Abstract
M.Sc.
The research question of this study was whether the two classification procedures namely a)
the a priori method based on the regional geographic site attributes to classify sites and b)
the a posteriori method, based on a multivariate analytical approach of the macroinvertebrate
data, generated an analogous categorization of biomonitoring sites in the Marico River
catchment. It was hypothesized that the two methods did not place the sites within the same
categories, thus the aim of this study was to establish how the classification of sites differed.
The objective was to comparatively analyse the site classification results using the two
approaches and provide alternative resource unit delineation for river management
purposes.
In the National River Health Programme the a priori approach, combining ecoregions and
longitudinal zonation, is used as a standardized site classification and resource delineation
tool. In this study the a priori approach consisted of superimposing a GIS layer of the Marico
River catchment over the ecoregion layers. This was done separately for level I and level II to
identify and delineate the ecoregions in the study area. The study sites were then
superimposed on the resulting ecoregion maps to derive their site classification at level I and
II separately. In this study the Department of Water Affairs’ 1: 500 000 GIS layer of the
longitudinal zones was used on which the study sites were superimposed. The coarse scale
of the GIS layer excluded the longitudinal classification of some sites located on tributaries
and these sites were manually characterised using diagnostic channel characteristics
including reach gradient and habitat features. The results obtained from the Marico River
catchment indicated that difficulties existed with this combined regional classification
approach. The difficulties included a) the indistinguishable main attributes of the ecoregion
diagnostic features limiting field verification and b) the need for manual longitudinal
classification of some tributaries due to scale exclusions.
The a posteriori that consisted of a multivariate approach in this study was based on
multivariate analysis which clustered sites into groups based on their macroinvertebrate
assemblage composition. Macroinvertebrate data collected from 2005–2010 according to the
standardized SASS5 methodology was analysed. Multivariate taxonomic composition
analysis of macroinvertebrate data formed the basis of the analyses of the regional
geographic site attributes and the following regional geographic site attributes were
analysed: Ecoregion level I, Ecotones, Longitudinal zonation and Combined E-regions with
longitudinal zones. The site attributes were assessed separately following the same
multivariate procedure. The macroinvertebrate site data were presence/absence transformed
using PRIMER Version 6 and the Bray-Curtis similarity co-efficient was applied. Hierarchical
agglomerative clustering, using group-average linking and NMDS were used to examine and
display similarities between sites. The NMDS stress coefficients were used to assess the
agreement of the respective ordinations. The site separation according to the respective
classification attributes was then assessed by the ANOSIM R statistic. Pairwise comparisons
were used to advise where the main between-group differences or similarities arise.
Multivariate analyses did not conclusively indicate that macroinvertebrate assemblages
clustered analogous to the ecoregion level I a priori approach. Transitional zones or
ecotones were not considered as part of the original ecoregion level I or II delineation
methodology. The inclusion of ecotones was investigated in this study as a refinement option
for the ecoregion delineations. Multivariate analyses confirmed the necessity of including
ecotones in the ecoregion delineations. The combination of ecoregion and ecotone
delineations was introduced as a new concept namely E-regions. Multivariate analyses did
not provide conclusive evidence that macroinvertebrate assemblages cluster according to
the longitudinal zone a priori site classification approach.
The most valuable results from the macroinvertebrate a posteriori analyses and longitudinal
zones showed the significant differences between the Mountain Stream and Lowlands River
zones; the Lowlands River and Source zones and the Mountain Stream and Source zones,
indicating that these longitudinal zones should be managed as separate resource units. The
combined E-regions with longitudinal zone multivariate analyses indicated no apparent
clustering of sites. The pairwise tests indicated that four of the combined classes could be
managed as a unit and that eight should not be combined but it was decided not to interpret
the pairwise results further as variables linked to ecoregions, E-regions or longitudinal zone
classification, including the lack of diagnostic features for ecoregions and environmental
variables such as flow variation could influence the combined classification. The derived
resource unit delineation for management purposes was based on the results of the a
posteriori analyses and resulted in two significant groups, separating the upper reaches,
which includes Source and Mountain zones from the lower reaches, which includes Lower
Foothills and Lowlands Rivers. These two significant groups should thus be managed as
separate resource units.
As hypothesized, the a priori method where site classification and resource unit delineation
procedures was based on regional geographic site attributes and the a posteriori approach
using aquatic macroinvertebrate multivariate analyses did not generate an analogous
categorization of biomonitoring sites in the Marico River catchment. The hypothesis that
stated: “The a priori site classification results are inconsistent with the a posteriori clustering
based on macroinvertebrate assemblages” was thus accepted. The a priori approach per se
did not result in homogenous units for river monitoring, ecological management and
conservation planning, specifically based on macroinvertebrate assemblage analyses. The a
priori approach should thus be refined with a combination of field verification and the a
posteriori approach to delineate resource units for the Marico River catchment.