Académique Documents
Professionnel Documents
Culture Documents
Assessing catchment-wide
mining-related impacts on sediment
movement in the Swift Creek catchment,
Northern Territory, Australia, using GIS
and landform-evolution modelling
techniques
G. S. Boggs†* , K. G. Evans‡ , C. C. Devonport† , D. R. Moliere‡ and
M. J. Saynor‡
The Swift Creek catchment, the first catchment to be affected should any impact occur as a result of mining
of the Jabiluka uranium ore deposit, is located partly within the World Heritage Kakadu National Park (KNP),
and partly within the Jabiluka Mineral Lease (JML) that has been excised from KNP. Preliminary linking of a
landform evolution model with a Geographic Information System (GIS) has been completed and tested on a
catchment-wide basis for long-term total catchment management. This project represents the first attempt
to apply the model on a catchment-wide basis in the region. Linking the model with a GIS enhances the
modelling process, as the GIS assists in the derivation, storage, manipulation, processing and visualisation
of geo-referenced data on a catchment-wide scale. This preliminary assessment of landform evolution in
the Swift Creek catchment demonstrates the complex process associated with the parameterisation of the
SIBERIA model, and illustrates the benefits of integrating GIS with landform evolution modelling techniques.
Additional research is required to develop a more integrated GIS and landform evolution modelling approach
to assessing the possible impacts of mining on catchment sedimentary and hydrological processes.
2000 Commonwealth of Australia
Keywords: GIS, SIBERIA landform evolution model, hydrology model, mining, impact
assessment, northern Australia.
the preliminary assessment of future land- Sediment loss and Rainfall and
form evolution in the Swift Creek catch- runoff data runoff data
ment.
Environmental models attempt to simu-
Sediment transport DISTFW
late spatially distributed, time-variable envi- equation (STM) hydrology model
ronmental processes (Steyaert, 1993). GIS,
through their ability to capture, manipu-
late, process and display geo-referenced data, Long-term Discharge–area
are able to describe the spatial environment runoff relationship
(Burrough and McDonnell, 1998). GIS and
environmental modelling are therefore com-
Long-term
plementary and the overlap and relationship sediment loss
between these technologies is clearly appar-
ent (Fedra, 1993). Since GIS and environmen-
tal modelling have evolved separately, they SIBERIA input
parameterisation
have different data structures, functions and
methods for the input and output of spa-
tial information (Maidment, 1996). Over the Landform evolution
past two decades there has been consider- simulation
able research into the integration of these
two methodologies to the extent that the syn- Figure 1. A flow diagram depicting the processes
involved in the parameterisation of the SIBERIA
thesis of spatial data representations and landform evolution model.
environmental models has been described as
the new ‘Holy Grail’ (Raper and Livingstone, SIBERIA is a complex landform evolution
1996). Currently, many different approaches model that requires extensive parameterisa-
to linking environmental models with GIS tion (Willgoose et al., 1991). Parameterisation
exist, ranging from the very simple, in which of the model requires the use of separate
the GIS is used for the analysis of model out- hydrology and sediment transport models to
put, to closely integrated systems (Charnock derive a discharge/area relationship, long-
et al., 1996). term sediment loss and a sediment transport
SIBERIA, the landform evolution model rate (Figure 1). An extensive field data col-
used in this study, is a sophisticated three- lection program provides data on catchment
dimensional topographic evolution model. sediment movement and hydrology of Swift
The model has been used to investigate post- Creek, allowing the direct calibration of these
mining rehabilitated landform design at the models.
Energy Resources of Australia (ERA) Ranger
Mine since 1993 (Willgoose and Riley, 1993;
Evans, 1997; Evans et al., 1998; Willgoose Study area
and Riley, 1998; Evans, 2000). To date, the
model has only been used to examine land- The Swift Creek catchment is located approx-
form evolution on post-mining rehabilitated imately 230 km east of Darwin and 20 km
landforms. This project is the first attempt to north-east of the town of Jabiru (Figure 2).
apply the model on a catchment-wide basis in The Swift Creek catchment lies partly in the
the region. Research will investigate whether Jabiluka Mineral Lease (JML) and partly
the shift from mine-site scale modelling of in the surrounding Kakadu National Park
landforms to catchment scale modelling of (KNP), and contains the ERAJM site in its
mining impact can be facilitated by the link- western section. The catchment is elongated
age of the landform evolution model with with a length of approximately 11Ð5 km, a
a GIS. Linking the model with a GIS will maximum width of approximately 7Ð5 km
greatly enhance the modelling process, as and a total area upstream from the most
the GIS can assist in the derivation, storage, downstream gauging site of almost 46 km2
manipulation, processing and visualisation (Figure 2).
of geo-referenced data on a catchment-wide Two distinct landform regions are repre-
scale. sented within the catchment. The upland
GIS and landform-evolution modelling 323
Figure 2. The location of the ERA Jabiluka Mine and Swift Creek catchment in the Northern Territory,
Australia.
plateau region occupies the eastern, southern is approximately 1450 mm. Perhaps more
and western sections of the catchment and importantly for landform evolution is the
consists of highly dissected sandstone, shal- high rainfall intensity associated with wet
low sandy soils and exposed rock. The cen- season storms, with events of 100 mm/h for a
tral and northern sections of the catch- duration of 10 min expected to occur annually
ment contain the Swift Creek floodplain, (Finnegan, 1993).
which is generally flat and covered by deep
sandy soils. Located in the monsoon trop-
ics climatic zone, the catchment experi-
Database establishment
ences a distinct wet season from October
to April, and a dry season for the remain- Data acquired from the field program are
der of the year. The average annual rainfall varied, and include information on catchment
324 G. S. Boggs et al.
Event 1: Rainfall/discharge 35
Discharge (cumecs)
6 16
Rainfall (mm)
5 14 30
12
Discharge (cumecs)
4 10
3 8 25
2 6
4 20
1 2
0 0 15
Time (8/1/99–9/1/99)
10
Event 2: Rainfall/discharge
Discharge (cumecs)
5 16 5
Rainfall (mm)
14
4 12
3 10 0 5 10 15
8
2 6 Time (h)
1 4
2 Figure 6. An example (Event 4) of a predicted
0 0
Time (15/1/99–16/1/99) hydrograph produced by DISTFW ( ) compared
with the observed hydrograph ( ).
Event 3: Rainfall/discharge
Discharge (cumecs)
10 35
Rainfall (mm)
8 30
25 Table 1. The observed and predicted discharges
6 20 for the 1998/1999 wet season in the Swift Creek
4 15 catchment
10
2 5
0 0
Observed Predicted Difference (%)
Time (30/1/99–31/1/99)
discharge (MI) discharge (MI)
7 30
6 25
5 20
4
15 for more conservative management of mining
3
2 10 impact. The final parameters were assessed
1 5
0 0 by comparing predicted total discharge and
Time (9/2/99–10/2/99) hydrograph for the entire 1998/1999 wet sea-
son with the observed total discharge and
Figure 5. Rainfall/discharge of selected events. hydrograph. The predicted total discharge
for the Swift Creek catchment at the down-
Calibration of the DISTFW hydrology stream gauging was found to be slightly less
model involves fitting parameters for selected than the observed values (Table 1), whilst the
storm events. The average rainfall, calculated hydrographs were similar in shape.
from the data collected at each of the gaug-
ing stations, was plotted with discharge for
the Swift Creek downstream gauging station Sediment transport model
for the 1998/1999 wet season. Two large and
two moderate discharge events were selected The sediment transport model (STM) to be
to be input for calibration of the hydrology used in this study is a standard equation
model (Figure 5). used by geomorphologists and soil scientists,
Parameter values were fitted to the selected relating discharge to total sediment loss. The
hydrographs for the observed rainfalls by STM has previously been used in the cali-
fitting a single parameter set that provided a bration of the SIBERIA landform evolution
good fit to the four hydrographs for each site model (Evans et al., 1998; Willgoose and
simultaneously. The predicted hydrographs Riley, 1998). The sediment transport model
compared reasonably well with observed data used is of the form (Evans, 1997):
for each event (Figure 6). There was some Z
over-prediction of the peak discharge of one TDK Qm1 dt .3/
of the events. However, the over-prediction
of runoff is preferred to under-prediction,
as this results in higher predicted sediment where,
movement, which in turn provides a basis K Db2 Sn1 .4/
GIS and landform-evolution modelling 327
R m
where T D total sediment loss, Q 1 dt D Discharge/area relationship
cumulative runoff over the duration of the 100
event (QDdischarge (l s 1 )), SDslope (m/m)
Discharge (cumecs)
and b2 , n1 and m1 are fitted parameters.
The parameters m1 and K are fitted using
multiple regression. However, although the 10
correlation was high (correlation coefficient of
0Ð838), the equation tended to underpredict
high sediment loss events. This is of partic-
1
ular concern as it is well recognised that 1 000 000 10 000 000 100 000 000
large events generally dominate sediment 2
Area (m )
loss. For example, of 81 runoff-producing
rainfall events measured on the eastern Dar- Figure 7. The peak discharge–area relationship
ling Downs of Queensland, six storms caused exhibited by the field data. ( ), discharge/area; ( ),
70% of the total soil erosion (Wockner and power discharge/area. R2 D0Ð9987; yD0Ð0004x0Ð6282 .
Freebairn, 1991). At ERA Ranger mine, 25%
of storms monitored on site removed 54% to empirical peak discharge data from the field
73% of total sediment removed during all monitoring program (Figure 7). The relation-
monitored events (Evans, 1997). As such, the ship derived using these data is:
final equation for predicting sediment loss in
the Swift Creek catchment was found to be: Qp D0Ð0004A0Ð63 .6/
Z
TD0Ð08 940 353 Q0Ð871 dt .5/ The acceptable range for m3 values is 0Ð5
to 1Ð0 (Willgoose et al., 1991). The value
determined here, 0Ð63, falls within that
It is important that large discharge events range.
are not underpredicted, as they are the most
erosive. Ferguson (1986) considered that
power curves of this form (Equation 5) under- Runoff series and long-term
predict sediment transport due to statistical sediment loss rate
bias. Therefore, a statistical bias correction
factor (1Ð119), has been incorporated in the The runoff series for the Jabiru historical
coefficient of Equation (5). rainfall record was used to determine the
long-term erosion rate .qs / in Equation (1) for
the Swift Creek catchment. The steps were:
Scale analysis–discharge area
(1) The fitted DISTFW model parameter
relationship values were used to generate long-term
runoff for the Swift Creek catchment
The parameters fitted here define how dis-
for several years of the Jabiru rainfall
charge used in the calculation of sediment
record. The sub-catchment model of the
transport rate varies with catchment area.
stand-alone version of the DISTFW model
The discharge–area relationship is described
was used because of the large amount
by Equation (2) (Willgoose and Riley, 1993).
of computer processing time required to
Huang and Willgoose (1992, 1993) inves-
generate a runoff series using DTM node
tigated the potential for using the DIS- data.
TFW hydrology model to determine the (2) The annual runoff determined in step 2
relationship between discharge and area. above was then used in the soil loss
Although this process has been used in pre- equation:
vious SIBERIA studies (Willgoose and Riley,
Z
1998; Evans et al., 1998), the methodology
was deemed unsuitable for investigating the TDbSn1 Qm1 dt .7/
larger Swift Creek, as it assumes that the R
rainfall in all parts of the catchment are where T Dtotal sediment loss (g), Qm1
the same. The area-dependence of discharge dtDcumulative runoff over the duration of
within Swift Creek was instead found using the event i.e. annual runoff [QDdischarge
328 G. S. Boggs et al.
Table 2. Long-term average soil loss, uncorrected for node scale, for the Swift Creek
catchment
Year Rainfall (mm) Soil loss mass rate .Mg y 1 / Soil loss volume rate (m3 y 1 )
(L s 1 )], SDslope (m/m) and b, n1 and m1 by SIBERIA and the correction factor is as
are fitted parameters. Equation (7) was follows:
used to determine an annual sediment 1 1
loss .Mg y 1 / which was converted to n
D n .9/
.DTMspacing/ 1 50 1
a volume .m3 y 1 / by dividing by the
bulk density of the surface material (1Ð38 The value b1 parameter used in SIBERIA
Mg m 3 ) (Table 2). Using the annual must include a multiplication by the correc-
sediment losses, a long-term average tion factor derived in Equation (8). Applying
sediment loss rate was then determined the correction factor to Equation (7), a value
.qs / for Equation (1) (Table 2). of 42Ð28 was solved for b1 . The values in
(3) The value of qs was then used to deter- Table 2 were derived assuming a slope, S,
mine b1 by substituting Equation (2) into of 1Ð0 m/m and therefore the qs values are
Equation (1) and transposing to give: not real. The internal algorithm in SIBERIA
corrects for true DTM node slope during sim-
qs
b1 D m1 m1 m3 n1 .8/ ulations.
b3 A S
model hydrologic systems. GIS can also assist approach, as described by Fedra (1993). This
in design, calibration, modification and com- level of integration will provide a common
parison of models. However, the acquisition user interface for both the GIS and the
and compilation of information required by a model, with the file or information shar-
GIS for hydrological modelling is often labour ing between the respective components being
intensive and is an issue commonly encoun- transparent to the end user. The DISTFW
tered in hydrologic applications of GIS (Hill hydrology model and the GIS will share the
et al., 1987). Linking the DISTFW hydrol- same database. There are various methods
ogy model with a GIS provided two major to implement this approach. This project will
objectives: (1) the development of a GIS tool- use higher-level application language asso-
box that will enable the automatic generation ciated with the GIS in the creation of links
of DISTFW input requirements; and (2) the between the GIS and DISTFW.
development of a GIS interface from which
the model can be launched. Objective 1 has
been achieved, with objective 2 to be com-
SIBERIA landform evolution model
pleted in the near future.
SIBERIA models the evolution of a catchment
The DISTFW hydrology model requires
through operations on cell-based (raster) dig-
the input of a significant amount of topo-
ital elevation data for the determination of
graphic information. Catchments are repre- drainage areas and geomorphology. GIS offer
sented within the model as being composed a wide range of raster data processing capa-
of a number of sub-catchments for which bilities, and a clear means for organising and
information must be derived describing their visualising data from a number of different
horizontal shape, vertical relief, conveyance formats (Rieger, 1998). Linking the SIBERIA
and flow relationships existing between the landform evolution with GIS therefore pro-
sub-catchments (Table 3). A significant chal- vides benefits not available in one or other of
lenge in this research project has been to these environments. The SIBERIA landform
develop a set of customised tools that auto- evolution model is computationally intensive,
matically generates this information from a and as a consequence, does not lend itself to
DEM. Six software tools have now been devel- interactive use. Integration of this model with
oped that extend the functionality of the GIS a GIS, therefore, requires the use of a rela-
to satisfy the topographic input requirements tively simple, loose coupling approach, which
of the DISTFW hydrology model. A descrip- involves transferring data from one system
tion of the tools developed for the derivation to another through the storage of data on file
of the required DISTFW inputs is shown in and the subsequent reading of that file by the
Table 3. other (Fedra, 1993).
It is proposed to link the DISTFW hydrology Although SIBERIA is based on relief,
model with the GIS using a ‘tight coupling’ the data formats used by the model are
Table 3. Descriptions of the tools developed to facilitate the automatic generation of the topographic input
requirements of the DISTFW hydrology model
Incidence tool Calculates the flow relationships between sub-catchments. Directly determines
‘maximum number of up slope sub-catchments’ and ‘sub-catchment
incidence’ for DISTFWs
Catchment-width tool Determines the average catchment width perpendicular to the central stream
channel. Directly determines ‘sub-catchment conveyance’ values for DISTFW
Stream-length tool Computes the length of a catchment, based on the central drainage channel.
Directly determines the ‘sub-catchment length’ values for DISTFW
MinMaxArea tool Calculates the minimum elevation, maximum elevation and area of each
sub-catchment within the catchment being studied. Directly inputs ‘UpSlope
Elevation’, ‘DownSlope Elevation’ and Sub-Catchment Area for DISTFW
Multi-point watershed Generates a grid of multiple watersheds. Where one point is downstream of
another, the intervening sub-catchment is automatically calculated.
Downstream tool Reduces the area of a sub-catchment where one sub-catchment is downstream
of another to the intervening area.
330 G. S. Boggs et al.
significantly different to those used by GIS. predicted by SIBERIA. It also shows a clear
There has been no direct methodology for differentiation in geomorphological activity
allowing the two to easily exchange data. between the less active floodplain areas and
Research has been directed at developing the more active upland plateau gorge bound-
GIS-based tools that provide for the direct aries of the Swift Creek catchment. No
formation of SIBERIA inputs and processing quantitative scale has been placed on the
of SIBERIA output data into a GIS readable grey scale in Figure 8, because of the diffi-
format. Tools have now been developed to culty in assigning spatial changes in model
exchange elevation data between SIBERIA parameters to the competent, very low erodi-
and the GIS. These help the user to prepare ble sandstone escarpment and uplands. The
DEM-based SIBERIA parameters within the highest incision occurs at the large change
GIS environment, and analyse SIBERIA of grade between the low gradient plateau
output using the spatial analysis capabilities and almost vertical escarpment. This pro-
of GIS. Further research is planned to allow cess has also been observed in the simulation
the user to prepare the input requirements, of the ERARM post-mining landform (Evans
launch the model and extract the model et al., 1998; Willgoose and Riley, 1998).
output without leaving the GIS environment, Use of one parameter value set applied to
thereby providing a relatively user-friendly the whole DTM surface results in greatly
front-end to this complex model. over-predicting erosion and deposition in
the vertical direction at the junction of the
cliff-face and plateau surface. The incorpora-
Application tion of spatially variable parameter values
to account for the highly resistant sand-
The evolution of the Swift Creek catchment stone escarpment and plateau surface will
was modelled for a period of 500 years, using be addressed in future research. Therefore
the parameters described above. Figure 8 interpretation of erosion and deposition at
shows the areas of erosion and deposition these areas can only be qualitative at this
Figure 8. Differences in elevation, indicating areas of erosion and deposition, between the Swift Creek
catchment at 0 years and after being modelled for a period of 500 years.
GIS and landform-evolution modelling 331
alternative such as waste-rock dump and anticipated that linking the landform evo-
infrastructure design variation. Impacts of lution model with GIS will provide a valu-
the alternative management scenarios on able tool for assessing the possible impacts
catchment evolution will be assessed over of mining impact on catchment sedimen-
both long and short-term time scales. Out- tary and hydrological processes. Additional
comes derived from these modelling sce- research is required to develop a more
narios can be used in the formation of fully integrated GIS and landform-evolution
management recommendations once final modelling approach that is beneficial for
decisions on mine development and design the proactive management of mining and
are made. more wide-ranging catchment management
scenarios.
Conclusions
Acknowledgements
The outcomes to date of this study pro-
vide a preliminary evaluation of integrated The authors thank Mr B. Smith and Mrs E. Crisp
hydrology and landform evolution mod- (eriss) for their assistance in the collection of the
elling techniques with GIS for assessing field data. We are also grateful to Drs G. Hancock
(University of Newcastle) and S. Timms (eriss) for
the possible impacts of mining on the their computer programming advice.
Swift Creek catchment, Northern Terri-
tory. A database has been established that
employs GIS as a framework for both spa-
tial and attribute datasets associated with
References
this project. This approach retains the flex-
ibility and functionality required to store Arkinstal, M., Willgoose, G. R., Loch, R. J. and
and manipulate each dataset independently, Pocknee, C. (1994). Calibration of DISTFW
Parameters for the QDPI Rainfall Simulator
whilst offering a central hub for the vari- Oaky Creek. Research Report No. 093.04.1994,
ous projects’ data. Hydrology and sediment Department of Civil, Surveying and Environ-
transport parameters were derived from mental Engineering, The University of New-
field data collected within the Swift Creek castle, New South Wales, Australia.
catchment. The derived hydrology param- Boggs, G. S., Evans, K. G., Devonport, C. C.
and Saynor, M. (2000). Developing a geo-
eters were used in the DISTFW hydrol- graphic information system for the geomorpho-
ogy model to predict annual hydrographs, logical assessment and management of mining
in order to determine long-term hydrol- impact. Journal of Environmental Management.
ogy parameters required by the SIBERIA In press.
landform evolution model. The predicted Bull, A. (1999). Past, Present and Future Devel-
opment of GIS at the Environmental Research
annual hydrographs were also used with Institute of the Supervising Scientist and the
the sediment transport parameters to derive Office of the Supervising Scientist. Internal
annual sediment loss values for SIBERIA. Report, Environmental Research Institute of the
This preliminary assessment of landform Supervising Scientist, Canberra. Unpublished
evolution in the Swift Creek catchment paper.
Burrough, P. A. and McDonnell, R. A. (1998).
demonstrates the complex process associated
Principles of Geographic Information Systems.
with the parameterisation of the SIBERIA Great Britain: Oxford University Press.
model. Charnock, T. W., Hedges, P. D. and Elgy, J. (1996).
Initial attempts to link the hydrology and Linking multiple process level models with GIS,
landform evolution models with GIS have IAHS publication no. 235. In Application of Geo-
indicated that the parameter derivation and graphic Information Systems in Hydrology and
Water Resources Management (K. Kovar and
modelling process can be simplified by the H. P. Nachtnebel, eds), pp. 385–393. Walling-
integration of these technologies. Linking ford: IAHS Press.
these models with GIS provides significant De Roo, A. P. J. (1996). Soil erosion assess-
advantages as the GIS assists in the deriva- ment using G.I.S. In Geographical Informa-
tion, storage, manipulation, processing and tion Systems in Hydrology (V. P. Singh and
M. Fiorentino, eds), pp. 339–356. London:
visualisation of geo-referenced data at a Kluwer Academic Publishers.
catchment-wide scale. Through the rapid pro- Desmet, P. J. J. and Govers, G. (1995). GIS-based
duction of modified input scenarios, it is simulation of erosion and deposition patterns
GIS and landform-evolution modelling 333
TR322. Ralph M. Parsons Laboratory, Erosion Model Calibration and Model Uncer-
Massachusetts Institute of Technology, Cam- tainty Assessment Suite User Manual. Research
bridge, Massachusetts. Report No. 108.03.1995. The University of New-
Willgoose, G. R., Bras, R. L. and Rodriguez- castle, Department of Civil, Surveying and En-
Iturbe, I. (1991). Results from a new model of vironmental Engineering.
river basin evolution. Earth Surface Processes Wockner, G. W. and Freebairn, D. M. (1991).
and Landforms 16, 237–254. Water balance and erosion study on the eastern
Willgoose, G. R., Kuczera, G. A. and Williams, B. J. Darling Downs: an update. Australian Journal
(1995). DISTFW-NLFIT: Rainfall-runoff and of Soil and Water Conservation 4, 41–47.