Skip to contents

hydroSense

hydroSense analyses water quality in freshwater systems potentially impacted by landfill leachate. It turns routine monitoring chemistry into ecologically interpretable, mixture-aware risk estimates, with explicit handling of below-detection data and local background conditions. It is part of the tidyWaste family of packages.

The package has three pillars:

  • Chemistry imputation. A domain-agnostic Bayesian multivariate model (fit_imputation_model() / impute_chemistry()) imputes below-detection and missing analyte concentrations while preserving cross-analyte correlation structure, and can return full posterior draws. Analyte groups are configured with impute_group(); leachate_impute_groups() is the bundled leachate preset.
  • Leachate detection (LMF). End-member mixing analysis (add_lmf()) estimates the leachate-mixing fraction of each sample from its major-ion signature, distinguishing leachate influence from natural variation.
  • Multi-substance toxicity (msPAF). Species sensitivity distributions (ssd_paf()) feed a multi-substance potentially-affected-fraction (add_mspaf()) that combines toxicants by concentration- and response-addition and subtracts the local background via the Added Risk Approach. The background can be a static reference or a contemporaneous one (fit_reference_model()) that predicts what the reference site would have shown at each sample’s moment from hydrology and season; a season-blind companion (fit_target_model()) predicts the impacted site between grabs, so mspaf_daily() can return a continuous daily risk series.

Toxicity is grounded in the ANZECC/ANZG freshwater guideline datasets, with bioavailability and physicochemical normalisations (hardness, pH, DOC, temperature) applied so that every sample is compared on the same index-condition basis.

Installation

You can install the development version from GitHub:

# install.packages("pak")
pak::pak("vorpalvorpal/hydroSense")

brms is a hard dependency (Imports) and installs with the package, but the Bayesian imputation step additionally needs a working Stan toolchain, which is not pulled in automatically; the toxicity and LMF pillars do not need Stan.

Usage

A typical msPAF assessment takes long-format monitoring data (one row per sample × analyte) and returns a per-sample multi-substance PAF, adjusted for local background:

library(hydroSense)

# Point the toxicity engine at the ANZG guideline-data folder once per session.
options(hydroSense.guideline_dir = "path/to/guideline data")

mspaf <- add_mspaf(
  df        = monitoring_long,   # sample_id, site_id, analyte, value, ...
  reference = reference_long     # local upstream / background chemistry
)

Single-substance PAF and SSD quantities are available directly:

# Fraction of species potentially affected by 9.3 mg/L total ammonia-N
# (at the pH 7.0 / 20 degC index condition):
ssd_paf("NH3-N", conc_ug_L = 9321)

# Concentration affecting 50% of species (used as the Toxic-Unit denominator):
ssd_hc50("Cu")

Learn more

  • Leachate Mixing Fraction — what fraction of a sample’s chemistry can be explained as a mixture of local reference water and landfill leachate.
  • Imputation — how missing and below-detection values are imputed.
  • Analyte normalisation — how bioavailability and physicochemical corrections bring concentrations to the SSD index condition.
  • msPAF — what the returned PAF values mean for an environmental assessment.
  • Function reference — all exported functions, grouped by pillar.

License

MIT © Robin Shannon. Toxicity data are derived from the ANZECC/ANZG freshwater guideline materials; see those sources for their own licensing terms.