Noise pollution, one of the most prominent features of urbanization, is an important factor influencing the vocal behavior of urban wildlife. Studies have reported that many songbirds raise their song minimum frequencies in response to urban noise. It has been proposed that this increased minimum frequency (IMF) of songs is an adaptation that allows urban populations to cope with the masking effect of noise pollution. However, urban populations of some songbirds do not exhibit significant IMF compared with nonurban populations; thus, the notion that IMF is an adaptation to urban noise has been questioned. Furthermore, the effects of IMF might be influenced by both noise levels and the acoustic structures of songs. Here, we employed dichotomous and gradient effect size approaches to investigate IMF regarding two distinct acoustic structures (whistled and harmonic) in songs of six grey-cheeked fulvetta (Alcippe morrisonia morrisonia) populations in Taiwan, three with high noise pollution and three with low noise pollution. We found that when using the dichotomous approach, paired populations with significant divergence in noise levels exhibited weak or insignificant divergence in the minimum frequencies for both whistled and harmonic phrases. In contrast, we found that when using the gradient approach, the effect size of noise-level divergence was strongly correlated with the effect size of divergence in the minimum frequency of the harmonic phrase and only moderately correlated with the effect size of divergence in the minimum frequency of the whistled phrase. These findings suggest that noise pollution has a more pronounced effect on the divergence in the minimum frequency of harmonic phrases used in short-range communication compared to the whistled phrases used in long-range communication. We conclude that for population comparisons on the IMF, adopting a gradient approach could provide insights into the impact of noise pollution on the acoustic structures of songs across various communication ranges.

For noise measurements:

The present study aimed to investigate the effect of noise from a population perspective rather than an individual territory perspective. Thus, we measured site noise instead of territory noise (Derryberry and Luther 2021). To represent the site noises that the population had been exposed to during the study period, we conducted noise level sampling at six sites in 2020, the middle year of the study period. Measurements of the maximum noise levels (maximum hold function, C-weighting function on a Sound Level Meter TES-1350, TES, Taiwan) were collected at locations that were at least 100 m apart along the song-recording route (BP site: 9 locations; CM site: 28 locations; MS site: 13 locations; NS: 11 locations; SM site: 10 locations; SP site: 18 locations) following the procedures detailed in Shieh et al. (2016). We held the sound level meter horizontally at approximately 1.5 m during each noise measurement. We turned the meter 360° in a clockwise direction to measure the noise from all directions and to measure the maximum noise levels within about 30 seconds. We measured the maximum noise levels at each location three times and averaged the three measurements for further analysis.

For minimum frequency measurements:

We applied a high-pass filter originally at 2.0 kHz and adjusted the high-pass frequency for each individual to produce spectrograms that minimized the unwanted noise components. From the recordings, we produced spectrograms with the following parameters: sampling frequency = 22.05 kHz, FFT = 512, Hamming window, frequency resolution = 43 Hz, and time resolution = 2.9 ms. We measured the minimum frequency of each phrase on these spectrograms using the following procedures. First, on the whole phrase level, we visually identified the syllable in each phrase with the lowest frequency and manually labeled the syllable on the spectrogram; if two syllables in the same phrase had close frequency ranges, a frequency cursor was added to the spectrogram and the syllable with the lowest frequency was quickly identified. Then, on the syllable level, the minimum frequency of the labeled syllable was automatically measured by the Automatic Parameter Measurements setup (peak frequency for the minimum parameter of entire element) in Avisoft-SASLab Pro v5.2.