The suggested next step (step three in figure 6.1) for water quality accounting is to establish the current state – that is, the loads and/or concentrations of contaminants in the FMU that you are accounting for. Determining the loads and concentrations already in the receiving water provides the comparison point (or reference) to which the sources and sinks (such as attenuation) are compared. The NPS-FM allows for the loads, concentrations and sources of relevant contaminants to be measured, modelled or estimated. However, because what is in the receiving waters provides a comparison point, they should be measured where possible.
In some cases, it is not possible or practical to measure water quality continuously, and it may be necessary to calculate loads and concentrations using some form of estimation. Thus, these loads and concentrations will have some levels of uncertainty associated with them. With a robust monitoring programme designed with the method of estimating loads (or concentrations) in mind, any bias and extraneous sources of error can be minimised, and the errors themselves known with a certain confidence.
Measurement and estimation is not always possible because of resourcing issues, in which case modelling is the only real alternative. While good modelling practice also deals with uncertainty, the difficulty of using modelling for estimating receiving water loads and concentrations in an accounting sense is that it is:
- difficult for stakeholders to understand
- likely that diffuse source loads and concentrations will also be modelled.
There is a risk that councils may end up comparing a modelled reference load/concentration (the load/concentration already in the receiving water) with modelled sources, possibly using the same model.
The use of measurement and modelling in establishing the current state is discussed in more detail below.
Measurement
All councils measure water quality to some extent, and have done so for extended periods. Some councils report on the state and trends in their water quality data through state of the environment reporting. [3]Others, such as Waikato Regional Council, provide similar information through regular updates of their websites.[4] Guidelines for setting up water quality monitoring networks are given by Davies-Colley et al (2011).
An existing water quality monitoring system should be able to meet the needs of water quality accounting. If setting up a system specifically for water quality accounting, the prime considerations would be:
- Relevant contaminants – these are what is required to be accounted for (see 6.2).
- Frequency – the frequency of sampling has implications for the precision of the estimate of load/concentration over the interval in question. There are good statistical texts (eg, McBride, 2005) and papers (Robertson and Roehrish, 1999) that will assist. A long record (more than 2 years) of monthly sampling is generally adequate for calculating annual medians and/or annual maximums (as stated in Appendix 2 of the NPS-FM) and is the duration recommended to councils for Environmental Monitoring and Reporting (EMaR) (Davies-Colley et al, 2012).
- Sampling strategy – deciding on a sampling strategy will involve considering the objectives of the programme and the flow distribution of the streams in question. Small ‘flashy’ streams will export most of their contaminants (especially particulate contaminants) during storms, which means estimating the load/concentration with high precision requires ‘storm chasing’ or use of flow-triggered automatic samplers. Such exercises may not be practical for routine monitoring. Some councils get around this problem by defining loads/concentrations that reflect river conditions ‘most of the time’, which is appropriate for most of the national values listed in the NPS-FM. For example, Horizons Regional Council uses flow-binning (case study 6.4) whereby loads are calculated according to particular flow ranges. Other councils use the fixed-interval approach (eg, sampling on the same day each month), which, providing the record is long enough, will theoretically sample the range of flow conditions in proportion to their flow distribution.
- Flows – measured or estimated? To calculate a load or concentration, you need a corresponding measurement or estimate of flow. Measured is desirable and, for existing water quality monitoring networks, councils may have a corresponding flow recorder. However, flow recorders can be expensive to install and maintain, and to get adequate spatial coverage for contaminant accounting it may be necessary to estimate flows from adjacent catchments. There are standard methods for making such estimates (see case study 6.3), but care needs to be taken to ensure the precision of the flow estimate is compatible with that of the load estimate.
- Estimation methods – there are a multiplicity of methods for load estimation and no ‘one size fits all’ method. For example, Diffuse Sources and NIWA (2013) reviewed estimation methods relevant to calculating nutrient loads to Waituna lagoon from diffuse sources. They recommended a range of regression approaches (‘rating curves’) for estimating contaminant loads. These regressions are based on concentrations varying significantly with flow rate and so the regressions employ a rating curve for load calculations.
Case studies 6.3–6.6 provide examples of methods that can be used to estimate contaminant loads.
Case study 6.3 – estimating flows for load calculations
One method of establishing flow records for streams with no flow recorder is the concurrent gauging technique. Griffiths and Horrell (2012) recently undertook a review for Environment Southland of the appropriateness of the use of the concurrent gauging technique to estimate flows at ungauged sites. A literature review was carried out and a set of guidelines for the technique distilled. These guidelines address the use of natural flows, tertiary site location, concurrent gauging runs, concurrent gauging range, flood flows, regression, errors and normalisation.
There are also other techniques for estimating flows in ungauged catchments. Woods et al (2006) describe the testing and development of a model that can be used to predict mean flows across New Zealand. Other models have been developed that can be used to optimise the ‘next best data’ by searching for the nearest flow recorder to the site of interest (see also Booker and Woods, 2014). There are often great uncertainties with using such methods.
Case study 6.4 – flow stratification methods
Horizons Regional Council uses flow stratification to address the issue of ‘what proportion of the total load of contaminants arises from each flow band?’.
Stratification includes defining 10 flow ‘bins’ based on subdivision of the flow frequency distribution (the flow duration curve). The component of the load associated with each flow bin is estimated as the product of the mean concentration and the flow associated with the bin. These loads are then summed to find the total load. An advantage of the method is that it automatically produces the proportion of the load associated with different parts of the flow hydrograph (eg, high flows). The flow-stratified averaging approach was found to be effective in reducing bias associated with monthly sampling that does not generally consist of either very high or low flows. However, loads at low flows (ie, below the 20th percentile) were calculated by removing the loads assigned to highest two flow bins.
See Roygard et al (2012) or Roygard and McArthur (2008), available from www.horizons.govt.nz/about-us/publications/about-us-publications/one-plan/technical-reports/.
Case study 6.5 – a tool for estimating sediment and other contaminant loads
SedRate is a tool developed more than 10 years ago to help address issues with predicting sediment loads and, specifically, for fitting rating curves and calculating time-averaged loads (ie, yields). It enables a sediment rating to be established for a particular site by means of one of four automated methods, or by user-fit with mouse-clicks on the rating plot. This sediment rating is then applied to the site flow distribution to obtain the annual average sediment yield, along with many other useful statistics. Flow and sediment information is obtained from an MSAccess database. Although originally designed just for suspended sediment, SedRate has been expanded to fit ratings to and calculate yields for any flow constituent – for example, nitrogen and phosphorous.
A feature of the tool is that results displayed enable the user to explore the sensitivity of the results to the way the rating is fitted and understand the uncertainties associated with the estimates produced.
SedRate was used to provide sediment yield data for Hicks et al’s (2011) national modelling of suspended sediment yields from New Zealand rivers and has also been used by Waikato Regional Council.
Case study 6.6 – review of methods for calculating loads
Recently, Aqualinc Research Ltd (2014) reviewed eight methods for calculating loads (including the regression methods discussed above) in an Envirolink-funded report prepared for Environment Southland. The purpose of the project was to provide advice on developing a nationally consistent approach to contaminant load calculation and software to calculate loads, by:
- identifying current methods and tools used
- assessing the accuracy and uncertainty of contaminant load estimates produced by these methods and tools.
Based on its review, Aqualinc found:
- Expertise, professional knowledge and judgement are required to evaluate contaminant loads produced. As such, a single load calculation method, or a simple rule-based procedure for selection of methods, may not be appropriate and is not promoted.
- There are existing appropriate tools to calculate loads (eg, SedRate tool and the Horizons Regional Council flow-stratified approach of Roygard et al, 2012), and further significant investment in load calculation software is not warranted.
- Training and support for council staff involved in load calculation should be provided, based on the databases (eg, Hilltop, Tideda) and existing tools (eg, SedRate and Excel) that analysts use.
- Further research into uncertainty of load calculations, its causes and implications should be carried out to provide improved future guidance associated with implementing the NPS-FM, and to provide guidance for future data collection for the purpose of load estimation, particularly for nutrients. (Note that Massey University is currently carrying out some research in this area with Horizons Regional Council.)
Modelling
As discussed above, we recommend that, for accounting purposes, measurement and estimation of loads and concentrations should be used as references against which sources and sinks can be compared. This should provide a robust measure against which other modelled and measured sources of contaminants contributing to that reference load or concentration can be compared. However, if measurement and estimation is not possible for whatever reason, then there are a range of models available to estimate loads and concentrations at both an activity scale and on a catchment basis.
Motu (2013) provides a review of the range of both activity scale and catchment scale models used in New Zealand for modelling nutrient losses from rural land and the concentrations and loads in receiving water bodies. Figure 6.3 outlines the catchment scale models currently used in terms of the areas that they cover and whether they include surface water or groundwater as well. In practice, it is likely that both activity scale models and catchment scale models will need to be used, with the activity scale outputs potentially feeding into the catchment scale model.
Figure 6.3: Classification of catchment scale water quality models in New Zealand
Source: Motu, 2013
Description of the figure
This shows the classification of different catchment scale water quality models used in New Zealand, based on scale (catchment, regional or national) and whether they cover surface and/or groundwater.
Motu (2013) made the point that there is no perfect model. All have their place and limitations, and the choice of one particular model over another will depend on the application and the skill of the modeller. None of the models in figure 6.3 were developed specifically with water contaminant accounting in mind, however, CLUES (Catchment Land Use for Environmental Sustainability) has been developed specifically to model nutrient, sediment andE. coliloads across the country. Recently, the source model developed by eWater[5] in Australia has been used in New Zealand by Horticulture New Zealand, which notes that, unlike CLUES, it is an open source model and able to be readily adapted for specific purposes (C Keenan, Horticulture New Zealand, pers. comm., 9 July 2014).
- [3] For example, www.horizons.govt.nz/managing-environment/state-of-the-environment-2013/, and www.es.govt.nz/environment/monitoring-and-reporting/state-of-the-environment/water-2010/.
- [4] See www.waikatoregion.govt.nz/Environment/Natural-resources/Water/Fresh-water-quality-monitoring/.
- [5] See www.ewater.com.au/
6.4 Establish the current (measured) state
March 2021
© Ministry for the Environment