Inter-individual consistency in habitat selection patterns and spatial range constraints of female little bustards during the non-breeding season

Abstract

Background: Identifying the factors that affect ranging behavior of animals is a central issue to ecology and an essential tool for designing effective conservation policies. This knowledge provides the information needed to predict the consequences of land-use change on species habitat use, especially in areas subject to major habitat transformations, such as agricultural landscapes. We evaluate inter-individual variation relative to environmental predictors and spatial constraints in limiting ranging behavior of female little bustards (Tetrax tetrax) in the non-breeding season. Our analyses were based on 11 females tracked with GPS during 5 years in northeastern Spain. We conducted deviance partitioning analyses based on different sets of generalized linear mixed models constructed with environmental variables and spatial filters obtained by eigenvector mapping, while controlling for temporal and inter-individual variation.

Results: The occurrence probability of female little bustards in response to environmental variables and spatial filters within the non-breeding range exhibited inter-individual consistency. Pure spatial factors and joint spatial-habitat factors explained most of the variance in the models. Spatial predictors representing aggregation patterns at ~ 18 km and 3-5 km respectively had a high importance in female occurrence. However, pure habitat effects were also identified. Terrain slope, alfalfa, corn stubble and irrigated cereal stubble availability were the variables that most contributed to environmental models. Overall, models revealed a non-linear negative effect of slope and positive effects of intermediate values of alfalfa and corn stubble availability. High levels of cereal stubble in irrigated land and roads had also a positive effect on occurrence at the population level.

Conclusions: Our results provide evidence that female little bustard ranging behavior was spatially constrained beyond environmental variables during the non-breeding season. This pattern may result from different not mutually exclusive processes, such as cost-benefit balances of animal movement, configurational heterogeneity of environment or from high site fidelity and conspecific attraction. Measures aimed at keeping alfalfa availability and habitat heterogeneity in open landscapes and flat terrains, in safe places close to breeding grounds, could contribute to protect little bustard populations during the non-breeding season.

Keywords: Inter-individual habitat selection; Non-breeding season; Ranging behavior; Spatial eigenvector mapping; Tetrax tetrax.

Fig. 1

Location and land-use composition of the study area. Boundaries are defined by the minimum convex polygon (MCP) including the pool of locations of tagged female little bustards. Lleida is the large urban area located on the left. The black star indicates the main breeding area for the species in the study area. Reference coordinates in UTM. Map derived from SIGPAC cartography

Fig. 2

Occurrence distribution of tagged female little bustard in the study area. For graphical purposes, the core area (50% kernel) of occurrences of each female in different seasons is shown. In parenthesis, number of patches of the core area and size

Fig. 3

Responses curves for the best habitat model explaining the occurrence probability of female little bustard in the Plana de Lleida. Values for terrain slope response were scaled by subtracting the mean and dividing the SD

Fig. 4

Univariable and independent contribution of habitat predictors in the best habitat model. Values are shown as percentage of the total explained variation by the best habitat model

Fig. 5

Spatial correlograms of spatial filters and their independent contribution to the best spatial model according to AIC. a Spatial correlograms of the 10 most important spatial filters (in order of importance: SF1, SF25, SF39, SF26, SF11, SF60, SF15, SF7, SF4 and SF23) in spatial models defined by Moran’s I coefficients in 5 distance classes, indicating links among points of the study area successively separated by 10 km. Spatial filters are represented in a blue gradient representing filters from broad (dark) to fine scales (light). In a the first distance at which Moran’s I values crosses the expected value in the absence of spatial autocorrelation (0) is shown as an estimate of the scale of the spatial pattern that each filter represents. And b, estimate of the spatial filters in backward stepwise explaining the occurrence probability of female little bustard in the Plana de Lleida

Fig. 6

Results of the best habitat (HAB), spatial (SPAT) and more complex models combining both habitat and spatial variables (HAB + SPAT) based on stepwise AIC. Deviance partitioning analysis for the probability of occurrence of female little bustard. The figure shows a conceptual diagram of variance partitioning: the two circles represent the total variance explained by models with the two components (habitat + spatial), while the left and right circles show the variance explained by the habitat and spatial models, respectively. Percentage of total variation in occupancy rate explained by the pure and joint effects of habitat and spatial filters is shown