Daily survival analysis
The daily survival of forest versus shoreline nests was compared for predation risk, based on 175 nests (88 forest and 87 shoreline nests surrounding in total of 46 water bodies) that survived the entire study period or with the known depredation time (68 nests depredated). Depredated nests where memory cards of the cameras failed were excluded because of unknown predation date (N = 2 in 2017). Furthermore, the daily survival of shoreline nests around permanent lakes and ponds was compared, based on 87 nests (41 with exact depredation time).
The GLMM framework was used to calculate daily nest survival probability by using modified logistic regression, which incorporates the number of exposure days (seven, each beginning at 12 pm) into the link function (Shaffer, 2004). The logistic exposure method is a modification of logistic regression and maximizes the use of nest survival data by treating each measurement day as a discrete trial. Daily nest fate was analysed as a binary response variable (1 = survived, 0 = depredated). In the forest-shoreline comparison explanatory variables were “DATE” (1,…7) and “HABITAT” (factorial: shoreline, forest). As nests were established in pairs around the water bodies, one in the shoreline and one further away from the shoreline, water body (“WETLAND_ID”) was used as a random factor. For larger lakes with more than one nest pair, an individual ID was given for every pair of nests. When comparing water body types, the explanatory variables were “DATE” and “TYPE” (factorial: lake, pond). “WETLAND_ID” was used again as a random factor, but this time it only meant shoreline nests. Field percentage (“FIELD”) around water bodies was used as an explanatory variable in both analyses. Year effect was found to be negligible during the data exploration and was thus discarded from the analysis.