This data package corresponds to a research paper by Giesbrecht et. al. (2025) in the journal Ecosystems with the title "Mapping the spatial heterogeneity of watershed ecosystems and water quality in rainforest fjordlands". https://doi.org/10.1007/s10021-025-00964-x

The data package contains:

1. A shapefile representing sampled watershed boundaries in .shp format ("G2025_Wts_boundaries.shp")
2. Table of watershed characteristics in .csv format ("G2025_Wts_data.csv")
3. Table of quality controlled water quality data in .csv format ("G2025_WQ_data.csv")
4. A README file with variable definitions for water quality data ("G2025_WQ_README.csv")
5. A README file with variable definitions for watershed characteristics ("G2025_Wts_README.csv")

In this study, we examined spatial controls on the quality of freshwater exported from diverse watersheds in fjordlands of a coastal temperate rainforest. Samples were collected about once per month for a year from the outlets of 56 watersheds spanning from high mountains with icefields to low islands with extensive wetlands. Watershed size ranged from < 1.5 km² to 5,782 km² (Homathko River), yet in the regional and global context, all are considered “small” coastal watersheds (< 10,000 km² following Milliman and Syvitski, 1992). The study watersheds were spatially distributed along two fjordland transects on the south-central coast of British Columbia, Canada (51°57' N to 50°07' N and 128°09' W to 123°44' W).

Watershed characterization and classification

This study takes advantage of a previous watershed classification effort, which used four widely available (open access) datasets and 12 easily computed watershed characteristics to define 12 types of small coastal watersheds with cluster analysis (Giesbrecht and others 2022). These clusters separated watersheds by characteristic water source (glacial, snowmelt, rain), topography (mountains, hills, lowland), climate and geographic location within the NPCTR (north, central, south).

For the present study, we assigned every sampled watershed to one of the 12 types of coastal watershed defined in Giesbrecht and others (2022). This assignment required a modelling step because 34 of our 56 sampled watersheds were smaller than the minimum size (20 km²) of well delineated watersheds (DW) used in the regional scale classification (2022). We used a random forest (Breiman, 2001) classifier (randomForest package (version 4.6-14) in R (R Core Team, 2020)) to assign class membership to newly delineated (very small) watersheds.The predictor variables were the 12 watershed characteristics originally used to define the regional watershed types via cluster analysis (Giesbrecht and others, 2022). The response variable was watershed type. The present analysis revised the previous regional watershed classification by better resolving the locations and extent of watersheds in the ~ 1 to 10 km² size range, particularly those with very low relief and extensive wetland cover.

Stream chemistry data

Water samples were collected from the watershed outlets roughly once every month for a year (March 2018 to March 2019), for a total of 405 observation site-days after quality control. Most watersheds were sampled eight to ten times. Each transect was surveyed over two to three consecutive days in order to sample under relatively similar weather and flow conditions. The two transects were surveyed as close together in time as feasible, but were often more than a week apart, thus not always capturing the same weather system.

From each water sample, we measured 22 aspects of riverine water quality, including DOC, alkalinity, cations, organic and inorganic nutrients, 𝛿¹⁸O-H₂O, 𝛿²H-H₂O, and handheld sensor (YSI ProDSS) readings of temperature, specific conductance, pH, and turbidity. Water samples and sensor readings were taken from the main flow, avoiding eddies, shallow water, loose substrates, or woody debris. Samples for dissolved constituents were field-filtered with a 0.45 µm Millipore® Millex-HP hydrophilic polyethyl sulfonate (PES) syringe filter. All samples were kept cool and dark during the field work. Samples were then preserved by freezing or acidification as appropriate, within 24 hours of field collection. The field and laboratory procedures for this study follow those of St. Pierre and others (2021) and Tank and others (2020). Laboratory results below the detection limit were replaced by ½ the detection limit, following common convention (e.g., EPA, 2000). In addition to direct measurements, we calculated several variables from the analytical laboratory results: the total concentration of dissolved inorganic nitrogen (DIN), dissolved organic nitrogen (DON), particulate nitrogen (PN) dissolved organic phosphorous (DOP), and particulate phosphorous (PP). Finally, we computed the mass ratio of sodium to calcium ions (Na:Ca) as a simple index of cation origin. High Na:Ca ratios can be caused by high inputs of cyclic marine salts (via precipitation) relative to cation inputs from rock weathering (Gibbs, 1970; Schlesinger, 1997) and by high inputs from silicate weathering relative to carbonate weathering (Gaillardet and others, 1999; Tank and others, 2012a).

Several quality control (QC) and data cleaning procedures were implemented prior to the analysis, using a combination of visual inspection and data-based criteria. For visual inspection, tables and plots of the water quality measurements were examined while cross referencing metadata from field notes and laboratory notes. We omitted any suspiciously high or low values that could be readily explained by a procedural anomaly such as a cracked sample vial. For data-based QC, outlier values of sensitive species (DIN species, TN, and SRP) were identified (mean ± 4SD) and omitted unless supported by independent measurements (e.g., high DIN supported by high TDN and high TN). Additional quality control procedures were applied to calculated values to avoid use of illogical results. For example, where DIN > TDN, the resulting negative DON value was replaced with ½ the detection limit of TDN to indicate a small non-zero quantity. We also omitted samples where specific conductance exceeded 200 µS cm^-1, which are suspiciously high for the geological conditions we sampled. These samples also had high concentrations of Na⁺, K⁺, Cl^-/SO₄^2-, or Sr²⁺ (where available), likely due to tidal mixing of brackish water. We identified seven such cases, representing five site-dates.

Please refer to the corresponding research paper for a more complete description of methods: https://doi.org/10.1007/s10021-025-00964-x