Predicted groundwater redox state in New Zealand


Denitrification (i.e. the attenuation of nitrate, a key groundwater contaminant in New Zealand) requires anoxic or low oxygen environments. Hence the redox status of groundwater is a useful indicator for identifying potential denitrification zones, as an ecosystem service for tailored land management and targeted environmental regulations.

This dataset was produced via a multivariate statistical model linking the redox status observed at wells to a range of physical variables associated with the well locations and available in nation-wide geospatial datasets (covering soils, geology, hydrology etc.). The trained model was then used to predict groundwater redox status in areas lacking observation data.

The dataset contains the following components for 5m, 15m, 30m and 50m depth below the groundwater table in New Zealand:

  • redox status (oxic, mixed or reduced)
  • difference in probability between the two most likely redox states, used as a measure of confidence in the redox prediction, where a small difference equates to low confidence
  • redox status (oxic, mixed, or reduced) where the data with low confidence is removed
  • model area split into above and below the confidence threshold


Date: Jan 2024 Version: v1

Owner: ESR, Lincoln Agritech

Contact: Dr Theo Sarris, ESR


Preview Image


Dataset attributes

Spatial extent New Zealand
Spatial resolution 250m
Temporal extent
Temporal resolution
Evaluation method (Validation)
Evaluation result (Numeric)
Evaluation result (Categorical)
Uncertainty method Refer to paper
Uncertainty data format (Numeric)
Uncertainty data format (Categorical)



The methodology is explained in the upcoming paper "Reducing uncertainty of groundwater redox predictions at a national scale" by Theo S. Sarris, Scott R. Wilson, Murray E. Close, Phillip Abraham, and Allanah Kenny.


Fitness for purpose / limitations

This table indicates whether the dataset is suitable for different types of questions at different scales.

Note: Users should carefully consider their purpose as this dataset may not be suitable.

Operational Absolute Relative Screening/scoping
Block/farm Maybe Maybe Maybe Maybe
Multi-farms(5+) Yes Yes Yes Yes
Catchment Yes Yes Yes Yes
National/regional Yes Yes Yes Yes
Caveat(s) Relatively coarse spatial resolution, and model is based on current data availability

Data and Resources