Skip to contents

Buoyancy frequency squared (N^2) of one CTD cast, TEOS-10

Source: R/ctd_derived.R
ctd_n2.Rd

Rasmus Swalethorp's request (email 2026-09-24, with his R draft "Boyancy frequency and pycnocline.r" and the legacy E_BuoyFreq_Calc.m): "Buoyancy frequency ... calculated just for the 1 m binned CTD data that could be served up on the CTD transect plotter. This product tells us something about the stability of the water column and is useful for identifying layers of mixing", modernized "following TEOS-10 standards using the Gibbs package". Absolute salinity (gsw_SA_from_SP) and conservative temperature (gsw_CT_from_t), then gsw_Nsquared() on the staggered mid-points, aligned back to the sample pressures by linear interpolation (rule = 2, as the draft does), then a centred triangular-weighted (Bartlett) running mean over smooth_m dbar.

Usage

ctd_n2(
  temperature,
  salinity,
  pressure,
  longitude,
  latitude,
  q_temperature = NA,
  q_salinity = NA,
  smooth_m = 10
)

Arguments

temperature

in-situ temperature (deg C, ITS-90), one cast (a scalar is recycled).

salinity

practical salinity (a scalar is recycled).

pressure

sea pressure (dbar).

longitude, latitude

the cast's position (scalar or per sample).

q_temperature, q_salinity

quality codes (NA = good).

smooth_m

running-mean window (dbar), default 10.

Value

numeric vector, N^2 (s^-2), the length of the input.

Details

The smoothing is the one departure from the draft, which fits smooth.spline(spar = 0.4) to the N^2 profile before searching for the pycnocline: a weighted mean over a stated window gives the same answer on every machine with no tuning parameter, and the legacy bottle script (F_PycnoclineDepth.m) used a 10 m centred difference of the interpolated density, which is the same window. The weights fall linearly from the centre to the window's edge, so a one-bin step smooths to a peak at the step (a flat box mean would make a plateau whose shallow edge wins the maximum). As in that script, the value is undefined (NA) within half a window of the profile's top and bottom, where a truncated window would inflate a surface step. smooth_m = 0 returns the aligned raw N^2 at every sample. A sample whose temperature or salinity is flagged 8/9 or missing takes no part in the computation and is NA in the result; the output is aligned to the input.

Examples

p <- 1:100
n2 <- ctd_n2(ifelse(p < 50, 16, 10), 33.5, p, -120, 33)
p[which.max(n2)]   # near the 50 dbar interface
#> [1] 49