Northern Prairie Wildlife Research Center
We examined the soil samples in two ways. First, we visually examined one duplicate of the three wetland and the three upland soil samples from each wetland for recalcitrant remains of invertebrates under a low-magnification dissecting scope after concentrating remains by sieving through a 0.5-mm-mesh screen. Second, we incubated the remaining soil samples in 37.9-L aquaria under standardized light (12-hour day length), specific conductance (700 ÁS cm-1), and temperature regimes (four weeks at 8°C followed by four weeks at 22°C). In order to promote the emergence of the maximum number of taxa, we used the above dual temperature regime meant to simulate temperatures of wetlands following early spring snowmelt and after major summer precipitation events. At the end of the first 4-week incubation, we siphoned the contents of each aquarium through a 0.5-mm-mesh screen and returned the sieved water to its original aquarium for a second incubation at the alternate temperature. Invertebrates retained by the 0.5-mm-mesh screen during each 4-week incubation were combined and processed as a single sample. Invertebrates were sorted into taxonomic groupings according to Pennak (1989) and enumerated from both the visually examined field samples and our incubated aquarium samples, hereafter termed field and incubated samples, respectively.
To facilitate use by those not very familiar with invertebrate identification, we also did an analysis with a simplified taxonomy. For this purpose, we combined into single groupings all planorbid snail shells, lymnaid snail shells, physid snail shells, cladoceran resting eggs (ephippia), ostracod shells, and trichopteran cases for the visually examined field samples, as well as all Cladocera, Copepoda, Ostracoda, Anostraca, and Conchostraca individuals for the incubated samples.
Data from the three transects were summed to provide a single value for each wetland and another for the adjacent upland. We computed means and standard errors of number of taxa and LogCount, unweighted and weighted by wetland-obligate status, for field and incubated samples, within location (upland, wetland), and by wetland class (seasonal, temporary). We tested for differences between upland and wetland sites and between wetland classes with a randomized-incomplete block analysis of variance using PROC MIXED (SAS Institute 1997), where habitat type (upland, temporary wetland, seasonal wetland) was the explanatory variable and sites were blocks. The response variables were number of taxa and LogCount; the explanatory variable was habitat type (upland, temporary wetland, seasonal wetland). We also computed least-squares means for each response variable by habitat type (SAS Institute 1997); separation among habitat types was performed using Fisher's protected LSD procedure following significant F-tests in ANOVAs (Milliken and Johnson 1984). This procedure was performed for both field and incubated samples and for complete and simplified taxonomies.
To classify wetland sites, we found a straight line in the Taxon RichnessLogCount plane (i.e., a two dimensional, flat surface) that best distinguished upland from wetland sites on the basis of those two variables (Figure 1). This was done through an iterative trial-and-error procedure by successively calculating lines and then determining the number of misclassified sites: the number of wetland sites below that line plus the number of upland sites above the line. A line that produced the minimum number of misclassifications was selected. This was done separately for the field and incubated samples. We chose this method because of its simplicity and the fact that the data did not meet assumptions of other straightforward classification methods, such as linear discriminant function analysis.