Sex drives intraspecific scaling of home range size in mammals

This dataset contains derived home range estimates from individual-level GPS tracking data of 349 resident mammals across 18 species in Brazil, as well as individuals' body mass, sex, and habitat quality estimations. The dataset was compiled to investigate intraspecific variation in home range (HR) size and its relationship with body mass (BM) and sex. The study addresses a key knowledge gap in movement ecology: while interspecific allometric scaling of mammal home ranges is well documented, it is unclear whether the same patterns hold within species and for both sexes. This dataset and the associated code allow researchers to reproduce the analyses made by Giroux et al., visualize results, explore sex-specific and species-specific patterns, and integrate these data into broader comparative studies.

Methods
Animal data collection

We conducted this study across savanna and grassland ecosystems in Brazil (S 58°11'40" - 46°9'50", W 16°9'27" - 32°20'54"), under a tropical climate. The studied landscapes were composed of mosaics of natural forested and non-forested areas (e.g., woody and shrubby savannas, and open natural grasslands) as well as anthropogenic land uses such as exotic forests, crops (mainly rice and soybean), pasture, and highways. From 2007 until 2024, we captured, sexed, weighed, and GPS tracked 349 adult, healthy, free-living individuals, encompassing 18 mammal species from 9 families and 5 orders. The monitored species were puma (Puma concolor; 5 females and 15 males), ocelot (Leopardus pardalis; 5 females and 6 males), Geoffroy’s cat (Leopardus geoffroyi; 17 females and 9 males), margay (Leopardus wiedii; 5 females and 6 males), jaguar (Panthera onca; 13 females and 8 males), crab-eating fox (Cerdocyon thous; 14 females and 19 males), hoary fox (Lycalopex vetulus; 9 females and 12 males), maned wolf (Chrysocyon brachyurus; 17 females and 20 males), coati (Nasua nasua; 10 females and 7 males), six-banded armadillo (Euphractus sexcintus; 8 females and 10 males), southern three-banded armadillo (Tolypeutes matacus; 9 females and 8 males), pampas deer (Ozotoceros bezoarticus; 19 females and 15 males), wild boar (Sus scrofa; 13 females and 13 males), giant armadillo (Priodontes maximus; 15 females and 6 males), brown brocket deer (Subulo gouazoubira; 2 females and 4 males), white-lipped peccary (Tayassu pecari; 7 females and 3 males), giant anteater (Myrmecophaga tridactyla; 8 females and 6 males), and capybara (Hydrochoerus hydrochaeris; 3 females and 3 males).

Home range estimation

We used the ninety-five percent area kernel density estimator corrected for autocorrelation (AKDEc 95%) to estimate HR (ctmm R package). AKDEc is a nonparametric HR estimator that assumes that movement data represent a sample from a nonstationary and continuous process. This estimator was designed to deal not only with the autocorrelated structure of movement data but also with irregularly sampled data, allowing the comparison of HR estimates from different sampling regimes and periods.

Habitat quality estimation

To estimate habitat quality, we relied on the 30 x 30 m MapBiomas land-use land-cover classification (LULC; Collection 7; https://mapbiomas.org/en). For each species, we determined the most used habitat type (forest or non-forest) based on the LULC classification of individuals' location points, reflecting species-specific habitat preferences. Within each individual home range, we then calculated the proportion of the habitat type most used by its species as a proxy of habitat quality.

Statistical model

We assessed intraspecific HR scaling by regressing individuals’ HR against individuals’ BM. However, because species vary substantially in HR and BM magnitude, we first standardized HR and BM by dividing these variables by their species-specific mean. Because greater variance in HR and BM can facilitate the detection of scaling relationships, we checked whether, within species, one sex consistently exhibited higher variance in these traits than the other. We then modeled the log of the standardized HR with a Gaussian multiple linear regression model, allowing for species-specific intercept and slope parameters. Covariates in the regression model were: habitat quality, sex (coded as a binary variable where 0’s were females and 1’s were males), log standardized BM, and the interaction between log standardized BM and sex. The residual variance consisted of the sum of an overall variance parameter and the variance of the log standardized HR. To calculate HR variance, we relied on the lower and upper 95% confidence intervals for each individual’s HR. Finally, we allowed for the species-specific intercepts and slopes to be phylogenetically correlated. To this end, we calculated phylogenetic distances using the mammal supertree and the “phytools” R package. We fitted this model within a  Bayesian framework using JAGS, within the R package jagsUI.

本数据集源自巴西境内18个物种共349只定居性哺乳动物的个体级GPS追踪数据，衍生得到其家域（home range, HR）估计值，同时包含个体体重（body mass, BM）、性别以及生境质量评估数据。本数据集的汇编旨在探究家域面积的种内变异及其与体重和性别的关联。本研究填补了运动生态学中的一项关键认知空白：目前哺乳动物家域的种间异速缩放规律已有充分记载，但同一物种内以及不同性别间是否存在相同规律仍不明确。本数据集及配套代码可供研究人员复现Giroux等人的分析流程、可视化研究结果、探索性别特异性及物种特异性模式，并将这些数据整合至更广泛的比较研究中。

## 方法
### 动物数据收集
本研究在巴西的稀树草原与草原生态系统（纬度范围：南纬58°11'40''至46°9'50''，经度范围：西经16°9'27''至32°20'54''）内开展，该区域属热带气候。研究景观由自然林与非林区域（如木本灌丛稀树草原、开阔天然草原）以及人为土地利用类型（如外来人工林、农作物（主要为水稻与大豆）、牧场及公路）镶嵌组成。2007年至2024年间，研究团队共捕获、鉴定性别、称量体重并对349只健康的野生成年个体进行GPS追踪，这些个体隶属于5个目、9个科的18个哺乳动物物种。监测物种包括：美洲狮（*Puma concolor*；雌5只、雄15只）、虎猫（*Leopardus pardalis*；雌5只、雄6只）、乔氏猫（*Leopardus geoffroyi*；雌17只、雄9只）、长尾虎猫（*Leopardus wiedii*；雌5只、雄6只）、美洲豹（*Panthera onca*；雌13只、雄8只）、食蟹狐（*Cerdocyon thous*；雌14只、雄19只）、灰狐（*Lycalopex vetulus*；雌9只、雄12只）、鬃狼（*Chrysocyon brachyurus*；雌17只、雄20只）、长鼻浣熊（*Nasua nasua*；雌10只、雄7只）、六带犰狳（*Euphractus sexcintus*；雌8只、雄10只）、南方三带犰狳（*Tolypeutes matacus*；雌9只、雄8只）、潘帕斯鹿（*Ozotoceros bezoarticus*；雌19只、雄15只）、野猪（*Sus scrofa*；雌13只、雄13只）、大犰狳（*Priodontes maximus*；雌15只、雄6只）、棕短角鹿（*Subulo gouazoubira*；雌2只、雄4只）、白唇西猯（*Tayassu pecari*；雌7只、雄3只）、大食蚁兽（*Myrmecophaga tridactyla*；雌8只、雄6只）以及水豚（*Hydrochoerus hydrochaeris*；雌3只、雄3只）。

### 家域估计
研究采用校正自相关的95%面积核密度估计器（AKDEc 95%）结合ctmm R包进行家域估计。AKDEc属于非参数家域估计方法，其假设运动数据来自非平稳连续过程。该估计器不仅可处理运动数据的自相关结构，还可应对不规则采样数据，支持不同采样方案与采样时段下的家域估计值对比。

### 生境质量估计
生境质量评估依托30米×30米分辨率的MapBiomas土地利用/土地覆被分类（land-use land-cover classification, LULC；第7版；https://mapbiomas.org/en）完成。针对每个物种，研究基于个体定位点的LULC分类结果确定其最偏好的生境类型（林地或非林地），以此反映物种特异性生境偏好。随后在每个个体的家域范围内，计算该物种最偏好生境类型的占比，以此作为生境质量的替代指标。

### 统计模型
本研究通过将个体家域对个体体重进行回归分析，评估家域的种内缩放规律。由于不同物种的家域与体重数值差异显著，研究首先以物种特异性均值对家域与体重进行标准化处理。鉴于家域与体重的方差越大越易检测到缩放关系，研究进一步检验了同一物种内是否存在某一性别在这两个性状上的方差显著高于另一性别。随后采用高斯多元线性回归模型对标准化家域的对数值进行建模，允许模型包含物种特异性的截距与斜率参数。回归模型的协变量包括：生境质量、性别（以二分类变量编码，雌性记为0，雄性记为1）、标准化体重的对数值，以及标准化体重对数值与性别的交互项。残差方差由总体方差参数与标准化家域对数值的方差两部分组成。家域方差的计算依托每个个体家域的95%置信区间上下限完成。最后，研究允许物种特异性的截距与斜率项存在系统发育相关性：基于哺乳动物超树与phytools R包计算系统发育距离，并在R包jagsUI提供的贝叶斯框架下使用JAGS拟合该模型。