Assessing the Roles of Patch Quality, Area, and Isolation in Predicting Metapopulation Dynamics

Two aspects of patch geometry—area and isolation—currently dominate the field of metapopulation dynamics. Under this area-and-isolation paradigm, models commonly assume that the probability of local extinction decreases as patch area increases and that the probability of colonization increases as patch connectivity increases. Environmental variables other than patch area and isolation are assumed to have relatively little effect on metapopulation dynamics. Our work on a metapopulation of the butterfly Speyeria nokomis apacheana highlights the need for a broader view of metapopulation dynamics. In this system, neither occupancy nor turnover patterns were best modeled as functions of patch area or isolation. Instead, other measures of habitat quality explained the most variance in occupancy and turnover. Our study also revealed temporal variation in the factors associated with occupancy and turnover. This variation can cause the results of analyses to vary with the temporal scale of analysis. For example, factors associated with turnover in this system differed among single-year and multiple-year analyses. We emphasize that factors other than patch geometry may drive extinction and colonization processes in metapopulations—especially in systems that experience substantial natural and anthropogenic environmental variability. Evaluación del Papel de la Calidad, Área y Aislamiento de Parches en la Predicción de la Dinámica Metapoblacional Resumen: Dos aspectos de la geometría de parches —área y aislamiento—dominan actualmente el campo de la dinámica metapoblacional. Bajo este paradigma de “área y aislamiento”, los modelos comúnmente suponen que la probabilidad de una extinción local disminuye si el área del parche aumenta y la probabilidad de colonización aumenta si la conectividad del parche aumenta. Se asume que otras variables ambientales diferentes al área y al aislamiento del parche tienen relativamente poco efecto en la dinámica metapoblacional. Nuestro trabajo en una metapoblación de la mariposa Speyeria nokomis apacheana resalta la necesidad de una visión más amplia de las dinámicas metapoblacionales. En este sistema, no se modelaron bien ni la ocupación ni los patrones de rendimiento como funciones del área o aislamiento del parche. Por lo contrario, otras medidas de calidad del hábitat explicaron la mayor parte de la variación en cuanto a ocupación y rendimiento. Nuestro estudio también reveló una variación temporal en los factores asociados con la ocupación y el rendimiento. Esta variación puede hacer que los resultados de los análisis varíen con la escala temporal del análisis. Por ejemplo, los factores asociados con el rendimiento en este sistema discreparon en los análisis entre años individuales y entre años múltiples. Señalamos que factores diferentes a la geometría del parche pueden conducir a procesos de extinción y colonización en metapoblaciones — especialmente en sistemas que experimentan una variabilidad ambiental natural y antropogénica substancial. § email [email protected] Paper submitted December 14, 2001; revised manuscript accepted July 25, 2001. Pr oo fs On ly 2 Prediction of Metapopulation Dynamics Fleishman et al. Conservation Biology Volume 16, No. 3, June 2002 Introduction Patch-based models of metapopulation dynamics are designed to efficiently predict patch occupancy, extinction, and colonization of local populations linked by limited dispersal ( Hanski & Gilpin 1997; Hanski 1999). Currently, these models focus primarily on the pivotal influences of patch area and isolation ( Thomas & Hanski 1997; Hanski 1998; Moilanen 1999). Empirical data and theory usually support assumptions that the probability of local extinction decreases as patch area increases and that the probability of colonization increases as patch connectivity increases ( Kindvall & Ahlén 1992; Thomas & Harrison 1992; Thomas et al. 1992; Thomas & Jones 1993; Moilanen et al. 1998). Although ecologists understand that populations may be influenced by many more than two variables, inclusion of environmental variables in addition to patch area and isolation arguably contributes little explanatory power to models of metapopulation dynamics ( Hanski 1994; Moilanen & Hanski 1998; but see Sjögren-Gulve & Ray 1996 ). The area-and-isolation paradigm ( Hanski 1998 ) has been validated by a small group of case studies on plants, vertebrates, and invertebrates (e.g., Burdon et al. 1985; Smith 1980; Verboom et al. 1991; Kindvall & Ahlén 1992; Hanski & Thomas 1994; Eber & Brandl 1996; Moilanen et al. 1998 ), and butterflies have played a major role in substantiating this paradigm ( Hanski 1999). Despite the frequent and legitimate assumption that patch “geometry” (area and isolation) is the dominant influence on metapopulation dynamics, other measures of habitat quality may also explain significant variance in patch occupancy and turnover (extinction and colonization) ( Hanski 1991; Hanski & Thomas 1994; Kindvall 1996; Sjögren-Gulve & Ray 1996; Gyllenberg & Hanski 1997; Harrison & Taylor 1997 ). Environmental variables such as food resources, topography, and vegetation structure can affect the dynamics of real metapopulations in significant ways ( Hanski & Gilpin 1991; Hanski et al. 1994; Singer & Thomas 1996; Thomas et al. 1996; Wahlberg et al. 1996; Harrison et al. 1998; Dennis & Eales 1999; Hanski 1998, 1999; Boughton 1999). If habitat variables other than area and connectivity have significant effects on metapopulation dynamics, then effective management may entail more than simply maintaining a suitable metapopulation geometry. Identifying habitat variables that are associated with patch occupancy and turnover may be especially useful when habitat and matrix qualities are affected by human activities (e.g., Sjögren-Gulve & Ray 1996). Understanding habitat quality and its ramifications for metapopulation dynamics may also clarify whether habitat restoration will be an effective management tool. Our research on the butterfly Speyeria nokomis apacheana ( S. nokomis ) addresses the relative influence of patch geometry and other patch variables on metapopulation dynamics. This research was initially motivated by the species’ rarity and by threats to remaining populations from urbanization and agricultural activities ( Britten et al. 1994 ). S. nokomis occupies seeps, springs, and riparian areas in the otherwise xeric Great Basin of North America. In desert landscapes, riparian areas attract disproportionately heavy use from numerous faunal groups, including humans. Thus, both riparian communities and S. nokomis are management priorities in the region (Kauffman & Krueger 1984; Armour et al. 1991; Thomas 1991; Dawson 1992). Changes in the quality of riparian habitat might have significant effects on the metapopulation dynamics of this riparian-obligate butterfly. To test this hypothesis, we identified a metapopulation of S. nokomis that meets criteria necessary for modeling patch dynamics ( Hanski 1994 ). This metapopulation includes a number of habitat patches (39), intermediate patch occupancy ( 33–90% over 6 years), and substantial observed turnover ( 17 colonizations and 31 extinctions). We used 6 years of data from this metapopulation to develop separate models of patch occupancy and population turnover. For each model, we explored patch area, isolation, and other measures of habitat quality as predictive variables. Choice and parameterization of metapopulation-occupancy models can be complicated by numerous factors ( Hanski 1999; Moilanen 2000; Sjögren-Gulve & Hanski 2000). For example, “state-transition” models require a sufficient number of observed turnover events for satisfactory parameterization ( Sjögren-Gulve & Ray 1996; Sjögren-Gulve & Hanski 2000). To yield accurate predictions, these observed turnover events also need to be representative of turnover patterns over time (e.g., Hanski 1999; Moilanen 2000). A single snapshot of patch occupancy is ostensibly adequate to parameterize some metapopulation models, including the incidence-function model (Hanski 1994, 1999), but parameters based on multiple years of data are expected to yield more-accurate predictions ( Hanski 1994, 1999; Moilanen 1999 ). To assess how models of patch occupancy and population turnover may be influenced by the period over which data are gathered, we also used the S. nokomis system to examine temporal patterns of metapopulation dynamics. We tested whether the same environmental factors were associated with turnover in different years and whether variables associated with turnover in 1 year were also significant in analyses that included data from multiple years.