5. Discussion
5.1.1. Representativeness of monitoring networks
In order to report on the state of New Zealand's rivers in general, we used the REC as a spatial framework for evaluating how well the selected monitoring sites represented New Zealand's rivers as a whole. At the Source of Flow level, the NRWQN was reasonably representative of New Zealand's 5th order rivers and above as classified by the REC. Therefore, we consider that the national state and trend results are a reasonable picture of river water quality in New Zealand rivers.
At the regional scale we found that the Southland (ESWQN) sites do not fully represent the range of rivers classified at the Land Cover level. In particular, the CD/H/HS/P class is poorly represented. This is regrettable as it is a significant class in terms of total length, making up 11% of Southland's mainstem rivers and potentially subject to intensification of land use. Another, environmentally similar class, CW/H/HS/P, which accounts for 11% of Southland's main stem rivers, was over represented by six sites.
5.1.2. Replication
The REC has shown that New Zealand's rivers are environmentally heterogeneous. Variation in environmentally important determinants of river water quality and biological communities can occur at relatively small spatial scales. For example, the classifications for two NRWQN sites on the same river are almost always different at the Land Cover level, and in many cases, at the Source of Flow level. This results in complex patterns of river environments in which sections of river network or 'patches' that are homogeneous with respect to river water quality and biological communities are generally small (i.e. 10-100 km). For example, geology and land cover change along the river network at a scale of 10's of kilometres. The fine-scaled environmental variation, combined with the scarcity of available sites, limited our power to detect differences in state or trends, particularly for the lower level REC classes.
State of environment assessment aims to determine the extent that resource use and management affect patterns in environmental quality. Ideally, our assessments of patterns in environmental state and trends would have included multiple sites that were environmentally similar (i.e. similar Climate, Source of Flow and Geology) but which differed with respect to Land Cover or other management 'treatments' (e.g. point source discharges). However, the variability of higher-level REC factors, reduced replication at the Land Cover level of the REC. In particular, Baseline classes (comprising REC classes with indigenous forest and tussock land cover) were poorly represented by the ESWQN and made up only 28 of 70 NRWQN sites.
5.1.3. Scope
The majority of sites in the NRWQN and ESWQN are located on mainstem rivers. To reduce the potential variability in water quality associated with wide variation in river size, we restricted our analysis to 5th order rivers and larger. Sites on high order rivers may not be representative of smaller streams in the network. However, smaller streams (order less than 5) make up a far greater proportion (approximately 95%) of the total waterway length. Patterns in water quality and biological variables in high and low order streams within networks need to be investigated. Such comparisons will guide interpolation of water quality conditions over river networks. River size does not strongly influence temporal variability patterns in rivers of the northeastern United States (Chetelat & Pick 2001), and the patterns in trends that we report for larger rivers may be transferable to the smaller streams in New Zealand.
5.2. Assessment of state of rivers
5.2.1. General patterns
The analysis of state indicated that the REC's hierarchy of environmental factors explains an increasing proportion of the total variation in river water quality and biological variables. The analysis indicated that climate is a dominant factor influencing many water quality and biological variables at a national level. Nutrients, BOD5 and E. coli concentrations vary among the REC Climate classes, with water quality tending to decrease as rainfall increases. The biological variables showed similar but less consistent relationships with climate. MCI scores generally decreased as rainfall decreased and periphyton cover generally increased as rainfall decreased.
Classifying sites at the Source of Flow level further increased the discrimination of differences in water quality and biological variables among monitoring network sites. Rivers in Dry climate categories and in Low Elevation Source of Flow categories generally have higher concentrations of SIN, SRP, Total Ammonia, BOD5 and E. coli than rivers in Wet and Extremely Wet climate categories and in Hill and Mountain Source of Flow categories. The mean of the maximum monthly filamentous periphyton cover showed a large variation among sites for each REC Source of Flow class. However, the median value of the maximum monthly filamentous periphyton cover for each class shows a pattern that is consistent with the water quality variables. For example, maximum cover was lower in Extremely Wet and Wet climate rivers than in Dry climate rivers, and, within climate classes, cover was lowest in Mountain Source of Flow rivers, followed by Hill and then Low Elevation sites. The mean MCI score showed a similar pattern among REC classes. Mean MCI was highest in Cool Extremely Wet climate rivers, followed by Cool Wet then Cool Dry. At the Source of Flow level MCI scores decreased from Mountain to Hill to Low Elevation rivers.
At the regional scale differences in geology and land cover were reflected in patterns of water quality state. However, at the Source of Flow level there was poor replication. As a result, we made relatively few comparisons of water quality variables for sites with different geology and land cover categories but similar climate and topography. However, the results suggested that poorer water quality was a feature of the catchments dominated by pastoral land cover.
A direct comparison of the effect of differences in land cover was only possible within one class Cool Wet Hill Hard Sedimentary (CW/H/HS). Further subdivision of sites in this class resulted in two land cover level REC classes; Baseline, (2 sites) and Pastoral, (6 sites). Baseline rivers had higher water clarity, and lower SIN and E. coli than rivers in the Pastoral class. Concentrations of total Ammonia and SRP were similar.
Some other differences in water quality variables among REC classes at the Geology level were evident from the regional scale analysis. Mean clarity for all sites in the Cool Dry Low Elevation Soft Sedimentary class (CD/L/SS) were lower than sites with Hard Sedimentary geology (CD/L/HS). This difference may be attributable to higher erosion and weathering rates associated with Soft Sedimentary geology. In addition, the CD/L/SS class has consistently higher mean phosphorus concentrations than the equivalent class with Hard Sedimentary geology. This probably reflects higher phosphorus levels in the underlying soft sedimentary geology.
Rivers in some New Zealand jurisdictional regions are primarily composed of single REC classes, and broad statements about water quality in these regions can be made. Most rivers of the West Coast of the South Island belong to the Cool Extremely Wet climate class and thus, the region has the highest overall water quality. Most rivers of the Northern North Island are in Warm Wet Low Elevation classes with relatively poorer water quality. Most other regions in New Zealand are heterogeneous. For example, most of the South Island's East Coast comprises a mix of three climate classes, all four Source of Flow Classes and various Geology and Land Cover categories.
The REC classes are expected to partition variation in a range of properties such as hydrology, water chemistry and biological communities. Thus the REC, like other regionalisations, is a hypothesis about causes of spatial patterns in rivers. Our analysis of water quality state provided a test of the assumptions used to develop the REC. The patterns we have identified are generally consistent with expectations underlying the theoretical basis for REC (see section 2). Statistical tests could be used to test the significance of differences for given variables among classes. However, because the replication of sites within the REC classes was limited in many cases we have not carried out such tests in this study. One option to increase replication, which is currently being investigated, is to aggregate data from other existing datasets (e.g. regional council datasets).
5.2.2. Evaluating state against guideline
High dissolved nutrient concentrations in rivers are a major issue nationally and in the Southland Region. Guidelines for SIN, and to some extent phosphorous, were often exceeded in the rivers selected for the national scale analyses of state. Guideline values for BOD5 and Total Ammonia were never exceeded. Recommend limits for clarity were exceeded in Low Elevation Source of Flow rivers. At the regional scale, rivers may exceed guidelines at a Land Cover level, whereas higher-level aggregations of rivers meet the criteria For example, Cool Dry Hill rivers (CD/H) are within the clarity and SRP guidelines at the national scale. However rivers in pastoral catchments within this class (i.e., CD/H/HS/P) in Southland exceed these guidelines.
SIN and SRP promotes the growth of periphyton and high periphyton biomass may be detrimental to aquatic animals and reduce cultural, natural character and recreation values. Rivers in many REC classes exceed the SIN guideline. However, SRP must also be available for periphyton growth. In rivers of some classes, SRP concentrations are lower than the guideline. In these rivers periphyton biomass may not become excessive as SRP availability may limit growth. A map of rivers that exceed SIN but are within SRP guidelines, and rivers that exceed guidelines for both nutrients, is shown in Figure 26. Rivers exceeding both SRP and SIN are potentially eutrophic although other factors such as light and periodic flood flows may prevent periphyton biomass from reaching nuisance levels (Biggs 2000). SRP concentrations have shown an upward trend nationally, particularly in two Low Elevation Source of Flow categories (CW/L and WW/L). This suggests that there is a risk of more widespread eutrophication if this trend were to continue beyond the period of analysis (i.e. 1989-2001).
Low Elevations rivers generally had the poorest water quality. Clarity was below the recommended limit of 1.6 meters (a contact recreation standard) in rivers of all low elevation categories, except for rivers in the Cool Extremely Wet climate class. Organic enrichment as indicated by BOD5 was highest at sites with a Low Elevation category. Total ammonia is rarely a problem, but rivers in two Low Elevation, pastoral Land Cover classes in Southland showed high Total Ammonia levels relative to the guideline. This result indicates the potential for ammonia toxicity problems in quite specific types of rivers and small 'patches' in the environment. E. coli concentrations exceeded the guideline in Low Elevation and some Hill Source of Flow classes in Southland. Exceedance is most pronounced where climate is driest, and where the catchment is dominated by pastoral land cover.
REC provides a spatial framework that allows the state of rivers, relative to the nominated guidelines, to be mapped (Figure 26). In general, New Zealand rivers are in relatively good condition, with impaired rivers confined to specific REC classes or individual rivers. Close & Davies-Colley (1990) and Smith & Maasdam (1994) reported similar findings. The lower section of the Tarawera River is in particularly poor condition.
Given the spatial variation in the environmental factors that affect water quality and biological variables, we suggest that expectations for river state should also vary spatially. In addition, different types of river environments are valued for different reasons and therefore conditions that are acceptable in one type of river may be considered unacceptable in another. Therefore, it would be of benefit to derive guidelines that are specific REC classes (Snelder, et al. 2000). The guidelines used here are applied across all REC classes and are, therefore, 'nominal', and have been used for the purpose of comparison. Exceedance of any particular guideline should be interpreted as signalling a risk of effects and not necessarily any actual effects.
5.3. Trends
5.3.1. Spatial patterns in trends
Nation-wide (i.e. using aggregated data for all sites), SRP showed an increasing trend in the period 1989 to 2001. Aggregating sites by Source of Flow classes showed that the trend in SRP was most pronounced in three Low Elevation categories; Cool Wet Low Elevation (CW/L), and Warm Wet Low Elevation (WW/L) and Warm Extremely Wet Low Elevation (WX/L). Increasing nutrient concentrations are consistent with the known effects of intensification of land use (e.g. Parkyn 2002), which tends to occur at the highest rate in lower elevation areas. Thus, the observed trends in SRP provide evidence of effects of intensification of agricultural production or changing land management. However, the finding that BOD5 has decreased in rivers nation-wide is difficult to explain in terms of human induced changes. Reduced BOD5 is consistent with improved management of organic contaminants loads to rivers. However, aggregating sites at the Source of Flow level, showed that BOD5 decreased in river classes that were least likely to have been subject to organic contaminant discharges (e.g. CX/M, CW/M, CX/Lk and CW/Lk). Thus, it is unlikely that the BOD5 trends can be attributed solely to improved water quality management.
Rivers in three Source of Flow classes (CD/L, CD/H and CW/Lk) showed decreasing trends in periphyton cover over the 1989 to 2001 period. Possible reasons for decreasing periphyton include changing hydrological conditions, temperature regimes, light regimes, changing grazer populations, or changing nutrient availability. However, the 1989 to 2001 period was not associated with major changes in mean flow, temperature, or clarity. There was an increasing trend in SRP, which would be expected to lead to increased periphyton. It is possible that changes in climate, in particular an increasing frequency of floods, may be the cause of the periphyton trend, although we have not tested this hypothesis. The reduction in BOD5 could be related to decreasing periphyton as oxygen demand will be partially associated with breakdown of sloughed algal cells.
SRP is taken up by periphyton in rivers, and increased SRP concentrations may be caused by reductions in periphyton biomass. However, while we have seen decreasing periphyton in several Source of Flow classes (CD/L, CD/H, CW/Lk) these are not classes that showed significant increases in SRP (CW/L, WW/L, WX/L). This suggests that the observed increase SRP may reflect differences in phosphorus input to rivers rather than differences in uptake within rivers. This strengthens the evidence that increases in SRP are due to with land use intensification, particularly the associated increase in fertiliser use.
We cannot exclude the possibility that patterns in water quality and biological variables are, at least partly, attributable to climatic variation. We decreased our analysis to the regional scale to minimise the potential variation associated with large-scale climate patterns, and at the same time find sites that shared the same class with respect to Climate, Source of Flow and Geology but which differed with respect to Land Cover. Our analysis detected one trend at the Land Cover level. Nitrate concentrations increased in rivers in the CW/H/HS/P class in the 1995 to 2001 period. This result was consistent the national trend in nitrate in the CW/H Source of Flow class detected from the NRWQN data.
When NRWQN sites were grouped on the basis of land cover we detected significant differences in trends derived by simple regressions of mean annual concentrations between Indigenous and Pastoral land cover categories. These results also provide evidence of decreasing trends in water quality associated with catchments that are dominated by pastoral land use. Even with flow-adjusted NRWQN data submitted to the Seasonal Kendall test, there were statistically significant differences in SRP and Total Ammonia between Indigenous and Pastoral categories with the net difference between the two categories indicating that degradation of water quality was associated with catchments that are dominated by pastoral land use.
Analyses of ESWQN data, grouped by land cover, mirrored the NRWQN trends to a large extent. Trends in nitrate and SRP concentrations differed significantly between Pastoral and Indigenous land cover categories. Thus there is evidence of increasing degradation of water quality in the period 1995 to 2001 associated with pastoral land use, both within the Southland Region and Nationally.
These water quality trends need to be considered alongside the state analysis. We found that the Cool Wet Hill class, which is a large class comprising 21% of New Zealand's main stem rivers (see Table 6), exceed the SIN guideline but comply with the SRP guideline. However, the analysis of trends over the period 1995 to 2001 indicated that SRP was trending upwards in the CW/H class both nationally and within Southland. Thus the increasing trend in phosphorus in this class has the potential to create conditions favourable to the growth of periphyton, possibly leading to nuisance algal blooms.
5.3.2. Climate variability and water quality patterns
Recent work (Scarsbrook et al., in review) has shown strong relationships between water quality variables, including SRP, and measures of climatic state such as the Southern Oscillation Index (SOI). Furthermore, McKerchar and Pearson (1994) showed that data from different parts of New Zealand exhibit quite marked differences in hydrological response to climatic variation such as the SOI. This spatial variation in climate variability has implications for interpretation of results from trend analyses. For example, our results indicate decreasing trends in water quality associated with Pastoral land cover categories. However, the Pastoral sites are heavily biased by Low Elevation Source of Flow catchments. Scarsbrook et al. (in review) showed that the association of SOI and temperature varies with site elevation. Therefore, we cannot exclude the possibility that trends in Pastoral and Indigenous categories observed in this report are influenced, at least partially, by temporal variability in climate.
5.4. Usefulness of the REC
The REC provides a useful framework for analysis and presenting information on state and trends in a spatially meaningful context. Previous state and trend analyses (e.g. Smith and Maasdam 1994; Smith et al. 1996) have been able to report for individual sites and nationally, but these levels are either too fine, or too coarse, to allow managers to obtain a clear regional or national picture of conditions. The REC provides a spatial hierarchy, allowing patterns to be detected at a variety of scales provided data is available.
Another benefit of the REC is the ability to interpolate site-specific data to other sections of the river network in the same class. This allows the 'extent' of issues to be more easily appreciated and thus adds spatial information to state of environment reports. In addition, REC provides a means of evaluating the effectiveness of monitoring networks. Classification of sites may help identify sites with poor water quality that belong to extensive classes, which require greater monitoring effort. Alternatively a large number of sites representing a single class that has low variability among sites may indicate over representation of that class in the monitoring network.
The REC also provides a framework for setting appropriate expectations for river water and biological quality. Previous work has suggested that such an approach could provide a useful structure for water quality planning carried out under the RMA (Snelder and Guest 2000).
