Gradual transitions in genetics and songs between coastal and inland populations of Setophaga townsendi

Setophaga townsendi is a species of wood-warbler (family Parulidae) in northwestern North America that has a geographic structure in the mitochondrial and nuclear genomes: while interior populations have differentiated mitonuclear ancestry from the sister species S. occidentalis, coastal populations have a mix of inland and S. occidentalis mitonuclear ancestries. This coastal-to-inland transition in genomic ancestry raises the possibility of similar geographic structure in phenotypic traits, especially those involved in mate choice. Using qualitative and multivariate approaches, we investigated whether there is a sharp transition between coastal and inland populations in both songs and nuclear DNA. We find there is a shallow geographic cline in the Type I song but not in the Type II song. Nuclear DNA shows a gradient between the coast and inland. There is little correlation between variation in song and the isolation-by-distance pattern in the nuclear DNA. The learned songbird song is shaped by both genetic and cultural processes. There has been a debate on whether song learning promotes or slows down population differentiation. By comparing the within-species variation in song and genetic structures, we can expand our understanding of the dynamic interplay between mating signals and population differentiation.

Methods
Data collection

Song recordings were collected at 30 locations across British Columbia from May to July of 2017, using a Marantz PMD660 digital recorder and an Audio-Technica 815a Shotgun microphone. Recordings were typically eight to ten minutes long and consisted of ten to forty songs. For songs recorded after June 25th, a playback of song recordings was used to encourage birds to sing. We designed playbacks to consist of three song variants from different regions of the S. townsendi range, to avoid playback matching. 

This dataset consists of songs of 249 birds (180 from field recordings, 39 from Xeno-Canto, and 30 from Macaulay Library). For each bird, songs were characterized into types based on visual similarity and the results of Janes and Ryker (2016) and Janes (2017). We classified the clear song as the Type I song (i.e., used more in female attraction), and the buzzy song as the Type II song (i.e., used more in territorial defense).

We randomly selected three numbers from the total in each recording. These three songs were analyzed as follows: Spectrograms were visualized in Raven Pro 1.4 using Hann spectrogram windows with 512 samples, discrete Fourier transform (DFT) size of 512 samples, hop size of 5.6 ms, sampling frequency of 44.1 kHz, and a time resolution of 11.6 ms. The boundaries of each selection were conducted using the power spectrum as outlined in Zollinger et al. (2012). 

This dataset shows the following for 3 songs per bird per song type (Type I song n = 195; Type II song n =74):

Twenty-one variables were measured from each song using Raven Pro 1.4(Figure 1): total number of notes, number of unique notes, duration, minimum and maximum frequencies, bandwidth, and aggregate entropy of the whole song and parts 1 and 2 of the song.

The scores for these songs were analyzed in a PCA and then the mean and standard deviation of each variable were calculated for the three songs of each type.

汤森德林莺（Setophaga townsendi）是分布于北美西北部的一种森莺（森莺科Parulidae），其线粒体与细胞核基因组存在地理结构：内陆种群与姊妹种西方林莺（S. occidentalis）具有分化的线粒体核基因组祖先，而沿海种群则兼具内陆种群与西方林莺的线粒体核基因组祖先成分。这种基因组祖先的沿海-内陆渐变模式提示，表型性状（尤其是与配偶选择相关的性状）可能也存在类似的地理结构。本研究采用定性与多变量分析方法，探究了沿海与内陆种群在鸣唱与核DNA层面是否存在显著分化。研究发现，I型鸣唱存在平缓的地理渐变群，而II型鸣唱则无此现象；核DNA在沿海与内陆种群间呈现梯度分化。鸣唱变异与核DNA的距离隔离模式之间相关性较弱。鸣禽的鸣唱为后天习得，其演化受遗传与文化过程共同塑造。学界此前围绕鸣唱学习会促进还是延缓种群分化存在争议，本研究通过对比鸣唱与种群遗传结构的种内变异，可深化我们对交配信号与种群分化间动态互作关系的认知。

一、研究方法
1. 数据采集
2017年5月至7月，我们使用马兰士（Marantz）PMD660数码录音机与铁三角（Audio-Technica）815a枪式麦克风，在不列颠哥伦比亚省的30个采样点采集了鸣唱录音。单条录音时长通常为8至10分钟，包含10至40段鸣唱。对于6月25日之后录制的录音，我们播放预先录制的鸣唱片段以诱导鸟类鸣唱。为避免回放匹配效应，我们设计的回放片段包含汤森德林莺分布区内不同区域的3种鸣唱变体。

本数据集涵盖249只个体的鸣唱录音（其中180段为野外直接录制，39段来自Xeno-Canto鸟类声音库，30段来自麦考利图书馆（Macaulay Library）声库）。针对每只个体的鸣唱，我们依据视觉相似度，结合Janes与Ryker（2016）及Janes（2017）的研究结果，将其划分为不同类型。我们将清亮型鸣唱归类为I型鸣唱（主要用于吸引雌性），将buzzy声型鸣唱归类为II型鸣唱（主要用于领地防御）。

我们从每条录音中随机选取3段鸣唱进行分析，具体流程如下：使用Raven Pro 1.4软件生成声谱图，采用汉宁（Hann）窗函数，采样点长度为512，离散傅里叶变换（Discrete Fourier Transform, DFT）尺寸为512，跳步时长为5.6 ms，采样频率为44.1 kHz，时间分辨率为11.6 ms。鸣唱片段的边界划定参考Zollinger等（2012）提出的功率谱方法。

本数据集为每只个体的每种鸣唱类型提供了3段鸣唱的相关数据（I型鸣唱样本量n=195；II型鸣唱样本量n=74）：

我们通过Raven Pro 1.4软件为每段鸣唱测量了21项变量（如图1所示）：总音符数、独特音符数、鸣唱时长、最低与最高频率、频带宽度，以及整段鸣唱与鸣唱第1、2部分的整体熵值。

我们对上述鸣唱的测量值进行主成分分析（Principal Component Analysis, PCA），随后计算每种鸣唱类型对应的3段鸣唱的各变量均值与标准差。