Skip to main content

Relationships between flow and river water quality monitoring data

Ton Snelder, Tim Kerr, LWP Land Water People
Auckland Council Research and Evaluation Unit, RIMU
Publication date:  

Relationships between flow and river water quality monitoring data and recommendations for assessing NPS-FM attribute states and trends

Appendix B: Hydrographs

Appendix C: Flow distributions


Extract from the Executive summary

Assessment of water quality state and trends are requirements of Section 35 of the Resource Management Act (RMA: NZ Government 1991) and the National Policy Statement for Freshwater Management 2020 (NPS-FM). The NPS-FM defines certain compulsory water quality variables to be attributes of the values of ecosystem health and human contact, the details of which are set out in the National Objectives Framework (NOF). NOF numeric attribute states are evaluated from water quality observations obtained by water quality monitoring and are generally percentiles of water observations (e.g., 50th percentile (median), 95th percentile). For each attribute, the NPS-FM also defines categorical attribute states, which are derived by assigning numeric attribute states in four (or five) “NOF bands”, which are designated A to D (or E).

The NPS-FM requires that the baseline attribute states are established as an initial step in the planning process and that attributes are used as a basis for setting target attributes states. The NPS-FM also requires that the condition of water bodies (the current attribute state) is systematically monitored and reported, and that action is taken where monitoring indicates deteriorating trends.

Flow influences water quality in rivers and streams across a range of timescales and therefore has an impact on attribute state and trends. Therefore, flow is often monitored at water quality monitoring sites and flow data is important supplementary information in the analysis of water quality state and trends. It is important for Auckland Council (AC) to consider how flow influences attribute state and trends, how to incorporate information about flow into assessment of state and trends and to ensure that continued monitoring of water quality and flow is consistent with the requirements of the NPS-FM. It is also important for AC to consider whether assessment of state and trends is consistent with the requirements of the NPS-FM and to be clear about the uncertainties associated with these assessments. AC therefore commissioned this study to provide guidance with respect to how to account for the influence of flow in both the sampling and analysis of water quality data and the implications of this specifically in relation to the requirements of the NPS-FM.

This study undertook a series of analyses of water quality and flow data associated with 15 sites and 10 water quality variables (of which five are NOF attributes) in the Auckland region. These analyses and the salient results are as follows:

1. Attribute state is estimated from water quality observations pertaining to an “assessment period”. AC samples water quality monthly, and in this study, the assessment period was five years, which are consistent with current practice. We assessed the current attribute state from the observations and estimated the precision of these assessments. Limited precision means the assessed state is not exact and arises because the observations represent a finite sample of the population (i.e., a subset of all possible water quality observations). The 95% confidence interval for the assessed state of some NOF attributes often extended over two, three or even four NOF bands.

2. We refer to instantaneous flow as the flow in a river or stream at the time the water quality was sampled. In this study the instantaneous flow was quantified by the mean daily flow observed at a nearby river flow gauging station. The flow regime refers to the characteristics of flows at longer than instantaneous timescales, including weeks, months and years. Flow regimes can be characterised by many statistics such as mean and median flows, variability, and seasonality. Variation in these types of statistics at timescales of weeks, months and years are all measures of flow regime variability. A drought year, for instance, will feature more frequent low flows than the long-term flow regime. This study showed that there is marked flow regime variation associated with different five-year assessment periods. Because flow influences water quality in rivers and streams, flow regime variation can cause differences in attribute states between assessment periods. This leads to uncertainty in state assessments that is additional to imprecision, and which is unquantified. We refer to this uncertainty as type A. Unquantified uncertainty of type A means there will be differences in assessments between assessment periods (such as between baseline state and current state) that is driven by flow regime variability that would occur even if there were no changes in the anthropogenic pressure in a catchment.

3. Water quality measures in rivers and streams are influenced by the instantaneous flow rate (i.e., discharge at the time the sample was taken). The strength of the relationship between observations and instantaneous flow differs across variables and sites and is associated with differences in the underlying mechanisms of mobilisation (“wash-off”) and dilution of the contaminants. Water quality observations need to be unbiased with respect to instantaneous flow if they are to represent the true attribute state. We found that the distribution of flows associated with AC’s water quality observations were not significantly different to the full flow distribution and can therefore be regarded as unbiased with respect to flow.

4. The relationships between water quality observations and instantaneous flow are commonly represented by bivariate (i.e., observation - flow) statistical models. These models are used in trend assessment in a statistical treatment known as flowadjustment. The purpose of flow-adjustment is to remove the confounding effect of flow so that the pattern of interest (the relationship between the observed water quality observations and time (i.e., the trend) can be more confidently inferred. This study showed that the definition of models describing observations - instantaneous flow is subjective and therefore there are unquantified uncertainties that arise due to procedural choices around flow adjustment that are likely to be made by individual analysts.

5. Trend analyses for all site variable combinations were used to demonstrate that procedural choices made in association with flow-adjustment have an appreciable influence on the assessment. These differences represent an unquantified uncertainty that is associated with trend assessment that we refer as unquantified uncertainty of type B.

6. We undertook rolling trend analyses with assessment periods of differing duration (5, 10 and 20-years) and the starting year incrementing by one year. These analyses indicate that site trend direction tends to oscillate (i.e., trend directions can reverse from increasing to decreasing over short time periods). The length of the reversal time decreases with decreasing trend period duration. This indicates that short-term trends (e.g., 5 and 10-year duration) are likely to be strongly influenced by flow regime variation. This was true even when trends were flow-adjusted. We showed that these reversals are associated with flow regime variability by employing a model that combines the entire flow record with the water quality observation know as Weighted Regression on Time, Discharge, and Season (WRTDS). This result indicates that water quality trends are at least partly associated with flow regime variation. The oscillations in the trends are evidence that the flow regime variation, and associated water quality variation, is partly attributable to quasi-periodic climate variation such as the El Niño- Southern Oscillation (ENSO). It is important to emphasise that this study has shown a link between water quality state and flow regime variation but did not directly investigate climate variation. However, there is direct evidence of the link between ENSO and water quality variation in New Zealand at interannual timescales. Therefore, when this report refers to the impact of “flow regime variability” on water quality, it is appropriate to consider that climatic variation is at least partly involved. ...

Appendix B: Hydrographs

Appendix C: Flow distributions

December 2022