pike_telemetry: Pike Telemetry

library(mgcv)
library(tidyverse)
library(lubridate)

dat <- readRDS("datf.rds")

dat <- dat %>% filter(!fish_id %in% c("NP-5", "NP-3", "NP-7", "NP-10", "NP-6"))

dat2 <- dat %>% filter(!is.nan(speed_mean))

dat$hour <- as.integer(dat$hour)
dat$fish_id <- as.factor(dat$fish_id)
dat$lightson_mean <- as.factor(dat$lightson_mean)
dat$month <- as.integer(dat$month)
dat$season <- as.factor(dat$season)

dat2 <- dat %>% filter(!is.nan(speed_mean))
dat2$julian_day <- yday(dat2$date)

hist(sqrt(dat$depth_raw_mean))


depth_mod <- bam(sqrt(depth_raw_mean) ~ s(hour, k = 24, bs = "cc") + season + lit_temp + lightson_mean +
                   + s(fish_id, k = 10, bs = "re"),
                 knots = list(hour = c(0, 23)), method = "REML", family = "gaussian", data = dat2)
summary(depth_mod)

plot(residuals(depth_mod))


hist(dat2$temp_mean)

temp_mod <- bam(sqrt(temp_mean) ~ s(hour, k = 24, bs = "cc") + season + lit_temp + lightson_mean +
                   + s(fish_id, k = 10, bs = "re"),
                 knots = list(hour = c(0, 23)), method = "REML", family = "gaussian", data = dat2)
summary(temp_mod)

hist(sqrt(dat2$speed_mean))

activity_mod <- bam(sqrt(speed_mean) ~ s(hour, k = 24, bs = "cc") + s(julian_day, k = 365, bs = "cc") + lightson_mean +
                      + s(fish_id, k = 10, bs = "re"),
                    knots = list(hour = c(0, 23), julian_day = c(1, 365)), method = "REML", family = "gaussian", data = dat2)

summary(activity_mod)


hist(sqrt(dat2$dis_2_shore_mean))

distshore_mod <- bam(sqrt(dis_2_shore_mean) ~ s(hour, k = 24, bs = "cc") + season + lit_temp + lightson_mean +
                      + s(fish_id, k = 10, bs = "re"),
                    knots = list(hour = c(0, 23)), method = "REML", family = "gaussian", data = dat2)
summary(distshore_mod)









### Tim's figures

# there are gaps in the data, apparently corresponding to when gear was offline
dat2 %>%
  ggplot(aes(x = date, y = dab_mean)) + 
  geom_point() +
  theme_bw()

# some individuals data is patchy or short, can we use all individuals
dat2 %>%
  ggplot(aes(x = date, y = dab_mean)) + 
  geom_point() +
  facet_wrap(~fish_id) +
  theme_bw()

dat2 %>%
  ggplot(aes(x = date, y = dis_2_shore_mean)) + 
  geom_point() +
  facet_wrap(~fish_id) +
  theme_bw()

dat2 %>%
  ggplot(aes(x = date, y = speed_mean)) + 
  geom_point() +
  facet_wrap(~fish_id) +
  theme_bw()

dat2 %>%
  ggplot(aes(x = date, y = bathy_mean)) + 
  geom_point() +
  facet_wrap(~fish_id) +
  theme_bw()

dat2 %>%
  ggplot(aes(x = date, y = temp_mean)) + 
  geom_point() +
  facet_wrap(~fish_id) +
  theme_bw()

dat2 %>%
  ggplot(aes(x = date, y = depth_raw_mean)) + 
  geom_point() +
  facet_wrap(~fish_id) +
  theme_bw()

# pike temperature corresponds pretty well to littoral water temp, but they are clearly deeper at some points
dat2 %>%
  ggplot() + 
  geom_point(aes(x = date, y = temp_mean), color = "grey") +
  facet_wrap(~fish_id) +
  geom_point(aes(x = date, y = lit_temp), color = "black") +
  theme_bw()


dat2 %>%
  ggplot() + 
  geom_point(aes(x = date, y = depth_raw_mean), color = "black") +
  scale_y_reverse()+
  facet_wrap(~fish_id) 





# scales data for each fish individually

ld <- split(dat2, dat2$fish_id)

new <- NULL

for(i in 1:length(ld)){
  df <- ld[[i]]
  df$act_scale <- scale(df$speed_mean)
  new <- rbind(new, df)
}

# plot of scaled data

new %>% group_by(fish_id, month, hour) %>%
  summarise(mean_act = median(act_scale)) %>%
  ggplot(aes(x = hour, y = mean_act, colour = fish_id)) + geom_point() + facet_wrap(~month)


new %>% group_by(fish_id, month, hour) %>%
  summarise(mean_act = median(speed_mean)) %>%
  ggplot(aes(x = hour, y = mean_act, colour = fish_id)) + geom_point() + 
  geom_smooth() + facet_wrap(~month)

new %>% group_by(fish_id, month, hour) %>%
  summarise(mean_depth = median(depth_raw_mean)) %>%
  ggplot(aes(x = hour, y = mean_depth, colour = fish_id)) + geom_point() + 
  geom_smooth() + facet_wrap(~month)

new %>% group_by(fish_id, month, hour) %>%
  summarise(mean_temp = median(temp_mean)) %>%
  ggplot(aes(x = hour, y = mean_temp, colour = fish_id)) + geom_point() + 
  geom_smooth() + facet_wrap(~month)

new %>% group_by(fish_id, month, hour) %>%
  summarise(mean_dist2shore = median(dis_2_shore_mean)) %>%
  ggplot(aes(x = hour, y = mean_dist2shore, colour = fish_id)) + geom_point() + 
  geom_smooth() + facet_wrap(~month)