• About this site|
• Site index|
• Contact us|
• Jobs

Lake Water Quality in New Zealand

Acknowledgements

Executive summary

1 Introduction

2 Methods

3 Results and discussion

4 Conclusions

References

Appendix 1: Glossary of common limnological terminology used in this report

Appendix 2: Morphological and Catchment Characteristics of the 121 Lakes

Appendix 3: Maps of Water Quality Parameters

Appendix 4: Map of Water Quality Trends (1990-2006) based on Percent Annual Change (PAC) using the Burns et al (2000) Methodology

3 Results and discussion continued

3.3 Multiple regression and classification

In this section we relate differences in lake water quality to all the possible drivers (eg, climate, land use, lake morphology) to identify the importance of different drivers and how they interact to control water quality. The regression tree analysis in section 3.3.1 uses the water quality data from the 121 lakes as dependent variables to generate groups of lakes with similar water quality, explaining them in terms of the possible drivers of water quality. In section 3.3.2 we look at how water quality differs among classes of lakes in the two existing classification schemes (Irwin, 1975 and Snelder, 2006). Note that in this exercise the classes are specified in advance on the basis of geological formation and environmental parameters, whereas in section 3.3.1 the groups of lakes are generated by the water quality data itself.

3.3.1 Regression tree analysis

We used regression tree analysis to relate TN, TP and chla to climate, lake morphology and land cover variables. Z_SD was excluded from the regression tree analysis because of the large number of lakes without Z_SD data.

Tree analysis chose a seven-leaf tree (ie, one in which the lakes were split into seven groups on the basis of regression) to best explain the water quality variation (Figure 12). The seven groups are groups of lakes that differ significantly in TN, TP and chla concentrations. Variables related to climate, land cover and lake area were those that best explained the variation in TN, TP and chla throughout the tree. Table 7 summarises the number of lakes in each group, the environmental conditions identified by the tree, how the groups differ in water quality, and gives examples for each group. The groups are highly consistent with the patterns observed in the land class and simple regression analyses presented earlier. The highest water quality is in very cold (ie, alpine and far southern) lakes with low pasture, and the poorest water quality is in warm lakes with high pasture. Climate was the most important factor in lake water quality, with lakes in cold climates being identified as significantly lower in trophic level than warmer regions.

Land cover was the next most important factor, with group 3 including lakes in warmer climates with high native cover, and the lowest water quality occurring in groups 6 and 7 (> 50% pasture). The very lowest water quality in group 7 includes the large, shallow lakes at low altitudes with high pasture cover, whereas group 6 includes the smaller pasture lakes. Groups 4 and 5 include those lakes in which both native and pasture land are individually less than about 50% cover. For many of these lakes, exotic forest is the dominant land use. In other cases, pasture or native cover may still be the greatest single land-use category (eg, Lake Waipori, in Otago, with cover of 40% native, 38% pasture and 21% exotic forest).

The latitude branch splits off a subset of the Aupori Peninsula lakes north of Houhora (group 4), in which chla is much lower than would be expected from their TN and TP data. Although the tree has picked latitude for this split, most of the other Aupori lakes are in groups 5 and 6 and it is unlikely that any climatic factor is responsible. The reason for the distinct water quality pattern in these lakes is not clear: their Z_SD ranges from 1.5 m to 6 m, which should allow considerable algal productivity, and no other likely features are evident. These are a cluster of the northernmost lakes in New Zealand, and there may be aspects of local geology leading to high dissolved organic N and P concentrations that are not available to algal productivity.

Table 7: Characteristics and examples of the regression tree groups of lakes with TN, TP and chla as the variables

Notes: n = number of lakes in each group. Total number = 121.

One initially surprising feature of the tree was that depth was not selected for any of the splits. This is because depth is strongly correlated with other parameters that were chosen in the regression (climate, area and land cover). Most of the large, deep lakes are in cold areas, with high proportions of their catchments in native cover. Of the 121 lakes, only 20 had maximum depths over 50 m, none of which had more than 40% pasture or exotic forest cover. Only one of these lakes, the urban Lake Pupuke, had low native cover.

We also experimented with trees for lakes for which we had data for all four main parameters (TN, TP, chla, and Z_SD), which reduced the number of lakes in the regression to 75 and produced a tree with six groups instead of seven (Table 8). It should be noted that the 75 lakes for which we did not have Z_SD included most of the best-quality microtrophic and oligotrophic lakes. Because this tree does not represent the full range of water quality in the 121 lakes, the tree output in Figure 12 and Table 7 is probably more suitable for management purposes, and we have not analysed the output in Table 8 in as much detail. However, it has used similar factors in generating its groups (ie, land cover and climate, but also altitude). Altitude was probably not chosen by the larger tree for the same reason as depth; ie, it is strongly correlated with climate and native cover. The single regressions above also emphasised that land cover, climate and altitude are all important factors for water quality.

Table 8: Characteristics and examples of the regression tree groups of lakes using all four TLI parameters as variables

Notes: n = number of lakes in each group. Total number = 75 m asl = metres above sea level.

The classifications emphasise that large, shallow lakes are highly likely to be supertrophic or hypertrophic, that lowland lakes have worse water quality than upland lakes, and that higher native cover is associated with higher water quality.

3.3.2 Relationship to lake classification

Ministry for the Environment (2006) compared the TLI data from councils across the Irwin (1975) categories and found that dune lakes and peat lakes were most likely to be in eutrophic or higher classes, whereas glacial lakes were most likely to be oligotrophic or microtrophic. As he noted, this is because dune and peat lakes are mostly small and shallow, whereas glacial lakes are usually larger and/or deeper. The mode of lake origin is not itself a direct cause of trophic state; rather, lake morphometry is the important variable within mode of origin that affects susceptibility of lakes to different land uses. Dune lakes include many oligotrophic lakes, because they often do not have surface inflows and so have very low nutrient inputs if their catchments are unmodified. Dune lakes also include some of the most eutrophic examples because they are shallow and poorly flushed, making them susceptible to increased nutrient load.

The Snelder (2006) classification recognises that climate, surface area and depth are important factors controlling lake ecosystems, because of their effect on mixing patterns and phytoplankton productivity. Table 9 compares water quality among the seven national-scale classes. Note that most of the lakes for which we had data were in the large, deep class (19 lakes); the medium, warm class (29); or the small, warm class (62).

Table 9: Median values of the component TLI indices for chla, Z_SD, TN and TP in the seven lake classes

Notes: n = number of lakes in each class. Z_SD data were absent from all but four classes. No estimates were made for small cool and medium cool classes (< 3 lakes).

All four component TLI indices show significant differences between these classes. Table 9 confirms that the large, deep lakes in New Zealand generally have good water quality (median values generally oligotrophic), medium-sized lakes have intermediate water quality, and small lakes have poor water quality. The large, shallow lakes had very poor water quality. Despite the small number of lakes in some classes, there is also a clear tendency in Table 9 for higher water quality in cooler lakes of similar size, consistent with the patterns detected in the regression tree analyses. We found poor water quality in cool lakes with low native and high pasture cover in the regressions; this suggests that higher water quality in cold climates reflects the historically lower human pressures in their catchments rather than any strong resilience to degradation in colder climates. Small lakes in cold regions therefore appear to be equally at risk of degradation to those in warm regions, given similar pressures.

Figure 13 summarises the differences in water quality using TLI. Bearing in mind the limitations of numbers of lakes in the different classes, the pattern of deteriorating water quality on the gradients from deep to shallow, and cool to warm, is evident in this comparison.

Notes: Horizontal lines within boxes show median values, boxes show 25–75% data ranges, whiskers show 5–95% ranges, and pluses outliers outside the 5–95% ranges.

3.4 Relationships between water quality and LakeSPI data

There were 39 lakes for which both LakeSPI and water quality data were available, from three regions (Northland, Waikato and Bay of Plenty). Although the geographic coverage and range of catchment land uses is therefore limited, these lakes included a range of trophic states from oligotrophic to hypertrophic.

There was a highly significant relationship between TLI and the plant colonisation depth for these lakes (see Figure 14). Plant depths in lakes are well known to be strongly correlated with light attenuation in the water, which is captured in the TLI by Z_SD (Schwarz et al, 2000). High nutrient concentrations and their associated high algal biomass can be clearly linked to shallower depth limits in the macrophyte community. Plant depth was also significantly correlated with those factors shown earlier to be related to water quality, especially catchment land use. Figure 14 shows how low native cover in the catchment was strongly related to shallow depths in the plant community.

However, there were no significant relationships between any of the catchment and water quality variables with either the native or exotic condition indices in the lakes. The presence of introduced plant species in New Zealand lakes is changing with time and is largely a function of boat access and boating activity (Johnstone et al, 1985). Native diversity is often affected by pressures such as the impacts of exotic species and water-level management, as well as water quality. These may all be reasons why species composition was not related to land use or water quality in this data set.

3.5 Trends in lake water quality

3.5.1 Long-term comparisons with data from 1983 and 1996

Comparisons with historical data need to be interpreted cautiously. The precision, accuracy and detection limits of analytical methods change over time, as do sampling sites and reporting methods. These caveats need to be kept in mind when looking at Figures 15 and 16 below, especially with the 1983 data, which were compiled from a range of studies carried out with different timeframes and data quality control. In general, the 1983 data are from occasional sampling, usually in summer, whereas the modern data are three-year averages of monthly or quarterly measurements. The 1996 data are more directly comparable with the 2006 data because similar sampling protocols were followed (Burns and Rutherford, 1998).

Nevertheless, a comparison of this type is helpful for developing a general picture of changes in lake water quality over the last 20 years, especially given the long-term nature of change in lake water quality and the relatively small number of lakes that have had regular, detailed monitoring over this time. Although Figures 15 and 16 do not lend themselves to detailed lake-by-lake comparisons or trend analysis, they provide a useful visual overview of how our lakes have changed since the 1980s.

Figures 15 and 16 strongly suggest differences in water quality between the three time periods, with large increases in TN, TP and chla apparent in both comparisons. Note that the axes in Figures 15 and 16 are logarithmic, so many of the changes between the time points are relatively large.

There are 26 lakes in total across the two comparisons. Only eight of these have data for 1983, 1996 and 2006, all but one of which are Bay of Plenty lakes. Because of this regional bias, we have not made direct comparisons between 1983 and 1996 or involving all three time periods. Nevertheless, most of the lakes have undergone large deteriorations in water quality in both comparisons, and these lakes are broadly representative of the range of lakes that have combinations of pastoral catchments, warmer climates and lower altitudes.

3.5.2 Recent trends according to TLI methodology

The 49 lakes in the data set with sufficiently detailed records for trend analysis included lakes in Northland (3), Auckland (7), Waikato (14), Bay of Plenty (12), Manawatu (Lake Horowhenua), Wellington (Lake Wairarapa), Otago (10), and West Coast (Brunner). Table 10 identifies those lakes with significant trends in water quality. In general, it presents a picture of northern lakes decreasing in water quality and more stable water quality further south. Although there was some variability in trends among parameters in some lakes, probable and definitely declining quality were most common in shallow, lowland lakes (eg, Rotokauri, Waikare) or in lakes such as Rotoiti (Bay of Plenty) that have well-documented point source inputs.

No change was detected for most of the lakes in Table 10, although there are two important qualifications to this finding. First, many of these lakes have already become supertrophic and hypertrophic, so additional nutrient run-off is unlikely to increase chla and decrease Z_SD further because algal growth is already nutrient-saturated. Second, it needs to be reiterated that deterioration in lake water quality is a long-term process, and that it is often non-linear (ie, there may be no apparent change in water chemistry for several years after changes in land use, until critical thresholds are passed or the food web structure changes). Some of these lakes may therefore be undergoing deterioration but the current data sets are still too short to reveal this.

Regression tree analysis was conducted to relate Percent Annual Change to climate, lake morphology and land cover. Relatively few lakes showed large, consistent changes in all parameters, and no clear regression trees could be produced following the usual one standard error regression approach. Altitude was the one parameter that could robustly split the trend data, with the lakes less than 50 metres above sea level (masl) having high PAC increases (median PAC > 6% for TN, TP and chla), and lakes with more than 50 m asl generally having no significant PAC.

Table 10: Recent trends in lake water quality for the 49 lakes (listed from north to south) with sufficient data for trend analysis, calculated using the Burns et al (2000) TLI methodology

View recent trends in lake water quality for the 49 lakes (listed from north to south) with sufficient data for trend analysis, calculated using the Burns et al (2000) TLI methodology (large table).

None of the 49 lakes in Table 10 were in alpine catchments, only one (Rotokakahi) was in a predominantly exotic forest catchment, and two (Pupuke and Hamilton Lake) were in urban catchments. Table 11 compares these trends for the remaining lakes in pasture and native forest catchments. Although the number with no change detected is similar in both categories, most of the pasture lakes with no change are already eutrophic, supertrophic or hypertrophic. A map of water quality trends is given in Appendix 4.

Table 11: Comparison of numbers of lakes in Table 10 in native catchments and pasture catchments in relation to water quality trends

Note: Names of lakes showing change are identified. Lake Rotokakahi (exotic forest catchment) was excluded.

The one native catchment lake with definite deterioration is Rotoiti (Bay of Plenty); the reasons for this were discussed earlier. Lakes Taupo and Brunner are two large lakes with predominantly native catchments within which conversion to pasture and increased nutrient run-off have recently been linked to decreased water quality (Gibbs, 2006; Kelly, 2006). In the case of Lake Taupo, other indicators such as hypolimnetic nitrate accumulation and changes in littoral ecology are providing further evidence of the early stages of deterioration. There are examples of both deterioration and improvement in the pasture lakes in Tables 10 and 11. The deteriorating lakes are all shallow Waikato lakes with a well-documented history of agricultural nutrient loading. However, an encouraging feature of Table 11 is the three pasture lakes showing probable improvement. Lakes Okaro and Hayes have catchment management plans in place involving actions such as restoration of marginal wetlands and riparian retirement of grazing activity, which may have contributed to their improvement. In Hamilton Lake and Lakes Whangape and Kereta, recovery of submerged aquatic vegetation has been linked to recent improvements in water quality.

3.6 Nationwide extrapolation of state and pressure

We applied the regression tree of Figure 12 to all lakes in the nationwide database to develop a picture of likely water quality in New Zealand’s full lake inventory (Table 12). Approximately 60% of New Zealand lakes occur in regions with very cold winters and have high native and low pasture cover, and probably retain excellent water quality. However, approximately 30% are likely to have water quality that is poor or very poor. Large, lowland lakes almost all have extremely poor water quality.

Table 12: Distribution of 3,820 lakes in New Zealand larger than 1 ha in each of the water quality groups*

* As defined by the regression tree groups (Figure 12; Table 6).

The Snelder (2006) classification can be used to identify the likely future trends in lake water quality, based on the strong relationships between catchment land cover, size and climate found in the water quality data (Table 13). Note the large proportion of small, warm and small, medium-temperature lakes (73%) that have non-natural catchment classes (exotic, pasture or urban), and the large proportion of lakes in total with non-natural catchment classes (1,871 lakes, or 50% of the total), and that there is a much higher proportion of medium and large lakes than small lakes with native catchments. Table 13 suggests that the water quality of small lakes is likely to be under high pressure, that lowland and northern lakes are likely to be under high pressure, and that many New Zealand lakes have predominant catchment land cover leading to degraded water quality.

Table 13: Number of lakes in each of the seven primary classes of the Snelder (2006) classification with alpine, exotic forest, native and urban catchment classes

Note: Some of the 3,820 lakes for which catchments cannot be clearly defined (mainly small warm lakes) are excluded.

[ Previous|Next