Data from: Ecological and intrinsic drivers of foraging parameters of Eurasian lynx across Europe
Data files
Nov 21, 2024 version files 996.14 KB
-
dataset_GLC_classification.csv
140.34 KB
-
handling_t.csv
355.78 KB
-
inter_kill.csv
490.80 KB
-
README.md
9.21 KB
Abstract
The estimation of foraging parameters is fundamental for understanding predator ecology. Predation and feeding can vary with multiple factors, such as prey availability, presence of kleptoparasites, and human disturbance. However, our knowledge is mostly limited to local scales, which prevents studying effects of environmental factors across larger ecological gradients. Here, we compared inter-kill intervals and handling times of Eurasian lynx (Lynx lynx) across a large latitudinal gradient from subarctic to the Mediterranean ecosystems using a standardised dataset of predicted adult ungulate kills from 107 GPS-collared lynx from nine distinct populations in Europe. We analysed variations in these two foraging parameters in relation to proxies reflecting prey availability, scavengers’ presence, and human disturbance, to improve our understanding of lynx predation at a continental scale. We found that inter-kill intervals and handling times varied between populations, social status and in different seasons within the year. We observed marked differences in inter-kill intervals between populations, which do not appear to be driven by variation in handling time. Increases in habitat productivity (expressed by NDVI, used as a proxy for prey availability) resulted in reduced inter-kill intervals (i.e. higher kill rates). We observed less variation in handling (i.e. feeding) times, although presence of dominant scavengers (wild boars and brown bears) and higher human impact led to significantly shorter handling times. This suggests that kleptoparasitism and human disturbance may limit the energetic input that lynx can obtain from their prey. We also observed that the human impact on foraging parameters can be consistent between some populations but context-dependent for others, suggesting local adaptations by lynx. Our study highlights the value of large-scale studies based on standardised datasets, which can aid the implementation of effective management measures, as patterns observed in one area might not be necessarily transferable to other regions. Our results also indicate the high degree of adaptability of these solitary felids, which enables them to meet their energy requirements and persist across a wide range of environmental conditions despite the constraints imposed by humans, dominant scavengers and variable prey availability.
This README file describes the provided dataset for the random forest classification analysis, and variation of foraging parameters (handling time and inter-kill interval).
Description of the data and file structure
- According to our paper, this dataset contains the information required to run the random forest models. It includes the type of cluster (class), animal social status (social_status), and cluster characteristics (latitude, cluster duration, number of cluster locations, fidelity, maximum foray, average cluster distance, maximum cluster radius, and night proportion fixed and automatic). All these parameters were extracted from GPS location clusters, using the GPSeqClus R package (https://doi.org/10.1111/2041-210X.13572).
- Format(s): .csv
- Size(s): 138 KB
- Dimensions: 3435 rows x 11 columns
-
Variables:
*class - Type of cluster; 0 - non-kills and small prey kills, 1 - adult ungulate kills
*latitude - Latitude of the cluster
*clus_dur_hr - Cluster duration (hours), from the first cluster point to the last
*n_clus_locs - Number of locations within a cluster
*fid - Fidelity to cluster, calculated as the total number of cluster locations minus the locations away from the cluster. If equal to 0, it has the same number of locations inside and outside of the cluster. If it is more than 0, more locations fall outside the cluster, while if it is less than 0, more locations are inside the cluster.
*max_foray - Maximum foray from cluster centroid (m), as the maximum distance from any location to the centroid
*max_clus_radius - Maximum cluster radius (m) based on the fixes within the cluster
*avg_clus_dist - Average cluster radius (m) based on the fixes within the cluster
*night_proportion_fixed - Proportion of night locations, as the number of night points divided by the number of cluster location. The night period is fixed between 5pm and 7 am.
*night_proportion_auto - Proportion of night locations, as the number of night points divided by the number of cluster location. Each GPS location is classified as night (or not) automatically from the suncalc package (https://cran.r-project.org/web/packages/suncalc/index.html).
*social_status - Social status of the lynx; 0 - adult males; 1 - adult females; 2- family groups (females with offspring) ; 3- juvenile males ; 4- juvenile females - Missing values - there are no missing values in this dataset
- This file is the direct input to perform random forest classification, with binary classification. Methods of data collection/generation: see publication for details.
Dataset Title : Data for modelling the variation of handling time of Eurasian lynx GLCs reflecting adult ungulates across Europe
Description of the data and file structure
- This dataset contains the information required to run the GAM models related to feeding time, accordingly to our paper. It includes cluster id (animal_cluster_id), lynx id (animals_id), population (population_id_cat), social status (social_status), handling time (feeding_t), month, and environmental characteristics (distance to settlements, proportion of wild boar, bear, and wolverine, and scavenger count). These characteristics had different origins and methods of extraction; please check the publication for details.
- Format(s): .csv
- Size(s): 348 KB
- Dimensions: 4425 rows x 12 columns
-
Variables:
*animal_cluster_id - Unique ID for each cluster
*animals_id - Individual lynx id
*population_id_cat - Population of each lynx
*social_status - Social status of the lynx; ‘a_m’ - adult male, ‘a_f’ - adult female, and ‘fam’ - family groups
*feeding_t - Handling time, estimated as the time, in days, between the start of kn and its abandonment (first and last GPS fixes of the given cluster, respectively)
*month - Month of the first fix of the GLC
*h_mod_index_17_kill - Human modification index extracted for each GLC (more details: Appendix S3- Table S3)
*dist_stlm_kill - Distance to settlements for each kill (more details: Appendix S3- Table S3)
*scavengers_count - Number of scavenger species potentially present at the GLC according to occurrence maps
*wildboar_period_prop - Proportion of the area of each tracking sequence (95% MCP) overlapping with wild boar occurrence map
*bear_period_prop - Proportion of the area of each tracking sequence (95% MCP) overlapping with bear occurrence map
*wolverine_period_prop - Proportion of the area of each tracking sequence (95% MCP) overlapping with wolverine occurrence map - Missing values - there are no missing values in this dataset
- This file is the direct input to run the GAMs with the model structure provided in Appendix S4: Table S4. Methods of data collection/generation: see publication for details.
Dataset Title : Data for modelling the variation of inter-kill intervals of Eurasian lynx GLCs reflecting adult ungulates across Europe
Description of the data and file structure
- This dataset contains the information required to run the GAM models related to handling, accordingly to our paper. It includes cluster id (animal_cluster_id), lynx id (animals_id), population (population_id_cat), social status (social_status), inter-kill interval (inter_kill), month, number of daily fixes used per tracking sequence (n_locs_original), and environmental characteristics (distance to settlements, proportion of wild boar, brown bear, and wolverine, scavenger count, forest edge density, NDVI, and MCP 85%). These characteristics had different origins and methods of extraction; please check the publication for details.
- Format(s): .csv
- Size(s): 348 KB
- Dimensions: 4166 rows x 16 columns
-
Variables:
*animal_cluster_id - Unique ID for each cluster
*animals_id - Individual lynx id
*population_id_cat - Population of each lynx
*social_status - Social status of the lynx; ‘a_m’ - adult male, ‘a_f’ - adult female, and ‘fam’ - family groups
*inter_kill - Inter-kill interval, estimated as the time elapsed, in days, between the start of a kill kn (considered as the first GPS fix of the cluster) and the start of the next kill kn+1.
*month - Month of the first fix of the GLC
*n_locs_original - Number of daily fixes used per tracking sequence
*h_mod_index_17_period - Human modification index calculated across each tracking sequence (95% MCP; median value); (more details: Appendix S3- Table S3)
*esm_prox_period - Distance to settlements calculated across each tracking sequence (95% MCP; median value); (more details: Appendix S3- Table S3)
*scavengers_count - Number of scavenger species potentially present at the GLC according to occurrence maps
*wildboar_period_prop - Proportion of the area of each tracking sequence (95% MCP) overlapping with wild boar occurrence map
*bear_period_prop - Proportion of the area of each tracking sequence (95% MCP) overlapping with bear occurrence map
*wolverine_period_prop - Proportion of the area of each tracking sequence (95% MCP) overlapping with wolverine occurrence map
*dens_f_edge_period - Forest edge density calculated for each tracking sequence (95% MCP; median value)
*NDVI_annual_mean_period - NDVI values (annual mean) extracted across each tracking sequence (95% MCP)
*mcp85_hr- 85% MCP of all GPS data for an individual with more than 10 months of tracking data. When an individual did not fit in this criterion, we attributed, as its home range size, the average HR size within the population and sex. - Missing values - there are no missing values in this dataset
- This file is the direct input to run the GAMs with the model structure provided in Appendix S4: Table S4. Methods of data collection/generation: see publication for details.
Contact Information
- Name: Teresa Oliveira
- Affiliations: Biotechnical Faculty, University of Ljubljana, Slovenia
- ORCID ID: https://orcid.org/0000-0001-9751-5198
- Email: mteresaoliveira92@gmail.com
Additional Dataset Metadata
Acknowledgements
- We obtained tracking data through the EUROLYNX network (https://euromammals.org/eurolynx/). This is the paper number 7 in the EUROLYNX series. We would like to thank all EUROLYNX members for the stimulating discussions. We thank Andrea Corradini and Anja Molinari-Jobin for their contributions in the earlier stages of this manuscript. We are extremely grateful to all the personnel involved in data collection (veterinarians, field technicians, researchers, hunters, rangers, volunteers, and students), including lynx capturing and collaring, as well as field-checking of GPS clusters.
Dates and Locations
- Dates of data collection: Field data for all study areas collected in several periods between 2008 and 2022.
- Geographic locations of data collection: Fieldwork conducted in several areas across a large latitudinal gradient in Europe (see publication and Figure 1 for more details).
We used GPS and kill-sites data from nine populations across Europe. Data were collected through the EUROLYNX network, a collaborative bottom-up platform of lynx researchers across Europe for sharing data and expertise (Heurich et al, 2021). We first developed a predictive model to classify GPS location clusters into adult ungulate kills or other (non-kill and/or small prey kill), following Oliveira et al. (2022) - see Appendix S2 for more details. The dataset used for this part of the analyses is entitled "dataset_GLC_classification".
Secondly, we used these predicted GPS location clusters to calculate inter-kill intervals and handling times. For each parameter, we extracted environmental covariates trelated to prey availability, human disturbance, and scavengers presence. We provide two datasets, one for inter-kill intervals ("inter_kill"), and another for handling times ("feeding_t").
For more details, please see the README document ("README_dataset_randomforest_classification.txt") and the accompanying published article. Oliveira et al., 2022. Predicting kill sites from an apex predator from GPS data in different multi-prey systems. Ecological Applications. Accepted