3.2 Pair and brood numbers, and invertebrate food abundance
Overall brood, but especially pair densities were higher in Maaninka than in Evo (for pairs, N = 117, U = 378, P < 0.001; for broods, N = 117, U = 863, P < 0.001; Table 3). The same pattern was also observed if only teal densities were considered (pairs N = 117, U = 495, P< 0.001; broods, N = 117, U = 986, P< 0.001, Table 3). Mallard pair, but not brood density differed between the areas (pairs N = 117, U = 884, P< 0.001; broods, N = 117, U = 1267, P = 0.242, Table 3). However, both mallard and teal per capita brood production was significantly higher in Evo than in Maaninka (mallardG = 20.7, df = 3, P < 0.001; teal G = 52.2, df = 3, P < 0.001; Table 3).
Three best models explained numbers of pairs within ΔAIC <2 (Table 4). The base model (SHORE + FIELD; these variables were included in all model) had the lowest AIC-value. The null model (intercept only) had the poorest fit. Pair numbers increased with shoreline length and field percentage, and ponds had fewer pairs than lakes. Food index appeared not to contribute (Table 5).
Two well-fitting brood models were within ΔAIC < 2 and both included “FOOD” (Table 4). The null model (intercept only) had again the poorest fit. Brood numbers increased with food abundance (Table 5), but less so with number of pairs. Field percentage had only weakly positive, but significant effect on brood numbers.
Water body type had a significant effect for the invertebrate food index. Ponds had higher index than permanent lakes (Table 6), indicating that ponds are more food rich habitats than lakes. Shoreline length or field percentage around the water body did not explain the invertebrate index.
DISCUSSION
Our results indicate that duck densities were higher in the agricultural landscape. However, ducks appear to face a trade-off, because the agricultural landscape also had a higher nest predation rate than the barren forest landscape as revealed by experimental nests equipped with wildlife cameras. Brood production per pair reflected this trade-off: production was higher in the barren forest landscape than in the agricultural landscape, which suggests the occurrence of a high nest predation rate and/or high brood mortality in the agricultural landscape. The results imply that the predator detection at artificial duck nests with wildlife cameras actually reflects their visitation to natural nests. However, our results only concern the early egg-laying period, while predation later in the nest period, during incubation, likely incurs a higher cost to the female.