Introduction

The anthropogenic sounds (from now on, anthropophony) are a significant component of the acoustic landscape in several ecosystems (Pijanowski et al. 2011). The impact of anthropophony is so pervasive that, in a recent review, some form of effect has been found in all studied animal groups (Kunc and Schmidt 2019). These effects go from temporary hearing loss observed in humans (Alexander 1968) to developmental delays or increased mortality in marine invertebrates (Nedelec et al. 2014; Pine et al. 2016) to show some examples. Despite these well-known effects, there is a significant gap in our understanding of how antropophony affects freshwater ecosystems (Gammell and O’Brien 2013). To date, there are fewer than a dozen articles that have focused on studying the effects of anthropophony on freshwater animals (for example Amoser and Ladich 2010; Desjonquères et al. 2020; Rojas et al. 2023). This lack of knowledge emerges prominently as freshwater ecosystems are among the most threatened by human activity (Dudgeon et al. 2006).

One of the main consequences following exposure to anthropophony is the potential for acoustic masking, which hinders proper signal perception by recipient organisms (Warren et al. 2006). In response to this scenario several mechanisms -such as the use of selective filters, spatial evasion, temporal evasion, or spectral evasion- have been proposed (Warren et al. 2006; Schmidt and Römer 2011; Schmidt and Balakrishnan 2015). Among these, spectral evasion has been reported as the most common process in terrestrial animals, which implies modifications in the frequency of acoustic signals to utilize a different acoustic niche than the one occupied by other sounds (Duquette et al. 2021; Gomes et al. 2022), particularly low-frequency sounds associated with human activity. As for the metrics to measure antropophony, one of the most utilized is the Normalized Difference Sound Index (NDSI), which assesses the energy relationship between high-frequency sounds (typically associated with biophony) and low-frequency sounds (mainly linked to motors, thus indicating anthropophony) (Kasten et al. 2012). This index has proven sensitive to urbanization, resulting in positive values in areas with low human density and vice versa along rural/urban gradients (Machado et al. 2017; Fairbrass et al. 2017; Ross et al. 2021). Consequently, in a sprawling city with diverse management zones like Mexico City (Núñez 2021), it is expected to observe distinct acoustic landscapes across different sites.

We have investigated how the prevalence of low-frequency sounds, as determined by the NDSI, influences the stridulation ability of Krizousacorixa femorata (Hemiptera: Corixidae), a common aquatic insect inhabiting water bodies across Mexico City. We used this species as a model organism to explore the effects of anthropophony on freshwater aquatic biota. K. femorata stridulates within a frequency band centered around 2 kHz (all authors’ pers. obs.), raising the possibility that these sounds may be masked by the lower-frequency sounds associated with anthropogenic activity between 1 and 2 kHz. The aims of this study were (1) to characterize the acoustic landscape of six sites in Mexico City along the year to generate a soundscape gradient evaluated through the NDSI, (2) to characterize the stridulation of K. femorata in different sampling sites by its peak frequency (also called dominant frequency; hereafter called Pf), and (3) to identify any relationship between the NDSI-evaluated acoustic landscape and the Pf. We predicted the existence of a soundscape gradient across the various sites in Mexico City, such that locations with a higher dominance of human activities would exhibit negative NDSI values. These values were expected to vary throughout the year, reflecting different patterns of human activity at different sites. Also, we anticipated distinct Pf values for K. femorata in different locations, with these values being associated with the natural history of each site. By last, we expected a negative relationship between Pf and NDSI, indicating that insect sounds would be higher-pitched in areas with greater human activity. This corresponds to the most reported response process in the presence of spectral masking (Duquette et al. 2021; Gomes et al. 2022).

Materials and methods

Study subject

Krizousacorixa femorata is a bug that has only been recorded in Mexico (Berner and Hungerford 1977) and inhabits water bodies throughout Mexico City and its surroundings (Contreras-Rivero et al. 2001, 2005, 2008; Farfán-Beltrán et al. 2023). Adult males of K. femorata produce stridulatory sounds by rubbing their front legs against their face, which has been identified as a secondary sexual trait (Peters 1962). This insect can survive and reproduce even under poor water quality conditions (Farfán-Beltrán et al. 2023), and it is considered culturally important because its eggs, known as “ahuautle”, have been consumed by the native people of the basin since pre-Hispanic times (Ancona 1933).

Study areas

Six sites with water bodies were selected in Mexico City (Fig. 1), where we were able to leave of equipment working for a minimum of 72 h. Each of these sites exhibits several management practices, leading to exposure to distinct sounds (Table 1). All sites were situated at altitudes ranging from 2200 to 2300 m and were separated by a minimum distance of 1.6 km (Fig. 2).

Fig. 1
figure 1

Pipeline diagram of the methods followed. Blue circled terms are the subdivisions of this section. Subsequent terms provide details about each step

Full size image
Table 1 Key social characteristics of the sampling sites. Letters in parentheses indicate how the sites are referenced throughout the text
Full size table
Fig. 2
figure 2

Sampling sites located in Mexico City. Sites selected covers the main water bodies of the municipality and the actual local distribution of K. femorata

Full size image

Recording methodology

At each site, automatic recordings were made using two recording systems (Fig. 1|): a song meter mini recorder from Wildlife Acoustics (sensitivity of -35 ± 4 dB re 1 V/Pa at 1 kHz, using a sampling rate of 44.1 kHz and a sample depth of 16 bits, hereafter aerial recorder) and a custom-made recorder assembled and configured according to the specifications of a SOLO unit (hereafter aquatic recorder) (Whytock and Christie 2017). For the latter, an Aquarian Audio H2a hydrophone (maximum sensitivity of -60 dB re 1 V/Pa in the 2.0 to 2.2 kHz band) was used to directly record insect stridulations underwater, with a recording configuration of 44.1 kHz of sampling rate and a sample depth of 16 bits. Despite its low sensitivity, we consider the use of this equipment suitable for recording stridulations because insects in this family tend to produce intense sounds, reaching up to 99.2 dB SPL ref 20 µPa in the case of Micronecta scholtzi (Sueur et al. 2011).

The first recording was made on June 17, 2021, and each weekend work was conducted at a different site. Each site was revisited every three months, aiming to capture acoustic landscape variability throughout the year. The recorders started operation at 12:00 p.m. on Saturdays and remained active until 11:00 a.m. on Tuesday for the aerial recorder, or 11:45 a.m. on Tuesday for the aquatic recorder. The selection of these days was carried out to ensure obtaining the greatest variability of activities at the sites. This is because recreational sites typically have a higher density of people on Saturdays and Sundays, while Mondays practically lack visitors.

The Song meter mini recorder was configured with a program that records the first five minutes of each hour, resulting in a total of 360 min of recording per site per quarter (hereafter, by quarter we mean a three-month period). The SOLO unit was configured to record five minutes every 15 min, resulting in 20 min of recording per hour, and a total of 1440 min per site per quarter. This was done to increase the likelihood of capturing bug stridulations.

Acoustic diversity analysis

Due to the limited sensitivity of the aquatic recorder and its non-flat response within the bandwidth, aerial recordings were chosen to estimate the aquatic acoustic landscape. Since dB are logarithmic units, a 25 dB difference in recording equipment implies that a sound must be 77.78 times more intense to produce the same analog signal. In our recording system, this means that by technical issues, aquatic recordings were not able to capture sounds that the aerial recording did. Although we are aware that we are not capturing the entirety of the aquatic soundscape, it has been demonstrated that aerial sounds comprise a significant portion of the aquatic acoustic landscape, as they can capture sounds of birds, aircraft, or even terrestrial traffic, which can impact aquatic biota (Rountree et al. 2020). Also, there are no differences of frequencies in sounds that propagates from air to water, as the differences in wavelength are offset by the differences in sound velocity (Au and Hastings 2008). Each aerial recording of five minutes was processed using the “seewave” package in R (Sueur et al. 2012) to estimate the NDSI, since it has been demonstrated that a five-minute recording is sufficient to characterize urban environments according to their acoustic parameters (Lavandier and Barbot 2003). This index, developed by Kasten and collaborators (2012), compares energy within different spectral frequency bands associated with human activities (anthrophony) and sounds produced by other animals (biophony). By default, the index assumes that human-origin sounds are in the 1 to 2 kHz band and sounds above 2 kHz up to 11 kHz are generated by non-human animals.

The index is calculated with the following expression:

$$NDSI= \frac{b-a}{b+a}$$

Where b corresponds to the spectral density of sounds considered biophony, and a corresponds to the spectral density of sounds considered anthrophony. Thus, the smallest possible value is -1 (indicating exclusive anthrophony in that recording), and the highest is 1 (indicating exclusive biophony in that recording). However, since values are often intermediate, we refer to dominant anthrophony or biophony throughout the text.

Characterization of K. femorata stridulation

Aquatic recordings from each site were individually analyzed using Raven Pro software (K. Lisa Yang Center for Conservation Bioacustics at the Cornell Lab of Ornithology 2022). In sites where insects were present in all files, signals distinct from others were selected (i.e., not overlapping with sounds from other individuals). For recordings with sparse stridulations, all events were selected.

Due to the lack of control over the distance between insects and the hydrophone and the possibility of acoustic attenuation or absorption phenomena that could hinder obtaining true maximum or minimum stridulation frequencies, only the Pf was chosen as the response variable. Pf, being the frequency with the highest energy within the stridulation, can travel a greater distance and stand out more among ambient sounds (Duquette et al. 2021).

For three sites (BA, DEX, and PEX), only values from the first three samplings are shown due to a technical issue with the hydrophone, which prevented data collection during the fourth sampling.

Relationship between NDSI and Pf

To assess the effect of anthrophony evaluated through NDSI on Pf, a mixed lineal model was conducted through “lme4” package (Bates et al. 2015). To do so, we averaged the NDSI of each site/sampling quarter and used this new value as the independent variable. Also, we used the Pf of each analyzed stridulation as the dependent variable. Sites and sampling quarters were considered as random effects. Selection of the best model was carried out through the Akaike Information Criterion.

Analyses were carried out using R Statistical Software (v4.2.2, R Core Team 2021).

Results

NDSI characterization of the lakes

Two recreational sites (BA and DEX) had negative NDSI values over time, implying anthropophony dominance throughout the year. The agricultural and livestock, and a recreational site (CH and PEX) exhibited values indicating biophony dominance in these locations. Both a recreational and an agricultural site (NN and CT) showed inconsistent values over time, which indicates the presence of both anthropophony and biophony dominance throughout the year (Fig. 3).

Fig. 3
figure 3

(A) Mean NDSI values and their standard deviation in the different studied sites throughout a year BA = Bosque de Aragón; CH = Lago de Chalco; CT = Chinampas de Tláhuac; DEX = Deportivo Ecológico Xochimilco; NN = Nuevo Nativitas; PEX = Parque Ecológico Xochimilco. Different sampling quarters are represented by distinct colors, encompassing: the first quarter from June to August, the second quarter from September to November, the third quarter from December to February, and the fourth quarter from March to May. (B) A spectrogram fragment showing what a soundscape looks like with higher energy between 1–2 kHz than at frequencies from 2–11 kHz. (C) A spectrogram fragment showing what a soundscape looks like with higher energy at frequencies between 2–11 kHz than in the band of 1–2 kHz

Full size image

Stridulation characterization

A total of 2085 stridulation events were recorded throughout the sampling period, showing variation among different sites over the year (Fig. 4). The lowest average Pf value was 2.21 kHz and took place in an agricultural zone (CT); the highest was 3.37 kHz and was observed in a recreational zone (PEX). Examples of high and low-pitched stridulations are provided in supplementary materials (Online Resource 1 and 2, respectively).

Fig. 4
figure 4

(A) Mean Pf values and their standard deviation in the different studied sites throughout a year BA = Bosque de Aragón; CH = Lago de Chalco; CT = Chinampas de Tláhuac; DEX = Deportivo Ecológico Xochimilco; NN = Nuevo Nativitas; PEX = Parque Ecológico Xochimilco. Different sampling quarters are represented by distinct colors, encompassing: the first quarter from June to August, the second quarter from September to November, the third quarter from December to February, and the fourth quarter from March to May. (B) A spectrogram fragment showing the appearance of stridulation with a Pf around 2.5 kHz. (C) A spectrogram fragment showing the appearance of stridulation with a Pf around 3 kHz

Full size image

Relationship between NDSI and stridulation

Overall, there is a negative and significant (p < 0.000) correlation between NDSI and Pf. Nevertheless, there is a strong effect of random effects (marginal R2 = 0.19; conditional R2 = 0.69), so sites and quarters of samplings affects the magnitude of the relationship (Fig. 5).

Fig. 5
figure 5

Relationship between NDSI and Pf for each study site. BA = Bosque de Aragón; CH = Lago de Chalco; CT = Chinampas de Tláhuac; DEX = Deportivo Ecológico Xochimilco; NN = Nuevo Nativitas; PEX = Parque Ecológico Xochimilco. Different sampling quarters are represented by distinct line styles, encompassing: the first quarter from June to August, the second quarter from September to November, the third quarter from December to February, and the fourth quarter from March to May

Full size image

Discussion

Although the NDSI was useful in differentiating the acoustic landscape of our water bodies, in some cases, the obtained values were inconsistent with expectations based on the activities occurring at each location, particularly in Nuevo Nativitas and Chinampas de Tlahuac. Nuevo Nativitas is known for having the highest tourist activity among the selected sites, receiving up to 45,000 visitors per month (Alquicira-González 2020). Traditional musicians (mariachis) often perform on boats called trajineras, and loud music is also played through portable devices. As for the Chinampas de Tláhuac, there are only around 450 active farmers for an area of 164 ha, who access the area mainly by canoe, and to a lesser extent, by car (González-Pozo 2016), and there are usually no gatherings of people. Additionally, in contrast to Nuevo Nativitas, this site is located on the outskirts of the city. This means that the acoustic landscapes in these places are more complex than we expected, as there are sounds contributing more energy in the biophony and anthropophony band bands than anticipated, respectively.

Previously, it had been described that the NDSI can yield negative values in locations distant from cities when analyzing recordings from places with waves (Portuguez-Brenes et al. 2021) or rain (Sánchez-Giraldo et al. 2020). Furthermore, it has been reported that some aquatic animals can emit sounds in spectral frequencies typically associated with human activities (Rountree et al. 2020), so their activity may lead to an overestimated anthropophony in terms of NDSI. For this reason, we consider that the results we obtained should be understood not only as an analysis of the relationship between anthropophony and animal response, but also as a sample of the effect of the dominance of low spectral frequencies in an acoustic landscape, regardless of its origin. While it is true that most low-frequency sounds in acoustic landscapes correspond to anthropophony (Rountree et al. 2020; Ferguson et al. 2023), the approach we propose can provide insights into what may occur in systems distant from cities where low spectral frequencies are generated for some reason.

The acoustic signal of K. femorata exhibited a higher Pf in places dominated by low spectral frequencies. In insects, this has been described for cicadas (Shieh et al. 2012) and grasshoppers (Lampe et al. 2012). For cicadas, it was evident that along a continuous transect with varying levels of anthropophony, areas with higher human activity were those where insects communicated with higher-pitched tones. In this instance, the authors proposed that cicadas might minimize the size of their abdominal cavity, to generate higher-pitched sounds. This adaptation could allow them to evade sound masking. How higher-pitched sounds are produced by K. femorata remains to be investigated.

Although the explanatory power of the model used was relatively low, it was highly significant. This indicates that while the dominance of low frequencies in the acoustic landscape is a factor affecting the Pf of K. femorata, it is not the only one playing a role in this process. Other studies on insects have demonstrated that variables such as seasonality (Beckers and Schul 2008), atmospheric pressure or photoperiod (Zagvazdina et al. 2015), as well as temperature (Desjonquères et al. 2020), can induce changes in the acoustic signals of these animals. Furthermore, there appear to be specific conditions at each site that lead to variations in stridulation frequencies. Some mechanisms that could explain this include geographical differences in female preferences to specific frequencies (Olvido and Wagner Jr 2004), the presence of specific acoustic filters by receptors that restrict the frequency range in which emitters can be detected (Schmidt et al. 2011), or even the existence of predators that locate prey emitting sounds at certain frequencies (Goerlitz et al. 2008).

In summary, our study established a gradient in the acoustic landscape that encompasses the predominance of low spectral frequencies (associated with anthropophony) and the predominance of high spectral frequencies (associated with biophony). It seems that K. femorata individuals exhibit differences in the values of Pf at different locations, and that these differences may be linked to the gradient in the acoustic landscape, although there may be other variables that influence the stridulatory frequencies. Finally, our work can be used to understand the different stressors that currently threat freshwater animals. How other biotic components are affected by antropophony and how such stress affects the viability of the entire freshwater biota, remains to be uncovered.