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

Spatial patterns in state and trends of environmental quality in New Zealand rivers: an analysis for State of Environment reporting

Executive Summary

1. Introduction

2. Methods

3. Results of analysis of physico-chemical and biological state

4. Results of analysis of trends in water quality and biological variables

5. Discussion

6. Conclusions

7. Acknowledgements

8. References

Appendix 1

Appendix 2: State analysis results summary table

4. Results of analysis of trends in water quality and biological variables

We predicted that groups of rivers with similar environmental attributes (i.e. REC classes) would exhibit similar trends for selected variables. Hence, our aim was to describe differences in trends among REC classes, particularly at the Land Cover level, to provide more powerful assessments of spatial extent of trends than is possible using individual sites.

4.1. National trends in the period 1989 - 2001

4.1.1. Water quality trends

Nation-wide (i.e., data aggregated from all 70 NRWQN sites), median concentrations of SRP increased significantly from 1989 to 2001 (Figure 27). Median values of BOD₅ decreased significantly over the same time period (Figure 27). Values for flow, temperature, clarity and SIN did not show significant changes.

The next scale of spatial aggregation was the Source of Flow level. Results of trend analysis for five variables (flow, clarity, SRP, SIN and BOD₅) at this level are summarised in Table 8. Flows during the period 1989 to 2001 did not show a significant change in any of the Source of Flow classes. Clarity showed increasing trends in rivers of two classes (CX/L and WW/L). In addition, the slope of the regression line was positive in 11 of the 13 classes. BOD₅ decreased strongly and significantly in many of the classes. Significant upward trends in SIN and SRP occurred in Low Elevation Source of Flow classes. For example, concentrations of SIN in rivers of the CX/L class show an increasing trend, whereas concentrations decreased significantly in CX/M rivers. Concentrations of SRP increased significantly in rivers in three Low Elevation classes (CW/L. WW/L and WX/L). These differences in SRP trends are illustrated in Figure 28, and indicate that the overall national trend in SRP over the 1989-2001 period is influenced by increases in the two Low Elevation classes. The result may reflect land use or management effects on Low Elevation catchments, which generally contain the greatest concentration of agricultural land.

Table 8: Slopes and p-values for regressions of annual medians on time for flow and four water quality variables for the period 1989 to 2001

View slopes and p-values for regressions of annual medians on time for flow and four water quality variables for the period 1989 to 2001 (large table)

4.1.2. Trends in biological variables

Periphyton cover was measured at 68 NRWQN sites from 1989 to 2001. The Macroinvertebrate Community Index (MCI) was calculated from invertebrate data collected at 59 NRWQN sites from 1989 to 1998.

There was a national trend of decreasing mean annual percentage periphyton cover (Figure 29). However, there is an indication that changes in periphyton cover are not well represented by the linear relationship shown in Figure 29. Despite the significant linear regression relationship the levels of periphyton cover in 1989 and 2001 are similar, and periphyton cover was lowest in 1996.

Analysis of trends for sites aggregated at the Source of Flow level (Table 9) shows decreasing trends in periphyton cover in three classes CD/L, CD/H and CW/Lk. Furthermore, the slope of the regression line in all seven classes is negative, suggesting that the observed national trend (Figure 29) is indicative of changes occurring across a wide range of river types.

The reasons for the strong decreases in periphyton cover are not clear. We note that SIN showed an increasing trend in the time period 1989 to 2001 in the CD/L class. Levels of SIN in the CD/L class are also very much higher than guidelines (Figure 8). This tends to discount improvements in management of nutrient discharges as a possible reason for the downward trends in mean periphyton cover and suggests that other factors, such as the frequency of bed scouring flows may be responsible for the observed trends.

Table 9: Regression line slopes and p-values of mean periphyton cover versus year for each of the Source of Flow classes

Bold = Slope of regression is significantly different to 0.

The nation-wide median value of MCI did not show a significant trend for the period 1989 to 1998 (Figure 30).

At the Source of Flow level, a positive trend in MCI values was observed only in the CW/M class (Table 10). This class is unlikely to be affected by changes in land use or management suggesting that the invertebrate trend may be induced by natural temporal variability.

Table 10: Regression line slopes and p-values of median MCI versus year for each of the Source of Flow classes

Bold = Slope of regression line is significantly different to zero.

The significant decreases in BOD₅ and periphyton cover in rivers of some Source of Flow classes would be expected to be associated with increasing MCI scores. However, the only river class that has a positive trend in MCI (the CW/M class) had the weakest trend in BOD₅. Decreasing trends in mean periphyton cover were detected for three Source of Flow classes but these classes do not show corresponding trends in MCI scores. Thus, the MCI trends show no clear relationships with the trends observed for water quality or periphyton and the reason for changing invertebrate community composition over time is unclear.

4.2. National and Regional trends in the period 1995-2001

The trend analysis reported above indicated that some water quality and biological variables showed statistically significant changes from 1989 to 2001. However, some trends were observed in river classes where anthropogenic influences should be minor. For example, BOD₅ showed decreasing trends in many classes including Mountain Source of Flow classes, which are dominated by indigenous land cover. These results indicate that at least some of the observed trends may be influenced by natural variability. This makes the linking of trends to changes in environmental management difficult.

Variation in flow at the time of water quality sampling is a potential confounding effect in the analysis and interpretation of trends. For example, an increasing trend in SRP concentrations may reflect increasing inputs of the nutrient, or a reduced dilution associated with decreasing flows. The process of flow adjustment minimises this potential confounding effect.

Another natural source of variation is associated with rainfall and temperature patterns driven by large-scale climatic phenomena (e.g. El Nino Southern Oscillation). It is known that such climate features affect different parts of New Zealand in different ways (Salinger & Mullan 1999). A pragmatic approach to dealing with such large-scale climate variation is to zoom to smaller spatial scales where climate should be more homogenous. For example, climatic variation could be controlled by selecting sites on rivers that are identically classified at the Source of Flow (and Geology) level, while providing for replicate sites in both impact (i.e. Pastoral) and non-impact (e.g. Indigenous Forest) Land Cover categories.

We have reduced the spatial scale of the trend analyses by using water quality data from the Southland region (ESWQN). The available data covers the period 1995 to 2001. This reduced temporal coverage reduces the power to detect trends, particularly when using regression analysis of annual median values. Hence we have used the non-parametric Seasonal Kendall test to detect trends in the monthly water quality data at individual sites. This test also allows for the removal of the influence of varying flows.

To provide comparisons of trends at national and regional scales we have also applied the Seasonal Kendall test to the NRWQN data over a comparable time period (1995-2001). This also provides a more powerful assessment of national trends than was possible with the linear regression analysis.

4.2.1. Seasonal Kendall test of national trends (NRWQN) for period 1995 -2001

The median slope of flow trends at the national scale (RSKSSE = -4.96; n=70) was significantly different from zero (binomial test), indicating that the period 1995 to 2001 was characterised by declining river flows over much of the country (Figure 31). This trend emphasises the need for flow adjustment of water quality data. Hence we report only flow-adjusted trend results for water quality data.

For flow-adjusted data there were no nation-wide trends observed for any of the selected variables (Table 11). However, concentrations of Nitrate, SRP and Total Ammonia increased significantly in rivers of the CW/H class. Nitrate also increased significantly in CW/L rivers.

Table 11: Median slope (RSKSSE) of flow-adjusted data aggregated for all 70 NRWQN sites and by Source of Flow class

4.2.2. Seasonal Kendall Test of regional trends (ESWQN) for period 1995 -2001

Mirroring the pattern observed at the national scale, flows in Southland showed a decreasing trend over the period of study (median RSKSSE = -5.54; n=21; p <0.05). The median slopes for all Southland sites (Table 12) showed similar patterns to the nation-wide slopes (Table 11). In both datasets the median slopes for clarity, Nitrate and SRP were positive, and the median slope for Total Ammonia was negative.

Reduced spatial heterogeneity allowed us to zoom our analyses to the REC Land Cover level in Southland. The only trend observed was of increasing Nitrate concentrations in rivers of the CW/H/HS/P class.

4.3. Aggregation by Land Cover category

To further investigate trends associated with anthropogenic effects we analysed water quality data grouped by dominant catchment land cover (Figure 32) without including higher-level factors such as climate and source of flow. This was done to increase the level of replication. Only two land cover groups were designated; Pastoral and Indigenous, which included tussock & indigenous forest. This approach meant that we were better able to detect trends that may be attributable to land use or land management, but reduced our spatial resolution.

We analysed trends in the two land cover groups at both the national (NRWQN) and regional (ESWQN) scales. The 70 NRWQN sites were grouped into 38 pastoral sites and 28 indigenous sites. Two methods of trend analysis were used at the national scale. Linear regression analysis provided analysis of trends for 1989 to 2001. RSKSSE for flow-adjusted data were used to analyses trends for the 1995 to 2001 period. For the regional dataset (ESWQN) we only compared median trends in flow-adjusted data (1995-2001). The ESWQN sites were grouped into 5 indigenous and 16 pastoral classes.

4.3.1. National dataset

Clarity showed a significant linear increase at indigenous sites over the period 1989 to 2001, but did not change significantly at pastoral sites (Figure 33). SRP concentrations increased significantly at pastoral sites, Total Ammonia concentrations decreased significantly at indigenous sites and nitrate concentrations did not change significantly.

Results for the comparison of mean slopes (RSKSSE for flow-adjusted data), indicated that trends in clarity were similar in both classes (Figure 34). Slopes for SRP concentrations were significantly greater at sites dominated by pastoral land cover than those dominated by indigenous land cover. Trends in Total Ammonia also differed significantly with land cover. Sites under indigenous land cover showed a negative mean slope, whereas sites in pastoral land cover had a mean slope close to zero. Mean RSKSSE values for nitrate were similar under both land cover classes.

4.3.2. Regional Dataset ESWQN

The mean of the slopes (RSKSSE for flow adjusted data) for SRP and nitrate concentrations differed significantly for sites under pastoral and indigenous land cover (Figure 35). In both cases results indicate a worsening of conditions in pastoral streams relative to streams under indigenous land cover. Differences in Total Ammonia and clarity were not statistically significant.