castr: castr: An R package to process CTD casts.
In jiho/castr: Process CTD casts

This package contains functions to process data collected during CTD casts:

smooth(), despike(), or integrate() a variable;
compute the depth of clines with clined(), of the deep chlorophyll maximum with maxd(), of the bottom of the mixed layer with mld();
compute a stratification index with stratif()

It also provides utilities to apply a function in a moving window along the cast (slide()), hence enabling the computation of moving averages (to smooth() the data), medians (to detect spikes()), etc. Finally, two small datasets are provided to test the functions: one with a single cast (d) and another one with 41 casts (ctd).

All functions in castr work on vectors and make few assumptions regarding your data. Therefore, they fit in the tidyverse philosophy and work well with dplyr verbs, as shown in the example below.

## Not run: 
# We have a data.frame containing a time series of 41 CTD casts collected
# at the same station over two years, identified by the date of sampling
head(ctd)

# Display the data
library("ggplot2")
ggplot(ctd) + geom_path(aes(x=temp, y=-depth, group=date), alpha=0.6)
# there are a few spikes in temperature

# Despike each profile
library("dplyr")
ctd_clean <- ctd %>% group_by(date) %>%
  mutate(temp=despike(temp, mult=3))
ggplot(ctd_clean) + geom_path(aes(x=temp, y=-depth, group=date), alpha=0.6)
# the spikes are gone! Do the same for all variables
# NB: the despiking parameters should really be tested and adjusted
ctd_clean <- ctd %>% group_by(date) %>%
  mutate(
    temp=despike(temp, mult=3),
    sal=despike(sal, mult=3),
    chla=despike(chla, k=2, mult=4),
    # NB: chlorophyll is more spiky so be less stringent
    sigma=despike(sigma, mult=3)
  )
ggplot(ctd_clean) + geom_path(aes(x=sal, y=-depth, group=date), alpha=0.6)
ggplot(ctd_clean) + geom_path(aes(x=chla, y=-depth, group=date), alpha=0.6)

# Now compute summary statistics on each despiked profile
stats <- ctd_clean %>% group_by(date) %>%
  summarise(
    thermocline = clined(temp, depth, n.smooth=2, k=2),
    pycnocline = clined(sigma, depth),
    strat_index = stratif(sigma, depth, min.depths=0:5, max.depth=60:65),
    DCM = maxd(chla, depth, n.smooth=2, k=3),
    MLD = mld(sigma, depth, ref.depths=0:5, default.depth=80),
    # it is even possible to use variables computed above to make the
    # following computations adapted to each cast:
    # average tempeature in the mixed layer only
    temp_avg = integrate(temp, depth, from=0, to=MLD, fun=mean),
    # stock of Chl a within 10 m of the DCM
    chla_dcm_stock = integrate(chla, depth, from=DCM-10, to=DCM+10)
  )
# Inspect the results
ggplot(stats) + geom_path(aes(x=date, y=-thermocline))
# thermocline depth is still messy despite the smoothing, probably
# related to sharp changes in this coastal region.

ggplot(stats) + geom_path(aes(x=date, y=strat_index))
ggplot(stats) + geom_path(aes(x=date, y=-MLD))
# the stratification index and the MLD show a nice seasonality in mixing,
# with shallow MLDs and strong stratification in summer and deep mixing
# in winter.

ggplot(stats) + geom_path(aes(x=date, y=temp_avg))
# average temperature in the mixed layer shows the expected seasonality.

ggplot(stats) + geom_path(aes(x=date, y=chla_dcm_stock))
# the stock of chlorophyll a (estimated by fluorescence) within 10 m on
# either side of the DCM suggests that 2016 had a more productive DCM.

## End(Not run)