1994 Proceedings
North American Conference on Savannas and Barrens

Long-term monitoring studies can provide important insights into patterns of natural variation in species abundances. It is necessary to recognize the extent and degree of short-term fluctuations resulting from natural ecological processes to distinguish these from long-term trends that may require conservation and management strategies. Patterns of temporal variation in abundance can also indicate how prone to extinction a species might be under conditions such as habitat fragmentation that result in small population sizes and isolation.

Comparisons of the temporal patterns of population dynamics among species can also indicate important factors affecting these dynamics. Similarity in population dynamics among species is expected if species abundances are strongly influenced by variation in the same general set of extrinsic environmental factors. In contrast, independence or negative temporal associations among species could indicate that species are responding to different environmental variables, that species abundances are regulated by intrinsic mechanisms, or that species interactions such as competition strongly influence the dynamics of coexisting species. Although it is difficult to exclude the possibility of strong interspecific interactions when population parameters are highly correlated, this similarity at least indicates that environmental variation is the stronger determinant of year-to-year fluctuations in abundance.

STUDY AREA

The study was conducted at the Hastings Natural History Reservation (HNHR) of the Museum of Vertebrate Zoology, University of California, Berkeley, in the Carmel Valley, Monterey County, California. The reserve includes 809 ha of the Santa Lucia Range and is typical of the central Coast Range of California. The area consists of steep rolling hills covered by a mosaic of forest, open woodland, chaparral, and grassland habitats. Elevation ranges from 467 to 953 m a.s.l. Data used in this study were collected in three distinct habitat types: open oak woodland, chamise chaparral, and valley grassland.

METHODS

A permanent transect with trap stations spaced 10 m apart was established in each of the above sites in 1967. The chaparral transect was 35 stations long, with 11 stations along the chaparral edge and the remainder in dense brush. The grassland and oak woodland transects were each 20 stations long. Snap-trapping censuses were conducted along each transect each summer (June) and fall (November) from 1967 through 1979. Censuses were conducted only in the fall from 1980 through 1988. For each census, three Museum Special snaptraps baited with peanut butter and oatmeal were set within 2 m of a fixed marker stake at each trap station. Traps were checked each morning and evening and reset as needed for two days and nights.

Two other long-term data sets were available for this period. First, annual growth of grasses and forbs was measured at the grassland site from 1968 through 1987. Each year at the end of the growing season (late May), the above-ground vegetation in 0.25-m2 quadrats randomly offset from 10 of the grassland transect trap stations was clipped at ground level, separated into mulch and new growth, and weighed. The 20-yr. record of mean standing crop biomass (g/m2) of new growth was used as an index of primary production. Second, weather records, including measurements of daily precipitation, have been taken at HNHR since 1939. As in other areas with a Mediterranean-type climate, most precipitation occurs during a winter rainy season. We used three measures of precipitation as indices of year-to-year climatic variation: Annual Rain was the total precipitation recorded for a "biological year" beginning in November and ending in October; Summer Rain was the amount of rainfall recorded from April through October, and reflected the intensity of the summer drought; Winter Rain was the amount of precipitation recorded from October through April and reflected conditions during the normal winter growing season.

RESULTS

Rodent abundance in the fall was significantly positively correlated with Annual Rain, but only weakly positively correlated with Summer Rain. Rodent abundance in the spring was positively correlated with primary production and Winter Rain, but the relationships were not statistically significant. Rodent population dynamics were generally positively correlated among the different habitat types, indicating that annual climatic variation on a regional scale was more important than local habitat differences in its effect on rodent abundance.

A total of 11 species of small mammals was captured; nine of these were captured frequently enough to be examined independently. Rodent population dynamics was generally positively correlated among species, although many correlations were weak (not statistically significant) and a few were negative. In the correlation matrix based on fall censuses, 30 correlations were positive with 16 of these having p-values < 0.05; only six correlations were negative with all p-values > 0.33. In the correlation matrix based on spring censuses, only 15 out of 36 correlations were positive, and only three of those had p < 0.05; 20 correlations were negative and one was zero, but all of these had p-values > 0.15.

Most rodent species were associated primarily with either the grassland or the chaparral habitats in this area. The open oak woodland had lower rodent abundance overall, and no species was the most abundant there. Oaks in this particular woodland site were mostly hybrids and poor acorn producers; the oak woodland may therefore have acted as a population sink rather than as prime habitat for these rodent species. Rodent species were clustered by similarity in habitat use and by similarity in population dynamics using UPGMA algorithms. Similarity in population dynamics among rodent species did not correspond closely to similarity in habitat use or to taxonomic affinity. Most species seemed to respond to environmental variation in an individualistic manner, although there was some evidence of interspecific interactions between rodent species in the chaparral.

We used the standard deviation of the base-ten logarithm of N+1 (to compensate for zero values) as an index of variability. There was considerable variation among species in this index. Spring censuses were distinctly more variable for some species, but fall censuses were more variable for others. No habitat was strikingly more variable than any other in terms of total rodent numbers for either spring or fall censuses. All values of the index fell comfortably within the range previously reported for a variety of small mammals and other vertebrates.

DISCUSSION

Rodent abundance in this open oak woodland, chaparral, grassland habitat mosaic was significantly influenced by year-to-year climatic variation. The similarity in population dynamics among species and among assemblages within all three habitats supports the hypothesis that extrinsic factors such as environmental variation exert a stronger influence on rodent abundance than intrinsic mechanisms. Much residual variation remains unexplained, however, leaving the possibility of an important role for biotic interactions in at least some cases. Lack of strong correspondence between similarity in population dynamics and similarity in habitat use implies that local differences in environmental variation among habitats were not the reason for observed differences in population fluctuations among species.

There were few habitat generalists in this system. Most species were strongly associated with either the grassland or the chaparral, and secondarily with the oak woodland. Thus, rodent species in this savanna-dominated system were microhabitat specialists rather than just "savanna species."

It has been suggested that temporal variability in abundance can be used to predict the risk of local extinction, especially when populations become small or isolated. Most of the species examined were not captured in at least one census, but none was identified by an index quantifying variation in abundance as particularly extinction-prone due to high interannual variability in abundance. One species, Dipodomys venustus, appeared to be strongly associated with only a single habitat type, however, and may therefore be the most sensitive to area and isolation effects caused by habitat fragmentation.

ACKNOWLEDGMENTS

Many people collected the data on which this analysis was based, and we appreciate their graciousness and generosity. In particular, we thank J. Davis, W. Z. Lidicker, Jr., J. L. Patton, and O. P Pearson for their many years of work collecting the rodent data, and J. R. Griffin for his measurements of grass production. M. R. Stromberg and W. D. Koenig provided us with access to the trapping records and weather data files for HNHR.