Northern Prairie Wildlife Research Center
On each study area, we delineated 2 blocks of six, 11-m diameter treatment plots. Blocks within a study area were separated by a minimum of 300 m, and outer perimeters of adjoining plots within a block were separated by a minimum of 3 m. Plots within a block were visually selected for similar topography, elevation, exposure, and vegetative cover, and plot centers were visually placed at the highest density of spurge. Each of 6 treatments was randomly assigned to 1 plot within each block. Treatments consisted of: beetles only (beetles released, no burning); fall preburn (burned in fall preceding beetle release); spring preburn (burned in spring preceding beetle release); fall postburn (burned in fall after beetles were confirmed to have established); spring postburn (burned in spring after beetles were confirmed to have established); and control (unburned, no beetles released).
We established 4 equally spaced permanent transects radiating from the center to the perimeter of each plot. In July, 1993, we estimated leafy spurge height and stem density, depth of litter layer, and percent cover of leafy spurge, grasses, shrubs, forbs, and bare ground (Daubenmire 1959) at distances of 1, 3, and 5 m from the plot center on each of the transects. Leafy spurge stem density was estimated again at the same sampling points in July 1994 and 1995.
We used analysis of variance (ANOVA) to examine vegetation data collected in 1993. Our ANOVA model was a split-plot in a randomized block design. Study area by block combinations were considered random blocks, with plots being whole-units and distance from the center of the plot the sub-unit (Steel and Torrie 1980). Within distance classes, data from each plot were averaged across the transects. The general linear models procedure (PROC GLM) of SAS (SAS Institute Inc., 1989b) was used to conduct ANOVAs throughout this study. Exploratory distribution plots of all response variables did not indicate any violations of normality assumption; therefore all analyses were conducted in original units of measurement. Fisher's LSD procedure was used for multiple comparisons (Milliken and Johnson 1984). Reported means are least-squares means (SAS Institute, Inc. 1989b) unless stated otherwise. Statistical tests were considered significant at the 0.05 level.
On 27-29 June 1994, we released 150 A. nigriscutis at the center of each fall- and spring-preburn, fall- and spring-postburn, and beetle-only plot. Beetles used in the study were collected by the USDA Agricultural Research Station in Sidney, Mont., from the "Rugg" insectary (Glendive, Mont.) and were refrigerated until their release within 5 days after capture.
Beginning 13 June 1995, shortly after beetles began to emerge in A. nigriscutis insectaries at Northern Prairie Wildlife Research Center (Stutsman Co., N.D.), plots were swept once for beetles on each of 3 days at intervals of about 1 week during suitable weather (sunshine, dry vegetation, temperature ≥25 °C, wind ≤17 km/hr). A standard sweep sample consisted of 5 sweeps through the upper 25-30 cm of vegetation with a 39-cm diameter sweep net on each of 5 equally spaced transects from the perimeter to the center of the plot, for a total of 25 sweeps. Captured beetles were counted and released immediately at the center of the plot. The greatest number of beetles captured on a plot in a standard sample during the 3 days was used as a population index for that plot.
In 1996, all plots on which beetles were collected in 1995, except those burned in experiment 2 (see below), were again swept for beetles 3 times beginning on 18 June to assess size of the second generation. Between 8-11 July 1996, a single auxiliary sample, consisting of 45-50 sweeps in a spiral pattern from the center to the edge of the plot, was taken on plots on which no beetles were found in standard sweep net samples in 1996. An auxiliary sample was also taken on plots where no beetles had been found in 1995 on the Storhoff, Walsh, Arrowwood, and Kemmer study areas.
Frequency of occurrence (number of plots with beetles established vs not established) was analyzed with chi-square tests of homogeneity (Dowdy and Weardon 1983) in 2-way tables. The categorical data modeling procedure (PROC CATMOD) of SAS (SAS Institute, Inc. 1989a) was used to conduct chi-square tests. ANOVA in a randomized block design (Steel and Torrie 1980) was used to assess the effects of treatments on beetle population index in 1995 because distance from center of plot was not taken into account for these data. Again, study area by block combinations were considered random blocks. We used split plot ANOVA to test for differences in vegetation between plots in which beetles became established and plots in which they did not. Each study plot was treated as a whole-plot. Observations within whole-plots were assigned to sub-plots according to their distance from the plot center. To examine joint treatment effects of fire and beetles on leafy spurge stem density from 1993 to 1994 and 1993 to 1995, ANOVA was also used, with the model being as described for pre-treatment vegetation data above.
Because few colonies were found on the postburn and beetle only plots in 1995, the study was modified in order to test the effect of burning on established colonies. We selected the 12 plots with the greatest indicated Aphthona populations on the Storhoff, Kemmer, and Arrowwood study areas and randomly assigned 1 of 3 treatments (fall-burn, spring-burn, and unburned reference) to each of 4 plots. Most of these 12 plots were fall- and spring-preburn plots. Burns were conducted 11-23 Oct. 1995 on the fall-burn plots and 13-16 May 1996 on the spring-burn plots; spurge phenology and results of burns were similar to those in 1993-94. Beginning on 6 June 1996, we swept all plots 3 times as described above. Changes in beetle population indexes from 1995 to 1996 by treatment were analyzed with ANOVA in a 1-way design, with plots nested within burn or reference groups.