Flexible use of visual and acoustic cues during roost finding in Spix’s disc-winged bat

Statistical analysis

Author

Miriam Gioiosa, Marcelo Araya-Salas, Cristian Castillo-Salazar, Silvia Chaves-Ramirez, Maurizio Gioiosa, Nazareth Rojas, Mariela Sanchez, Dino Scaravelli, Gloriana Chaverri

Published

February 26, 2023

Source code and data also found at https://github.com/maRce10/Roost-finding-behavior-in-Thyroptera-tricolor

Data analysis for the paper

Gioiosa, Miriam; Araya-Salas, Marcelo; Castillo Salazar, Cristian; Chaves-Ramírez, Silvia; Gioiosa, Maurizio; Rojas, Nazareth; Sanchez, Mariela; Scaravelli, Dino; Chaverri, Gloriana. 2023. Flexible use of visual and acoustic cues during roost finding in Spix’s disc-winged bat. Behavioral Ecology.

Abstract

The ability of an animal to detect environmental cues is crucial for its survival and fitness. In bats, sound certainly plays a significant role in the search for food, spatial navigation and social communication. Yet, the efficiency of bat’s echolocation could be limited by atmospheric attenuation and background clutter. In this context, sound can be complemented by other sensory modalities, like smell or vision. Spix’s disc-winged bat uses acoustic cues from other group members to locate the roost (tubular unfurled leaves of plants in the order Zingiberales). Our research focused on how individuals find a roost that has not been yet occupied, considering the urge to find a suitable leaf approximately every day, during nighttime or in daylight. We observed the process of roost finding in T. tricolor in a flight cage, manipulating the audio/visual sensory input available for each trial. A broadband noise was broadcast in order to mask echolocation, while experiments conducted at night reduced significantly the amount of light. We measured the time needed to locate the roost under these different conditions. Results show that with limited visual and acoustic cues, search time increases significantly. In contrast bats seemed capable of using acoustic and visual cues in a similarly efficient manner, since roost search times were equally short when bats could use only sound, only vision, or both senses at the same time. Our results show that non-acoustic inputs can still be an important source of information for finding critical resources in bats.

Load packages

First install/load packages:

Code

source("https://raw.githubusercontent.com/maRce10/sketchy/main/R/load_packages.R")

load_packages(c("readxl", "brms", "viridis", "ggplot2", "pbapply",
    "cowplot", "kableExtra", "warbleR", github = "maRce10/brmsish"))

Read and prepare data

The data file can be downloaded at the paper’s dryad repository or at the github repository (link)

Code

total_dat <- read.csv("Supp_Mat_1-time_entering_roost_thyroptera.csv")

total_dat$sensory_input <- factor(total_dat$sensory_input, levels = c("Sound & vision",
    "Noise control", "Sound", "Vision", "Lessen input"))

Sensory input treatments:

Sound & vision: diurnal experiments with no noise playback
Noise control: diurnal experiments with control (no echolocation-masking) playback
Sound: nocturnal experiments with no playback
Vision: diurnal experiments with echolocation-masking playback
Lessen input: nocturnal experiments with echolocation-masking playback

1 Sample sizes

First, exclude individuals with 1 experiment:

Code

tab <- table(total_dat$individual[!duplicated(paste(total_dat$sensory_input,
    total_dat$individual))])

dat <- total_dat[!total_dat$individual %in% names(tab)[tab == 1],
    ]

Experiments per date:

Code

# Dates
table(dat$date)


2019-11-06 2019-11-09 2019-11-10 2019-11-11 2019-11-12 2019-11-13 2019-11-14 
         4          3          4          3          4          7          4 
2019-11-16 2019-11-17 2019-11-19 2019-11-25 2019-11-26 2019-11-27 2019-11-28 
         3          7          6          3          6          9          4 
2019-11-29 2019-12-01 2019-12-07 2020-01-24 2020-01-28 2020-01-29 
         6          3          2          4          4         10

Test per individual:

Code

table(dat$individual)


982.126051278486 982.126051278498 982.126051278508 982.126051278509 
               3                7                2                4 
982.126051278515 982.126051278518 982.126051278520 982.126051278530 
               3                3                4                3 
982.126051278531 982.126051278554 982.126051278558 982.126051278559 
               6                3                3                3 
982.126051278588 982.126057845184 982.126057845188 982.126057845230 
               3                4                3                3 
982.126058484283 982.126058484309 982.126058484311 982.126058484344 
               3                3                3                6 
982.126058484345 982.126058484348  982126052945834  982126052945878 
               3                3                2                2 
 982126052945887  982126052945892  982126052945893  982126052945904 
               2                2                2                2 
 982126052945905  982126058484333  982126058484340 
               2                2                2

Number of individuals by number of tests:

Code

aggregate(individual ~ sensory_input, dat, function(x) length(unique(x)))

sensory_input	individual
Sound & vision	22
Noise control	21
Sound	9
Vision	22
Lessen input	9

Experiments per individual:

Code

table(dat$individual[!duplicated(paste(dat$sensory_input, dat$individual))])


982.126051278486 982.126051278498 982.126051278508 982.126051278509 
               3                3                2                3 
982.126051278515 982.126051278518 982.126051278520 982.126051278530 
               3                3                3                3 
982.126051278531 982.126051278554 982.126051278558 982.126051278559 
               3                3                3                3 
982.126051278588 982.126057845184 982.126057845188 982.126057845230 
               3                3                3                3 
982.126058484283 982.126058484309 982.126058484311 982.126058484344 
               3                3                3                3 
982.126058484345 982.126058484348  982126052945834  982126052945878 
               3                3                2                2 
 982126052945887  982126052945892  982126052945893  982126052945904 
               2                2                2                2 
 982126052945905  982126058484333  982126058484340 
               2                2                2

Number of individuals by number of experiments:

Code

table(table(dat$individual[!duplicated(paste(dat$sensory_input, dat$individual))]))


 2  3 
10 21

Tests per treatment:

Code

table(dat$sensory_input, dat$time_of_the_day)

                
                 day night
  Sound & vision  24     0
  Noise control   30     0
  Sound            0     9
  Vision          24     0
  Lessen input     0     9

Experiments per treatment:

Code

table(dat$sensory_input[!duplicated(paste(dat$sensory_input, dat$individual))],
    dat$time_of_the_day[!duplicated(paste(dat$sensory_input, dat$individual))])

                
                 day night
  Sound & vision  22     0
  Noise control   21     0
  Sound            0     9
  Vision          22     0
  Lessen input     0     9

31 individuals were tested
The 2 individuals with only 1 treatment were excluded so the final individual sample size was 31
The mean number of tests per individual (after excluding those with 1 treatment) was 3.1 (range = 2, 7)
The mean number of experimental treatments in which each individual was tested (after excluding those with 1 treatment) was 2.68

2 Effect of sensory input on the time to find the roost

Bayesian generalized linear models on time (in s) to enter the roost, with individual as a random effect and sensory input treatment as predictors. An intercept-only (null) model was also included in the analysis:

Sensory input as a categorical variable: \[Time\ to\ enter\ roost \sim + categorical\ input + (1 | individual)\]
Null model with no predictor: \[Time\ to\ enter\ roost \sim 1 + (1 | individual)\]

A loop is used to run these models, 4 chains per model, 5000 iterations each. Models were run on the complete data set and on a subset including only trials in which individuals entered the roost.

2.1 Models using all the data

Raw data plot

Code

cols <- viridis(10)

agg_dat <- aggregate(time_to_enter ~ sensory_input, dat, mean)
agg_dat$n <- sapply(1:nrow(agg_dat), function(x) length(unique(dat$individual[dat$sensory_input == agg_dat$sensory_input[x]]))) 
agg_dat$n.labels <- paste("n =", agg_dat$n)
agg_dat$sensory_input <- factor(agg_dat$sensory_input)

# raincoud plot:
fill_color <- adjustcolor("#e85307", 0.6)

ggplot(dat, aes(y = time_to_enter, x = sensory_input)) +
  # add half-violin from {ggdist} package
  ggdist::stat_halfeye(
    fill = fill_color,
    alpha = 0.5,
    # custom bandwidth
    adjust = .5,
    # adjust height
    width = .6,
    .width = 0,
    # move geom to the cright
    justification = -.2,
    point_colour = NA
  ) +
  geom_boxplot(fill = fill_color,
    width = .15,
    # remove outliers
    outlier.shape = NA # `outlier.shape = NA` works as well
  ) +
  # add justified jitter from the {gghalves} package
  gghalves::geom_half_point(
    color = fill_color,
    # draw jitter on the left
    side = "l",
    # control range of jitter
    range_scale = .4,
    # add some transparency
    alpha = .5,
    transformation = ggplot2::position_jitter(height = 0)  
  ) +
   ylim(c(-30, 310)) +
  geom_text(data = agg_dat, aes(y = rep(-25, 5), x = sensory_input, label = n.labels), nudge_x = -0.13, size = 6) + 
   scale_x_discrete(labels=c("Control" = "Noise control", "Sound vision" = "Sound & vision", "Vision" = "Vision", "Lessen input" = "Lessen input")) +
  labs(x = "Sensory input       ", y = "Time to enter roost (s)") + theme(axis.text.x = element_text(angle = 15, hjust = 1))

Code

# ggsave(filename = "./output/time_to_enter_by_treatment.tiff", dpi = 300, width = 3000, height = 1500, units = "px")

Code

model_formulas <- c("time_to_enter ~ sensory_input + (1 | individual)",
    "time_to_enter ~ 1 + (1 | individual)")

iter <- 5000
chains <- 4

priors <- c(set_prior("student_t(10,0,1)", class = "sigma"), set_prior("student_t(10,0,1)",
    class = "sd"))


# Run loops with models
brms_models <- lapply(model_formulas, function(x) {

    mod <- brm(formula = x, iter = iter, thin = 1, data = dat, family = lognormal(),
        silent = 2, chains = chains, cores = chains, prior = priors)

    mod <- add_criterion(mod, c("loo"), save_pars = save_pars(all = TRUE))

    return(mod)
})

names(brms_models) <- model_formulas

saveRDS(brms_models, "./data/processed/regresion_models_brms.RDS")

Code

brms_models <- readRDS("./data/processed/regresion_models_brms.RDS")

comp_mods <- loo_compare(brms_models[[1]], brms_models[[2]], model_names = names(brms_models))

df1 <- kbl(comp_mods, row.names = TRUE, escape = FALSE, format = "html",
    digits = 3)

cat("Compare models:")

Compare models:

Code

kable_styling(df1, bootstrap_options = c("striped", "hover", "condensed",
    "responsive"), full_width = FALSE, font_size = 12)

	elpd_diff	se_diff	elpd_loo	se_elpd_loo	p_loo	se_p_loo	looic	se_looic
time_to_enter ~ sensory_input + (1 \| individual)	0.000	0.000	-510.740	16.182	20.972	1.890	1021.481	32.364
time_to_enter ~ 1 + (1 \| individual)	-7.225	3.742	-517.965	16.101	19.136	1.764	1035.930	32.202

Code

cat("Best model:\n")

Best model:

Code

cat(paste("-  ", rownames(comp_mods)[1], "\n"))

time_to_enter ~ sensory_input + (1 | individual)

Code

# best model
if (!grepl("1 +", rownames(comp_mods)[1], fixed = TRUE)) extended_summary(fit = brms_models[[rownames(comp_mods)[1]]],
    gsub.pattern = "sensory_input", gsub.replacement = "", fill = "#e85307",
    trace.palette = terrain.colors, print.name = FALSE)

	priors	formula	iterations	chains	thinning	warmup	diverg_transitions	rhats > 1.05	min_bulk_ESS	min_tail_ESS	seed
1	Intercept-student_t(3, 3.5, 2.5) sd-student_t(10,0,1) sigma-student_t(10,0,1)	time_to_enter ~ sensory_input + (1 \| individual)	5000	4	1	2500	0	0	7632.569	7283.589	652773362

	Estimate	l-95% CI	u-95% CI	Rhat	Bulk_ESS	Tail_ESS
b_Intercept	3.228	2.643	3.827	1	7632.569	7360.967
b_Noisecontrol	0.208	-0.424	0.854	1	13237.365	7884.717
b_Sound	0.476	-0.647	1.636	1	8482.597	7968.398
b_Vision	0.604	-0.075	1.269	1	13132.351	7283.589
b_Lesseninput	2.407	1.284	3.534	1.001	8279.037	7899.409

Compare all contrasts

Code

# contrasts
contrasts(fit = brms_models[[rownames(comp_mods)[1]]], predictor = "sensory_input",
    n.posterior = 2000, level.sep = " VS ", fill = "#e85307", gsub.pattern = c("Lesseninput",
        "Soundvision", "Noisecontrol"), gsub.replacement = c("Lessen input",
        "Sound & vision", "Noise control"), html.table = TRUE, plot = TRUE,
    sort.levels = c("Lessen input", "Vision", "Sound", "Sound & vision",
        "Noise control"))

	Hypothesis	Estimate	Est.Error	l-95% CI	u-95% CI
1	Lessen input VS Vision	1.803	0.571	0.669	2.918
2	Lessen input VS Sound	1.931	0.570	0.811	3.062
3	Lessen input VS Sound & vision	2.407	0.572	1.284	3.534
4	Lessen input VS Noise control	2.198	0.561	1.079	3.262
5	Vision VS Sound	0.128	0.577	-1.022	1.257
6	Vision VS Sound & vision	0.604	0.344	-0.075	1.269
7	Vision VS Noise control	0.395	0.331	-0.245	1.044
8	Sound VS Sound & vision	0.476	0.574	-0.647	1.636
9	Sound VS Noise control	0.267	0.568	-0.859	1.405
10	Sound & vision VS Noise control	0.208	0.325	-0.424	0.854

Code

ggsave(filename = "./output/contrasts_between_treatments.tiff", dpi = 300,
    width = 2000, height = 1000, units = "px")

2.1.1 Takeaways

“lessen input” was the only sensory input treatment in which a increase in time was detected: it show a longer time to enter the roost than all other treatments

2.2 Models on the data excluding bats that did not enter the roost

Raw data plot

Code

agg_dat <- aggregate(time_to_enter ~ sensory_input, dat[dat$time_to_enter < 300, ], mean)
agg_dat$n <- sapply(1:nrow(agg_dat), function(x) length(unique(dat$individual[dat$sensory_input == agg_dat$sensory_input[x] & dat$time_to_enter < 300]))) 
agg_dat$labels <- c("a", "a", "a", "a", "b")
agg_dat$n.labels <- paste("n =", agg_dat$n)
agg_dat$sensory_input <- factor(agg_dat$sensory_input)

# raincoud plot:

ggplot(dat[dat$time_to_enter < 300, ], aes(y = time_to_enter, x = sensory_input)) +
  # add half-violin from {ggdist} package
  ggdist::stat_halfeye(
    fill = fill_color,
    alpha = 0.5,
    # custom bandwidth
    adjust = .5,
    # adjust height
    width = .6,
    .width = 0,
    # move geom to the cright
    justification = -.2,
    point_colour = NA
  ) +
  geom_boxplot(fill = fill_color,
    width = .15,
    # remove outliers
    outlier.shape = NA # `outlier.shape = NA` works as well
  ) +
  # add justified jitter from the {gghalves} package
  gghalves::geom_half_point(
    color = fill_color,
    # draw jitter on the left
    side = "l",
    # control range of jitter
    range_scale = .4,
    # add some transparency
    alpha = .5,
    transformation = ggplot2::position_jitter(height = 0)
  ) +
   ylim(c(-30, 310)) +
  # geom_text(data = agg_dat, aes(y = rep(340, 5), x = sensory_input, label = labels), size = 7) +
  geom_text(data = agg_dat, aes(y = rep(-25, 5), x = sensory_input, label = n.labels), nudge_x = -0.13, size = 6) + 
   scale_x_discrete(labels=c("Control" = "Noise control", "Sound vision" = "Sound & vision", "Vision" = "Vision", "Lessen input" = "Lessen input")) +
  labs(x = "Sensory input       ", y = "Time to enter roost (s)") + theme(axis.text.x = element_text(angle = 15, hjust = 1))

Code

model_formulas <- c("time_to_enter ~ sensory_input + (1 | individual)",
    "time_to_enter ~ 1 + (1 | individual)")

iter <- 5000
chains <- 4
priors <- c(set_prior("student_t(10,0,1)", class = "sigma"), set_prior("student_t(10,0,1)",
    class = "sd"))


# Run loops with models
brms_models <- lapply(model_formulas, function(x) {

    mod <- brm(formula = x, iter = iter, thin = 1, data = dat[dat$time_to_enter <
        300, ], family = lognormal(), silent = 2, chains = chains,
        cores = chains, prior = priors, control = list(adapt_delta = 0.99,
            max_treedepth = 15))

    mod <- add_criterion(mod, c("loo"), save_pars = save_pars(all = TRUE))

    return(mod)
})

names(brms_models) <- model_formulas

saveRDS(brms_models, "./data/processed/regression_models_brms_subset.RDS")

Code

brms_models <- readRDS("./data/processed/regression_models_brms_subset.RDS")

comp_mods <- loo_compare(brms_models[[1]], brms_models[[2]], model_names = names(brms_models))


df1 <- kbl(comp_mods, row.names = TRUE, escape = FALSE, format = "html",
    digits = 3)

cat("Compare models:")

Compare models:

Code

kable_styling(df1, bootstrap_options = c("striped", "hover", "condensed",
    "responsive"), full_width = FALSE, font_size = 12)

	elpd_diff	se_diff	elpd_loo	se_elpd_loo	p_loo	se_p_loo	looic	se_looic
time_to_enter ~ sensory_input + (1 \| individual)	0.000	0.000	-398.054	14.044	18.956	1.898	796.107	28.087
time_to_enter ~ 1 + (1 \| individual)	-2.318	2.771	-400.372	14.170	15.192	1.660	800.744	28.340

Code

cat("Best model:\n")

Best model:

Code

cat(paste("-  ", rownames(comp_mods)[1], "\n"))

time_to_enter ~ sensory_input + (1 | individual)

Code

# best model
if (!grepl("1 +", rownames(comp_mods)[1], fixed = TRUE)) extended_summary(fit = brms_models[[rownames(comp_mods)[1]]],
    gsub.pattern = "sensory_input", gsub.replacement = "", fill = "#e85307",
    trace.palette = terrain.colors, print.name = FALSE)

	priors	formula	iterations	chains	thinning	warmup	diverg_transitions	rhats > 1.05	min_bulk_ESS	min_tail_ESS	seed
1	Intercept-student_t(3, 3.1, 2.5) sd-student_t(10,0,1) sigma-student_t(10,0,1)	time_to_enter ~ sensory_input + (1 \| individual)	5000	4	1	2500	0	0	5171.14	6488.326	1582467647

	Estimate	l-95% CI	u-95% CI	Rhat	Bulk_ESS	Tail_ESS
b_Intercept	3.018	2.417	3.634	1	5171.140	6488.326
b_Noisecontrol	0.124	-0.554	0.803	1	8844.558	7129.572
b_Sound	0.693	-0.388	1.814	1	6254.619	6711.689
b_Vision	0.518	-0.186	1.225	1	9433.443	7726.126
b_Lesseninput	2.948	0.985	4.924	1	7428.895	7424.103

Compare all contrasts

Code

# contrasts
contrasts(fit = brms_models[[rownames(comp_mods)[1]]], predictor = "sensory_input",
    n.posterior = 2000, level.sep = " VS ", fill = "#e85307", gsub.pattern = c("Lesseninput",
        "Soundvision", "Noisecontrol"), gsub.replacement = c("Lessen input",
        "Sound & vision", "Noise control"), html.table = TRUE, plot = TRUE,
    sort.levels = c("Lessen input", "Vision", "Sound", "Sound & vision",
        "Noise control"))

	Hypothesis	Estimate	Est.Error	l-95% CI	u-95% CI
1	Lessen input VS Vision	2.429	1.008	0.472	4.434
2	Lessen input VS Sound	2.255	1.008	0.256	4.221
3	Lessen input VS Sound & vision	2.948	1.000	0.985	4.924
4	Lessen input VS Noise control	2.824	0.992	0.906	4.778
5	Vision VS Sound	-0.175	0.555	-1.277	0.922
6	Vision VS Sound & vision	0.518	0.356	-0.186	1.225
7	Vision VS Noise control	0.394	0.345	-0.285	1.081
8	Sound VS Sound & vision	0.693	0.560	-0.388	1.814
9	Sound VS Noise control	0.569	0.552	-0.516	1.656
10	Sound & vision VS Noise control	0.124	0.343	-0.554	0.803

2.2.1 Takeaways

Results remain qualitatively similar after excluding individuals that did not enter the roost

3 Treatment diagram

Code

reverse.viridis <- function(...) viridis(..., direction = -1)

wv <- read_wave("https://github.com/maRce10/Roost-finding-behavior-in-Thyroptera-tricolor/raw/main/data/raw/thyroptera_echolocation_clip.wav")


wv <- ffilter(wv, from = 40000, to = 170000, bandpass = TRUE, output = "Wave")

wv1 <- cutw(wv, from = 0.005, to = 0.009, output = "Wave")
wv2 <- cutw(wv, from = 0.036, to = duration(wv) - 0.004, output = "Wave")

subwv <- pastew(wv1, wv2, output = "Wave")

ns <- seewave::noisew(f = subwv@samp.rate, output = "Wave", d = duration(subwv))


mask <- ffilter(ns, from = 0, to = 45000, bandpass = FALSE, output = "Wave")

no_mask <- ffilter(ns, from = 0, to = 45000, bandpass = TRUE, output = "Wave")

subwv@left <- subwv@left[1:length(mask)]

ovlp <- 99
cex <- 0.8  # try cex 2 for saving plot
res <- 200
bl <- 4
hgh <- wdh <- 480 * 4

lf <- c(0, 2/14, 6/14, 10/14, 0, 1/6, 1/2, 5/6)
rgh <- c(2/14, 6/14, 10/14, 1, 1/6, 1/2, 5/6, 1)
btm <- c(1/2, 1/2, 1/2, 1/2, 0, 0, 0, 0)
tp <- c(1, 1, 1, 1, 1/2, 1/2, 1/2, 1/2)
m <- cbind(lf, rgh, btm, tp)

# graphics.off()
invisible(close.screen(all.screens = TRUE))

# tiff(filename = './output/diagram_experimental_design.tiff',
# res = 300, width = 3500, height = 2000)

sc <- split.screen(figs = m)

# sound and vision
screen(2)

par(mar = c(3, 6, 1, 1))
warbleR:::spectro_wrblr_int2(subwv, grid = FALSE, collev.min = -35,
    wl = 120, palette = reverse.viridis, ovlp = ovlp, zp = 1000, tlim = c(0.001,
        0.0085), axisX = FALSE, tlab = NULL, axisY = FALSE, flab = NULL,
    flim = c(0, 220))

box(lwd = bl)

axis(side = 2, cex.axis = cex, labels = c(0, 100, 200), at = c(0,
    100, 200))

mtext(side = 2, text = "Frequency (kHz)", line = 4, cex = cex)

mtext(side = 1, text = "Sound & vision  ", line = 1.5, cex = cex)


# vision
screen(3)
par(mar = c(3, 6, 1, 1))
warbleR:::spectro_wrblr_int2(subwv + mask * 1000, grid = FALSE, collev.min = -35,
    wl = 120, palette = reverse.viridis, ovlp = ovlp, zp = 1000, tlim = c(0.001,
        0.0085), axisX = FALSE, tlab = NULL, axisY = FALSE, flab = NULL,
    flim = c(0, 220))

box(lwd = bl)

axis(side = 2, cex.axis = cex, labels = c(0, 100, 200), at = c(0,
    100, 200))

mtext(side = 1, text = "Vision", line = 1.5, cex = cex)

# Noise control

screen(4)

par(mar = c(3, 6, 1, 1), new = TRUE)
warbleR:::spectro_wrblr_int2(subwv + no_mask * 1000, grid = FALSE,
    collev.min = -35, wl = 120, palette = reverse.viridis, ovlp = ovlp,
    zp = 1000, tlim = c(0.001, 0.0085), axisX = FALSE, tlab = NULL,
    axisY = FALSE, flab = NULL, flim = c(0, 220))

box(lwd = bl)

axis(side = 2, cex.axis = cex, labels = c(0, 100, 200), at = c(0,
    100, 200))

mtext(side = 1, text = "Noise control", line = 1.5, cex = cex)


# sound
par(bg = "black", new = TRUE)
screen(6)

par(mar = c(3, 8.5, 1, 1))
warbleR:::spectro_wrblr_int2(subwv, grid = FALSE, collev.min = -35,
    wl = 120, palette = reverse.viridis, ovlp = ovlp, zp = 1000, tlim = c(0.001,
        0.0085), axisX = FALSE, tlab = NULL, axisY = FALSE, flab = NULL,
    flim = c(0, 220))

box(lwd = bl, col = "white")

axis(side = 2, cex.axis = cex, labels = c(0, 100, 200), at = c(0,
    100, 200), col.ticks = "white", col = "white", col.axis = "white")

mtext(side = 1, text = "Sound", line = 1.5, cex = cex, col = "white")


# lessen input
par(bg = "black", new = TRUE)
screen(7)

par(mar = c(3, 8.5, 1, 1))

# par(mar = c(1, 9, 1, 1), bg = 'black', new = TRUE)
warbleR:::spectro_wrblr_int2(subwv + mask * 1000, grid = FALSE, collev.min = -35,
    wl = 120, palette = reverse.viridis, ovlp = ovlp, zp = 1000, tlim = c(0.001,
        0.0085), axisX = FALSE, tlab = NULL, axisY = FALSE, flab = NULL,
    flim = c(0, 220))

box(lwd = bl, col = "white")

axis(side = 2, cex.axis = cex, labels = c(0, 100, 200), at = c(0,
    100, 200), col.ticks = "white", col = "white", col.axis = "white")

mtext(side = 1, text = "Lessen input", line = 1.5, cex = cex, col = "white")
# dev.off()

par(bg = "black", new = TRUE)
screen(5)
par(mar = c(0, 0, 0, 0))
plot(1, frame.plot = FALSE, type = "n")
text(x = 1, y = 1.05, "Nighttime", srt = 90, cex = 1.2 * cex, col = "white")

par(bg = "black", new = TRUE)
screen(8)

par(bg = "white", new = TRUE)
screen(1)
par(mar = c(0, 0, 0, 0))
plot(1, frame.plot = FALSE, type = "n")
text(x = 1, y = 1.05, "Daytime", srt = 90, cex = 1.2 * cex)

Code

# dev.off()

Session information

R version 4.1.0 (2021-05-18)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 20.04.2 LTS

Matrix products: default
BLAS:   /usr/lib/x86_64-linux-gnu/atlas/libblas.so.3.10.3
LAPACK: /usr/lib/x86_64-linux-gnu/atlas/liblapack.so.3.10.3

locale:
 [1] LC_CTYPE=pt_BR.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=es_CR.UTF-8        LC_COLLATE=pt_BR.UTF-8    
 [5] LC_MONETARY=es_CR.UTF-8    LC_MESSAGES=pt_BR.UTF-8   
 [7] LC_PAPER=es_CR.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=es_CR.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] brmsish_1.0.0      warbleR_1.1.28     NatureSounds_1.0.4 knitr_1.42        
 [5] seewave_2.2.0      tuneR_1.4.1        kableExtra_1.3.4   cowplot_1.1.1     
 [9] pbapply_1.6-0      ggplot2_3.4.0      viridis_0.6.2      viridisLite_0.4.1 
[13] brms_2.18.0        Rcpp_1.0.9         readxl_1.3.1      

loaded via a namespace (and not attached):
  [1] backports_1.4.1      systemfonts_1.0.4    plyr_1.8.7          
  [4] igraph_1.3.5         splines_4.1.0        crosstalk_1.2.0     
  [7] TH.data_1.1-0        rstantools_2.2.0     inline_0.3.19       
 [10] digest_0.6.31        htmltools_0.5.4      fansi_1.0.3         
 [13] magrittr_2.0.3       checkmate_2.1.0      RcppParallel_5.1.5  
 [16] matrixStats_0.62.0   xts_0.12.2           sandwich_3.0-1      
 [19] svglite_2.1.0        prettyunits_1.1.1    colorspace_2.0-3    
 [22] signal_0.7-7         rvest_1.0.3          ggdist_3.2.0        
 [25] textshaping_0.3.5    xfun_0.36            dplyr_1.0.10        
 [28] callr_3.7.3          crayon_1.5.2         RCurl_1.98-1.9      
 [31] jsonlite_1.8.4       lme4_1.1-27.1        ape_5.6-2           
 [34] survival_3.2-11      zoo_1.8-11           glue_1.6.2          
 [37] gtable_0.3.1         emmeans_1.8.1-1      webshot_0.5.4       
 [40] distributional_0.3.1 pkgbuild_1.4.0       rstan_2.21.7        
 [43] abind_1.4-5          scales_1.2.1         mvtnorm_1.1-3       
 [46] DBI_1.1.1            miniUI_0.1.1.1       dtw_1.23-1          
 [49] xtable_1.8-4         diffobj_0.3.4        proxy_0.4-27        
 [52] stats4_4.1.0         StanHeaders_2.21.0-7 DT_0.26             
 [55] htmlwidgets_1.5.4    httr_1.4.4           threejs_0.3.3       
 [58] posterior_1.3.1      ellipsis_0.3.2       pkgconfig_2.0.3     
 [61] loo_2.4.1.9000       farver_2.1.1         utf8_1.2.2          
 [64] labeling_0.4.2       tidyselect_1.2.0     rlang_1.0.6         
 [67] reshape2_1.4.4       later_1.3.0          munsell_0.5.0       
 [70] cellranger_1.1.0     tools_4.1.0          cli_3.6.0           
 [73] generics_0.1.3       ggridges_0.5.4       evaluate_0.20       
 [76] stringr_1.5.0        fastmap_1.1.0        ragg_1.1.3          
 [79] yaml_2.3.7           processx_3.8.0       nlme_3.1-152        
 [82] mime_0.12            projpred_2.0.2       formatR_1.11        
 [85] xml2_1.3.3           compiler_4.1.0       bayesplot_1.9.0     
 [88] shinythemes_1.2.0    rstudioapi_0.14      gamm4_0.2-6         
 [91] tibble_3.1.8         stringi_1.7.12       highr_0.10          
 [94] ps_1.7.2             Brobdingnag_1.2-9    lattice_0.20-44     
 [97] Matrix_1.5-1         nloptr_1.2.2.2       markdown_1.3        
[100] shinyjs_2.1.0        fftw_1.0-7           tensorA_0.36.2      
[103] vctrs_0.5.2          pillar_1.8.1         lifecycle_1.0.3     
[106] bridgesampling_1.1-2 estimability_1.4.1   bitops_1.0-7        
[109] httpuv_1.6.6         R6_2.5.1             promises_1.2.0.1    
[112] gridExtra_2.3        codetools_0.2-18     boot_1.3-28         
[115] colourpicker_1.2.0   MASS_7.3-54          gtools_3.9.3        
[118] assertthat_0.2.1     rjson_0.2.21         rprojroot_2.0.3     
[121] withr_2.5.0          shinystan_2.6.0      multcomp_1.4-17     
[124] mgcv_1.8-36          parallel_4.1.0       gghalves_0.1.3      
[127] grid_4.1.0           coda_0.19-4          minqa_1.2.4         
[130] rmarkdown_2.20       shiny_1.7.3          base64enc_0.1-3     
[133] dygraphs_1.1.1.6