| jupytext |
|
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| kernelspec |
|
+++
+++
Analysis and visualization was done using R and various packages. The following is the script used to generate 2 scatterplot graphs.
library(tidyverse)
library(lubridate)
library(ggplot2)
library(plotly)
options(repr.plot.width=10, repr.plot.height=6)
+++
Temperature data {cite}Dewees2021 are found in the "data" folder, while coordinates (and the time recorded) are in the "data/nav" folder.
All of the temperature data was cleaned using a Python script, located in the data_cleaning folder. No packages were used, and can be used as long as a v3.9 Python is installed (anything above or below is untested) and the scripts are pointed to the right data sources.
data/temp_processed_summarized.csv: mean temp over mindata/nav/nav_processed.csv: mean GPS over mindata/nav_temp_joined_processed.csv: joined mean temp & GPS over min
+++
The dates are all of type character, meaning extracting any use without it being a proper date type is hard. Therefore, time and date must be formatted.
format_datetime <- function(df) {
df_new <- df %>%
# https://www.neonscience.org/resources/learning-hub/tutorials/dc-time-series-subset-dplyr-r
mutate(date = as.Date(date, format = '%m/%d/%Y')) %>%
# https://www.tidyverse.org/blog/2021/03/clock-0-1-0/
mutate(datetime = as.POSIXct(date, "America/Vancouver")) %>%
mutate(datetime = datetime +hour(time)+ minute(time))
return(df_new)
}
The following are all of the functions needed to clean up 1 temperature file. However, we have quite a few files, and trying to clean and instantiate each by hand is cumbersome. Therefore, we will iterate through all of the files and summarize.
Please keep in mind that the following code blocks will take pretty long to run.
clean_SBE45_data <- function(x) {
read <- read_delim(x, delim = ",",
col_names = c("date",
"time",
"int_temp",
"conductivity",
"salinity",
"sound_vel",
"ext_temp"),
col_types = cols(
date = col_character(),
time = col_time(format = ""),
int_temp = col_double(),
conductivity = col_double(),
salinity = col_double(),
sound_vel = col_double(),
ext_temp = col_double()
)) %>%
select(date, time, ext_temp)
return(read)
}
clean_STT_TSG_data <- function(x) {
read <- read_delim(x, delim = ",",
col_names = c("date",
"time",
"type",
"diff",
"ext_temp",
"int_temp",
"extra")
) %>%
select(date, time)
return(read)
}
clean_temp_data <- function(x) {
# https://stackoverflow.com/questions/10128617/test-if-characters-are-in-a-string
if(grepl("SBE45-TSG-MSG", x, fixed = TRUE)) {
return(clean_SBE45_data(x))
} else {
return(clean_STT_TSG_data(x))
}
}
This is the data before it is summarized by the minute.
# https://stackoverflow.com/questions/11433432/how-to-import-multiple-csv-files-at-once
all_temperature_loaded <- list.files(path = "data/",
pattern = "*.Raw",
full.names = T) %>%
map_df(~clean_temp_data(.))
head(all_temperature_loaded)
summary(all_temperature_loaded)
all_temperature_cleaned <- all_temperature_loaded %>%
filter(!is.na(ext_temp)) %>%
filter(ext_temp > 2) %>%
format_datetime()
all_temperature <- group_by(all_temperature_cleaned, datetime) %>%
summarize(mean_ext = mean(ext_temp, na.rm = TRUE))
write_csv(all_temperature, "data/temp_processed_summarized.csv")
head(all_temperature)
summary(all_temperature)
All of the navigation files will also be cleaned up and summarized in similar manner to the temperature files.
clean_nav_data <- function(x) {
read <- read_csv(x,
col_names = c(
"date",
"time",
"type",
"time_num",
"lat",
"lat_NS",
"long",
"long_WE",
"gps_quality",
"num_sat_view",
"hort_dil",
"ant_alt",
"ant_alt_unit",
"geoidal",
"geoidal_unit",
"age_diff",
"diff_station"
)
) %>%
select(date, time, lat, long)
return(read)
}
Below is the raw ocean navigation data.
all_nav_loaded <- list.files(path = "data/nav/",
pattern = "*.Raw",
full.names = T) %>%
map_df(~clean_nav_data(.)) %>%
format_datetime()
summary(all_nav_loaded)
Since the latitude and longitude are in the degree minutes format, they must be converted.
all_nav <- all_nav_loaded %>%
group_by(datetime) %>%
summarize(mean_lat = mean(lat), mean_long = mean(long)) %>%
mutate(mean_lat = mean_lat/100, mean_long = mean_long/100) %>%
mutate(deg_lat_int = trunc(mean_lat, 0),
deg_long_int = trunc(mean_long, 0)) %>%
mutate(deg_lat_dec = round((mean_lat - deg_lat_int) * 10000),
deg_long_dec = round((mean_long - deg_long_int) * 10000)) %>%
mutate(mean_deg_lat = deg_lat_int + deg_lat_dec/(60 * 100),
mean_deg_long = deg_long_int + deg_long_dec / (60 * 100)) %>%
select(-deg_lat_int, -deg_long_int, -deg_lat_dec, -deg_long_dec, -mean_lat, -mean_long)
write_csv(all_nav, "data/nav/nav_processed.csv")
head(all_nav)
summary(all_nav)
Since we have the date and time (by the minute) of both the temperature and it's coordinates, we can match the two columns.
joined_temp_nav <- inner_join(all_temperature,
all_nav,
by = c("datetime" = "datetime"))
write_csv(joined_temp_nav, "data/nav_temp_joined_processed.csv")
head(joined_temp_nav)
summary(joined_temp_nav)
+++
Time and mean temperature plotted on a scatterplot to see temperature changes over time. Notice that since the ship moves in 1 way vs time, the shape of the 2D scatterplot and the 3D plot is very similar.
time_plot <- ggplot(all_temperature, aes(x = datetime,
y = mean_ext,
colour = mean_ext)) +
geom_point() +
scale_colour_gradient(low = "blue", high = "red") +
labs(x = "Date and Time PST",
y = "Mean (min) Ocean Temperature (°C)",
colour = "Mean Ocean Temperature (°C)")
time_plot
p<- plot_ly(joined_temp_nav,
x = ~mean_deg_lat,
y = ~mean_deg_long,
z = ~mean_ext,
color = ~mean_ext
) %>%
add_markers(size = 0.7) %>%
colorbar(title = "Mean Ocean Temp (°C)")%>%
layout(title = "Mean Ocean Temperature and Coordinates",
scene = list(
xaxis = list(title = "Mean Latitude (°)"),
yaxis = list(title = "Mean Longitude (°)"),
zaxis = list(title = "Mean Ocean Temp (°C)")
)
)
embed_notebook(p)