mao angua amis - adu.org.za · pdf filemao angua amis msc. conservation biology programme, ......
TRANSCRIPT
ASSESSING THE ECOLOGICAL INTEGRITY OF FRESHWATER
ECOSYSTEMS IN SOUTH AFRICA
By
Mao Angua Amis
MSc. Conservation Biology Programme,
Percy FitzPatrick Institute of African Ornithology,
University of Cape Town
and
GISI Conservation Planning Unit
Kirstenbosch Research Centre
South African National Biodiversity Institute
Report submitted to the Council for Scientific and Industrial Research (CSIR)
February 2005
The copyright of this thesis rests with the University of Cape Town. No
quotation from it or information derived from it is to be published
without full acknowledgement of the source. The thesis is to be used
for private study or non-commercial research purposes only.
Univers
ity of
Cap
e Tow
n
ABSTRACT
Freshwater ecosystems are highly imperiled the world over, yet conservation planning is
mainly focused on terrestrial and marine ecosystems. One of the reasons is that, few
criteria exist to assess the ecological integrity of rivers for conservation planning.
Another cause for the disparity between terrestrial and freshwater conservation planning
is that, data for assessing ecological integrity of rivers generally have limited
geographical coverage. Here, I use a fine-scale dataset from South Africa to explore
how well measures of ecological integrity can be predicted from readily available
remotely sensed data. I built a spatial statistical model that uses broad land use
variables for predicting the ecological integrity of rivers (subdivided into riparian and
instream integrity). I also tested the importance of the spatial scale of land use variables
in predicting ecological integrity in order to identify what scale such predictive models
should be built. Results showed that riparian and instream integrity of river systems
could be reasonably predicted, although riparian integrity was more accurately predicted
than instream integrity. The total area under natural cover is the most significant
variable for assessing riparian integrity. Riparian integrity is best predicted by land use
activities at the catchments level rather than more locally. This GIS-based model
provides a fine-filter approach to supplement landscape-level conservation plans of river
systems. The model represents a significant contribution towards the monitoring
component of the River Health Program (RHP), which reports on the state of rivers in
South Africa.
KEYWORDS
Freshwater ecosystems, ecological integrity, conservation planning, predictive modeling
INTRODUCTION
The conservation status of freshwater ecosystems worldwide is poor and declining fast,
with rivers and wetlands among the most threatened ecosystems (Vitousek et al. 1997,
Revenga et al. 2000). This trend is attributable to aquatic systems having being
severely altered by human activity (Moyle & Williams 1990, Jensen et al. 1993).
Although a comprehensive global assessment of the status of freshwater
biodiversity is difficult to carryout (Revenga et al. 2000), the few estimates that exist are
shocking. About 30% of freshwater vertebrate species in the world have become
extinct, threatened or endangered (World Conservation Union 2000). The situation is
even more serious for other freshwater faunal groups (Abell 2002). The challenge to
conservation biologists is to acknowledge the freshwater biodiversity crisis, and to
reflect it through focused research on topical issues in freshwater. Current research
does not seem to however, reflect the concern for this crisis (Abell 2002).
Despite their obvious importance and the seriousness of the threats that they
face, freshwater ecosystems remain poorly understood and inadequately represented in
biodiversity assessments (Higgins 2003). One reason for the disparity between
freshwater conservation and terrestrial conservation is the lack rapid assessment
techniques of ecological integrity of freshwater ecosystems. Assessment of river
ecological integrity is resource intensive, and so there is a dearth of information on
many freshwater ecosystems. Those data that are available for conservation planning
are often too coarse for fine-scale planning.
In South Africa, the status of freshwater ecosystems matches trends worldwide,
with recent assessments showing that 44% of South Africa's mainstem rivers are
3
critically endangered (Nel et al. 2004). Sixty Three percent of freshwater fish in South
Africa are extinct, threatened or endangered (Moyle & Leidy 1992). These statistics
reflect the dire need to incorporate freshwater ecosystems into the mainstream
conservation planning process. However, there are very limited ecological data on
South Africa's rivers, with the assessment of freshwater ecosystems having focused to
date on mainstem rivers e.g. the National Spatial Biodiversity Assessment (Nel et al.
2004). Yet most of the mainstems are highly transformed by human impacts. With many
of these already seriously damaged, conservation attention could usefully focus on
tributaries, though data on these is extremely limited.
Fortunately for South Africa, there are relatively good Geographical Information
Systems (GIS) data on terrestrial biodiversity and land cover. Recent terrestrial
conservation plans developed in South Africa, which use these data for systematic
planning, are probably the most detailed and explicit for any part of the developing world
(Balmford 2003). The availability of this information presents an opportunity for
freshwater systems too: if available fine-scale information on the ecological integrity of
rivers can be predicted with reasonable confidence using remotely-sensed data, then in
principle GIS data could be used to estimate freshwater integrity more widely, and
hence accelerate freshwater conservation planning.
The overall aim of this study is therefore to use South African data to test the
feasibility of GIS-based assessment of the ecological integrity of rivers. Specifically my
objectives are to: -
1. Review the different criteria for assessing river health and ecological integrity in
South Africa and the suitability of these criteria for conservation planning;
4
2. Develop a GIS-based model for predicting river ecological integrity;
3. Determine the spatial scale at which GIS data best predict ecological integrity;
4. Test whether the relationship between land use and ecological integrity holds at
a finer scale than the entire catchments so as to predict the integrity of
tributaries.
RIVER HEALTH AND ECOLOGICAL INTEGRITY
Definitions of river health and ecological integrity and methods for their assessment are
still a subject of considerable debate (Wicklum & Davies 1995, Norris & Norris 1995,
Norris &Thoms 1999, Quigley et al. 2001). Ecological integrity may be defined as" the
ability to support and maintain a balanced, integrated, adaptive community of organisms
having a species composition, diversity, and functional organization comparable to that
of natural habitat of the region" (Angermeier & Karr 1994). Unlike ecological integrity
that is based purely on ecological criteria, river health refers to the ecological condition
of a river and its ability to provide ecosystem service (Boulton 1999, RHP 2003). River
health is thus the incorporation of ecological values of a river system and the human
values associated with such a river system (Boulton 1999). Here, my primary concern is
to assess the ecological status of riverine ecosystems, regardless of their potential to
provide ecosystem services, and so in the context of this paper I use both terms
interchangeably.
In order to design a conservation plan for freshwater ecosystems and to
determine the impact of anthropogenic activities, it is important that river integrity is
mapped across the planning domain (Nel et al. 2004). In South Africa, various indices
5
have been developed for assessing ecological integrity and water quality e.g. the South
African Scoring System (SASS) (Chutter 1998), the index of habitat integrity (Kleynhans
1996), the riparian vegetation index, (Kemper 1998), and the fish assemblage index
(Kleynhans 1999). In the following section, I review these indices, their usefulness for
freshwater conservation planning, and how they have been applied in the River Health
Program.
Methods of Assessing River Ecological Integrity
Several methods and indices have been developed in South Africa for assessing
ecological integrity of riverine ecosystems (Table 1). Most of the criteria are data
intensive, and most of them do not have a wide geographic coverage. Together with the
lack of manpower, it has made their use unsuitable for landscape scale conservation
planning. Most of the indices are biotic indices that use the presence of organisms as
indicators of river integrity. The use of broad surrogates may offer a more robust and
less data intensive means of assessing the ecological integrity of freshwater
ecosystems in cases where data from the field is not readily available. Thus the choice
of suitable surrogates that represent both the abiotic and biotic characteristics of aquatic
ecosystems is critical.
Fish Assemblage Integrity Index (FAil)
This index measures the biotic integrity of river systems based on attributes of native
fish (with the presence of alien fish reducing the FAil score). The index is based on the
relative intolerance of fish occurring in a given river segment (Kleynhans 1999). Four
6
components are taken into consideration in estimating intolerance - habitat preferences
and specialisation, food preferences and specialisation, requirements for flowing water
during different life stages, and association with habitats with unmodified water quality
(Kleynhans 1999). The shortfall of this index is that "experimental information on the
tolerance of various South African fish species is lacking" (Kleynhans 1999). As a result
the assessment of fish intolerance is based entirely on field observation. Since the
representativeness of FAil is dependent on sampling intensity, its wide application for
freshwater conservation planning is again limited.
South African Scoring System (SASS)
SASS is an index of water quality based on the presence of families of aquatic macro
invertebrates (Chutter 1998). Aquatic macro-invertebrates form one of the most
commonly assessed components of the freshwater biota in South Africa (Dallas & Day
2004). Each macro-invertebrate taxon is allocated a sensitivity or tolerance score,
according to the water quality conditions it is known to tolerate (Dallas 1997). SASS was
developed in response to a need by resource managers for a rapid and cost-effective
means of assessing water quality (Dallas 2000). The major shortfall in the application of
SASS is that it is resource intensive.
Water Quality Index (Diatoms)
This index uses diatoms as an indicator of water quality (Bate et al. 2004). Diatoms are
widely used throughout the world as bioindicators of water quality (Bate et al. 2004).
The shortage of expertise in the identification has made the use of this index unsuitable
7
for use in South Africa (Uys 1996). However, it is in the process of being developed for
estuaries and wetlands (Day, J. personal comm.)
PESC (present ecological status)
The PESC index was derived in a scoping exercise during the planning phase of the
river health program (Kleynhans, C.J. personal comm.). It was used to give a very
rough idea of the river systems of South Africa, and was mainly based on expert
opinion. PESC is still in use in some instances for aquatic reserve determination. There
is a national coverage of PESC, but the data is too coarse (only available for mainstem
rivers of quaternary catchments), and it was only based on expert knowledge.
Riparian Vegetation Index (RVI)
The Riparian Vegetation Index was developed by Kemper (1998). The index uses
physical and biological attributes of the riparian zone to assign an ecological integrity
class for river segments. The attributes assessed for the determination of the riparian
index are: - vegetation decrease, alien vegetation encroachment, bank erosion, channel
modification, water abstraction, inundation, flow modification, and water quality. Each
attribute is assigned a weight, and a multicriteria decision analysis (MCDA) is used to
arrive at a final integrity score. Deriving this index requires a thorough knowledge of all
the river systems under review. It is thus impractical for developing a conservation plan
for large areas or in cases where there is lack of expert knowledge. The riparian
vegetation index is used to determine riparian integrity while compiling the index of
habitat integrity discussed below.
8
Index of Habitat Integrity (/HI)
The IHI index was developed by Kleynhans (1996). It assesses the impact of
disturbance both in the riparian zone and in the instream habitat (RHP 2003). The index
of habitat integrity (IHI) is the overall index used most frequently for determining river
integrity in South Africa. It assesses the impact of disturbance both in the riparian zone
and in the instream habitat (RHP 2003).
Instream habitat integrity
In the determination of instream integrity, both measures of physical and biotic
characteristics are used to estimate overall instream integrity (Table 2). Fish
assemblages and macro-invertebrates are used to estimate the biotic integrity. Other
biotic attributes include, the presence of exotic macrophytes and fauna (Kleynhans
1996). For measures of physical characteristics the following attributes are measured:
water abstraction, flow modification, bed modification, channel modification, water
quality, inundation, and solid waste disposal.
Riparian habitat integrity
Riparian habitat integrity is determined using the RVI, where vegetation removal,
cultivation, flow modification, channel modification, inundation, erosion, alien vegetation,
and water quality are assessed (Kleynhans
1996). Most of the measures of riparian integrity are physical rather than biotic (Table
2).
9
The River Health Program (RHP) of South Africa
The RHP is a national biomonitoring program, whose mandate is to assess and monitor
the ecological integrity of riverine ecosystems in South Africa (Roux et al. 1999). It was
established in 1997, and incorporates several components of the biota (Dallas 2000).
The RHP uses instream and riparian integrity monitoring to characterize the response of
aquatic environments to multiple stressors (Roux et al. 1999). The indices used in the
RHP represent the most widely available indices for assessing ecological integrity in
South Africa (Fig. 1).
In the RHP, an overall integrity score of the river system was calculated based on
riparian and instream integrity. However, this was done only for the purposes of
reporting on the state of rivers (SOR), and there was no scientific basis for this
derivation (Kleynhans, C.J personal comm.). Scores of river integrity (instream and
riparian) were categorized in the following manner, (90-100)= A; (80-90)= B; (60-80)= C;
(40-60)= D; (20-40)= E; (0-20)= F. A category of (A) represents a natural unmodified
river system, while a category of (F) represents a very highly modified system with
almost a complete loss of natural habitat.
Land use as a Measure of River ecological Integrity
Ideal indicators of river health should have features that are easy to quantify and are of
direct relevance to river health (Boulton 1999). In the absence of suitable indicators, a
method needs to be developed for rapid assessment of the ecological integrity of river
systems.
10
Anthropogenic activities impose a huge stress on river systems (Allan & Flecker
1993, Dynesius & Nilsson 1994). And therefore land use could be a potential predictor
of integrity. By modeling potential predictors of ecological integrity on a GIS platform, it
may be possible to develop criteria for assessing the ecological integrity of river
systems. In this study, I explore the suitability of variables associated with land use, and
others like population density as proxy for assessing river ecological integrity for the
purposes of conservation planning.
METHODS
Study Area
The Crocodile (West) and Marico Water Management Area (WMA) (Fig. 2 & 3), is one
of 19 WMAs in South Africa. It covers an area of 4.7 million hectares, and forms part of
the Limpopo River basin, which spans four countries (Botswana, Zimbabwe, South
Africa and Mozambique). The main rivers in the water WMA are the Crocodile and
Marico rivers and their tributaries. Other important rivers include the Elands, Pienaars
and the Bierspruit (DWAF 2002a).
The location of the Crocodile and Marico, covering South Africa's economic
centre, has bearing on both water requirements and the ecological integrity of its rivers.
The WMA has a semi-arid climate, and water requirements are very considerable.
Irrigation accounts for 37% of water requirements, the rest being attributable to urban,
industrial and mining needs (DWAF 2004). Agriculture is the most predominant land
use, accounting for 53% of area; other major economic activities include mining and
light industries (Fig. 3) ( DWAF 2002b). The WMA supports a human population of 4.9
11
million (1995 census), of which about 70% live in urban areas with the remainder living
in rural areas (DWAF 2004). I chose this study area for my research because it has
excellent data on river integrity, and because of the diverse land use in the area.
Delineation of the study area
The study area was divided into 25 regions (hereafter referred to as 'catchments'),
based on Level 2 ecoregions (Roux 1999) and the location of monitoring sites.
Ecoregions denote areas with similarity in ecosystems and environmental resources.
The classification of ecoregions is based on a hierarchical procedure involving the
delineation of spatial units, with a progressive increase in detail at each higher level of
hierarchy (Roux et a/1999). In South Africa, level 1 and level 2 ecoregion typology have
been done for the entire country. The level 1 ecoregion was based on physiography,
climate, geology, soils and potential natural vegetation (Roux et a/2002, DWAF 1999).
While the level 2 ecoregion was a refinement of the level 1, and was based on
variations in geology, natural vegetation, altitude, as well as expert knowledge of the
rivers that flow through the planning domain (Low & Rebelo 1996, Roux et a/1999).
The 25 catchments were subdivided into smaller units based on tributaries
(hereafter referred to as 'sub-catchments') giving rise to 47 sub-catchments. Sub
catchments were created to isolate tributaries independent of the mainstem rivers.
Present Ecological Integrity of the River System
Data on the ecological integrity of the river systems in the study area were obtained at
RHP monitoring sites and expert knowledge (RHP in press). The overall scores of river
12
ecological integrity were based at the catchments level (RHP in press). Scores
comprised two measures, riparian zone habitat integrity, and instream habitat integrity.
Riparian and instream habitat integrity were based on proxy variables to define habitat
integrity (Table 2). I chose to build the model based on the RHP assessments, because
they have the most fine scale data on river ecological integrity available for any WMA in
South Africa.
In order to explore potential predictors of ecological integrity as a whole, I next
identified a suite of potential predictors (Table 3), which were available in GIS format,
and which I considered might be linked to these individual components of river integrity
(whilst bearing in mind that some components may not have suitable remotely-sensed
surrogates) .
Quantification of potential predictors of ecological integrity
I chose a series of land use variables that might be suitable surrogates for assessing
ecological integrity (Fig, 3) from the land cover/land use map of 2000 (CSIR in press)
and incorporated separate GIS surfaces of population and number of mines. Together
with the total area of catchments, 14 variables were used to predict ecological integrity
(Table 3); waterbodies were excluded from the analysis. I derived the land use variables
by reclassifying the initial 49 land use types from the land cover map. For instance, I
lumped all naturally occurring vegetation types together to form a single land use type.
This reclassification was necessary so as to avoid too many independent variables for
subsequent modeling.
13
Because river integrity may be influenced not just by local conditions but also by
more distant patterns of landuse (Allan et al. 1997), I attempted to predict river integrity
scores using land use data at a range of spatial scales. Riparian strips were created on
both sides of the river to capture land use patterns up to 100m, 500m, 1000m, 2000m,
and 3000m from the river of interest, using Arcview 3.2a and Arclnfo (ESRI, Redlands,
CA). The choice of the size of riparian strip was arbitrary, and the riparian strips covered
the entire length of the river in each of the 22 catchments. I also compiled catchments
wide summaries of land use, so that together with the riparian strips, analyses were
carried out at six spatial scales.
The Model
The model predicting river ecological integrity was derived based on 22 of the
catchments because three did not have sufficient data. The model was then applied to
predict the integrity of the 47 sub-catchments. I chose a Generalized Linear Modeling
(GLM) approach (Chambers & Hastie 1992) to predict integrity scores because I was
looking for simple relationships and also due to my small sample size of catchments
(22), I wanted to avoid over-fitting the model, which might tailor the current data too
much to the detriment of future predictions (Vanderpoorten et al. 2005). With more data,
a generalized additive model (GAM) could have been more suitable. I assessed
the potential predictors of riparian and instream integrity independently of each other,
using the overall riparian and the instream integrity scores as dependent variables in a
series of GLMs. I predicted the components of riparian integrity using the best model
14
out of the six models that were built. I did not attempt to predict the components of
instream integrity.
Fitting the Model
I built six models for each of the different spatial scales of analyses in S-Plus software
(Mathsoft 1999) using GLM with Poisson error distribution (link
function =log (mu), variance =mu), because non-negative variables are best modeled
using a Poisson-error distribution (Crawley 1993). Due to limited sample size I
constrained the models in various ways. No attempt was made to mix the spatial scales
at which predictor variables were measured within a given model (i.e. the same model
did not include land use variables derived from 100m and 500m riparian strips). I also
did not consider any possible interactions between the various predictor variables, even
though it is possible that the magnitude of the effect of one variable on river integrity
may depend on the effect of other variables (Chambers & Hastie 1999). Interactions
were not considered because I had a small sample size (22), with too many variables
already. Thus interactions could result in over fitting the model, yet they are difficult to
interpret.
In all instances where the model output contained several predictors, I limited the
number to the best four explanatory variables, based on Akaike Information Criterion
(AIC) scores (Akaike 1974). The maximum number of parameters used in the model
was limited to four due to the small sample size, and to enable comparison of
competing explanatory models compiled at differing spatial scales.
15
Testing the accuracy of the models
To test for the strength of each model, r2 values were calculated. I also measured how
well the integrity categories (A to F) assigned to catchment in the RHP compared to the
category (A to F) predicted by a model.
Application of the Model at a finer Scale (sub-catchments)
I used the best model among the six developed to predict riparian and instream integrity
of the sub-catchments. The choice of this model was based on its ability to predict
integrity accurately according to the categories assigned in RHP and whether there is a
logical explanation of the potential predictors used in the model. An expert who is well
versed with the study area evaluated the results of the predictions by giving his opinion
on what he thinks is the actual ecological integrity of the sub-catchment.
RESULTS
Overall predictive power of the models
Modeling riparian and instream integrity in the 22 catchments of the study area revealed
that riparian integrity (Fig. 4) could be predicted more accurately (Table 5 &6) than
instream integrity (Fig. Sa), and that land use at the catchments scale is a better
predictor (~ = 0.79, P < 0.001) than at the smaller scales of riparian strips (r2 = 0.5, P <
0.001) (Table 5). Instream integrity is also somewhat better predicted at the catchments
scale than at smaller riparian strips (Table 6). When the model for instream integrity
was run for headwaters only; it yielded a better fit (~ =0.97, p< 0.001) (Fig. 5b) using
four land use variables only (natural vegetation, population density, area of mines, and
16
density of dams), than the fit for models in all the catchments (Fig. 5a), where upstream
effect was not taken into consideration.
Most important predictor variables of river integrity
The relative extent of cover by natural vegetation was the single most powerful land use
variable for predicting riparian integrity scores, with a positive correlation in all the
spatial scales of analyses. Regardless of the spatial scale of analyses, natural
vegetation cover emerged as the key determinant of riparian integrity (Table 5). Another
important and independent predictor was the presence of a mine in the vicinity, which
appeared to be more important than the relative area occupied by mines. With the
exception of the scenarios with the riparian strip nearest the river (100 meters), all other
spatial scales featured the number of mines as a significant predictor of riparian
integrity. Other land use variables important for predicting riparian integrity, but
restricted to specific spatial scales include the relative extent of land under degradation,
the relative area occupied by mines, the relative area of rural dwellings and eroded
areas (Table 5). The rather surprising result was that the density of roads, the
proportion of cultivated area, plantations, and the relative extent of dams, industry,
eroded areas, and built-up residential or commercial areas were not important
predictors of riparian integrity (Table 5).
The most important predictors of instream integrity are not very different from
those of riparian integrity. Natural vegetation cover was again the key determinant of
instream integrity regardless of the spatial scale of analyses (Table 6). The relative
extents of plantations, of rural dwellings, and of industrial developments were also found
17
to be important predictors of instream integrity at most scales, while relative area of
mines, population density and the density of dams are important predictors at the
catchments scale only (Table 6). The relative extents of cultivated lands, road density,
and the total area of the planning unit do not feature as significant predictors of instream
integrity in most spatial scales (Table 6).
Predicting the components of riparian integrity
Attempts to predict the components of riparian integrity (i.e. vegetation decrease, bank
erosion, channel and flow modification, water abstraction, inundation and water quality)
yielded poor results because fitting the model involved too many predictors, making it
difficult to explain their effect. However, when the model was restricted to the best four
variables (based on AIC scores) in predicting each of the eight components of riparian
integrity there was an improvement, although it was still difficult to explain the effect of
the potential predictors on the components (Table 7). The most important variables in
the model output were: - natural vegetation cover, the relative area of mines, eroded
areas and population. Channel modification yielded the best prediction (r2 = 0.89, p<
0.001). The most important variable for determining channel modification was the total
area of mines. Other important variables for determining channel modification were
population, natural vegetation cover, and eroded areas.
Model application for State of the Rivers (SOR) reporting
Most of the rivers were predicted in the appropriate category of ecological integrity
according to the River Health Program categorization mentioned earlier. Riparian
18
integrity was estimated with an accuracy of up to 77% (Table 5). Incorrect Predictions
were either overestimated or underestimated by only one class (Fig. 6). For instream
integrity, 41% of predictions were in the correct category of classification according to
the river health program (Table 6). Some predictions were overestimated or
underestimated by more than one category (Fig. 7).
Model application for predicting ecological integrity at a finer scale (sub
catchments)
The results of the predictions of the sub-catchments based on the best model were
mapped (fig. 8).
Discussion
Prediction of riparian and instream integrity
My analyses suggests that both riparian and instream integrity can be predicted with
reasonable accuracy using a series of variables covering land use, settlement, and
infrastructure. Riparian integrity was predicted more accurately than instream integrity
however, because riparian integrity is primarily a reflection of physical characteristics
such as bank condition and riparian vegetation cover. Adjacent land cover has a direct
impact on these characteristics. In contrast, land-use data are relatively poor predictors
of instream integrity, which in large part is linked to the structure of the biotic community
of the river, itself only indirectly impacted by adjacent land use. It appears that activities
upstream of the catchments maybe better predictors of instream integrity than is
adjacent land use. This finding is consistent with Roth et al (1996) and Allan et al.
19
(1997) who found that habitat condition and index of biotic integrity correlated strongly
with regional land use upstream of a site.
In the derivation of riparian and instream integrity, numerous characteristics were
used (Table 2) and thus, the potential of the model to estimate the magnitudes of the
components of integrity was determined by the choice of predictors. Some components
e.g. channel modification (Table 7) was better predicted because it was directly related
to potential predictors such as dams and mines. In general, the application of the model
to predict components of integrity was unsuitable. This was the only case throughout my
analyses where the model was over-fitted, with more than eight potential predictors
used in the initial attempt to predict the components. I had to limit the number of
predictors to four and run the model again.
The most important predictor variables of river ecological integrity
Natural vegetation cover was the most important predictor of both riparian and instream
integrity. This was not surprising because, natural vegetation may influence stream
processes like regulation of sediment transport (Osborne & Kovacic 1993),
development of channel morphology (Naiman et al. 1993), and regulation of erosion
(Berkman & Rabeni 1987). In the absence of other data on the state of a river,
assessing the natural vegetation cover alone can provide a fairly accurate prediction of
integrity. Such rapid assessments are very useful for purposes of prioritising river
systems for conservation interventions.
The surprising result from the prediction of ecological integrity is that potential
predictors like road density, relative coverage of dams, industries, eroded areas and
relative area covered by cultivated lands were not important in predicting river integrity.
20
This may imply that even if these are major factors in the alteration of ecological
integrity of river systems (Roth et al. 1996, Quigley et al. 2001, Higgins 2003), they are
not necessarily the best broad surrogates for predicting river integrity.
The application of land cover as a potential surrogate for developing biodiversity
plans have been widely used by the Nature Conservancy of the US and World Wildlife
Fund (WWF). The surrogates they use remain largely untested (Abell et al. 2002). For
example, they place a lot of emphasis on the importance of road density as a potential
predictor of river integrity, but this study has shown that road density is not an important
predictor of integrity, instead natural vegetation cover, and population were more
important in predicting river integrity.
Spatial scale of analysing potential predictors of river integrity
The influence of land use on river integrity is scale dependent (Allan et al. 1997,
Sponseller et al. 2001), and I found that land use at the catchments level provided the
best scale for predicting overall river integrity. In a similar study, Morelyand Karr (2002)
found that the "effectiveness of localized patches of riparian corridor in maintaining
biological integrity varies as a function of basin-wide urbanization". Instream processes
and patterns such as organic matter input may depend on local vegetation cover, while
nutrient supply, channel characteristics and hydrology are influenced by activities at the
catchment scale. The fact that some potential surrogates were only useful in predicting
integrity at specific spatial scales implies that, the scale of analyses also affects
individual predictors. However, this needs to be investigated further. GIS is very
appropriate for modeling such varying scenarios, because it enables assessments to be
21
carried over a large geographic area and different spatial scales, although there is
always a tradeoff between the geographic extent and the resolution of analyses.
Limitations of the model
The model needs validation on an independent dataset preferably in an entirely
different region to ascertain if the findings in this study hold. A similar approach, with
reasonable sample of field-based data, which span a broad range of ecological and land
use conditions (e.g. all those in a country), which can then be divided into reasonably
sized calibration and validation datasets could be useful. Also more experts need to
evaluate the result of the predictions in cases where there is no independent dataset
available, as was the case in this study.
The challenge for such predictive modelling is to use potential predictors that
closely mimic both, the abiotic and biotic characteristics of aquatic ecosystems,
because the explanatory power of the model is limited by the choice of predictors.
Another important determinant of the robustness of a predictive model is the sample
size. In this study, the sample size of 22 catchments was quite small. A small sample
size affects the ability of the model to explain the relationship between potential
predictors and the dependent variables, and also makes validation of the model difficult
(Stockwell & Peterson 2002).
Implications for conservation planning
Tributaries contribute significantly to the overall conservation status of a river, because
in many cases mainstem rivers are highly disturbed. The potential of this model to
22
predict the integrity of the sub-catchments is thus a significant contribution to freshwater
conservation efforts. The resource intensive nature of freshwater assessments means
that very often data is unavailable for certain river systems or some areas are
inaccessible. In such a situation, predictive modelling offers an opportunity for
assessing the ecological integrity of the river systems.
The dependency of river integrity on catchment-wide activities has considerable
implications for conservation planning. In the past there has been strong focus on
protecting riparian corridors (Gregory et 81. 1991), as key to conserving freshwater
ecosystems. But it appears that catchment-wide activities are vital in assessing river
integrity as well. In South Africa, buffer strips of 500m on both sides of the river have
been widely used by freshwater conservation planners in assessing ecological integrity
(Roux. D. personal comm.). My findings show that this is perhaps not an appropriate
scale, a wider buffer zone of at least 1000m maybe more useful.
It should be emphasised that predictive modelling cannot substitute field-based
surveys, because field assessments still offer the best means of assessing ecological
integrity (Roth et 81. 1996). But with the advances in GIS technology and increasing
focus on landscape scale conservation planning, predictive modelling may offer the only
solution in bridging the gap between freshwater conservation planning and terrestrial
conservation planning.
ACKNOWLEDGEMENT
I am grateful to my supervisors Dr. Mathieu Rouget, Dr. Andrew Balmford, and Prof.
Jenny Day for their close guidance and excellent advice throughout the course of this
23
research project. I'm thankful to Jeanne Nel for conceiving this project idea and making
sure it goes all the way. This project was shaped in many ways through extensive
discussions and logistical support from C.J. Kleynhans, Dirk Roux, Gillian Maree, and
Juanita Moolman. Wilfried Thuiller provided statistical and modelling expertise, for that
I'm grateful. I acknowledge my colleagues at the GIS/conservation planning unit
(SANBI) for assistance with GIS analyses. Funding and logistical support came from the
Tropical Biology Asscociation (TBA), South African National Biodiversity Institute
(SANBI), and the Council for Scientific and Industrial Research (CSIR). Finally I would
like to thank Prof. Morne du Plessis and all the staff at the FitzPatrick Institute for
making sure nothing went wrong during the course of my studies.
References
Abell R., M. Thieme, E. Dinerstein, D. Olson .2002. A sourcebook for conducting
biological assessments and developing biodiversity visions for Ecoregion
conservation. volume II: Freshwater Ecoregions. World Widlife Fund,
Washington, D.C. USA.
Abell, R. 2002. Conservation biology for the biodiversity crisis: a freshwater follow up.
Conservation Biology. 16 (5): 1435-1437.
Akaike, H. 1974. A new look at statistical model identification. IEEE
transactions on automatic control. AU-19: 716-722.
Allan, J.D., A.S. Flecker. 1993. Biodiversity conservation in running water. Bioscience.
43: 32-43.
24
Allan, J.D., D.L. Erickson, J. Fay .1997. The influence of catchment land use on stream
integrity across multiple scales. Freshwater Biology 37: 149-161.
Angermeier, P.L., J.R. Karr. 1994. Biological integrity versus biological diversity as
policy directives. Bioscience 44: 690-697.
Balmford, A. 2003. Conservation planning in the real world: South Africa shows the way.
Trends in Ecology and Evolution. 18 (9): 435-438.
Bate, G., P. Smailes, J. Adams. 2004. A water quality index for use with diatoms in the
assessment of rivers. Water SA. Vol. 40 (4): 493- 498.
Berkman G.C., C.F. Rabeni. 1987. Effect of siltation on stream fish communities.
Environmental Biology of Fishes. 18: 285-294.
Boulton, A.J. 1999. An Overview of river health assessment: Philosophies, practice,
problems and prognosis. Freshwater Biology. 41: 469- 479.
Chambers, J.M., T.J. Hastie. 1992. Statistical Models in S. Wadsworth & Brooks/Cole
Advanced Books & Software Pacific Grove, California, USA
Chutter, F.M.1998. Research on the Rapid Biological Assessment of Water Quality
Impacts in Streams and Rivers. Water Research Commission Report No.
422/1/98. Water Research Commission, Pretoria, South Africa. 23pp.
Council for Scientific and Industrial Research (CSIR) (in press). National Land cover
Map pf South Africa. CSIR, Pretoria.
Crawley, M.J., 1993. GUM for ecologists. Blackwell, Oxford.
25
Dallas, H.F. 1997. A preliminary evalutation of aspects of SASS (South African Scorig
System) for rapid bioassessement of water quality in rivers, with particular
reference to the incorporation of SASS in a national biomonitoring programme.
South African Journal of Aquatic Sciences. 23 (1): 79-94.
Dallas, H.F. 2000. Ecological reference conditions for riverine macroinvertebrates and
the River Health Program, South Africa. 1st WaterNet Symposium: Sustainable
use of water resources; Maputo, Mozambique.
Dallas, H.F., J.A. Day. 2004. The effect of water quality variables on aquatic
ecosystems: A review. Water Research Commission. WRC Report No. TT
224/04.
Department of Water Affairs and Forestry (DWAF). 1999. Resources directed measures
for protection of water resources. Volume 3. River Ecosystems, Version 1.0.
Department of Water Affairs and Forestry, Pretoria, South Africa.
Department of Water Affairs and Forestry (DWAF). 2002a. Proposed first edition
national water resource strategy. Department of Water Affairs and Forestry,
Pretoria. Ref. D3.
Department of Water Affairs and Forestry (DWAF). 2002b. Integrated water
management strategies, guidelines, and pilot implementation in three water
management areas, South Africa. Department of Water Affairs and Forestry,
Pretoria. Ref. J. No. 123/138-0154.
26
Department of Water Affairs and Forestry (DWAF). 2004. Internal strategic perspective:
Crocodile (West) and marico water management area. Department of Water
Affairs and Forestry. Report No. P WMA 03/00010010303.
Dynesius, M., C. Nilsson. 1994. Fragmentation and flow regulation of river systems in
the northern third of the world. Science 266: 753-762.
ESRI. 1987. Environmental Systems Research Institute. Redlands, California, USA.
Gregory, ev., F.J. Swanson, W.A. McKee, K.W. Cummins. 1991. An ecosystem
perspective on riparian zones. Bioscience. 41, 540-51.
Higgins, J.v .2003. Maintaining the ebbs and the flows of the landscape. Conservation
planning for freshwater ecosystems. Chapter 10 in: Groves, C. R. and
contributors. 2003. Drafting a Conservation Blueprint: a Practitioner's Guide to
Regional Planning for Biodiversity. Washington, D.C.: Island Press.
Jensen, D.B., M.S. Torn, J.Harte. 1993. In our own hands: A strategy for conserving
California's biological diversity. University of California Press, Berkely.
Kemper N.P.1998. Riparian vegetation index development. Progress report to the Water
Research Commission for Project K5/580.
Kleynhans, C.J. 1999. The development of a fish index to assess the biological integrity
of South African rivers. Water SA vol. 25. NO.3: 265-278.
Kleynhans, C.J.1996. A qualitative procedure for the assessment of the habitat integrity
status of the Luvuvhu river (Limpopo systems, South Africa). Journal Of Aquatic
Ecosystem Health 5: 41-54.
27
Low, AB., AG. Rebelo. 1996. Vegetation of South Africa, Lesotho and Swaziland.
Department of Environmental Affairs and Tourism, Pretoria, South Africa.
Mathsoft Inc. 1999. S-PLUS statistical software. Mathsoft Inc., Cambridge, UK.
http://www.mathsoft.com/.
Morley S.A, J.R. Karr. 2002. Assessing and restoring the health of urban streams in
Puget sound basin. Conservation Biology. 16: 1498-14509.
Moyle, P.B., J.E. Williams. 1990. Biodiversity loss in the temperate zone: Decline of the
native fish fauna of California. Conservation Biology 4: 275-284.
Moyle, P.B., RA Leidy. 1992. Loss of biodiversity in aquatic ecosystems: Evidence
from fish faunas. Pages 128-169. In P.L. Fiedler and S.A Jain, editors.
Conservation Biology: The Theory and Practice of Nature Conservation,
Preservation, and Management. Chapman and Hall, New York.
Naiman, RJ., H. Decamps, M.Pollock. 1993. The Role ofriparian corridors in
maintaining regional biodiversity. Ecological Applications. 3: 209- 212.
Nel, J. G. Maree, D.Roux, J. Moolman, N. Kleynhans, M. Silberbauer, A Driver. 2004.
South African national spatial biodiversity assessment 2004: Technical report.
Vol. 2: River component. CSIR Report Number ENV-S-I-2004. Council for
Scientific and Industrial Research. Stellenbosch, South Africa.
Norris RH., K.R Norris. 1995. The need for biological assessment of water quality.
Australian Journal of Ecology. 20: 1-6.
Norris. R.H., M.C. Thoms. 1999. What is river health. Freshwater Biology. 41: 197-209.
28
Osboren, L.L., D.A Kovacic.1993. Riparian vegetated buffer strips in water quality
restoration and stream management. Freshwater Biology. 29: 243-258.
Quigley, T.M., R.W. Haynes , W.J. Hann .2001. Estimating ecological integrity in the
interior Columbia river basin. Forest Ecology and Management 153: 161-178.
Revenga, C., J. Brunner, N. Henninger, K. Kassem, R. Payne. 2000. Pilot analysis for
global ecosystems. Freshwater Systems . Washington DC, USA: World
Resources Institute.
River Health Programme. 2003. State-of-rivers report: Hartenbos and Klein Brak Rivers.
Department of Water Affairs and Forestry, Pretoria
River Health Programme. In press. State-of-Rivers Report: Crocodile (West) and Marico
Water Management Area . Department of Water Affairs and Forestry , Pretoria.
Roth, N.E., J.D. Allan, D.L. Erickson. 1996. Landscape influences on stream biotic
integrity assessed at multiple spatial scales. Landscape Ecology . Vol. 11: 3
pp141- 156.
Roux, D., F. De Moor, J. Cambray, H. Barber-James.2002. Use of landscape-level river
signatures in conservation planning: A South Africa case study . Conservation
Ecology 6(2): 6. (Online:http://www.cosecol.org/voI6/iss2/art6)
Roux, D.J., C.J.kleynhans, C.Thirion, L.Hill, J.S. Engelbrecht, AR. Deacon, and N.P.
Kemper . 1999. Adaptive assessment and the management of riverine
ecosystems: The Crocodile/Elands river case study. Water SA Vol.25 NO.4:
501-512.
29
Sponseller R.A, E.F. Benfield, H.M. Valett. 2001. Relationship between landuse, spatial
scale and stream macroinvertbrate communities. Freshwater Biology 46, 1409
1424
Stockwell , D.R.B., AT. Peterson. 2002. Effects of sample size on accuracy of species
distribution models . Ecological Modelling. 148: 1-13.
Uys, M.C., P.A Goetsch , J.H. O'Keffe. 1996. National biomonitoring programme for
riverine ecosystems: Ecological indicators , a review and recommendations.
NBP Report Series NO.4. Institute of Water Quality Studies , Department of
Water affairs and Forestry. Pretoria. 92 pp.
Vanderpoorten, A, A Sotiaux, P. Engels. 2005. A GIS survey for the conservation of
bryophytes at the landscape scale. Biological Conservation. 121: 189- 194.
Vitousek, P.M., H.A Mooney, J. Lubchenco, J.M. Melillo. 1997. Human domination of
earth's ecosystems. Science 277: 494-499.
Wicklum . D. & R.W. Davies. 1995. Ecosystem health and integrity. Canadian Journal
and Botany. 73: 997-1000.
World Conservation Union (IUCN). 2000. The 2000 IUCN red list of threatened species.
IUCN, Gland, Switzerland . Available from http://www.redlist.org/
30
LIST OF TABLES AND FIGURES
Table 1. Indices presently used for assessing river ecological integrity in South Africa;
parameters measured, indicator, type of index, potential surrogates for the indices and the
possible GIS measures! measures of estimating the potential surrogates
Index Measure Indicator Type PotentialSurrogatesfor theindices
Possible GIS ReferencesAnalyses/measures
Bate (2004)
Chutter(1998)
Kleynhans(1996)
Kemper(1998)
Unpublished
Level of Kleynhanstransformation in (1999)riparian zone
None
None
None
Land use
None
Biotic
Biotic
Biotic
Abiotic Land use, Transformation,/biotic erosion, dam coverage,
impoundmentdigital elevations, models, mining,
agriculture,urbanization,alien vegetation
Habitatsuitability &water quality
WaterqualityHabitatsuitability
Assemblage of Wateraquatic qualityinvertebratesTrophic levelpreference,requirements forflowing water,habitat integrity,health of fish,Assemblage ofdiatomsVegetationdecrease, exoticvegetation,channelmodification, waterabstraction,inundation, flowmodification, waterquality,
PESC (present Bed modification, Habitat Abiotic/ Land use, Transformation,ecological flow modification, suitability & biotic impoundmentdam coverage,status inundation, water water quality s digital elevationcondition) quality, introduced models, mining,
alien biota, agriculture,riparian vegetation urbanization,conditions alien vegetation
Index of HabitatWater abstraction, Habitat Abiotic/ Land use, Transformation,Integrity (IHI) flow modification, sultabillty biotic impoundmentdam coverage,
Channel s digital elevationmodification, solid models, mining,waste, exotic agriculture,organisms, urbanization, andinundation, water alien vegetationquality, alienvegetation
Water qualityindexRiparianvegetationindex (RVI)
SASS (SouthAfrica Scoringsystems)FAil (fishassemblageintegrity index)
31
Table 2. Parameters used as indicator of riparian and instream Integrity, and the
weights of their relative contribution to the overall river integrity, as used in the RHP
assessments (RHP in press)
Riparian habitat integrity ParameterVegetation decrease
Exotic vegetation
Bank erosion
Channel modification
Water abstraction
Inundation
Flow modification
Water quality
Instream habitat integrity ParametersWater abstraction
Flow modification
Bed modification
Channel Modification
Water quality
Inundation
Exotic macrophytes
Exotic fauna
Solid waste
Weights13
12
14
12
131112
13
Weights14
131313
14
10
9
8
6
32
Table 3. Variables used as potential surrogates for predicting integrity, and a description
of the measures used in their quantification
Predictors Description
Natural areas Area under natural vegetation cover (%)
Plantations Area under forest & grassland plantations (%)
Degraded Areas Area under man-induced low vegetation cover (%)
Cultivated areas Area under all forms of agriculture (%)
Rural clusters Area under clustered rural dwellings (%)
Residential/commercial areas Area under residential or commercial usage (%)
Industrial Area built up of industries (%)
Mines Area used for mining (%)
Eroded areas Area under bare rock/soil (gully or sheet erosion) (%)
Roads Length of roads per Km2 (%)
Dams Area under dams (%)
Population* Population density (KM2)
Number of mines Total number of active mines
Area Total area of catchments/planning unit (m2)
* Assessed only at the scale of regions
33
Table 4. Correlation matrix between potential predictors summarized at the catchments
scale
rJJ"0 3 c ......
C "0 rJJ .9 0.9 Q) ..... Q)
Q) ..... ~
"@ ..... "0 ce C .;:: ...... "0..... Q)ce
~ ce ce .::: Q) ..... 0 rJJ Q) rJJ rJJ "3 ..0 rJJ..... "@ "0 rJJ;:l C
~ ..... ;:l ce Q) "0 "0 6 6 Q)..... eo "3 ~ 'Vi Q) c 0 ce 0.. Cce ce Q) ;:l Q) "0 ~ ·s 0 ce 0 ;:l ·SZ is:: 0 c <t:~
0 Zu ~ ~ - ~ ~ P-.
Natural
Plantations
Degraded
Cultivated
Rural
Residential
Industries
Area of mines
Eroded
Roads
Dams
PopulationNumber of mines
1
-0.66 1
-0.23 -0.39 1
-0.30 -0.11 0.11 1
-0.20 -0.18 0.31 0.29 1
-0.86 0.84 -0.10 -0.12 -0.08 1
-0.78 0.86 -0.20 -0.19 -0.11 0.97 1
-0.38 0.16 0.10 0.16 0.69 0.22 0.14 1
0.09 -0.09 -0.01 -0.30 0.00 0.03 0.03 -0.14 1
-0.67 0.81 -0.29 0.08 0.10 0.73 0.73 0.33 -0.12 1
-0.03 0.22 -0.22 -0.13 0.39 0.06 0.15 0.51 -0.08 0.46 1
-0.17 0.41 -0.21 -0.20 -0.18 0.30 0.42 -0.02 -0.17 0.45 0.61 1-0.11 0.17 0.25 -0.18 -0.05 0.08 0.07 0.22 -0.04 0.06 0.20 0.29 1
34
Table 5. Output of the six different models showing the effect of potential predictors on
riparian integrity, the strength of prediction (r2) , and the % accuracy of the predictions
according to the State of the River (SOR) categorization. The (+) and (-) signs signify
the effect of the parameter on integrity according to the model (e.g. increase in natural
cover (+) results in higher integrity). Blank columns show that the variable did not
appear in the model output as a predictor of integrity. All the predictions in the different
spatial scales (models) were significant at p < 0.001
I/)(1)
I/)c
RIPARIAN (1) 'E cI/) ro c .... 0C "0 "0 I/) 'E c
0 1:50 (1) (1) +:i (1) 0..... C +:i .... :0ro +:i "0 ro (1) 'C .... "0 ro (1).... ro ro > "0..... 0 (1) I/)
I/) "5 .0 (1)..... ro I/) "0::J C....
+:i .iii ::J ro "0 E E ro ......... ro 0>::J
.... (1) 0 ro 0. (1) 0-ro (1) ::J (1) "0 0 ro 0 ::J0- c .... .... .... <'L :::RZ 0 U 0:::: 0:::: « w 0:::: 0 0- Z « 0
1OOm riparian strip + + + + 0.5 45.5
500m riparian strip + + 0.76 63.6
1000m riparian strip + + 0.68 77.3
2000m riparian strip + + 0.71 72.7
3000m riparian strip + + + 0.78 59.1
Catchments + + + 0.79 68.2
35
Table 6. Output of the six different models showing the effect of potential predictors on
instream integrity, the strength of prediction (r2) , and the % accuracy of the predictions
according to the SOR categorization. The (+) and (-) signs signify the effect of the
parameter on integrity according to the model (e.g. increase in natural cover (+) results
in higher integrity). Blank columns show that the variable did not appear in the model
output as a predictor of integrity. All the predictions in the different spatial scales
(models) were significant at p < 0.001
l/)Q)
l/)c
INSTREAMQ) 'E c
l/) CllC
C ..... 0C "0 "0 :p l/) 'E 0 0 ~0 Q) Q) c Q) :p ....:p "0 ..... 'C ..... "0 Q) "0Cll Cll Q) 0 l/) CllCll
~..... Q) l/) .c Q).... ..... > Cll
"0 l/) "0 :::::l ....:::::l C C) :p .iii :::::l Cll "0 Cll E c.. E Cll o,..... Cll :::::l....
~ 0 ~Cll Q) :::::l Q) "0 .... 0 Cll 0 :::::l "LZ a.. 0 0 0::: 0::: c <! LU 0::: 0 o, z <! :::R0
100m riparian strip 0.5 50
500m riparian strip + 0.47 45
1000m riparian strip + 0.52 41
2000m riparian strip + + 0.64 68
3000m riparian strip + + 0.64 64
Catchments
Headwaters
+
+ +
+ 0.68 55
0.97
36
Table 7. Model of catchment showing the effect of potential predictors at the catchments
scale on the components of riparian integrity, and the strength of prediction (r2) . The (+)
and (-) signs signify the effect of the parameter on the component of riparian integrity
according to the model (e.g. increase in natural cover (+) results in higher integrity).
Blank columns show that the variable did not appear in the model output as a predictor
of integrity. All the predictions in the different spatial scales were significant at p < 0.001
Components(/)
of riparianQ)
(/)c:
Q) 'E(/) m c: c: -integrity c: "0 "0 :;::::; (/) 'E 0 00 Q) Q) c: Q) :;::::; -ן
m :;::::; "0 ..... 'C - "0 Q)ro ro ro Q) ..... 0 Q) rn ro
.0I- ..... > m "0 (/) 0(/) "5:::::l I- ro "0 E E roc: 0> :;::::; '00 :::::l ttl 0-.... ro "5
I- Q) 0 Q)ro Q) :::::l Q) "0 l- I- 0 ro 0 :::::l I-
Z a.. 0 0 0::: 0::: c .g:: W l::: 0 o, z <! "L
Vegetation
Decrease + + 0.65
Exotic
Vegetation + + + 0.62
Bank erosion + + 0.82
Inundation + + 0.86
Channel
Modification + + 0.89
Water
Abstraction + + 0.8
Flow
Modification 0.5
Water Quality + + 0.8
37
INSTREAM HABITATINTEGRIT
INDEX OF HABITATINTEGRITY llHI\
RIPARIAN ZONEHABITAT
FISH ASSEMBLAGElFAII\
DIATOMINDEX
MACRO·INVERTEBRATERESPONSE lSASS)
INTEGRATEDINSTREAM BIOTIC
RIPARIAN VEGETATIONINDEX lRVI\
INTEGRATED RIVERBIOLOGICAL
Figure 1. A schematic representation of the different biotic indices developed in South
Africa, and their application in the determination of river health
38
• Monitoring sitesRivers
. I Catchments
Figure. 2. Monitoring sites and major river systems in the Crocodile (West) and Marico
WMA.
39
N
D Cidc hllleutsL.11H11IS e types
C10 11 " .lIod0 0 09,.111..1o Eroded tl lellS
. lndlO slIyMines_(llI,lI liesPl oll1 t.l 1l0ll S
• Resl dollHolOlOll O' cialD RIlI.,1Chiste..o W.11 0,h odles• Roads. 0~II1 S
o 50 100
Figure. 3. Major land use types in the Crocodile (West) Marico water management area
40
a0:>
0
aco
0
UJ @
0:: 00 ao I"-(J)
0 0
~ 0
0:: a 0 0o <00
UJ 0I-z0 a 0L()UJ 0 00::~
~ a 0"'":2 0
aC')
0
0
30 40 50 60 70 80
PREDICTED INTEGRITY SCORE
Figure 4. Accuracy of predicted riparian integrity based on the model at the scale of
catchments (r2 =0.68)
41
a)
a0 00 000
aw a a~
0UC/) 0 a~
coa
0::(') aw.... a~ a0w 0 a a a~ ... a::::JC/) a "i'1i:;; 00
1') a
a
b)
20 30 40 50 60
PREDICTED INTEGRITY SCORE
70
a0 a a a00
aw a a~
0UC/) 0
~co
~(')w....;:::0w 0 a~ ...::::JC/)
i'1i:;; a a
0
'"a
20 40 60 80
PREDICTED INTEGRITYSCORE
Figure 5. Prediction of instream integrity for; a) all the catchments (r2= 0.68), b) headwater
catchments (r2= 0.97) only. The Predictions were based on the model forentire catchment
(Table 6).
42
A 8 c
Inlegrity CID~U~
ABCoEF
• OVER ESTIMATED
• UI JDERESTIMATED
D E F
Figure 6. Crocodile (West) and Marico WMA showing predictions of riparian Integrity
using different spatial scales (Table 5), (A=100m, 8= 500m, C= 1000m, D= 2000m, E=
3000m , F= entire catchment). The map in the center depicts measured integrity. The
dots show whether the model has over estimated riparian integrity or underestimated
based on the measured integrity.
43
A
D
B
E
c
, -fIJI ly 1.1~~~ .. \
AelCo[
J
• OVER ESTMATED
• I ~lrH< F TI Tf r ,
F
Figure 7. Crocodile (West) and Marico WMA showing predictions of instream Integrity
using different spatial scales (Table 6), (A=100m, B= SOOm, C= 1000m, D= 2000m, E=
3000m, F= entire catchment). The map in the center depicts measured integrity. The
dots show whether the model has over estimated instream integrity or underestimated
based on the measured integrity.
44
N"-et'IIlrEGAll ' CLo' 5E5
~
CoE•Sow ( I::h rnrt ,
/'N
Figure 8. Predicted riparian integrity of sub-catchments and river systems in the
Crocodile (West) and Marico. The predictions were based on the best model
(Catchments model) . ( integrity classes:- A= intact, unmodified habitat; F= highly
modified habitat due to anthropogenic activities).
45