Skip to main content.

Methods

Data sources

The fish data came from the New Zealand Freshwater Fish Database (NZFFD) maintained by the National Institute of Water and Atmospheric Research (NIWA) and contains records of fish distribution for around 100 years . The data has been supplied by many individuals and institutions and in 2008 contained more than 23,000 records. Each entry includes the site location details and the species of fish and large crustaceans found there, using a number of survey methods. The amount of detail varies from only presence/absence and no habitat details, to complete site descriptions and detailed abundance and fish size measures. However, due to the differences in survey methods and measures of abundance all data were converted to presence/absence, as comparison requires consistent levels of data accuracy. To analyse the trends in the database related to land use the River Environment Classification (REC)2 land-cover classes were applied to all sites. The major land-cover classes; pastoral, urban, indigenous forest, scrub, and exotic forest were used.

Statistical analyses

The relationship between land-cover and fish communities were analysed by comparing the mean IBI scores using an analysis of variance Proc ANOVA . ANOVA compares the means between two or more samples; the result is an F–value, which is the test statistic and a P-value, which is the statistical significance of any differences. Temporal trends in fish community structure were analysed by comparing IBI scores over years and decades for all sites and then by individual REC classes to find the land-cover types underlying the differences and trends. Statistical analysis of temporal trends was done using general linear regression models PROC GLM . To visualise these trends, mean decadal scores and variances were plotted. General linear models incorporate a number of statistical models including ANOVA and, when there is just one dependant variable, as in this case, then they can also be referred to as a multiple regression. As with the ANOVA, the F value is the test statistic and the P value is the measure of the statistical significance.

The reaches that had multiple sampling events were assessed for changes in IBI scores over time using a Spearman Rank Correlation (PROC CORR ). The results were reported as the number of significant positive or negative relationships. To further investigate the changes at sites sampled more than once, sites that had been sampled before 2000 – 2007 were compared with the 2000 – 2007 period by counting the number of reaches that had more or less species.

Data limitations and interpretation of IBI scores

The fish distribution data used in this report were not collected expressly for this analysis, rather they are a collection of sites sampled for many reasons by many different operators. Thus, there will be differing levels of sampling intensity and ability of operators; notwithstanding this the large number of sites should override this limitation to some extent. Furthermore, it is likely that sampling efficiency has improved over the 40 years, but the effect of this would be a tendency to increase scores over time. To help get around the shortcomings mentioned above of variable sampling intensity and ability, only presence/absence (p/a) data were used, and abundance data where available was converted to p/a data. However, the limitation of using p/a data is that there is a tendency to underestimate changes to fish communities because the reduction in a species abundance happens long before local extinction.

The use of the land-cover classes for sites that may have changed land-cover over time is a limiting factor that is difficult to quantify. However, sites are less likely to go from pasture into other land-cover classes e.g., scrub or indigenous forest, but the reverse is more likely, and given the recent intensification of farming the change from scrub or indigenous forest land cover into pastoral land cover is possible.

The IBI score can range between 0 for no fish and 60. The IBI was developed using data from the NZFFD for the period 1980 to 2002. The scoring process involves the top 33% of the sites that score the maximum within each of the six IBI metrics. This ensures that the expectations and scoring are reasonable for the area it is developed for (Joy & Death 2004). For New Zealand, 5% of the sites scored the maximum of 60. This is the level recommended by Karr (1981) with the upper species richness line in the original development of the IBI in America. For this national analysis the emphasis was on changes over time and within land-use types rather than what the scores mean in relation to condition, as this is better done at a regional or catchment scale. However, for clarity a higher IBI score indicates higher quality.

 

2 http://www.niwa.cri.nz/ncwr/rec

 

[ |