Data from: Differences in distress: variance and production of American crocodile (Crocodylus acutus) distress calls in Belize

We recorded American Crocodiles on Ambergris Caye from May – August 2015, December 2015 – January 2016, and March of 2016. Concurrent with ongoing population surveys, we collected recordings on Caye Caulker in January, March, and August 2016. We collected acoustic recordings from BAL June – August 2016. We recorded distress calls ad libitum in the field during capture events as recording conditions and difficulty of capture determined the actual number of recordings collected. Independent of acoustic recording, we noted incidence of call production for each individual capture as not all captures resulted in call production or successful recording. We collected distress call recordings from May 2015 – January 2016 using a Marantz PDM661 or Roland R-26 digital recorder coupled with a Senheiser ME67 shotgun directional microphone. During the March–August 2016 field seasons, we employed a Sony Zoom H5 digital recorder with an XY modular microphone capsule. We recorded all calls in .wav format at a sampling rate of 44.1 kHz and 24 bits per sample. Although equipment differed, sampling protocol remained consistent. We held hatchling and juvenile crocodiles in hand and collected recordings approximately 50 cm from the microphone. We recorded sub-adult and adult crocodiles from 1–2 m from the microphone to ensure safety of personnel. We also recorded any calls or responses from nearby conspecifics, noting size class if possible, when we recorded distress calls of captured crocodiles. Sound Analysis We performed acoustic analysis to determine the structure of distress calls for each size class. We analyzed five call per individual, in one case only 3 calls were analyzed due to heavy background noise, and measured seven acoustic variables, two temporal and five spectral, using Raven Pro 1.5 acoustic analysis software (Bioacoustics Research Program, 2014). We used spectrographic analysis (window size 1024, overlap 80%) of the fundamental frequency to determine maximal frequency (Fmax, Hz), frequency at end of first quartile (F1/4, Hz), and final frequency (Fend, Hz; Fig. 3A). Using call oscillograms, we measured temporal properties for total duration (DT, s) and duration of the first quartile (D1/4, s; Fig. 3B). Measurement windows were drawn around the fundamental frequency and the maximal frequency values at the beginning, end, and first quartile were recorded. We used frequency and temporal measurements to calculate call modulation of the first temporal quartile slope (Slope 1, Hz/s, calculated as (F1/4−Fmax)/D1/4), and the slope of the remaining three temporal quartiles (Slope 2, Hz/s, calculated as (Fend−F1/4)/( DT−D1/4)) (Vergne et al. 2012). Concurrent to call measurements, we recorded number of calls produced by each individual for 10, 20, and 30 second intervals as total recording time varied between individual crocodiles. We began call counts at the first recorded call for each individual We used size designation to organize and analyze distress call recordings by overall size class. We performed statistical analyses using RStudio version 0.99.902 (RStudio Team, 2015). Our call parameter data did not meet the assumptions of normality or homogeneity of variance (P < 0.05)(Boucher 2017). Thus, we analyzed call parameter means (DT, D1/4, Fmax, F1/4, Fend, Slope 1, Slope 2) among size classes (hatchling, juvenile, sub-adult, adult) by performing non-parametric Kruskal-Wallis (H) tests. We used Mann-Whitney post-hoc testing with Bonferroni correction to determine pairwise variance of call parameters between size classes following a significant Kruskal-Wallis test (P < 0.05). We tested size class differentiation by call parameters using a principal component analysis cross-validated by discriminant function analysis which assigned each recorded individual to a size class based on call parameters. To determine variance in number of calls produced (10, 20, 30 second intervals) we performed one-way analysis of variance (ANOVA) tests as these data met assumptions of normality (P > 0.05) and equality of variance (P > 0.05). We analyzed differences in call production by determining the total number of American Crocodiles captured throughout the study for each location and compared them by size class. To account for small sample size, we aggregated data for Caye Caulker and BAL sites. We performed a 2-sample test for equality of proportions with continuity correction (Newcombe, 2008) on total call production proportions among size classes. Designed for small sample sizes (< 5), we performed Fisher’s Exact Test (Agresti, 2002) to determine inequality in call production by size classes between Ambergris Caye, and the combined Caye Caulker and BAL sites. As a compliment to proportional tests, we performed Mann-Whitney tests to determine variance of spectral parameters of hatchling and juvenile calls between Ambergris Caye and aggregated Caye Caulker and BAL sites.

本研究于2015年5月至8月、2015年12月至2016年1月以及2016年3月期间，在安伯格里斯岛（Ambergris Caye）对美洲鳄（American Crocodile）开展了记录工作。同步开展种群监测工作，我们于2016年1月、3月和8月在考尔克岛（Caye Caulker）采集了声学数据，并于2016年6月至8月在BAL采集了声学录音。

在野外捕获过程中，我们根据录音条件与捕获难度灵活录制遇险叫声，实际采集的录音数量随现场情况而定。除声学录音外，我们还记录了每一次捕获事件中叫声的产生情况——并非所有捕获过程都能产生遇险叫声或获得有效录音。

2015年5月至2016年1月期间，我们采用马兰士（Marantz）PDM661或罗兰（Roland）R-26数字录音机，搭配森海塞尔（Senheiser）ME67枪式指向麦克风，采集遇险叫声录音。2016年3月至8月的野外季，我们改用索尼（Sony）Zoom H5数字录音机及XY模块化麦克风拾音头。所有录音均采用.wav格式，采样率为44.1 kHz，每样本位深为24比特；尽管设备存在差异，但采样流程保持一致。

我们手持刚孵化幼体与幼年期鳄类，将麦克风置于约50 cm处进行录音；对于亚成体与成体鳄类，为保障野外人员安全，我们将麦克风置于1~2 m外进行录制。在录制捕获鳄类的遇险叫声时，我们同时记录附近同类发出的叫声或回应，并尽可能标注其体型等级。

声学分析
我们开展声学分析以明确不同体型等级鳄类的遇险叫声结构。每只个体选取5份叫声样本进行分析，仅1只个体因背景噪音过大仅分析了3份样本。我们使用Raven Pro 1.5声学分析软件（生物声学研究项目组，2014），测量了7项声学变量，其中2项为时域参数（temporal）、5项为频域参数（spectral）。

针对基频（fundamental frequency）开展声谱分析（spectrographic analysis，窗口大小1024，重叠率80%），以确定最大频率（Fmax，Hz）、第一四分位频点频率（F1/4，Hz）与最终频率（Fend，Hz；见图3A）。通过叫声振荡图（oscillograms），我们测量了总时长（DT，s）与第一四分位时长（D1/4，s；见图3B）两项时域属性。我们在基频信号周围划定测量窗口，记录起始、结束及第一四分位处的最大频率数值。

我们结合频率与时域测量结果，计算得到第一四分位斜率的叫声调制率（Slope 1，Hz/s，计算公式为(F1/4−Fmax)/D1/4），以及剩余三个四分位的斜率（Slope 2，Hz/s，计算公式为(Fend−F1/4)/(DT−D1/4)）（Vergne等，2012）。

同步开展叫声测量工作期间，我们针对每只个体，按10秒、20秒、30秒的时间间隔统计其发出的叫声总数——因不同鳄类个体的总录音时长存在差异。我们以每只个体的第一声录制叫声作为计数起点。

我们按照体型等级对遇险叫声录音进行分组，开展后续分析。本研究使用RStudio 0.99.902版本（RStudio团队，2015）开展所有统计分析工作。

本研究的叫声参数数据不符合正态性与方差齐性假设（P < 0.05，Boucher，2017），因此，我们采用非参数克鲁斯卡尔-沃利斯（Kruskal-Wallis, H）检验，分析不同体型等级（刚孵化幼体、幼年期、亚成体、成体）间的各项叫声参数均值（DT、D1/4、Fmax、F1/4、Fend、Slope 1、Slope 2）。当克鲁斯卡尔-沃利斯检验结果显著（P < 0.05）时，我们采用经邦弗朗尼（Bonferroni correction）校正的曼-惠特尼（Mann-Whitney）事后检验，以确定不同体型等级间叫声参数的成对差异。

我们采用经判别函数分析（discriminant function analysis）交叉验证的主成分分析（principal component analysis），基于叫声参数将每只录制个体归类至对应体型等级，以检验不同体型等级间的声学特征区分度。

为分析10秒、20秒、30秒时间间隔内的叫声数量差异，我们开展单因素方差分析（one-way analysis of variance, ANOVA）——此类数据符合正态性（P > 0.05）与方差齐性（P > 0.05）假设。

我们通过统计整个研究期间各采样点捕获的美洲鳄总数量，并按体型等级进行比较，以分析叫声产生情况的差异。考虑到部分采样点样本量较小，我们将考尔克岛与BAL采样点的数据进行合并。

我们针对不同体型等级的总叫声产生比例，开展了带连续性校正（continuity correction）的双样本比例相等检验（Newcombe，2008）。针对小样本量（<5）的场景，我们采用费希尔精确检验（Fisher’s Exact Test，Agresti，2002），以检验安伯格里斯岛与合并后的考尔克岛、BAL采样点之间，不同体型等级的叫声产生情况是否存在显著差异。

作为比例检验的补充分析，我们采用曼-惠特尼检验，对比安伯格里斯岛与合并后的考尔克岛、BAL采样点之间，刚孵化幼体与幼年期鳄类叫声的频域参数差异。