study area and field methods
Fieldwork was carried out at Lake Tuzla (36°43′ N, 35°03′ E), southern Turkey (Székely & Cuthill 1999; Székely, Cuthill & Kis 1999) in 4 years (1996: 13 April−16 July; 1997: 15 April−30 June; 1998: 8 April−10 July; 1999: 14 April−8 July). The Kentish plovers reared their chicks on the north shore of Lake Tuzla (‘shore’ henceforward), and in alkaline grassland of Salicornia europaea and Antrochnemum fruticosum (Uzun et al. 1995) situated in a 50–800 m strip on the north side of the lake (‘saltmarsh’ henceforward). Shore also included mudflats along the lake edge and edges of shallow temporary pools. Saltmarsh also included cultivated fields with short vegetation that bordered the study area on the north. The distance between the two habitats was variable, the shortest distance was about 50 m while the longest was about 1 km. The distribution of these two habitats is not related to field sites mentioned in previous studies of the same population (see Székely & Cuthill 1999, 2000; Székely et al. 1999).
Both parents were caught and ringed with a metal ring and a unique combination of colour rings. Parents were caught by funnel traps either during late incubation or just after hatching of the clutch. Chicks were ringed either in the nest scrape, or if they had already left the nest we ringed them at the first encounter. The age of the latter broods was estimated using the length of their right tarsus (see Székely & Cuthill 1999). We studied 144 broods altogether, and we visited each brood every other day until the chicks died or reached the age of 25 days. Kentish plover chicks fledge at about 29 days of age (Cramp & Simmons 1983). During these visits the number and sex of attending parent(s), and the type of brood-rearing habitat were recorded. Behavioural records were taken for 1–2 h every fourth–sixth day by scanning the behaviour of parents and of chicks every 30 s (see Székely & Cuthill 1999). Brood-rearing habitat was also recorded during behavioural observations. In 1998 we estimated the density of plovers by recording the number of Kentish plovers (including adults and chicks, except the focal brood) in a 25 m radius around the focal brood every 5 min. Ambient temperature was measured to the nearest 0·1 °C at ground level at the end of behavioural observations.
We considered each brood as the unit of analysis. If more than one record was available for a brood, we took the mean of these records. Varying samples sizes in the analyses are due to missing values. Dates were calculated as the number of days since 1 March each year. Hatching dates and dates of behavioural observations did not differ between years (one-way anovas, hatching date: F3,140 = 0·746, P = 0·527; date of behavioural observations: F3,140 = 0·405, P = 0·749), thus we did not consider year-effects in the analyses. Nevertheless, we tested the robustness of our conclusions by including year in all models of temperature, plover density and behaviour (results not shown); in all cases our conclusions remained unaltered.
Brood-rearing habitat was calculated in two comparable ways. First, in the analysis of movement of broods and duration of biparental care, we treated it as a continuous variable, i.e. brood-rearing habitat was the proportion of time each family spent on the shore during brood visits. Second, habitat was treated as a dichotomous variable in the analyses of behavioural samples. For those broods that spent some time in both habitats, the habitat where they spent more time was selected, and we included only those behavioural records in the analyses that were collected in the corresponding habitat.
Kentish plovers are insectivorous (Cramp & Simmons 1983), feeding with a characteristic run-and-peck style by taking food items from the ground, vegetation or water. They may also catch prey, such as flies, from the air. We use feeding efficiency (no. of pecks × 102 × (no. of pecks + no. of runs)−1) as a measure of food abundance, where the number of pecks and number of runs were calculated from the behavioural records (see above). This measure appears to be a better indicator of foraging success than percentage of time pecking, because it reflects the effort the plover puts into gaining a prey item. Furthermore, observations of foraging rates may provide a better estimate of resource abundance than standard arthropod sampling methods, as the arthropod samples do not necessarily reflect prey availability (Palmer, Lane & Bromley 2001). Plover density was calculated as number of plovers × ha−1. Brooding by a parent was defined as per cent of time the parent brooded at least one chick, whereas brooding of a chick was defined as per cent of time when the chick was brooded by a parent. Fighting and feeding were defined as per cent of time a bird spent on fighting and feeding, respectively.
We analysed plover density and behavioural data by General Linear Models (GLMs). For the analyses of plover density and feeding efficiency of parents we included habitat (factor), observation date (covariate), and their interaction in the models. Feeding behaviour, fighting, brooding and chick feeding efficiency may also be influenced by the age of the chicks, thus brood age (covariate) was also included in GLMs of these variables in addition to habitat, observation date and the habitat × observation date interaction. Behavioural variables (brooding, fighting, feeding and feeding efficiency) were arcsine-square root transformed (Sokal & Rohlf 1995). The variance of plover density was not constant between the habitats; therefore, we used a gamma error distribution in the analyses of plover density (Crawley 2002). Brooding, fighting and feeding were also analysed by manovas; we quote the F- and P-values associated with Wilks’λ statistic, and the probability from univariate F-tests (Norušis 1994).
Over half (55·5%) of the broods (n = 110 broods) were biparental, the female deserted in 40·9% of broods and the male deserted in 3·6% of broods. Desertion by males was thus rare, so we investigated the duration of parental care only for females. Duration of biparental care (response variable, in days) was analysed by a single GLM whereby both brood-rearing habitat (proportion of time on the shore, covariate) and observation date (covariate) were included in the model. All broods in which either the female deserted the brood, or both parents stayed with their chicks for at least 25 days were included in the GLM. The latter analysis, however, may be biased because it did not include those broods that died while the female still attended the brood. Thus we also analysed duration of biparental care by Cox regression (Norušis 1994), in which the response variable was the duration of biparental care, the terminal event was desertion by the female, and both habitat and observation date were covariates. Censored cases included (1) broods that were cared for by both parents over 25 days, and (2) broods that died while both parents attended. Note that in Cox regression a negative sign of the parameter estimate indicates a decrease in the probability of the event, i.e. desertion by female.
We carried out several experimental manipulations during the study (e.g. Székely & Cuthill 1999, 2000), and these manipulations may potentially influence our results presented here. However, these manipulations did not influence the behaviour of parents (manovas, 1996–99, males: Wilks’λ = 0·671–0·918, P = 0·373–0·980; females: Wilks’λ = 0·705–0·751, P = 0·094–0·595), the behaviour of their chicks (Wilks’λ = 0·832–0·959, P = 0·427–0·972), and the use of brood-rearing habitat (one-way anovas, 1996–99, P = 0·201–0·924) in the current dataset, and thus manipulations are not analysed further.
We also tested whether the location of brood (‘sites’ in the terminology of Székely & Cuthill 1999, 2000; Székely et al. 1999; unrelated to the saltmarsh/shore dichotomy) may influence the movement of broods and the behaviour of parents and their chicks. However, by including sites in statistical analyses, our conclusions were not changed (results not shown).
Ambient temperature may be different between the two type of habitats, and may influence the movement of families especially when the chicks are young and do not have well developed thermoregulation (Visser & Ricklefs 1993). Nevertheless, ambient temperature is unlikely to explain the movement of broods, because neither the slopes nor the intercepts of temperature over the breeding season were different between the habitats (GLM on log10(x + 1) transformed data, habitat: F1,140 = 0·234, P = 0·629; observation date: F1,140 = 44·748, P < 0·001; habitat × observation date: F1,140 = 0·109, P = 0·742).
Statistical analyses were carried out using R 1·7·1 for Windows (Ihaka & Gentleman 1996) and SPSS for Windows 8·0 (Norušis 1994). In all GLMs we used Type III sum of squares. Two-tailed probabilities and mean ± SE are given.