Physicochemical and Microbiological Water Quality Sampling

Where practical, individual rivers water quality and ecology (RWQE) monitoring sites are sampled at the same time of the month (and usually at the same time of the day) and all sites on a river or stream are sampled on the same day. Field meters are calibrated on the morning of the day of sampling and on the return. Water samples are collected in mid-stream (where possible), typically in run-type habitat from a representative reach of stream. Samples requiring laboratory analysis are placed in chilly bins with ice and couriered overnight to RJ Hill Laboratories in Hamilton. Water samples for heavy metal and dissolved nutrient analysis were all laboratory filtered.


Nutrients: Nitrogen & Phosphorous

Ammoniacal- and nitrate-nitrogen are toxicants in freshwater that can cause lethal or sub-lethal effects for aquatic species. In many cases, nitrate concentrations need to be managed at considerably lower than toxic levels to avoid excessive periphyton and macroalgae growth.

Dissolved reactive phosphorus (DRP), when substantially elevated above natural reference conditions, can negatively impact ecological communities. In combination with other conditions favouring eutrophication, DRP enrichment drives excessive primary production and significant changes in macroinvertebrate and fish communities, as taxa sensitive to hypoxia are lost.

Benchmarking

Results for pH adjusted ammoniacal nitrogen, nitrate nitrogen, and dissolved reactive phosphorus are rated against the Ministry for the Environment (MfE) National Objectives Framework (NOF) guidelines. See the National Policy Statement for Freshwater Management (NPS-FM) document for more information.

Model Estimates

The predicted water quality values were generated using Random Forest (RF) models. The RF empirical modelling method predicts the values of response variables using a suite of predictor variables and a dataset of observations (the ‘training data’). RF models are an advanced form of regression-tree models.

The observational data used in the RF models consisted of site median values from monthly and quarterly measurements for the period 2009-2013. These data came from 354-586 monitoring sites (depending on the variable). The sites are reasonably well distributed across the North and South Islands, with some gaps in inaccessible areas.

The RF models performed well in predicting median water quality state, based on the amount of variation in the observational data explained, the congruence between observed and predicted values, low model bias (tendency to over- or underestimate), and low prediction uncertainty. See section 4 of the Larned, S, Snelder, T, & Unwin, M (2017) for more details on the model results and performance.


Metals: Copper & Zinc

Metals can have toxicant effects on aquatic life in both a dissolved state and when attached to sediment particles. Zinc and copper have been adopted throughout the Te Awarua-o-Porirua Whaitua Implementation Plan (WIP) as proxies for the suite of other urban contaminants (e.g., polycyclic aromatic hydrocarbons, other toxic metals (such as cadmium and chromium), detergents/surfactants and other chemicals). Copper is approximately 5 to 10 times more toxic to aquatic life than zinc but occurs in lower concentrations.

Benchmarking

In the absence of NPS-FM/NOF attribute tables for zinc or copper, an interpretation table was developed for zinc and copper that follows the same rationale as the toxicity attributes in the NPS-FM, that is, it includes two sets of state band thresholds for chronic and acute exposure (see Appendix 1 of the Whaitua Implementation Plan (WIP)). The chronic exposure thresholds adopt the figures for 99%, 95% and 80% species protection given in the ANZECC (2000) guidelines.

The application of the framework is limited by not having a second set of toxicity data that enabled the acute thresholds to be derived for the NOF toxicity attributes. Instead, this table has adopted lower species protection thresholds for the A and B attribute states (i.e., 95% and 90% for A and B states respectively), while the bottom of the C attribute state is defined from United States Environmental Protection Agency (USEPA) acute toxicity thresholds (USEPA 1996, USEPA 2007). Because these thresholds are uncertain proxies for acute toxicity thresholds, it is suggested to set objectives for 95th percentile concentrations rather than the more stringent maximum.


Microbiology: E. coli

Escherichia coli (E. coli) is a type of bacteria commonly found in the intestines of warm-blooded animals, including people. E. coli in river waters is one of five indicators that provide an overview of New Zealand’s river water quality and how it is changing over time.

E. coli in fresh water can indicate the presence of pathogens (disease-causing organisms) from animal or human faeces. The pathogens can cause illness for anyone who ingests them. Campylobacter is one of the most common pathogens associated with animal and human faeces, but it is difficult to measure. We use E. coli concentrations measured as colony forming units (cfu) to infer Campylobacter infection risk in waterways.

Benchmarking

E. coli results are rated against the Ministry for the Environment (MfE) National Objectives Framework (NOF) guidelines designed to help guide decisions related to the protection of human health. See the National Policy Statement for Freshwater Management (NPS-FM) document for more information.

Model Estimates

The predicted water quality values for each of the four statistical metrics (median, 95th percentile, % >260 cfu/100ml, and % >540 cfu/100ml) were generated using Random Forest (RF) models. The RF empirical modelling method predicts the values of response variables using a suite of predictor variables and a dataset of observations (the ‘training data’). RF models are an advanced form of regression-tree models.

The observational data used in the RF models consisted of State of the Environment (SoE) data from monthly and quarterly samples collected throughout New Zealand from 1990 at some sites until the end of 2013.

The RF models for median, 95th percentile and % >260 cfu/100ml had generally good performance and the model for % >540 cfu/100ml had satisfactory performance. All four models had very low bias. See section 4.2 of Snelder, T., Wood, S., Atalah, J. (2016) for more details on the model results and performance.

Modelled grades were subsequently adjusted where the predicted estimates did not accurately represent monitoring sites. These adjustments were based on:

  • expert opinion from freshwater scientists.
  • fact-checking with regional councils.
  • actual data at a monitoring site.

See page 56 and Appendix A of MfE (2018) for further information on subsequent development.


Sediment: Suspended & Deposited, Water Clarity

Sediment discharged into rivers, streams and harbours can negatively impact a range of values, including ecosystem health and the way people use water for recreational, cultural, and spiritual purposes. Sediment affects ecosystem function in rivers and streams by:

  • reducing light penetrating the water, affecting the ability of plants to grow
  • impacting the health of fish by abrading skin and gills and making predators and prey difficult to see
  • filling the interstitial spaces (spaces between rocks and pebbles, etc.) in stream beds, making these spaces less suitable for macroinvertebrate communities to survive and thrive.

In estuaries and harbours, sediment:

  • alters and degrades habitat and the composition of ecological communities by smothering invertebrates, shellfish, and seagrass
  • changes the depth of water and flow patterns.
  • reduces clarity.
  • changes the feel of substrate underfoot.

Benchmarking

Results for deposited fine sediment and water clarity are rated against the Ministry for the Environment (MfE) National Objectives Framework (NOF) guidelines. See the National Policy Statement for Freshwater Management (NPS-FM) document for more information.

Model Estimates

Sediment cover at each reach is estimated from a contemporary boosted regression tree (BRT) model of Clapcott and Goodwin (2017) accessed from the MfE data service. This model estimated sediment cover based on land cover (such as native vegetation, exotic vegetation, and pastoral heavy) and environmental variables (such as slope, geology, and rainfall days). Testing over observed data at 8482 sites showed fair to good model performance and effectively no bias. More details of the data and model development can be read in sections 2 and 4 of Clapcott and Goodwin (2017).

Water clarity estimates are obtained from an RF model developed by Larned et al. (2017) using data from 454 sites. This model performed well though slightly underestimates at high values and overestimates at low values.


Ecology


Macroinvertebrates

Macroinvertebrates play a central role in stream ecosystems by feeding on periphyton (algae), macrophytes, dead leaves and wood, or each other. They are extremely important for processing terrestrial and aquatic organic matter, and in turn, are an important food source for animals further up the food chain, such as wading birds and fish. When the insects become adults, they leave the water and become food for animals such as birds, bats, spiders, etc.

The Macroinvertebrate Community Index (MCI) is based on the presence or absence of invertebrate species (taxa) with different tolerances/sensitivities to organic pollution and nutrient enrichment. For this reason it is regularly used as an indicator of river or stream ecosystem health.

Sampling

A single macroinvertebrate sample is collected at RWQE water sampling sites during summer/early autumn. The timing of sampling is determined at random, although macroinvertebrate sampling is, where practicable, avoided within two weeks of any flood event (flood events are defined as flows greater than three times the median river flow).

Samples are collected with the use of a kick-net (0.5 mm mesh size) following Protocol C1 of the national macroinvertebrate sampling protocols (Stark et al. 2001) for the 42 sites with hard substrate (in riffle habitat) and Protocol C2 for the 9 sites with a soft substrate. All samples are processed in accordance with Protocol P2 (Stark et al. 2001).

Benchmarking

Macroinvertebrate Community Index (MCI) scores are assessed against quality classes recommended for the Greater Wellington Region and Greater Wellington Natural Resources Plan (NRP) plan outcomes (Clapcott and Goodwin, 2014).

These thresholds have been developed based on regional data for six Freshwater Ecosystems of New Zealand (FENZ) river classes and were defined from statistical distributions of data from a mix of:

  • observed reference sites.
  • modelled ‘reference’ conditions using reaches with land use restrictions that indicate low disturbance.
  • all contemporary observed sites.
  • all modelled contemporary conditions.

Limits for MCI class and NRP outcomes in the table below refer to the latest three year median MCI score.

MCI score quality class
NRP outcomes
River class D C B A All rivers Significant rivers
1 (Steep, hard sedimentary) < 110 110-120 120-130 ≥ 130 ≥ 120 ≥ 130
2 (Mid-gradient, coastal and hard sedimentary) < 80 80-105 105-130 ≥ 130 ≥ 105 ≥ 130
3 (Mid-gradient, soft sedimentary) < 80 80-105 105-130 ≥ 130 ≥ 105 ≥ 130
4 (Lowland, large, draining ranges) < 90 90-110 110-130 ≥ 130 ≥ 110 ≥ 130
5 (Lowland, large, draining plains and eastern Wairarapa) < 80 80-100 100-120 ≥ 120 ≥ 100 ≥ 120
6 (Lowland, small) < 80 80-100 100-120 ≥ 120 ≥ 100 ≥ 120

Model Estimates

This regional specific model estimated MCI score based on land cover and environmental variables such as slope, geology, climate. Model performance diagnostics indicated a very good predictive model, with 95th percent confidence intervals of < 29 MCI units, and effectively no bias (< 0.1 MCI unit). More details of the data and model development can be read in section 2.1.1 of Clapcott J, Goodwin E (2014).


Periphyton & Cyanobacteria

Periphyton is algae/slime that attaches to hard surfaces such as rocks and tree roots in freshwater environments. It is an important food source for invertebrates and some fish, and can absorb contaminants from water (e.g., nitrate, ammonia, phosphorus, and metals). However, too much of it can limit the food sources and/or habitat of macroinvertebrates (e.g., insects, snails, and worms), affect the ability of fish to find food, and cause harmful water quality effects such as daily fluctuations in dissolved oxygen and pH (acidity). Periphyton blooms can also be visually unappealing and can make access to streams difficult (slippery).

Cyanobacteria (commonly known as blue-green algae) are photosynthetic prokaryotic organisms that are integral parts of many terrestrial and aquatic ecosystems. In aquatic environments, under favourable conditions, cyanobacterial cells can multiply and form planktonic (suspended in the water column) blooms or dense benthic (attached to the substrate) mats. An increasing number of cyanobacterial species are known to include toxin-producing strains. These natural toxins, known as cyanotoxins, are a threat to humans and animals when consumed in drinking water or by contact during recreational activities. The mechanisms of toxicity for cyanotoxins are very diverse, ranging from acute unspecified intoxication symptoms (e.g., rapid onset of nausea and diarrhoea), to gastroenteritis and other specific effects, such as hepatotoxicity (liver damage) and possibly carcinogenesis (MfE & MoH 2009).

Sampling

Formal periphyton & cyanobacteria assessments are limited to the 42 RWQE sites with hard substrates.

Monthly Assessment of Visible Streambed Cover

Periphyton cover is determined by estimating the percentage of mat (>1 mm thick), cyanobacterial mat (>1 mm thick) and filamentous (>2 cm long) periphyton present on the stream or riverbed. Note that cover of mat and cyanobacterial mat-periphyton are mutually exclusive (ie, cyanobacterial mat cover >1 mm thick will be counted as separate from mat-periphyton). A total of 20 observations are taken at each site from two transects of ten observations, or, if the stream or river is not wide enough or too swift to wade across more than half of the river’s width, four transects of five observations. Each observation is typically made with an underwater viewer and covers an approximate area of a 30 cm diameter circle.

Visible streambed periphyton cover assessments are carried out equally in both run and riffle-type habitats if these are present at a sampling site/reach.

Monthly Assessment of Biomass

Periphyton samples for quantitative biomass assessments (chlorophyll a) are collected on a monthly basis. During 2020/21, chlorophyll a samples were collected from 15 of the 42 RWQE sites with hard substrates. Sampling protocols involved collecting samples from a run habitat and following modified versions of quantitative methods 1b (QM-1b) and 3 (QM-3) as outlined by Biggs and Kilroy (2000). This involves pooling periphyton samples from 10 rocks into a single composite sample for analysis (see Greenfield (2016) for further details).

Benchmarking

Monthly observations of percent streambed periphyton cover (filamentous and mat-forming periphyton) are compared against the periphyton composite cover guidelines (Matheson et al. 2012). The threshold for nuisance mat cover is twice that for filamentous periphyton cover, so the periphyton weighted composite cover (WCC) can be defined as filamentous periphyton cover + (mat periphyton cover / 2) with a nuisance guideline of ≥ 30%.

Results for periphyton biomass are rated against the Ministry for the Environment (MfE) National Objectives Framework (NOF) guidelines. See the National Policy Statement for Freshwater Management (NPS-FM) document for more information.

Model Estimates

Periphyton biomass state has been estimated for each river reach by comparing modelled median total nitrogen (TN) and dissolved reactive phosphorus (DRP) concentrations from Larned et al. (2017) to DRP thresholds in Snelder et al. (2019) and revised TN thresholds in MfE (2019).

These nutrient thresholds relate to each NOF state where increasing levels of estimated TN and DRP (see nutrients model estimates above where TN is modelled using the same approach) correspond to higher risk of increased periphyton biomass. In the case that TN and DRP thresholds estimated different periphyton states for the same river reach the higher risk state has been used.


Habitat Quality

Habitat assessments are undertaken annually at RWQE sites during summer/early autumn when invertebrates samples are collected following the updated methods outlined in Clapcott (2015). This assessment provides an indication of the condition of the physical habitat and its ability to support stream biota, and incorporates the following variables: deposited sediment cover, invertebrate habitat abundance and diversity, fish habitat abundance and diversity, hydraulic heterogeneity, bank erosion and vegetation, and riparian width and shade. Each category is scored between 1 (‘poor’) and 10 (‘excellent’). Summation of individual scores provides an overall total habitat quality score for each site (lowest and highest possible scores are 10 and 100, respectively).

This methodology was developed with a focus on wadeable hard-bottomed streams (Clapcott, 2015) and hence its applicability to other stream/river types has not been explored.


Note on COVID-19

Two to three field samples were missed for several of water quality variables due to lockdown periods.