NOW-Project

This repository contains the data processing and analysis pipeline supporting the manuscript "The North Water Polynya in Transition: Multi-Decadal Changes of Hydrographic and Nutrient Dynamics" (Schiffrine & Tremblay, submitted to Journal of Geophysical Research: Oceans). The North Water Polynya (NOW), located at the confluence of Nares Strait and Baffin Bay, is a biologically productive Arctic region where Pacific-origin and Atlantic-origin waters interact. Understanding how upstream hydrographic changes propagate through this system is essential for predicting shifts in nutrient supply, primary production, and carbon cycling under ongoing Arctic transformation.

This project implements a complete analytical workflow for processing multi-decadal (2005–2023) hydrographic and biogeochemical observations collected during ArcticNet expeditions along a zonal transect across the NOW (~76.2–76.5°N). The pipeline includes: (1) standardization of nutrient databases across multiple expedition years, (2) geographic extraction and hierarchical station clustering for spatial consistency, (3) water mass classification using a σ₀–spicity methodology adapted from Huang et al. (2024), which provides more robust discrimination than traditional θ–S diagrams in high-latitude systems, and (4) derivation of geochemical tracers including N* (Gruber & Sarmiento), Pacific Water fraction (fPW), and Arctic N–P indices.

Five water masses are identified and characterized: Polar Mixed Water (PMW), Upper Halocline Water (UHW), Baffin Bay Polar Water (BBPW), Subpolar Mode Water (SPMW), and Canadian Basin Atlantic Water (CBAW). The classification framework includes sensitivity analysis to optimize discrimination thresholds and ensure robust detection of transitional waters.

Je vois maintenant le fichier réel. Voici la documentation corrigée :

---

README — 20260209_NOW.R

Purpose

20260209_NOW.R is the master orchestration script that constructs the final analysis-ready dataset NOW. Crucially, there is no separate parameters.R file — all global constants are defined inline at the top of this script before any function is sourced.

---

Workflow

---

STEP 0 — Environment Setup

rm(list = ls())
setwd(dir = "/Users/nicolas/.../")

Clears the workspace and sets the working directory to the OneDrive root containing the DataBase_Nut/ folder.

---

STEP 1 — Inline Configuration

Defined directly in 20260209_NOW.R, before any source() call:

parameter <- c("Temperature", "Salinity", "Nitrate", "Nitrite", 
               "Ammonium", "Oxconc", "Phosphate", "Silicate")

NOW_BOUNDS <- list(
  longitude = c(-80, -65),
  latitude  = c(76.2, 76.5)
)

parameter defines the biogeochemical variables extracted from raw files. NOW_BOUNDS defines the geographic bounding box delimiting the NOW transect (76.2–76.5°N, 80–65°W). Both objects are global constants available to all downstream functions.

Note: Clustering parameters (n_clusters = 15, min_stations_per_cluster = 7) and classification parameters (threshold = 0.6, comparable_ratio = 1.3) are passed directly as arguments in the relevant function calls rather than stored as named constants.

---

STEP 2 — Package Loading

source("DataBase_Nut/20260209/20260209_Package.R")

Loads load_now_packages() and immediately calls it. This function iterates over required_packages — a vector of ~30 packages including tidyverse, gsw, ggplot2, ggOceanMaps, metR, lubridate, trend — and installs any missing packages automatically when install_missing = TRUE.

---

STEP 3 — Function Sourcing

source("DataBase_Nut/20260209/20260209_DataLoading.R")
source("DataBase_Nut/20260209/20260209_Extract_Transect.R")
source("DataBase_Nut/20260209/20260209_ClusterStation.R")
source("DataBase_Nut/20260209/20260209_sw_pspi.r")
source("DataBase_Nut/20260209/20260209_WaterMass_Classif.R")
source("DataBase_Nut/20260209/20260209_PacificWater.R")

All project-specific functions are loaded into the global environment. No computation occurs at this stage.

---

STEP 4 — Data Loading

raw_data <- load_data()

load_data() is sourced from 20260209_DataLoading.R. It reads and standardizes three tab-delimited ArcticNet source files:

File	Coverage
R_Database_dec2021.txt	2005–2021
R_Data_2022.txt	2022
R_Data_2023.txt	2023

For each file the function:

1. Reads raw data with empty strings coerced to NA
2. Forward-fills station metadata to resolve orphan rows
3. Converts Western-positive longitudes to negative-East convention
4. Extracts year, month, day, and day-of-year from date strings
5. Coerces biogeochemical fields to numeric (non-parseable values → NA)
6. Selects only the variables listed in parameter

The three processed data frames are row-bound and deduplicated into a single standardized database.

Output: raw_data

---

STEP 5 — Transect Extraction

raw_now <- extract_transect(data = raw_data)

extract_transect() is sourced from 20260209_Extract_Transect.R. It applies a between() filter on decimalLongitude and decimalLatitude using the bounds defined in NOW_BOUNDS, retaining only observations within the NOW transect spatial domain. A console report logs original count, retained count, and retention rate.

Output: raw_now

---

STEP 6 — Station Clustering

now_clusters <- cluster_stations(
  data = raw_now,
  n_clusters = 15,
  min_stations_per_cluster = 7
)

now_renamed <- rename_clusters(now_clusters)

now_geo <- raw_now %>%
  left_join(now_renamed, by = c("Station","decimalLongitude",
                                 "decimalLatitude","year","month","day")) %>%
  select(Cruise, Station, Station_Cluster, everything(), -Cluster) %>%
  filter(!is.na(Station_Cluster))

Two functions are sourced from 20260209_ClusterStation.R:

cluster_stations() Addresses positional offsets between nominally identical stations across expedition years. The function:

1. Extracts unique (Station, lon, lat) combinations
2. Computes a pairwise geographic Euclidean distance matrix
3. Applies Ward's minimum variance agglomerative clustering (ward.D2)
4. Cuts the dendrogram at n_clusters = 15
5. Discards clusters with fewer than min_stations_per_cluster = 7 temporal instances

rename_clusters() Assigns canonical west-to-east station identifiers:

1. Computes median longitude per cluster
2. Ranks clusters geographically and assigns IDs (100, 101, 103, …, 116)
3. Detects and resolves naming conflicts when multiple historical station names map to a single cluster

Cluster assignments are joined back to raw_now; observations outside any valid cluster are discarded via filter(!is.na(Station_Cluster)).

Output: now_geo

---

STEP 7 — Oceanographic Property Calculation

now_hydro <- now_geo %>%
  mutate(
    sigma0  = gsw_sigma0(SA = Salinity, CT = Temperature),
    spicity = sw_pspi(S = Salinity, temp = Temperature,
                      temp_unit = "conservative", sal_unit = "SA",
                      longitude = ref_coords$longitude,
                      latitude  = ref_coords$latitude, pr = 0)
  )

Reference coordinates are first computed as the median longitude and latitude of now_geo, then passed to both functions.

gsw_sigma0() from the TEOS-10 Gibbs SeaWater toolbox computes potential density anomaly (σ₀, kg m⁻³) referenced to surface pressure.

sw_pspi() is sourced from 20260209_sw_pspi.R. It computes potential spicity (π₀, kg m⁻³) using the 41-coefficient polynomial of Huang et al. (2011). Spicity varies along isopycnals and provides enhanced water mass discrimination in Arctic stratified systems where contrasting T–S combinations share similar densities. Together, σ₀ and π₀ define the two-dimensional classification space used in Step 8.

Output: now_hydro

---

STEP 8 — Water Mass Classification

wm_ref <- create_endmembers(
  reference_lon = ref_coords$longitude,
  reference_lat = ref_coords$latitude
)

now_wm <- classify_watermass(
  now_hydro, wm_ref,
  threshold        = 0.6,
  comparable_ratio = 1.3,
  return_long      = FALSE
)

Two functions are sourced from 20260209_WaterMass_Classif.R:

create_endmembers() Defines five Arctic water mass endmembers by their canonical T–S properties and converts each to (σ₀, π₀) coordinates using the reference position:

Water Mass	Code
Polar Mixed Water	PMW
Upper Halocline Water	UHW
Baffin Bay Polar Water	BBPW
Subpolar Mode Water	SPMW
Canadian Basin Atlantic Water	CBAW

classify_watermass() For each observation the algorithm:

1. Computes scaled Euclidean distances to all five endmembers in (σ₀, π₀) space, normalized by each endmember's uncertainty ellipse
2. Identifies the nearest (d₁) and second-nearest (d₂) endmember
3. Applies a two-rule decision:
   • d₁ < threshold (0.6) → assign to nearest endmember
   • d₁/d₂ < comparable_ratio (1.3) → label Mixed (observation lies in a transitional zone equidistant between two water masses)
   • Otherwise → assign to nearest endmember

Output: now_wm with classification column (PMW | UHW | BBPW | SPMW | CBAW | Mixed)

---

STEP 9 — Geochemical Tracer Derivation and Final Assembly

NOW <- now_wm %>%
  mutate(
    Nstar    = 0.87 * (Nitrate - 16 * Phosphate + 2.9),
    TIN      = Nitrate + replace_na(Nitrite, 0) + replace_na(Ammonium, 0),
    ANP      = ANP(Phosphate, TIN),
    fPW      = fpw(Phosphate, TIN, "Jones1998", "Yamamoto.Kawai2008"),
    NO       = 9   * Nitrate   + Oxconc,
    PO       = 135 * Phosphate + Oxconc,
    NO_PO    = NO / PO,
    POs_star = POs_star(Phosphate, Oxconc, Salinity),
    classification  = factor(classification, 
                             c("PMW","UHW","BBPW","SPMW","CBAW","Mixed")),
    Station_Cluster = factor(Station_Cluster, 
                             c("100","101","103","105","107","108",
                               "110","111","113","115","116"))
  ) %>%
  select(Cruise:Salinity, sigma0:spicity, Nitrate:Silicate,
         TIN, Nstar, NO:POs_star, classification, distance, ANP, fPW) %>%
  filter(!year %in% c(1997, 1998, 1999, 2007, 2008))

Three functions are sourced from 20260209_PacificWater.R:

fpw() Estimates the Pacific Water fraction (fPW, 0–1) using a two-endmember PO₄–N mixing model. The Atlantic endmember follows Jones et al. (1998) and the Pacific endmember follows Yamamoto-Kawai et al. (2008). Values approaching 1 indicate predominantly Pacific-origin waters; values near 0 indicate Atlantic-origin dominance.

ANP() Computes the Arctic N–P tracer index following Newton et al. (2013). The function calculates orthogonal distances from each observation in (PO₄, TIN) space to both the Atlantic and Pacific N–P regression lines, returning the normalized ratio d_AW / (d_AW + d_PW). ANP = 0 indicates pure Atlantic affinity; ANP = 1 indicates pure Pacific affinity. This approach discriminates water mass origin independently of absolute concentration levels.

POs_star() Computes a salinity-normalized phosphate–oxygen combined tracer adapted for Arctic waters. It combines phosphate and dissolved oxygen signals with a salinity normalization to account for freshwater dilution effects characteristic of Arctic shelf and halocline environments.

The inline mutate() block additionally computes:

Tracer	Formula	Physical Meaning
TIN	NO₃ + NO₂ + NH₄	Total inorganic nitrogen; reduced forms set to 0 when missing via replace_na()
Nstar	0.87 × (NO₃ − 16×PO₄ + 2.9)	Deviation from Redfield N:P stoichiometry after Gruber & Sarmiento (1997); negative values flag Pacific-origin denitrified waters
NO	9 × NO₃ + O₂	Broecker (1974) semiconservative tracer; 9:1 weighting reflects stoichiometric O₂ consumption per mole of NO₃ produced during remineralization
PO	135 × PO₄ + O₂	Analogous semiconservative tracer using Redfield O₂:P ratio of 135:1
NO_PO	NO ÷ PO	Deviations from unity signal non-Redfield processes or mixing of water masses with contrasting preformed nutrient ratios

Temporal filtering excludes years 1997–1999 (pre-ArcticNet data with heterogeneous protocols) and 2007–2008 (anomalous sea-ice conditions precluding standard transect occupation).

Factor ordering enforces:

• classification: PMW → UHW → BBPW → SPMW → CBAW → Mixed (surface-to-deep vertical structure)
• Station_Cluster: 100 → 101 → … → 116 (west-to-east geographic ordering)

Output: NOW — the final analysis-ready data frame consumed by 20260209_NOW_Figure.R

---

Script Dependency Map

20260209_NOW.R
  │
  ├── 20260209_Package.R          → load_now_packages()
  ├── 20260209_DataLoading.R      → load_data()
  ├── 20260209_Extract_Transect.R → extract_transect()
  ├── 20260209_ClusterStation.R   → cluster_stations(), rename_clusters()
  ├── 20260209_sw_pspi.R          → sw_pspi()
  ├── 20260209_WaterMass_Classif.R → create_endmembers(), classify_watermass()
  └── 20260209_PacificWater.R     → fpw(), ANP(), POs_star()

---

Output Object Structure

Group	Variables
Metadata	Cruise, Station, Station_Cluster, year, month, day, doy, decimalLatitude, decimalLongitude, Depth
Physical	Temperature, Salinity, sigma0, spicity
Nutrients	Nitrate, Phosphate, Silicate, Oxconc, TIN
Tracers	Nstar, NO, PO, NO_PO, POs_star, ANP, fPW
Classification	classification, distance