1 Introduction
1.1 Overview
Groundwater quality is monitored in every region of New Zealand. This report focuses on groundwater quality data collected for routine state of the environment (SOE) monitoring. The SOE monitoring programmes conducted in different regions vary in terms of the number of sites sampled, the frequency of sample collection, the sampling methods, and the groundwater quality parameters measured. Regional councils (and unitary authorities) report regularly on the results of SOE monitoring within their own regions, but with the exception of occasional national surveys of nitrate contamination (Burden, 1982; Bright et al, 1998), SOE data have not been collated to provide a national overview of groundwater quality in New Zealand.
The purpose of this report is to provide an assessment of the current state and trends in groundwater quality at a national scale. This assessment is based on data from regional council SOE monitoring programmes, and from the National Groundwater Monitoring Programme (NGMP) operated by GNS Science. All data relate to the period between 1995 and 2006.
The remainder of this report is divided as follows:
The rest of the Introduction contains a description of the data, data limitations and methods used for data analysis, and then a brief background discussion of the general chemical characteristics of groundwater.
Section 2 gives a summary of groundwater quality status in New Zealand from 1995 to 2006, based on the medians and percentiles of 33 different indicator parameters. Section 3 provides a summary of changes in groundwater quality in New Zealand over the period from 1995 to 2006, based on calculated temporal trends for 33 indicator parameters. Section 4 gives an assessment of the importance of factors that might influence groundwater quality status and trends, such as well depth, aquifer characteristics and surrounding land use
Section 5 then compares the results from this study with previous regional, national and international assessments of groundwater quality, and the final section discusses the results in relation to managing New Zealand’s groundwater resources.
1.2 Methods
Groundwater quality data
This investigation used groundwater quality data from 16 different databases (see Table 1). SOE groundwater monitoring data were extracted from 15 different regional council databases by regional council personnel. In addition, data collected through the National Groundwater Monitoring Programme (Rosen, 2001; Daughney and Reeves, 2005) were extracted from the Geothermal-Groundwater database housed at GNS Science. For the remainder of this report the data from the NGMP sites are grouped together with the SOE data from the relevant region.
Table 1: Sources of groundwater quality data and abbreviations used in this report
|
Abbreviation |
Data source |
|---|---|
|
ARC |
Auckland Regional Council |
|
EBOP |
Environment Bay of Plenty |
|
ECAN |
Environment Canterbury |
|
ES |
Environment Southland |
|
EW |
Environment Waikato |
|
GDC |
Gisborne District Council |
|
GWRC |
Greater Wellington Regional Council |
|
HBRC |
Hawke’s Bay Regional Council |
|
MDC |
Marlborough District Council |
|
MWRC |
Manawatu-Wanganui Regional Council |
|
NRC |
Northland Regional Council |
|
ORC |
Otago Regional Council |
|
TDC |
Tasman District Council |
|
TRC |
Taranaki Regional Council |
|
WCRC |
West Coast Regional Council |
|
NGMP |
National Groundwater Monitoring Programme |
The entire data set used in this study included groundwater quality data from 1,068 SOE monitoring sites, most of which are privately owned wells. Data on the end use of water (eg, as drinking-water, irrigation, industrial use) at each monitoring site were not compiled for this study. The median well depth is 20 m, with a quarter of all wells less than 9m deep and a quarter more than 45m deep. The sites were divided into categories describing aquifer lithology and confinement, based on information provided by regional councils (see Table 2). The sites were also divided into categories describing the surrounding land use and land cover, based on information provided by regional councils and the Land Cover Database (TerraLink, 2007), respectively (see Table 3). Specific details on each site are presented in map form in Appendix 1 and in table form in Spreadsheet 1 (all spreadsheets are on the CD accompanying this report).
This investigation considered groundwater quality data from 23,675 individual sampling events between 1 January 1995 and 31 December 2006. This amounts to an average of 22 sampling events at each site over the period of interest, or roughly one sampling event per site every six months. Some sites were sampled once a month over the entire 11-year period considered in this study, whereas some were sampled just two or three times.
Table 2: Number of sites considered in this investigation, categorised by programme, aquifer lithology and aquifer confinement for each regional council
Table 3: Number of sites considered in this investigation, categorised by surrounding land use and surrounding land cover for each regional council
Across the 16 databases considered in this study, more than 500 different names are used for analytical parameters (see Spreadsheet 2). To facilitate a national assessment of groundwater quality, analytical results from the different databases were condensed into 33 parameter categories (see Table 4), by comparing parameter names and considering ancillary information related to sampling and analytical methods. (This operation assumes that analytical results from different laboratories are directly comparable, even though for some parameters different laboratories might employ different analytical procedures.)
This data compilation also involved combining results relating to:
- “dissolved” and “total” concentrations specific to a particular major, minor or trace element
- “field” and “lab” measurements of pH, conductivity, etc
- various microbiological parameters.
Note that dissolved versus total and field versus lab results are not directly comparable for many parameters (on a site-specific basis). Thus tests to compare separate analyses of dissolved” versus total concentrations and field” versus lab measurements were conducted as part of this investigation (results not compiled). These tests indicated that in most cases the combination analytical results, as described above, would not unduly bias the national-scale assessment of groundwater quality.
Many of the parameter categories defined in this report are represented in the base suite of groundwater chemical determinands for routine SOE monitoring, recently developed by the Regional Groundwater Forum (Environment Waikato, 2006). The base suite includes major cations and anions, nitrate and ammoniacal nitrogen, dissolved iron and manganese, pH, electrical conductivity, and calculated total dissolved solids (TDS), all of which are among the parameter categories considered in this report (see Table 4). The base suite also includes the trace elements boron and arsenic, which it suggests should be measured on at least two occasions and then at regular intervals afterwards if concentrations are greater than 50% of the relevant drinking-water guideline value. Most other trace elements considered in this report are not included in the base suite because they are usually derived from corroded pipework rather than from the aquifer itself.
Table 4: Parameter category names and units, and the abbreviations used for the remainder of this report (see also Spreadsheet 2)
Data analysis methods
An automated spreadsheet program (Daughney, 2005) was used to compute descriptive statistics for each of the 33 parameter categories defined above, based on methods described in detail elsewhere (Helsel and Cohn, 1988; Helsel and Hirsch, 1992; Daughney and Reeves, 2003, 2005, 2006). In all cases, outliers were identified (on a per-site and per-parameter basis) and excluded from subsequent calculations, and if fewer than three results were available for the parameter at the site in question, the calculations were not performed (no results tabulated). The statistics are compiled on a per-site and per-parameter basis in Spreadsheet 1, including:
-
median, calculated using log-probability regression to deal with “censored” results (ie, those reported as being below some analytical detection limit)
-
trend significance and magnitude, quantified using Sen’s slope estimator for all trends that are detectable with the Mann-Kendal test at the 95% confidence interval (positive numbers indicate increasing trends), or tabulated as “N” for non-significant trends.
The site-specific medians and trend magnitudes were then used to compute national-level statistics for each of the 33 parameter categories. For medians, this entailed determining:
- the total number of sites for which a site-specific median had been calculated for the parameter in question
- the percentage of sites at which the median value was “censored” and could only be listed as being less than some detection limit
- the 5th, 25th, 50th, 75th and 95th percentiles
- the maximum
- the median absolute deviation (MAD).
The MAD is a measure of the spread in the analytical results that is analogous to the standard deviation but less subject to biasing by extreme values (Helsel and Hirsch, 1992). Similar national-level statistics were calculated for trends, with the exception of the MAD, and the addition of the minimum and the percentage of sites at which a significant trend could be detected relative to the total number of sites at which a trend assessment could be performed. National-level statistics are compiled in Spreadsheet 3.
The national-level statistics were then used for further analysis, with methods described below in relevant sections of this report.
-
Hierarchical categorisation was used to assign each monitoring site to one of six categories based on current status of groundwater quality (section 2.1) and one of five categories based on trends in groundwater quality (section 3.1).
-
Water quality standards were used as a benchmark for comparison to the calculated median values (section 2.2).
-
The Kruskal-Wallis test was used to assess the influence of various categorical factors (eg, aquifer lithology or confinement, surrounding land use, etc.) on calculated medians and trends (section 4). The Kruskal-Wallis test is similar to the more common t-test and analysis of variance (ANOVA), but it is non-parametric and so less biased by extreme values in the data set (Helsel and Hirsch, 1992).
Data limitations
Representativeness of monitoring sites
Although this study involves the analysis of groundwater quality data from over 1,000 monitoring sites across the country, it is not clear whether or not these sites provide an accurate representation of groundwater quality in New Zealand.
First, this is because many monitoring programmes target contaminated or at-risk regions or aquifers, and so the data set used in this study may lack sites in pristine areas to provide highly relevant baseline (background) groundwater quality data. Other monitoring programmes might specifically target aquifers that are important for water supply. For example, a large proportion of wells in Auckland’s SOE monitoring programme tap into deep confined aquifers used for water supply, and so groundwater quality issues in shallower groundwater systems, if they exist, would be difficult to detect in this investigation.
Second, the source area (capture zone) of groundwater that reaches each monitoring site is usually not known. Thus monitoring sites must be displayed as points on maps, which makes groundwater quality hard to characterise at the scale of an aquifer, a region, or a nation with a limited amount of point data.
A third difficulty is that most monitoring wells in New Zealand have fairly long screened or open intervals, which allow them to extract or pump water from several different flow lines within the aquifer. Any sample from such a well is therefore a composite from flow lines with potentially different ages and different groundwater quality. Little is known about the distribution of flow lines that reach monitoring wells in New Zealand, although it is likely to be site-specific and dependent on well construction and aquifer lithology. Without this information highly contaminated groundwater could go unnoticed if it represents only a small proportion of the composite sample extracted from the well, or if most of the water extracted from the well is old and pre-dates the contamination event.
Site details
Important information about some monitoring sites is either not known or could not be provided within the timeframe for this study. Thus 22% of monitoring sites have unknown aquifer lithology, 53% have unknown aquifer confinement and 30% have unknown surrounding land use (see Tables 2 and 3). Even for sites where this information was provided, it might be known with only limited confidence. Feedback provided by regional council personnel on a draft version of this report suggested that this is often the case with reported aquifer lithology and confinement.
It is also very difficult to supply meaningful land-use information, because it is the land use in the capture zone at the time the groundwater was recharged that has the potential to influence current groundwater quality at any given monitoring site, but the capture zone and groundwater age are unknown for most monitoring sites in New Zealand.
Interdependencies in the data set
The data set used in this investigation has some features that can cloud statistical interpretation. For example, well depth is statistically related to confinement: there are more shallow wells (less than 10 m) in unconfined than confined aquifers. This is to be expected, but it means that any statistically significant relationship between well depth and groundwater quality might in fact be determined by aquifer confinement.
There are weak relationships between well depth and surrounding land use. When testing for the effect of land use on groundwater quality, it is therefore important to consider only the wells within a limited range of depths.
There are also significant relationships between depth and aquifer lithology for the sites used in this study: wells in sand and gravel aquifers are shallower than wells in sandstone, ignimbrite or shellbed aquifers (well depths in other lithologies are intermediate between these two groups).
Finally, the Chi-square test shows that aquifer lithology is statistically related to confinement. Specifically, for the sites considered in this study, sand, gravel and pumice aquifers are most likely to be unconfined, whereas basalt, greywacke, ignimbrite, lignite, limestone, sandstone and shellbed aquifers are most likely to be confined.
Coverage of parameters
The data set used in this investigation did not include some important groundwater quality parameters; for example, pesticide concentrations, which previous work has shown can be present at low but detectable levels in about 10% of shallow wells across the country (Close and Flintoft, 2004). This study also did not consider volatile organic compounds, petroleum hydrocarbons (eg, benzene, xylene), or other potentially important classes of organic contaminants (eg, pharmaceuticals, endocrine-disrupting compounds) simply because these are not routinely analysed in SOE groundwater monitoring programmes.
1.3 General chemical characteristics of groundwater
Before discussing the specific results from this study it is useful to understand the general chemical characteristics of groundwater. A global perspective on this subject is provided by Turekian (1977), Freeze and Cherry (1979), Hem (1985) and Langmuir (1997). In New Zealand, Daughney and Reeves (2005) have determined that groundwater chemistry is governed both by natural processes and by human influence. Natural processes are reflected primarily by changes in total dissolved solids (TDS), cation and anion ratios (ie, water type), and redox state (ie, oxidation-reduction potential). In New Zealand, human influence (including the effects of agricultural activities) is usually indicated by above-background concentrations of NO3-N.
Total dissolved solids
Natural changes in TDS can in some cases be used as a proxy for groundwater source and age. Rainwater is dilute, with TDS around 10 mg/L. Reaction between rainwater and rock results in an increase in TDS, with the global average for rivers being around 120 mg/L (Langmuir, 1997). Groundwater that has been recently recharged from a river may have a similar TDS, especially near the recharge zone. In comparison, recently recharged groundwater that is sourced primarily from rainfall typically has slightly higher TDS (around 160 mg/L), due to the accumulation of salts during passage through the soil zone.
The TDS of groundwater increases with time and distance along the flow path (the global average TDS for groundwater is 350 mg/L; Langmuir, 1997) due to prolonged contact with aquifer minerals and lack of dilution by rainfall. It is not uncommon for older and/or more chemically evolved groundwater to have TDS well over 1,000 mg/L. In some coastal aquifers, high TDS is an indication of saltwater intrusion (for comparison, seawater has TDS of 34,500 mg/L).
Major cation and anion ratios
Natural changes in the concentration ratios of major cations and anions (ie, water type) can also be used to infer the origin and extent of evolution of a particular groundwater (Freeze and Cherry, 1979; Langmuir, 1997). Rainwater usually has Na as the dominant cation and Cl as the dominant anion, because its composition is governed primarily by equilibration with salt aerosols of marine origin. In contrast, river waters usually have Ca as the dominant cation and HCO3 as the dominant anion.
The shift from Na-Cl to Ca-HCO3 dominance is caused by the rainwater reacting with carbonate minerals. Carbonates are common, they dissolve rapidly relative to other minerals, and as little as 1% carbonate in a rock will dominate the chemistry of the solution during water-rock interaction. Groundwater that has been recently recharged from a river will often retain Ca and HCO3 as the dominant cation and anion, respectively. Groundwater that has been recently recharged from rain will often be a Ca-Na-HCO3-Cl type, due to the accumulation of Na and Cl during passage through the soil zone. As a groundwater ages, there is often a gradual transition from Ca to Na as the dominant cation, and from HCO3 to SO4 to Cl as the dominant anion, due to the dissolution of silicate minerals (a slow process) and sulphate- and chloride-bearing minerals (which are quite rare).
Redox state
Natural changes in redox state (ie, oxidation-reduction potential, or EH) control the form and behaviour of many elements (Langmuir, 1997). Recently recharged groundwaters are usually (but not always) oxygen-rich (ie, oxidised) due to contact with the atmosphere. In contrast, it is common (but not universal) for more evolved groundwaters to be oxygen-poor (ie, reduced), due to natural processes of microbial respiration which deplete oxygen. Oxidised groundwater can contain NO3-N, but would not contain significant concentrations of any elements or compounds that only accumulate in groundwater in their reduced forms, such as Fe, Mn, As or NH4-N. Conversely, a reduced groundwater might contain significant concentrations of Fe, Mn, As and/or NH4-N, but would not contain NO3-N. In highly reduced groundwaters SO4 is also usually absent and methane can accumulate.
Human influence
In New Zealand, human influence on groundwater quality is usually indicated by relatively high concentrations of NO3-N, sometimes also with relatively high concentrations of K, SO4 and/or Cl (Daughney and Reeves, 2005). It is certainly possible that human influence can give rise to other types of groundwater quality degradation (eg, contamination by petroleum hydrocarbons or heavy metals), but these are not typically detected in SOE monitoring programmes in New Zealand and so are not discussed in this report. Note that as described above, elevated NO3-N concentrations arising from human influence would only accumulate in an oxidised groundwater; under reducing conditions the NO3-N is removed by microbial respiration (denitrification).
