CHAPTER 26 Patch-based Models to Predict Species Occurrence: Lessons from Salmonid Fishes in Streams Jason B. Dunham, Bruce E. Rieman, and James T. Peterson nvironmental heterogeneity often produces patchy variation and biological responses can produce weak Eor discontinuous distributions of organisms. Even or misleading inferences (Goodwin and Fahrig 1998). broadly distributed species show localized peaks of Our focus in this chapter is on definition of patches abundance (Maurer 1999). This is particularly obvi- suitable for supporting local breeding populations. ous in stream ecosystems, where patch dynamics is a This is a key prerequisite for applying ideas from dominant theme (Pringle et al. 1988). Features of the metapopulation and landscape ecology to predicting environment that may influence species occurrence in species occurrence. streams are believed to result from a hierarchy of Here, we review our attempts to develop patch- physical processes operating within drainage basins. based classifications of aquatic habitat and models to This idea has formed the basis of several classification predict occurrence of salmonid fishes in streams. We schemes for stream habitats (e.g., Frissell et al. 1986; begin with a brief overview of the concept of patchi- Hawkins et al. 1993; Imhof et al. 1996; Naiman ness. Next, we outline criteria to define the biological 1998; see Morrison and Hall (Chapter 2) for defini- response of interest: occurrence of local populations. tion of “habitat”). These classifications provide a use- We then describe models to predict the distribution of ful framework for understanding physical processes local populations within stream basins. These models that generate stream habitat over areas of varying size allow delineation of patches of suitable habitat within and spatial resolution, but they do not explicitly con- watersheds and definition of patch structuring. Pat- sider how individual species actually perceive or uti- terns of patch structuring and characteristics of indi- lize these patchy environments. vidual patches provide the basis for modeling occur- To be most useful, patches should be clearly defined rence of local populations. We compare patch-based by associations between a biological response (e.g., re- models of occurrence for two threatened salmonids: production, migration, feeding) and environmental bull trout (Salvelinus confluentus) and Lahontan cut- variability (Addicott et al. 1987; Kotliar and Wiens throat trout (Oncorhynchus clarki henshawi). Finally, 1990). Classifications of aquatic habitat based purely we compare our results to alternative approaches to on physical characteristics or subdivision of water- predict occurrence of salmonids and discuss implica- sheds into arbitrary segments may not adequately de- tions of a patch-based approach that should be scribe “patchiness” from an organism’s point of view. generally relevant for developing models of species Lack of attention to realistic scaling of environmental occurrence. 327 328 PREDICTING SPECIES OCCURRENCES The Concept of Patchiness plies some degree of reproductive (genetic) isolation. Spatial isolation of spawning and rearing habitat for The term “patch” has been applied in numerous con- salmonids is reinforced by strong natal homing texts in ecology (e.g., Pickett and White 1985; McCoy (Quinn 1993), and patches of suitable habitat may and Bell 1991; Pickett and Rogers 1997; Morrison therefore support relatively discrete local populations. and Hall, Chapter 2). Our definition of patches paral- Ultimately, it would be desirable to use multiple lels the concept of ecological neighborhoods intro- sources of information to delineate local breeding duced by Addicott et al. (1987). Ecological neighbor- populations. Several studies have demonstrated the hoods are defined by a specific biological response and limitations of using limited genetic or demographic in- not by an arbitrary temporal or spatial scale or by a formation alone to infer population structuring (Ims perceived boundary or control imposed on the system. and Yoccoz 1997; see Utter et al. 1992, 1993 for A “patch” corresponds to limits or boundaries of en- salmonid examples). vironmental conditions that can support a biological Unfortunately, detailed genetic and demographic response. Patches of environmental conditions poten- data are not available for most systems. For salm- tially suitable to support local populations of a species onids, we have defined patches of suitable habitat by are often the focus in landscape and metapopulation modeling the distribution limits of smaller, presum- ecology. ably “premigratory” or resident individuals within Kotliar and Wiens (1990) provided a general streams. Larger juvenile and adult salmonids may framework for defining patch structure. Given that a adopt migratory life histories (Northcote 1997, Fig. biological response is observed to occur within a de- 26.1) and range far outside of spawning and rearing finable spatial frame, patch structure can be character- areas, but their existence ultimately depends on ized by (1) the degree to which patches can be distin- spawning and rearing habitat. Our delineation of guished from each other and the surrounding patches for salmonids, then, is based principally on environment (patch contrast), and (2) how patches are ecological information and an assumption of natal spatially aggregated. Patch structuring may be charac- homing. Population genetic analysis (e.g., Kanda terized by a nested or hierarchical pattern and may vary widely among biological responses. Patch structuring is not directly synonymous with a specific temporal or spatial scale. For example, patches defined here may vary by an order of magni- tude or more in size (patch area). It is definition of common biological responses and environmental crite- ria for determining patch structure, not spatial or tem- poral scale per se, that provides a foundation for patch-based models of species occurrence that may be generalized within and among species. Defining a Biological Response We were interested in predicting occurrence of fish in patches of habitat suitable for local breeding popula- Figure 26.1. Simplified schematic of life-history variation in tions. Patches suitable for local populations should salmonid fishes. Fish with a “resident” life history spend their correspond to locations where population growth can entire lives within spawning and rearing areas. Migratory fish use habitats outside of spawning and rearing areas but return be attributed primarily to in situ reproduction, rather faithfully (homing) to breed in natal areas. Some dispersal is than immigration (Addicott et al. 1987). Limited de- possible among both life-history types. Our definition of mographic interaction among local populations im- patches corresponds to the extent of spawning and rearing. 26. Patch-based Models to Predict Species Occurrence 329 1998; Spruell et al. 1999) for some systems indicates genetic divergence does correspond to juvenile distri- butions. Our approach to defining patches appears to be a reasonable approximation, but detailed demo- graphic and/or genetic data will be necessary to con- firm the structure of any system (Haila et al. 1993; Rieman and Dunham 2000). Models of Distribution Limits and Patch Delineation Unlike terrestrial habitats, streams are generally viewed as one-dimensional systems in terms of fish distributions and dispersal. Therefore, boundaries of habitat patches may be delineated in an up- and/or Figure 26.2. Map of study areas: the upper Boise River Basin, downstream direction. Many factors can potentially Idaho, and the eastern Lahontan Basin, Nevada. limit the distribution of spawning and rearing habitat for salmonids, including natural and artificial disper- sal barriers, water temperature, interactions with non- ture. Various researchers have classified thermally suit- native salmonids and other fishes, human disturbance, able habitat from variation in groundwater (Meisner and geomorphic influences. These factors are often 1990; Nakano et al. 1996), air (Keleher and Rahel not independent. For example, interspecific interac- 1996), and surface water temperatures (Eaton et al. tions mediated by water temperature may influence 1995; Rahel et al. 1996). Our approach is currently longitudinal distributions of species within streams based on modeling elevation gradients, which are cor- (De Staso and Rahel 1994; Taniguchi et al. 1998). related with temperature (Keleher and Rahel 1996). In the case of bull trout and cutthroat trout, spawn- Our attempts to delineate the amount and distribution ing and early rearing usually occur in upstream or of suitable habitat (i.e., patch structure) for salmonids headwater habitats (often fourth-order streams or have relied on empirical relationships between down- smaller), so we were particularly interested in factors stream distribution limits of juveniles and elevation or that determine downstream distribution limits of juve- geographic gradients (see also Flebbe 1994). Our niles. Our two study areas are located at the southern work has been with Lahontan cutthroat trout in the margin of the range for both species, where unsuitably eastern Lahontan Basin in southeast Oregon and warm summer water temperatures in streams are northern Nevada, and bull trout in the upper Boise probably an important factor limiting the amount of River Basin in southern Idaho (Fig. 26.2). suitable habitat (Rieman et al. 1997; J. B. Dunham Delineation of patches for bull trout in the Boise and B. E. Rieman