Experiments to assess the impact of light grazing were conducted on several sites in the Belgian Ardenne (Deifelt, Commanster, Shoenberg, Martelange, and Straimont). Within these sites, 14 plots larger than 2 ha were selected, eight lightly grazed (stocking rate < 0.2 LU/ha/year) for at least 5 years and six abandoned (control). Three distinct forms of light grazing were distinguished: (1) continuous grazing, cattle being present all the year round; (2) seasonal grazing, cattle being present only during late summer and autumn; and (3) alternate grazing, cattle being present every other year (for 3-9 months). Transect counts of adult butterflies were done during the flight period of both butterfly species in the summer of 2003 and 2004. We analyzed these frequency counts by log linear models with Poisson error, testing the effects of grazing regime (four classes: continuous grazing, seasonal grazing, alternate grazing, and abandoned), sampling year (two classes: 2003 and 2004), sampling effort (transect duration in minutes), and tree/edge density (three classes: some trees and/or edges, intermediate density of trees and/or edges, high density of trees and edges) with AlC-based model selection (Burnham and Anderson 2002; Johnson and Omland 2004; Richards 2005) implemented in SAS. To take into account model selection uncertainty, we report the AIC weight of each explanatory variable, expressing the probability that the variable influences the response (here, the frequency of butterflies), the model-averaged parameter estimates and the confidence limits (Burnham and Anderson 2002).

The frequency of P. eunomia adults was highly related to grazing regime, sampling effort, and sampling year, as shown by an AIC weight above 98% (Table 1). P. eunomia adults were more numerous in control plots than in grazed plots (Fig. 3a), but much rarer in continuously grazed plots (~75% loss) than in seasonally or alternately grazed plots (~25% loss). More P. eunomia butterflies were counted during longer transects, more in 2004 than in 2003. The abundance of L. helle adults was highly related to grazing regime and tree/edge density, showing no significant differences between years, and only a marginal increase with sampling effort (Table 1). A significant difference (~50% loss) was detected in the case of seasonal grazing (Fig. 3b). When considering both species together, the impact

Proclossianc, |
( eunomia |
Lycaena helle | |||||||

95% confidence interval |
95% confidence interval | ||||||||

Variable |
Modality |
AIC weight (%) |
Estimate |
Lower |
Upper |
AIC weight (%) |
Estimate |
Lower |
Upper |

Grazing Type |
Alternate |
100.00 |
-0.374 |
-0.661 |
-0.088 |
98.57 |
0.431 |
-0.145 |
1.006 |

Continuous |
-1.041 |
-1.521 |
-0.561 |
-0.185 |
-1.457 |
1.086 | |||

Seasonal |
-0.510 |
-0.799 |
-0.221 |
-1.239 |
-2.112 |
-0.366 | |||

Control |
0.000 |
0.000 | |||||||

Time |
Time |
100.00 |
0.020 |
0.008 |
0.032 |
50.84 |
0.025 |
-0.005 |
0.055 |

Year |
2003 |
98.97 |
-1.152 |
-1.357 |
-0.947 |
40.59 |
0.470 |
-0.217 |
1.157 |

2004 |
0.000 |
0.000 | |||||||

Tree/Edge |
Edgel |
43.44 |
0.242 |
-0.153 |
0.636 |
98.61 |
-0.714 |
-1.692 |
0.264 |

Edge2 |
0.084 |
-0.212 |
0.379 |
0.775 |
-0.048 |
1.597 | |||

Edge3 |
0.000 |
0.000 |

AIC weight expresses the probability that the variable influences the density. Parameter estimates and confidence intervals for each modality were model averaged to take model selection uncertainty into account (Burnham and Anderson 2002). Parameter estimates quantify the difference in the log of the frequency with the last modality of each variable, fixed at 0 by definition a □ Control plots □ Grazed plots

100%

Continuous Seasonal Alternate

Grazing regime

Continuous Seasonal Alternate

Fig. 3 Impact of grazing treatment on the frequency of the two butterfly species. Shown are the percentages of adult butterflies counted in grazed and neighboring control plots along transects at same dates for same durations, (a) for P. eunomia, (b) for L. helle. NB: In the case of alternate grazing (two localities), there was no strict control plot available in one locality (Deifelt); each plot being grazed for 3 months in autumn every second year, the temporarily ungrazed plot from the two adjacent plots was used as a control (in this graph only, for loss estimation - not in the log linear models of Table 1). Mean butterfly densities (measured by the number of individuals counted in 30 min) on these two Deifelt plots were very similar or higher after the start of grazing management than before, in the nineties (P eunomia : 23,9 ind./30 min. after vs 16,9 ind./30 min. before ; L. helle: 3,7 ind./30 min. after vs 3,8 ind./30 min. before), suggesting no detectable detrimental effect of this grazing regime on these species in this particular site

on populations was the least marked with alternate grazing regime, and higher with seasonal and continuous grazing (Table 1).

Was this article helpful?

## Post a comment