Are microhabitat preferences of coexisting species under selection and adaptive?

Ecology, March, 1998 by Thomas E. Martin

INTRODUCTION

Habitat preferences are assumed to be adaptive, such that fitness is higher in preferred habitats, causing natural selection to maintain preferences if they have a genetic basis (Jaenike and Holt 1991). Clear demonstration of adaptive habitat preferences, however, are uncommon. For example, some studies of the adaptiveness of host plant choice by insects have shown higher larval performance (a fitness component) on preferred hosts, but many other studies have shown lower or no change in performance (Thompson 1988, Jaenike and Holt 1991, Valladares and Lawton 1991, Etges 1993). Most of these tests with insects were conducted in laboratory environments where effects of enemies (i.e., predators and parasites) on fitness are understated. Tests in field situations where enemies are present may provide more appropriate tests of the adaptiveness of preferences, but such tests are rare and particularly so for vertebrates.

Studies of habitat preference in birds, for example, often do not examine fitness and instead measure density because density is expected to be positively correlated with habitat preference (Brown 1969, Fretwell 1972, Whitham 1980, Cody 1985, Petit and Petit 1996). However, density may not reflect preference or be positively correlated with fitness (van Horne 1983, Pulliam 1988). Fitness components need to be measured. Yet, when fitness is measured for birds, it is usually measured at the population level such that mean individual fitness is compared among habitats following Fretwell (1972). Such approaches ignore variation in fitness of individuals related to variation in microhabitat quality within habitats (Martin 1986). The clearest test of the adaptive basis of habitat preference is provided by examining the fitness consequences of habitat choices by individuals when each individual is given the full range of habitats from which to choose.

Here, I examine nest site preferences and whether they are adaptive in a study system with seven ground and shrub-nesting bird species. I focus on nest site preferences because nest sites represent a readily quantifiable microhabitat that can have clear fitness consequences. Indeed, nest predation accounts for 98% of nest mortality in the study system (Martin 1992; T. E. Martin, unpublished data) and nest predation is influenced by microhabitat (Martin and Roper 1988, Martin 1993). The study system used here is particularly appropriate for examining microhabitat preferences because vegetation varies along a microclimate gradient and territories of each species encompassed the entire vegetation gradient [ILLUSTRATION FOR FIGURE 1 OMITTED], thereby making the complete range of potential nesting microhabitats available within the territory of each individual. Thus, I assume that the frequency that a microhabitat type is chosen represents preference, such that the most frequently chosen microhabitat is the preferred one for each bird species in this study system. I define microhabitats based on vegetation, but birds could have been choosing another environmental feature (e.g., microclimate) associated with plants. Yet, nests were generally placed under or in specific plants, suggesting that birds were choosing specific vegetation, and vegetation is known to directly influence nest survival (see Martin 1992, 1993). Thus, I assume that vegetation was a reasonable approximation of phenotypic variation in nest site choice and in influencing reproductive success. Selection can only act on this phenotypic variation if it has a genetic basis. Few data are available showing genetic bases of habitat selection, but I assume a genetic basis because some evidence suggested that habitat preferences do indeed have a genetic basis (see Jaenike and Holt 1991). Given that nest site choice can vary over time or with prior experience within individuals, then selection might be acting on a reaction norm. I begin by showing that nest site choices were nonrandom and differed among species. I then document preferences by examining the frequency of use of vegetation types. Subsequently, I examine adaptiveness of preferences and whether selection gradients could favor and maintain preferences.

STUDY AREA AND METHODS

Study area

Study sites were 20 snowmelt drainages located on the Mogollon Rim in central Arizona at [approximately]2600 m elevation. Canopy trees were quaking aspen (Populus tremuloides), Douglas-fir (Pseudotsuga menziesii), white fir (Abies concolor), ponderosa pine (Pinus ponderosa, white pine (Pinus strobiformis), and Gambel's oak (Quercus gambellii). The understory included canyon maple (Acer grandidentatum), New Mexican locust (Robinia neomexicana), saplings of overstory tree species, golden pea (Thermopsis pinetorum), and various grasses. These drainages contrasted with surrounding forest, which was characterized by open ponderosa pine with Gambel's oak in the subcanopy and little understory vegetation.

The seven study species represented all species that nested on the ground or in shrubs in the understory on these sites: Virginia's Warbler (Vermivora virginiae), Dark-eyed Junco (Junco hyemalis), Red-faced Warbler (Cardellina rubifrons), Orange-crowned Warbler (Vermivora celata), Green-tailed Towhee (Pipilo chlorurus), MacGillivray's Warbler (Oporornis tolmei), and Hermit Thrush (Catharus guttatus).