State and Trends in the National River Water Quality Network (1989-2005)

• Executive Summary
• 1 Introduction
• 2 The National River Water Quality Network
• 3 Water Quality State (2005)
• 4 Changes in Water Quality State (1989-2005)
• 5 River Water Quality Trends (1989-2003)
• 6 Discussion
• 7 References

5 River Water Quality Trends (1989-2003)

Monthly water quality data from 77 sites was analysed for trends in individual parameters using Seasonal Kendall tests on flow-adjusted data. Detailed trend analysis of the NRWQN dataset is carried out every five years with the period 1989-2003 constituting the 15-year analysis. Trend analysis was carried out using S-ESTREND (Slack et al, 2003), a software package developed by USGS as an add-on to S-Plus 6.1. Flow adjustment used LOWESS smoothing with a 30 percent span. The Seasonal Kendall Slope Estimator (SKSE) was used to represent the magnitude and direction of trends in flow-adjusted data. Values of the SKSE were relativised by dividing through by the raw data median (RSKSE), allowing for direct comparison between sites.

The statistical significance of trends at the national scale was determined using a binomial test of the hypothesis that the true proportion of upward (or downward) slopes was half. If this hypothesis was rejected (P < 0.05), a national trend for the period was inferred. This analysis used slope values from all sites rather than just sites that returned a statistically significant Seasonal Kendall test.

The NRWQN dataset is sizeable. For any one parameter at a site the number of samples (N) can be as high as 180. As with most statistical analyses, the null hypothesis may be rejected more often when sample size is large, even when differences or changes are small (and especially when variability is low). In some cases, a small change may be assigned statistical significance but have little meaning in an ecological or management sense. Therefore, we need to protect against making inferences where the magnitude of trends is very small. To address this we have sought to identify a "meaningful" trend that is likely to be relevant in a management sense. In the figures below, we have summarised results of trend analysis in the following terms:

i. no significant change - the null hypothesis for the Seasonal Kendall test was not rejected (ie, P > 0.05)

ii. significant increase/decrease - the null hypothesis for the Seasonal Kendall test was rejected (ie, P < 0.05)

iii. meaningful increase/decrease - the null hypothesis for the Seasonal Kendall test was rejected (ie, P < 0.05) and the magnitude of the trend (SKSE) was greater than one percent per annum of the raw data median (ie, RSKSE > 1 % yr^-1).

Note that the choice of one percent to infer a meaningful increase/decrease is purely arbitrary. It may be possible to develop more rigorous indicators of actual versus statistical significance, but that is outside the scope of the current report.

There were few meaningful trends in water temperature over the 1989-2003 period, and significant increasing and decreasing trends occurred at only nine and six sites respectively. At the national scale there was no significant trend in temperature (binomial test; P = 0.181). The median RSKSE for temperature was 0.02% year^-1.

There were significant increasing trends in conductivity around the country, with Ohinemuri (HM6) and Hakataramea (TK5) rivers showing meaningful increasing trends. There was a significant increasing national trend in conductivity over the period 1989-2003 (binomial test; P < 0.001), although the median RSKSE was only 0.15% year^-1. This suggests that the changes are relatively minor and the cause is not known at this time.

There were no meaningful increases or decreases in pH over the period 1989-2003, but there were a large number of statistically significant decreases. At the national scale, there was a significant decreasing trend in pH (P < 0.001) but the magnitude of this change was very low (median RSKSE = -0.04% year^-1).

Increasing and decreasing trends in dissolved oxygen were observed throughout the country but no trends were large enough to be defined as meaningful. There was no significant trend in % DO at the national scale (P = 0.127) and the median RSKSE was only -0.02% year^-1.

A large number of meaningful trends in water clarity were seen at sites throughout New Zealand. Decreasing clarity was observed at northern sites whereas increasing clarity was a feature of central and southern sites. Overall, there was a significant national trend of improving water clarity during 1989-2003. Median RSKSE was 0.56% year^-1.

Meaningful increases in NO_x-N were observed at 26 sites during 1989-2003 whereas meaningful decreases were seen at 12 sites. However, at the national scale there was no statistically significant difference in the ratio of positive:negative slopes (P = 0.055). Median RSKSE was 0.47% year^-1.

Most sites showed meaningful decreases in NH₄-N during 1989-2003. The lower Mataura, Manawatu, Tarawera, Waihou and upper reaches of Mangakahia and Kawarau rivers all showed meaningful increases. Overall, ammoniacal nitrogen showed a significant decreasing trend at the national scale during 1989-2003 (P < 0.001). Median RSKSE was -4.4% year^-1.

There was a strong increasing trend in total nitrogen at the national scale during 1989-2003 (P < 0.001). This is consistent with results of moving median trends (Section 4). Median RSKSE was 0.98% year^-1. Increasing trends in TN were particularly evident in the South Island where 25 of 33 sites showed meaningful increases.

Meaningful increases in DRP tended to be observed in the lower reaches of rivers. Overall, there was a strong national trend of increasing DRP concentrations during 1989-2003 (P < 0.001). Median RSKSE was 0.72% year^-1. This result contrasts with the relatively weak trends observed in moving medians (Section 4). One explanation is that DRP may have decreased in 2004-05, and this could be tested using formal trend analysis.

Meaningful increases and decreases in TP were observed at sites throughout New Zealand. Increases tended to be concentrated in the South Island and decreases in the North Island. Overall, there was a significant national trend of increasing TP concentrations during 1989-2003 (P = 0.001). Median RSKSE was 0.44% year^-1.

BOD₅ showed consistent strong decreasing trends. Indeed, all sites showed negative values for RSKSE and, as a result, there was a very strong national-scale decreasing trend (P < 0.001). The median RSKSE was -2.14% year^-1. Note: BOD₅ analyses were suspended for 74 of the 77 sites in July 2002. Rangitopuni (AK2), and lower Tarawera (RO2) and Manawatu (WA9) rivers continue to be sampled monthly for BOD₅.

5.1 Summary

Data shown in the maps in the previous pages were summarised in a box plot (Figure 5.12) to show the relative levels of change observed in different parameters and highlight parameters where national-scale trends were observed.

National-scale trends were detected for pH and conductivity but the magnitude of these changes across all sites was small (Figure 5.12). This suggests the changes may not be a significant concern for water resource managers, although it should be noted that pH is measured on a log-scale, so even relatively small changes could have significant ecological effects. In contrast, the national trends observed in TN, DRP and TP involve some sizeable increases (ie, > 5% year^-1 at some sites; Figure 5.12). Such changes may signal marked deteriorations in water quality over relatively short periods and, as such, justify the attention of water resource managers. Changes in NO_x-N were also large at some sites although there was no statistically significant aggregate national trend for this parameter.

Note: Data covers period 1989-2003. Boxes in red are of concern (ie, indicate deteriorating water quality) whereas boxes in blue indicate improving water quality.

5.2 Links between trends and land use

Correlations between land use and trend magnitude were calculated using both the SKSE and RSKSE values. It was felt that reporting the RSKSE alone might hide some important associations. For example, sites with high median NO_x-N concentrations may have large SKSE values but, when divided by large median values, the reported trend magnitude might be very small.

With respect to SKSE values, highly and very highly significant positive correlations were observed for conductivity, NO_x-N, TN, DRP and TP (Table 5.1). Significant positive correlations were also found between the magnitude of RSKSE values for DRP and NH₄-N and percent pastoral land cover in the contributing catchment. Weaker positive correlations for RSKSE were observed for conductivity and NO_x-N. Dissolved oxygen and pH trends were weakly and negatively correlated with pastoral land cover for both RSKSE and SKSE values.

In Table 3.1, associations between percent pastoral land cover and current water quality state (2005) were reported. Table 5.1 reports on associations between percent pastoral land cover and the magnitude of change over time (1989-2003). Despite the differences in time periods, the information in these tables provides an indicator of changing pressures on river ecosystems. For example, high DRP concentrations are associated with sites draining highly developed catchments, and for the period 1989-2003 the magnitude of increases in DRP was strongly associated with percent pastoral land cover in the catchment. What this tells us is that streams draining highly developed catchments are coming under increasing pressure. While it is not possible to associate cause and effect with the data reported here, the patterns observed would be consistent with the increasing use of fertilisers associated with land-use intensification.

Table 5.1: Correlations for associations between percent pastoral land cover and values of SKSE and RSKSE at 77 NRWQN sites

Parameter	SKSE	RSKSE
Temperature	0.19	0.20
Conductivity	0.47***	0.40***
pH	-0.28*	-0.28*
Dissolved oxygen	-0.27*	-0.27*
Visual clarity	-0.26*	-0.11
Oxidised nitrogen	0.30**	0.23*
Ammoniacal nitrogen	0.29*	0.68***
Total nitrogen	0.35**	-0.01
Dissolved reactive phosphorus	0.59***	0.48***
Total phosphorus	0.31**	0.18

Note: '*' = significant (P < 0.05); '**' = highly significant (P < 0.01); '***' = very highly significant (P < 0.001).

Previous|Next