned. Landscape metrics that showed significance in late-successional and mature woodland stages include patch density, mean area and Simpson’s diversity. Additional keywords: FARSITE, fire models, FRAGSTATS, Idaho, Juniperus occidentalis Hook. var. occidentalis, Northern Great Basin, simulated fire, western juniper. Introduction Western juniper (Juniperus occidentalis Hook. var. occidentalis) encroachment in Oregon, Idaho, Washington and northern California has become a major concern to land managers because of the effect it is having on changing vegetation com- position, wildlife habitat, nutrient cycling and hydrologic pro- cesses, as well as fire potential and fire regimes (Miller and Rose 1999; Miller et al. 2005). The introduction of domestic live- stock, fire suppression, climate change and human disturbance are proposed explanations for this exponential expansion over the past 120 years (Burkhardt and Tisdale 1976; Shinn 1980; Miller and Wigand 1994; Miller and Rose 1995, 1999). State and transition models describing the development of western juniper woodlands in sagebrush steppe landscapes and effects on ecological processes have been developed by Miller et al. (2000, 2005), and the effects of succession on fuel composition and simulated flame length, fuel-bed depth and rate of spread have been quantified by Yanish (2002). Prescribed burning can be used to reverse western juniper encroachment up to a threshold or successional boundary where fires become difficult to start and carry across the landscape (Miller et al. 2000; Yanish 2002; Miller et al. 2005). Modern technology has been a driving force behind our understanding of wildfires and landscape patterns at different scales (Albright and Meisner 1999). Development of simple fire-behaviour models, applicable only at a local scale, into landscape-scale predictive simulation tools has been made possible because of the recent increase in computational power, availability of remotely sensed data, and advances in database management within geographic information systems (Green 1989; Albright and Meisner 1999; Miller and Urban 1999). Models not only predict fire rates of spread and intensity within one vegetation type at a time, but can also incorporate effects of environmental variables such as weather conditions, slope, aspect, elevation, fuel moisture and landscape composition (Finney and Andrews 1999; Andrews and Queen 2001; Finney 2002). Digital maps and custom fuel models increase the ability to predict fire behaviour at broader scales, and create unlimited opportunities to study fire patterns across the landscape (Green 1989; Finney and Andrews 1999). Landscape pattern metrics have been developed to quantify landscape fragmentation, patch density or the human effects on landscape composition (O’Neill et al. 1988; Riitters et al. 1995). Important information can be derived from landscape pattern CSIRO PUBLISHING www.publish.csiro.au/journals/ijwf International Journal of Wildland Fire 2011, 20, 69–77 Ó IAWF 2011 10.1071/WF08189 1049-8001/11/010069