Using a System Dynamics Model as a Boundary Object in an
Integrative Approach to Regional Water Schemes in South Africa
Rob van Waas, Jill Slinger and Sander van Splunter
Delft University of Technology
Faculty of Technology, Policy and Management
Jaffalaan 5
2628 BX Delft, the Netherlands
+31 (0) 651713610
rob@vanwaas.net
j.h.slinger@ tudelft.nl
s.vansplunter@ tudelft.nl
Abstract: This article explores the use of a System Dynamics model as a boundary object in a case
study regarding decision-making on water scarcity in South A frica. The model integrates expertise
from the hydrological and ecological sciences with socio-economic information for a specific area,
the Mossel Bay region. The model proved to be adaptable to multiple stakeholders, robust enough
to maintain identity across stakeholders, and succeeded in allowing different stakeholders to work
together without necessarily requiring consensus. This study supported communication between the
stakeholders and enhanced the democratization of the decision-making processes by improving
deliberation! on contentious issues. Further applications of boundary spanning activities using
system dynamics modeling in other cases is recommended.
Key words: System Dynamics, Boundary Objects, Coastal/Estuarine Negotiation, Policy Analysis,
Water Management
1. Introduction
South African water institutions have undergone major changes after the democratic elections in
1994 and the new National Water Actin 1998. The main pillars of the South A frican National Water
Act of 1998 are sustainability, efficiency and equity. The water law strives to maintain a balance
between utilizing and protecting the water resources. However, the current process for establishing
rationing schemes is unable to deal with the increased competition over the scarce resource (Hughes
& Mallory, 2009) and governmental authorities are struggling with this challenge. Influential
hydrologists such as Hughes and Mallory (2009) recognize that relying on technical knowledge
alone is not sufficient to address this challenge. They urge that social and economic scientists step
in and help to understand and address the complex South A frican water system (van Waas, 2015).
This paper represents a response by system dynamic modelers to this call. Seeking to work at the
interface between decision-making and society, a system dynamics model is developed and used to
cross both disciplinary boundaries and the policy-science interface.
The Mossel Bay region in the Western Cape province in South A frica (Figure 1) is struggling with
the challenge of decision making on allocation of water during a period of scarcity. The region is
mostly dependent on freshwater from river a runoff that is stored in four dams (the Wolwedans,
1 Besides being a representative democracy South Africa aspires to be a deliberative democracy in which
deliberation is central to decision-making (Cohen, 1989). The use of a system dynamics model as a boundary
object seeks to enable the deliberation.
Klipheuwel, Emest- Robertson and Hartbeeskuil Dam). The main users of water are the Mossel Bay
town (~60.000 inhabitants), the ecosystem of Great Brak estuary, the agricultural sector and a large
gas-to-liquids plant operated by South Africa’s national oil company: PetroSA.
Figure 1: Mossel Bay region in perspective to South-Africa
During dry spells the storage provided by the dams fails to supply the full water requirement of all
users and rationing is required. In recent years, multiple droughts occurred that required rationing
of water (Makana, 2013; Mokhema, 2013; Mossel Bay Advertiser, 2009; Mossel Bay Municipality,
2011; PE Herald, 2011; Steyn, 2013). In Mossel Bay the regional water scheme? forms a contentious
issue. The municipality disagrees with the trade-offs that have been made and desires a more
consultative process (Mossel Bay Municipality, 2012).
In this paper, we employ the idea that system dynamics modeling can support conversations
between actors (Ackermann, Anderson, Eden, & Richardson, 2010; Beall, Fiedler, Boll, & Cosens,
2011; Stave, 2003), and that system dynamics modeling facilitates policy analysis processes
(Mayer, Daalen, & Bots, 2004). These attributes are combined in this research on whether a system
dynamics model can successfully be used as a boundary object. A boundary object allows different
people or groups to work together without requiring consensus or the same level of expertise. By
using a system dynamics model in a boundary spanning manner across different disciplines and into
the domain of civil society, the deliberative process is enabled. Our interest is to establish the extent
to which a system dynamics model can be used to facilitate both content and process in managing
a contended resource within a complex socio-ecological system.
First, the theoretical concept of using a model as a boundary object is elucidated and the chosen
case study together with the methods used in this paper are presented (section 2). This is followed
by a description and specification of the System Dynamics model (section 3). Then we explain how
the model was used (section 4), and evaluate its use as a boundary object (section 5) before
concluding the paper (section 6).
? The Regional Water Scheme is the arrangement in which the rationing is determined. It contains operating
rules that determine rationing based on the current water level in the dams.
2
2. Models as Boundary Objects
2.1. Boundary objects
Boundary objects are constructs that can enable communication and collaboration between
heterogeneous groups of experts, (local) stakeholders and scientists, even in non-consensus groups.
A scale model of a skyscraper is an example of a boundary object, because each individual will
recognize it as a skyscraper, albeit from their own perspective: an architect recognizes its aesthetic
aspects, an engineer focuses on construction aspects and a local community member sees it
bringing shade to their backyard.
Boundary object is a term coined by Star and Griesemer (1989) in working with heterogeneous
groups of stakeholders. Three main attributes of boundary objects are: interpretive flexibility;
material/organizational structure of different types of boundary objects, and the question of
scale/granularity (Star, 2010). As such, “boundary objects are a sort of arrangement that allow
different groups to work together without consensus” (Star, 2010, p. 602). Benefits of using
boundary objects in heterogeneous stakeholder groups aim at collaboration and the enhancement of
the sensibility to other stakeholders through the generalization of findings (Star & Griesemer, 1989;
Star, 2010).
Accordingly as a boundary object, a model would need to: (i) be adaptable to multiple stakeholders,
(ii) be robust enough to maintain identity across stakeholders, and (iii) succeed in allowing different
stakeholders to work together without consensus (Star & Griesemer, 1989; Star, 2010). These are
the three requirements for evaluating the functioning of a model as a boundary object.
2.2. A system dynamics model as a boundary object
The strategic nature of the water scarcity decision making problem of Mossel Bay, the long time
horizon and the limited availability of (technical) data on the regional water scheme argue for a
system dynamics approach. Moreover, the problem situation requires the cooperation of experts
from different fields and a deliberative process with citizens from all ranks and classes. This argues
for a boundary object. Accordingly, the approach of building a system dynamics model for use as
a boundary object in Mossel Bay was chosen.
The modeling was undertaken in South Africa by means of an engaged process with experts from
different fields, and stakeholders from the Mossel Bay region. The modeling process is depicted in
its simplest form in Figure 2. The role of the System Dynamics modeler was to translate the
knowledge held by the experts and stakeholders into a single, connected model and to create an
implementation in Vensim (version 6.3). The knowledge and information of the experts and
stakeholders was accessed through a series of interviews. These interviews were conducted with
individuals, not in a group modelling process. The choice for individual interviews was made from
a practical and methodological viewpoint. First, experts were located at great geographical distance
from each other and second, they did not agree to meet and collaborate. Third, separate interviews
also allowed more time for exploration of the individual mental models of the experts and
stakeholders. An overview of the different experts and stakeholders consulted in the modeling
process is provided in Appendix A.
FF =
Figure 2: Simplified modeling process
3. A Model for Determining Regional Water Schemes in South Africa
The crux of the water scarcity problems in the Mossel Bay region was found to revolve around the
operation of the largest storage dam in the area, the Wolwedans Dam. During this research the
Dynamic Water Allocation Model (D-WAM) has been created and specified for the Wolwedans
Dam. The multiple subsections are connected as shown in Figure 3. The six most important
subsections of D-WAM are specified in detail: (i) the Wolwedans dam subsection, (ii) the Mossel
Bay municipality subsection, (iii) the downstream Great Brak estuary subsection, (iv) the local
Great Brak community subsection, (v) the PetroSA subsection, and (vi) the upstream agricultural
subsection.
Two additional sub-sections, the Klipheuwel dam subsection and the downstream agriculture
subsection, are adaptations of the Wolwedans dam and upstream agricultural subsections. Because
their structure is so similar to the aforementioned subsections, they are not described separately.
Other substructures such as forestry, evaporation and overflow are relatively small and described
in Appendix B. Appendix B contains a list of the D-WAM variables together with the uncertainty
space over which they can be simulated and references to the data sources used.
Runoff into
Dam
Upstream Agriculture
(wi)
(monthly constant
use)
Figure 3: Connected Sub-M odels in Dam Operation Model
3.1 The Wolwedans dam subsection
The volume of freshwater in the Wolwedans dam (x,) is influenced by the runoff into the dam from
the Great Brak river (x,,), the rainfall directly onto the surface of the Wolwedans Dam (x),
evaporation from the Wolwedans Dam (x,3), overflow of the Wolwedans Dam (x,,) and extraction
of water from the Wolwedans dam (x,;) for different uses downstream. This results in the following
equation:
da
ag U1 = Maa + X12 — X13 — X14 — Ms
The runoff into the dam from the Great Brak river (x,,) uses a time dependent runoff function
(runoffWDf(t)) and is affected by the upstream use of water for agriculture
(Useagricutture upstream) and the streamflow reduction by plants and trees (streamf lyorrest). The
streamflow reduction is calculated by making a simplified streamflow reduction per square
kilometer of forest and calibrating this to the data used in the RWS study (Mallory, Ballim, &
Forster, 2013). The rainfall directly onto the surface of the Wolwedans Dam (x, ) is determined by
a time dependent rain function (rainwD f (t)) which is based on hydrological data (see appendix
B). The evaporation of water from the dam (x,3) is determined by a time dependent evaporation
function evapW Df (t). The overflow of the dam (x,4) occurs when the current volume of water in
the dam (x,) exceeds the capacity of the Wolwedans dam (capyp) and more water comes in than
the sum of water extracted for use (x,;) out and evaporates (x,,) at that moment in time. The
extraction of water from the Wolwedans dam (x,5) is the sum of use by the estuary (useestuary),
water used by the Mossel Bay municipality(usemosseay), Water used by PetroSA (usepetrosa)
and water used by downstream irrigation (us€q gricuiture downstream):
X41 = runof fWDf (t) — useggricutture upstream — (Surfacerorrest * SfTyorrest)
X12 = rainWDf (t)
X13 = evapWDf (t)
Xy4 = MAaX(Xy1 — (X43 + X45) if x, > Capywp and 0 otherwise
X15 = USestuary + USCmosselbay + USCpetrosa + USCagricuiture downstream
3.2 The Mossel Bay municipality subsection
The population of the Mossel Bay municipality (x.) changes by the amount of births in Mossel Bay
(x21), the deaths in Mossel Bay (x22) and the net amount of people migrating to Mossel Bay (x23).
The equation for the population of Mossel Bay would then be:
d
ap 2 = 721 + X23 — X22
The amount of births (x,,) and deaths (x2,) are calculated by multiplying the population of Mossel
Bay (x,) with the birth rate (br,,,,) and the death rate (dr,,,) of Mossel Bay. The amount of people
migrating to and from Mossel Bay has been put in a single net migration that is calculated by
multiplying the population of Mossel Bay with a net migration rate (mr,,,).
Xo, = Xz * btiny
X22 = Xz * Uinp
X23 = X2 * Mnp
The total number of tourists residing in Mossel Bay(x3) changes by the arriving of tourists in Mossel
Bay (x3,) and tourists leaving Mossel Bay (x3).
zz X3 = X31 — &
at *3 31 32
The arrival of tourists in Mossel Bay (x3,) is calculated by multiplying an average number of
tourists (at,,,) with a seasonally oscillating function (touristf(t)). The departure of tourists is
dependent on the average staying time for tourists (astt) and the number of tourists that are
currently in Mossel Bay (x3).
X31 = Atm, * touristf (t)
2p
%a2 = astt
The domestic demand coming from the Mossel Bay municipality (demand,,,,) is then calculated
by multiplying the amount of people in Mossel Bay with a demand for water per person per month
(dpp).
demand» = (Xz + X3) * dpp
3.3 The downstream G reat Brak estuary subsection.
The Great Brak estuary subsection is based around the estuary with an indicator that represents the
estuarine health (x,). The health can either increase (x,,) at a certain pace, or deteriorate at a certain
pace (x42). This estuarine health is an abstract number in the case of this model. It has a range
between zero and two, zero representing a biologically degraded ‘dead’ estuary, two representing a
very healthy estuary and one representing the estuary in its present state.
X= Ky — X.
at ** 41 42
The increase and decrease are both dependent upon the fraction of water that is supplied (x43) and
the current level of health (x,). The fraction of water supplied (x,3) equals the water that is supplied
(averagesuppliedestyary) a8 a Tunning average over twelve months divided by the water that is
required to retain health (x,,). The amount of water that is required is calculated with a function
that is dependent on the current health of the ecosystem (waterrequiredf (x4)). The effect of
supplying enough water is larger if the estuary is further away from its maximum health
(health ax). And the increase effect is spread over several months by the delay in health increase
(delayneattnincrease). Analogously, for the decrease of health, supplying less water than required
will make the health decrease more strongly and if the health comes closer to zero, the decrease will
become less. This effect occurs over some time, the delay in health decrease (delaypeaitnaecrease)+
_ X43 *(healttmax—X4):
%aa = max(0, delayneatthincrease
_ (1x49)
*a2 = max(0, delaYnealthdecrease
averagesupplied estuary
x43 =
X44
X44 = waterrequiredf (x4)
3.4 The local Great Brak community subsection
The quality of living conditions for the people in Great Brak (LQ,,) is included as an index in the
model.
Xs + (1—Xx6) + *
LQgn = —_—_—
The living qualities are determined by the attractiveness of Great Brak to tourists (x,), the effect
that a flood has on the area (x,) and the health of the estuary (x,). The attractiveness to tourists
(xs) is modeled as a stock which restores (x;,) to a certain level after it has been decreased by the
effects of a low water quality (x52) or a flood (x;3). A flood also has a direct effect on the quality
6
of living conditions (x) this effect goes up after a flood occurred (x,,) and slowly dies out if time
passes after a flood (x2). The check to whether a flood occurs is based on the amount of water that
is spilling over the dam. This is a simplification, since in reality it would depend on the water level
in the estuary. There is a strong connection to the spillover and the water level of the estuary,
however tide and timely breaching also play a role.
a Xs = X51 — X52 — X53
dt
Xe = Xo — X
dt 6 61 62
1-x5
Xs, = ———_—
et delay;agp
X52 = X5 * (1— ef fectyarf (X4))
X53 =x; if ‘flood = yes' and 0 otherwise
Xo. = max (0,1 — x6 + X62) if ‘flood = yes’ and 0 otherwise
Xe
Hen = durationyiooa
flood = yesif x14 > floodoversiow
3.5 The PetroSA subsection
The PetroSA subsection is modeled relatively simple. The processes in the plant have not been
modeled, but a constant operation is assumed, requiring a constant monthly amount of water
(demand etrosa)- This demand can be met or not resulting in a certain utilization of the PetroSA
plant (x,). This is a running average of the fraction that the plant is in use (operatingyetrosa) OVET
a year. How much the plant is in use at a certain moment is a function of the amount of water that
is supplied to the plant (operatingpetrosaf(X7)). PetroSA also uses 1.000 ie of the Reverse
Osmosis plant that runs on Mossel Bay effluent.
3.6 The upstream agriculture subsection
Agriculture is practiced both upstream as well as downstream of the Wolwedans dam, however
mostly upstream. It therefore is difficult to ration in practice, since it extracts water before it is
inside the dam. There is also some agriculture downstream which is included in the model. Only
the upstream agriculture is specified in this article, since the structure is very similar.
Central in the agricultural subsection is the total area of land in use (xg). This changes when new
land is taken in use (xg,) or land is reduced for other uses (xg2).
d
a? = Xe1 — Xg2
New land is taken in use for agriculture when there is an attractiveness for agriculture (x.) and there
is area available for the construction (ta,,,). A certain period is taken into account for the
construction and abolishment of agricultural land (delayag;i).
(9 — 1) (aa = Xs)
delayagri
(1 = x0) (Xe)
delayagri
Xg, = max(0,
Xg2 = max(0,
The monthly demand that the agriculture has (demand,,,) is determined by an average for water
consumption of the crops that are grown (consumption,;op;), together with a seasonal factor for
irrigation (irrigationf (t)) multiplied by the amount of land on which agriculture is practiced (xg).
The attractiveness of agriculture upstream (x4) can rise (x9;) or fall (x92) due mostly by the amount
of water that is supplied compared to the desired amount of water (fractionsupplied,,). The
attractiveness has a ceiling (maxattr,,,) and a tipping point (tippingpointattr,,,) at which level
of rationing it becomes unattractive for farmers to have more agricultural land. The fraction that is
supplied to farmers (fractionsupplied q,,) is calculated over the period of the last twelve months.
The model uses the following formulas for this:
demandgy = Xg * consumptiongrop, * irrigationf (t)
d
a? = Xo1 — X92
X91 = max(0, ((fractionsupplied,, — tippingpointattr,,)
* (maxattray, — Xo))
Xo2 = max(0, ((tippingpointattr,,, — fractionsuppliedg,) * maxattrgy,)
t
f_,, demand,
fractionsuppliedg, = a
Sp12 Sau
This model as a set of differential equations has been instantiated and simulated in Vensim.
Appendix C shows how functions have been implemented as table functions in Vensim, appendix
D shows screenshots of the different sub-models in Vensim, appendix E shows behavior of runs in
graphs and finally appendix F depicts a small test to verify the Euler integration for this modeling
instance.
4. The use of the model as a boundary object
The use of the D-WAM model in boundary spanning is depicted in Figure 4 (next page). The
diagram shows that via model simulations the experts received feedback on how their sub-system
influences, and is influenced by, the other sub-systems. This provoked some interesting discussions,
particularly on the level of detail that needs to be included in such a boundary spanning model. For
instance, initially the ecologists wished to include a great deal of detail on the response of the
downstream estuary to different water allocations. Only by receiving feedback from D-WAM
simulations did they come to understand that the present level of detail of the model enables a
different discussion than is currently held. The new discussion focused at national level decision
making on the regional water schemes. So, the system dynamics model worked as a boundary object
to select and focus the discussion. This experience represents one of many examples of the experts
who gained new, interdisciplinary insights by engaging with the model.
The translation of model outcomes into scorecards provides the means by which citizens can
interact with the D-WAM model. The scorecards present the outcomes of interest on a colored scale.
These outcomes represent the effects of different combinations of dynamic allocation alternatives
and different run-off scenarios (including different water scarcity situations). The citizens can then
rank the combinations according to their own preferences. Using the simplified scorecards enables
citizens that are uncomfortable with quantitative models to participate in deliberations on water
allocations. This may apply to many citizens that are affected by the Mossel Bay regional water
scheme. By facilitating inclusive model-based decision making, the D-WAM boundary spanning
process potentially addresses the concem of the Mossel Bay municipality for more deliberation.
Further, there are two types of information flowing out of this process into national level decision-
making on regional water schemes. These include (i) interdisciplinary knowledge on the system
that is gained from the modeling process; this could be information on the effects of the operating
tules on the resilience of the ecosystem, and (ii) information on the values of citizens contained in
the trade-offs they make regarding water allocations. It should he stressed that this last step is very
meaningful given the South A frican ambition to be a participatory democracy.
National level decision-making
A A
Combined knowledge on Information on citizen and
system simulated over stakeholder values over the
uncertainty outcomes of interest
Boundary Object: System Dynamics Model
Translates into |— a
JOvtcome ui ifoutcome Wf
Joutcore u |
Purcome
rh oscars
| [Guisore wt vicome
IT Hydrology J
Rank the different
t altematives
ci
Expert Groups Citizens
Figure 4: Use of the Model as a Boundary Object in the C ontext of the Existing Process
5. Evaluating the use of the model as a boundary object
To recap, a boundary object exhibits the following three attributes: (i) it is adaptable to multiple
stakeholders, (ii) it is robust enough to maintain identity across stakeholders, and (iii) it succeeds in
allowing different stakeholders to work together without consensus. The questions that now need
answering are: “Did the D-WAM model function as a boundary object?” and “How well did it
function as a boundary object?”
D-WAM modeling sessions with different experts took place at different geographic locations, at
their convenience. This means that the experts and stakeholder were not constrained in expressing
their views by the presence of others, nor did they have to agree with each other. Instead, the systems
modeler travelled rather than the experts and stakeholders. By designing the consultation in this
way the process of model building and interaction succeeded in allowing stakeholders to work
together without forcing consensus (satisfying the third criterion above). During the sessions two-
way exchanges of knowledge occurred. This meant that the modeler gained knowledge on the
required structure and behavior of the sub-model to which the expert(s) were contributing. At the
same time, the experts gained knowledge on the appropriate level of detail required for connection
9
between the sub-models. The experts expressed that they leaned by having to make their
knowledge of dynamic interactions explicit, as the following example illustrates.
A modeling session with an ecologist, water specialist and hydrologist on the 30" of July (Appendix A). During
this session, insights on the functioning of the estuary were shared by the experts with the modeler and the experts
learned more about the connections of the estuary — their sub-system of interest — to tourism, of particular interest
to the Mossel Bay municipality.
When interacting with the model, different stakeholders can focus on different aspects: e.g., a
hydrologist may see infrastructural issues such as the capacities of pipelines, dams and reclamation
works, while a farmer may see seasonal patterns in the water availability that affect his irrigation
scheme. Finally municipal rep ives may bei 1 in water pricing for households. These
examples underline the adaptability of the system dynamics model to multiple viewpoints
(satisfying the first criterion).
The model also maintains its identity across multiple viewpoints (the second criterion). Some
aspects of the D-WAM model are recognized universally by all the different experts and
stakeholders: e.g., flood occurrences orvariations in the water price over time. This can be regarded
as the model maintain identity across stakeholders, despite its adaptability to multiple uses.
In summary, the Dynamic-Water Allocation Model proved to be:
e Adaptable to multiple stakeholders, in the sense that it allowed for experts and stakeholders
to contribute to the model, at their own convenience and level of understanding, and gain a
diversity of insights from the model.
e Robust enough to maintain identity across stakeholders, since the model is simulated in an
integrated fashion allowing interactions between the different sub- models.
e Successful in allowing different stakeholders to work together without consensus. The lack
of cc on model } was dealt with by specifying uncertainty ranges for the
parameters. This allowed the process to continue, while the participants can agree to
disagree and yet keep working with the model.
This means that the D-WAM model functioned as a boundary object within this research endeavor
and can act to facilitate further deliberation in the decision-making on water allocation. However,
amore extensive use of the D-WAM within the Mossel Bay region would require interaction with
a broader representation of stakeholders and citizen groups, rather than the experts consulted in this
development and initial application phase.
Our further interest is to establish the extent to which system dynamics modelling can facilitate the
integration of specialist knowledge into decision making processes in the management of contended
resources within complex socio-ecological systems. By conceptualizing and using a system
dynamics model as a boundary object it can serve as a catalyst for interactions that involve
individual stakeholders at multiple levels from decision makers to specialists to local citizens. This
integrates both the content and the processes within resource management, but needs to be validated
further in practice.
6. Concluding Remarks
Boundary objects and System Dynamics rarely coincide in the scientific literature, and there is little
research on the use of a System Dynamics model as a boundary object. Traditionally System
Dynamics has been used in a rational, advisory style and more recently in a consensus-seeking
Group Model Building style. The experience from this study reveals that a System Dynamics model
can be useful in eliciting experts’ knowledge and stakeholders’ perspectives. The model can act to
10
allow communication across disciplinary boundaries and can span the science-policy divide. In our
opinion using a System Dynamics model as a boundary object can help in democratizing and
improving decision-making processes in controversial policy areas. We recommend that in-depth
applications are performed to test whether the promise identified in this study holds true both in a
broader application within our case study and in other situations.
11
References
Ackermann, F., Anderson, D. F., Eden, C., & Richardson, G. P. (2010). Using a group
decision support system to add value to group model building. System Dynamics Review,
Vol. 25 No.4, 26(4), 335-346. doi:10.1002/sdr
Beall, A., Fiedler, F., Boll, J., & Cosens, B. (2011). Sustainable water resource management
and participatory system dynamics. Case study: Developing the Palouse basin
participatory model. Sustainability, 3(5), 720-743.
Cohen, J. (1989). Deliberation and Democratic Legitimacy. D ebates in Contemporary
Political Philosophy, 342-360.
DWA. (2013). Strategic Overview of the Water Sector in South Africa 2013. Pretoria.
Retrieved from http://nepadwatercoe.org/wp-content/uploads/Strategic-Overview-of-the-
Water-Sector-in-South-A frica-2013.pdf
Hughes, D., & Mallory, S. (2009). The importance of operating rules and assessments of
beneficial use in water resource allocation policy and management. Water Policy, 11(6),
731. doi:10.2166/wp.2009.035
Institute Water for Africa. (2014). Water Consumption figures webpage (datasource FAO).
Retrieved June 6, 2015, from http://www. water-for-africa.org/en/water-
consumption.html
Lehohla, P. (2013). Tourism 2012 (Vol. 02). Pretoria: Statistics South A fria.
Makana, C. (2013). Drought cripples Limpopo farmers. Retrieved from
http://www. iol.co.za/news/south-africa/limpopo/drought- cripples-limpopo-farmers-
1.1448228#V CV bkPna5Qs
Mallory, S., Ballim, F., & Forster, L. (2013). Mossel Bay Regional Water Supply Schemes
Report (Study Development of Water Supply and Drought Operating Rules for Stand-
Alone Dams and Schems Typical of Rural/Small Municipal Water Supply Schemes:
Southern Region. Pretoria.
Mayer, I. S., Daalen, C. E. Van, & Bots, P. W. G. (2004). Perspectives on policy analyses: a
framework for understanding and design. International J ournal of Technology, Policy
and Management, 4(2), 169. doi:10.1504/IJTPM.2004.004819
Mokhema, T. (2013). South Africa’s North West Province Is “Drought-Stricken.” Retrieved
from http://www.bloomberg.com/news/2013-09- 18/south-africa-s-north- west- province-
is-drought-stricken- html
Mossel Bay Advertiser. (2009). Dam Level Remains Low. Retrieved from
http://saweatherobserver.blogspot.nl/2009/10/dam-level-remains-low.html
Mossel Bay Municipality. (2011). Mossel Bay water situation rosier, but no room for
complacensy. Retrieved from http://saweatherobserver.blogspot.nl/2011/01/mossel-bay-
water-situation-rosier-but.html
12
Mossel Bay Municipality. (2012). Proposed DWA Operating Rules: Comments and
Recommendations. Mossel Bay.
PE Herald. (2011). Eastern Cape Dam levels up, but restrictions stay. Retrieved from
http://saweatherobserver.blogspot.nl/2011/06/eastem-cape-dam-levels-up-but.html
Smithers, J., & Schulze, R. E. (1995). ACRU agrohydrological modelling system user
manual. WRC Report TT 70/95. Pretoria.
Star, S. L. (2010). This is Not a Boundary Object: Reflections on the Origin of a Concept.
Science, Technology & Human Values, 35(5), 601-617. doi:10.1177/0162243910377624
Star, S. L., & Griesemer, J. R. (1989). Institutional Ecology, “Translations” and Boundary
Objects: Amateurs and Professionals in Berkeley’s Museum of Vertrebate Zoology.
Social Studies of Science, 19(3), 387-420.
Stave, K. (2003). A system dynamics model to facilitate public understanding of water
management options in Las Vegas, Nevada. Journal of Environmental Management,
67(4), 303-313. doi:http://dx.doi.org/10.1016/S0301-4797(02)00205-0
Steyn, A. (2013). South A frica: Drought Crisis - Gov Must Urgently Intervene. Retrieved
from http://allafrica.com/stories/201311271428.html
Van Waas, R. (2015). Designing an integrative approach to regional water schemes in South-
Africa (Master Thesis). Technical University of Delft, Delft. Retrieved from
http://repository. tudelft.nl/assets/uuid:456082a4- 1f3b-4d91-8922-
ab27a836a86a/Master_Thesis Rob_van_Waas_Repository.pdf
13
Appendix A List of interactions with stakeholders concerned with RWS decision-making
An extra character was added in the name field to prevent showing in search engines.
Date Name Field Purpose Type of interaction
THQ i geisoner Intemational Relations Networking Collegial Talk
T4201 | 0 reunke Public Policy Networking Collegial Talk
7% ry 5
TEI Te soyo Collegial Talk
THQ Tr cigaassen Ecology Networking Collegial Talk
“E A
WED Te Norsie Collegial Talk
BD | Student Networking Collegial Talk
WED Ts vatosry Hydrology Networking Tnterview
WED | rigot ‘Anthropology Gaining insight Excursion
T5204] 6 aisiory Hydrology Networking Interview
552018 Ts watiosiy Hydrology Gaining insight Interview
1 Networking Workshop Session
F420 | 1 Govgender Networking Workshop Session’
FADE | roses Networking Workshop Session’
FET | Oegeman Networking Workshop Session
£2014 Water Law Networking Workshop Session
A Pigke
FEIT |v resddy Networking Workshop Session’
E20 TS ugggunan Networking Workshop Session’
FEIT vr xghan Networking Workshop Session
£2 ; 7
FEIT sso Networking Workshop Session
£2 ; 7
E201 | van der$Meu Networking Workshop Session
“£2 ; vi
FEE sty Networking Workshop Session
L204 Networking Workshop Session
B Zwane
E20 | a vinsele Networking Workshop Session’
FEIT Ts sisngh Networking Workshop Session
FET | Sa Networking Workshop Session
52018 | costes Ecology Gaining insight Introduction Meeting
5204 | saris Ecology Gaining insight Brunch
520 | Kragne Water Stewardship Gemany | Gaining insight Brunch
25-20 | vangory Hydrology Gaining insight Tnterview
F6-201E | ss thomspson Water Law Gaining insight Tnterview
75-204 | a veisssner Intemational Relations Gaining insight Excursion
36-2014 Govemment Gaining insight and Validation | Interview / excursion
N Fougrie
19-6-2014 Citizen of Great Brakriver | Gaining insight Tnterview
H Vengter
196-2014 | > aewset Citizen of Great Brakriver | Gaining insight Tnterview
BEET atosry Hydrology Modeling Tnterview
256-2014 Coastal science.‘ & | Gaining insight Interview
P de Villgier
FEEDOM Ty suininsyga Coastal engineering Gaining insight Tnterview
256-2014 Water Quality: Estiary and | Gaining insight Tnterview
$ Taljaagrd Marine
EDIE | an Nieskerk Ecology Gaining insight Tnterview
IEE | Crargk Ecology Gaining insight Tnterview
14
276204] Weiagemann Govemment Gaining insight Tnterview
30-620 Te ruining Coastal engineering Modeling Modeling Session
30-6-2014 Water Quality: Estiary and | Modeling Modeling Session
$ Taljagard Marine
BOEDOE Tan Nsickerk Ecologist Modeling Modeling Session
30-6 7
306-20] rrerson Coastal engineering Validation Presentation & Discussion
30-6-2014 Validation Presentation & Discussion
B Gwesba
30-6: 7
30620 | o sboslambi Validation Presentation & Discussion
30-6-2014 Validation Presentation & Discussion
H Mp$e
36204 | Hydrodynamic modelling Validation Presentation & Discussion
30-6201 Te van Bgallegooyen Validation Presentation & Discussion
30-6-2014 Validation Presentation & Discussion
M Carstengs
WE TG Validation Presentation & Discussion
30-620 Ts rataagrd Validation Presentation & Discussion
— >
306-20 | an Nsiekerk Validation Presentation & Discussion
FTF Ecology Gaining insight Tnterview
E Merstz
F720 6 citnger Govemment Gaining insight Interview
F720 |» waistoo Govemment Gaining insight Interview
€7-2014 Intemational Relations Validation Presentation & Discussion
R Mesissner
87-2014 ublic Policy Validation Presentation & Discussion
N Fungke
Te 7
87-201 | a Mosyo ‘Anthropology Validation Presentation & Discussion
oT 7
E720 | a cragassen Ecology Validation Presentation & Discussion
€ 72014 ‘Anthropology Validation Presentation & Discussion
K Nortstje
B72 | Student Validation Presentation & Discussion
T6720 TO nesker ‘Anthropology Validation Presentation & Discussion
167-2014 Researcher HIE Validation Presentation & Discussion
Z Nkugna
16-7-2014 Candidate Researcher Validation Presentation & Discussion
E MamSakwa
TET | 6 apatgi Manager Validation Presentation & Discussion
T6-7-2014 Researcher Wash & Public | Validation Presentation & Discussion
B Map$osa Health
16-7-2014 Researcher Water Quality Validation Presentation & Discussion
E Ngorigma
16-7-2014 Researcher Numerical | Validation Presentation & Discussion
P Pagge modeler
15
Appendix B_ Variables and Uncertainties
The model variables are described, uncertainty ranges are provided and units for the variable are
provided in the table below.
Model variable Description Range Units
Reduction of runoff by | The amount of runoff that is reduced by forestry. Methods are | 7500 - 9500 m
upstream trees _—_etc. | available to assess this, currently the value is backwards engineered Saonihi
(Streamf lrorrest) form a more extensive study (Mallory et al., 2013, pp. 4-3).
Rain on Wolwedans dam | Currently notin model. In reality this should be a function of the water > m
(rainw Df (t)) surface as well, however this has been kept out of the current model. aon
South Africa does have rainfall data for the dams available. (see
appendix C)
Evaporation from | Currently notin model. In reality this should bea function of the water mm
Wolwedans dam | surface as well, however this has been kept out of the current model. ont
(evapw Df (t)) South Africa does have evaporation models for the dams available.
(see appendix C)
Runoff into Wolwedans | The runoff into the Wolwedans dam is just downstream from the 0,01-27,22 m
dam (runof fWDf (t)) quatemary catchment area K20A. The time dependent function that is inonih
used is based on simulated runoff for the period 1920 to 2010. The
unit for this is m* per unit of time (see 0 for more on the table
functions).
Capacity of the Wolwedans | The amount of million cubic meters of water can be contained in the 25,5 m
dam (capwo) dam at maximum capacity. This is found in (Mallory et al., 2013, pp.
3-2) and is relatively certain.
The population of the | Information taken from the Census (Census, 2011) 39430 person
Mossel Bay municipality
(x,)
Birth rate of Mossel Bay | Had difficulty finding accurate values, see migration for approach in person
(Ca) model. person = month
Death rate of Mossel Bay | Had difficulty finding accurate values, see migration for approach in person
(dnp) this model. person» month
‘Net migration rate Mossel | Since little data was found on birth, death and migration rates the 0,00187 BETSON
Bay (mrp) growth over ten years has been used to calculate a net growth rate for person = month
the three combined (Census, 2001 & 2011).
The average staying time | An estimate for the average time that tourists stay on theirholidayin | 0,10-1 ‘month
for tourists (astt) the area. No data was found on this, so an estimate is used.
The average amount of | The average amount of tourists that are staying. This value is | 15.000- person
tourists in Mossel Bay | multiplied by the seasonal impact function to get to how many tourists 25.000
region (aty,,) would nommally artive. No data was found on this, so an estimate is
used.
‘Average demand for water | The average water demand per parson in the Mossel Bay region. The | 0,75 —3,5 me
per person per month | basic reserve component is 25 liters per person per day (0,75 cubic
(dpp). meters per person per month) (DWA, 2013). The UN states 50 liter
per person per day is required (1,5 cubic meters per person per month)
and Germany uses 122 liter per person per day (3,6 cubic meters per
person per month) (Institute Water for Africa, 2014).
person * month
The surface of the forest | Itis found that this is 28,8 square kilometer (Mallory etal., 2013, pp. 288 km
area upstream of the | 4-3).
Wolwedans Dam
(Surface rorrest)
‘K streamflow reduction per | This is deducted from a deeper study into this (Mallory et al., 2013, 8622 me
square kilometer constant | pp. 4-3). That study used the 2006 streamflow reduction curves neem
(Sftrorrese) generated by ACRU (Smithers & Schulze, 1995).
The estuarine health (x,) | This in an arbitrary indicator for estuarine health. This should be 0-2 Dine
validated with the ecologists so that is captures the main behavior that
the estuary would exhibit given the water supplied. There should
always be a translation step by experts to make sense of this value
The maximum health the | The maximum value for the indicator for estuarine health. 2 pateisendess
estuary can have
(healthyyar)
The time over which an | The time the estuary needs to recover its health from being without 48 - 250 month
increase in health is spread | water for a certain period. This value needs to be calibrated using
(delay ) experts and data on the estuary.
The time over which a | The time the estuary will take to decrease in health when being 5-40 month
decrease in health is spread | supplied less than is required. This value needs to be calibrated using
(delay, ) experts and data on the estuary.
Recovery time of tourist | The time that the effects of a low water quality or flood diminishes 12-60 month
opinion on flood | for tourists. This is an estimate that should be validated.
(delayragp)
Duration of effect flooding | The duration a flood has a negative effect on a community. This isan 2 month
(durationfiooa) estimate that should be validated.
16
Variable for flood in
A flood occurs if the water level in the estuary rises. The water level 750.000 m
estuary (flood verpiow) is dependent on the amount of water in the estuary. In goes: overflow, ann
water served, rainfall and (some) runoff and out goes water into the
sea. In this case the variable is only measured using a certain overflow
of the dam. It provides a reasonable estimation for floods.
The demand of the PetroSA | PetroSA has an allocation of 5,6 million m3/annum from the 460.000 m
plant | Wolwedans Dam. This is being used fully in recent years (Mallory et Front
(demand yetrosa) al., 2013, pp. 4-2)
The total amount of land | Estimate — no reliable data available to me at this time. The area, 100.000 km?
available for agriculture | consumption per square kilometer have been reversed engineered
upstream (taqu) the figures.
Total area of agricultural | Estimate ~ no reliable data available to me at this time. The area, 10000 km?
land upstream (xx,) consumption per square kilometer have been reversed engineered
the figures.
Delay to construct or | Delay for farmers to respond to a change in the situation of water 36-60 month
abolish agricultural land | management. This is an estimate that needs validation.
(delaya,ri)
The average consumption | Estimate — no reliable data available to me al this ime. The area, 5 mi
of water for crops | consumption per square kilometer have been reversed engineered im «month.
( i ) from the figures.
tippingpointattra, The point in which farmers really start to get appalled by the water | _0,7-0,9 | Omensones
shortages. This is an estimate that needs validation.
maxattrn, The maximum value for the indicator for attractiveness of agriculture Z Dimensionless
upstream,
17
Appendix C Table functions in System Dynamics Model
In this appendix the table functions that have been used in the System Dynamics model will be
briefly introduced.
runof fWDf (t): Table function to determine the runoff into the Wolwedans dam. This function is
based on (simulated) hydrological data over a period from 1920 to 2010. In Figure 5 the table
function is presented as a graph. For rainWDf(t) & evapWDf (t) similar graphs will be used as
input. However these are presently not yet made available. In the current model therefore is assumed
that rainfall and evaporation cancel each other out. This is true over the span of a year, however can
make a difference on a monthly timespan.
Figure 5: Table function runoff into Wolwedans dam
tourist f(t): Table function to determine the number of tourists over time. This function is added
to account for the different seasons of the year regarding the number of tourists that reside in Mossel
Bay. Since a large share of the water is used by tourists this is added. The function is based on a
statistical study on tourism in South A frica (Lehohla, 2013, p. 13). In Figure 6 the table function is
presented as a graph. The x-axis (time) has a maximum of 12 in which each number represents a
month from January to December.
Gap Looky -SEASONAL INFLUENCE TOURISM MOSSEL BAY
Figure 6: Table function for tourists over time
18
waterrequiredf(x,): Table function for the water required for the estuary based on the current
level of health of the estuary. This is based on the expert session that was held at Stellenbosch on
30-06-2014 together with personal correspondence with Jill Slinger. This function might be
debatable and could be a good candidate for testing multiple table functions against each other. In
Figure 7 the table function is presented in a graph. At normal health (a value of 1 on the x-axis) the
requirement will be set at 800.000 cubic meters per annum. At low health this will increase to
1.100.0000 cubic meters per annum and at high health 600.000 cubic meters per annum. The
assumption hereby is that a healthy estuary is less ‘thirsty’ than an unhealthy estuary is.
‘Graph Loskp- EWR PER MONTH FOR ESTUARY
1 S106
0
a1 a
Figure 7: Table function for water required for estuary over estuarine health
ef fectyqrf (x4): Table function for the effect that a low water quality in the estuary has on the
attractiveness to tourists. The effect only occurs when the estuarine health gets below 1 and will
especially start having an effect if it gets below 0,5. In Figure 8 the table function is presented in a
graph.
4
Figure 8: Table function for the effect of water quality on the attractiveness for tourists
operating petrosaf (x7): Table function to determine the level of operation at PetroSA depending
on the fraction of its demand that is being met. Since PetroSA operates three units that can be
19
switched on or off the operating level will have three levels as well. In Figure 9 the table function
is presented in a graph.
(raph Lackap - FRACTION OF PETROSA OPERATIONAL
Figure 9: Table function for the level of operating at PetroSA depending on the fraction of demand for water supplied.
irrigationf (t): Table function to account for the seasonal variation in the demand for irrigation
for agriculture. At this moment this is just an estimate that should be further evaluated and validated
by experts from the region.
‘GephLoolup- SEASONAL FACTOR FOR IRRIGATION
Figure 10: Table function for the seasonal influence on irrigation water requirements
20
Appendix D Sub-models in Vensim
The following images show the structure of the model as implemented in Vensim.
‘MAXIMUM
WOLWEDANS DAM
VOLLUME
WOLWEDANS DAM
FILLED
Wolvvodana Dam Water Volume ‘evaporation from
wolwedans dara
runoff fom great
rake iver
‘water is extracted for
‘use wolwrodans
TURN OF UPSTREAM
USES GREAT BRAK
oid runoff great
Figure 11: Wolwedans Dam Sub-Model
Upstream Irrigation
‘Great Brak River
<azasonal
FACTOR FOR
ER.
DEMAND MET UPSTREAM
Figure 12: Upstream Agriculture Sub-Model
21
Estuary,
-RaTIONING VOLUME
— a Rees
ap pate ERT os
Ancor WATER TOMO NOL,
sdb we ot ay ees as
NSRARNEHELT, gPaCTUARNEMEADTE,
ee i
Figure 13: Great Break Estuary Sub-Model
fon gal
— oe -
me
Figure 14: Municipality of Mossel Bay Sub-Model
PetroSA.
Vout TO pee
PETROS
WATER FROM NOLWEDANG WATER
‘WOLWEDANS SERVED =ETROSA
a ee a a ee .
OPERATING ———7™
"RULE PETROS,
old uization
~ $ petnsa
WATER DEMAND ©
2 FRACTIONOE
PETROSA
war OPERATIONAL,
HARTENBOS REVERSE
‘OSMOSIS PLANT
Figure 15: PetroSA Sub-Model
22
Appendix E Preliminary Model Results
The following graphs show the preliminary model results. Since this article was mostly about the
use of the model as a Boundary Object rather than the model results or validity of the model the
graphs are left unexplained in this article. For more information contact the researcher.
Wolwedans Dam Water Volume
30M
22.5M
8 15m
e
900
Time (Month)
Wolwedans Dam Water Volume : Curent3
Figure 16: Graph of a Single Run for the Wolwedans Dam Water Volume
Health of Estuary
0 100 200 300 400 500 600 700 800 900 1000
Time (Month)
Health of Estuary : Current3
Figure 17: Graph of a Single Run for the Great Brak Estuary Health
23
Annual Consumption Mossel Bay
40M
0 100 200 300 400 500 600 700 800 900 1000
Time (Month)
Annual Consumption Mossel Bay : Cuent3_._ ——2--—-—___—\_
Figure 18: Graph of a Single Run for the Consumption by the Mossel Bay Municipality
Utilization of PetroSa Over A Y ear
I WH I
9
a
1
to
8
a
0 100 200 300 400 500 600 700 800 900 1000
Time (Month)
Utilization of PetroSa Over A Y ear : Curent3 $2.4
Figure 19: Graph of a Single Run for the Utilization of PetroSA over a year
24
Appendix F Testing of integration method
A small test was performed changing the time step of the Euler integrator method for solving the
differential equations. If changing the time step would cause different model behavior that would
be a problem. In Figure 20 test results on a running average created in the model has been done. It
did not show deviation for the time steps under 1. Therefore no clues were found that the Euler
integration method is not coping with the discrete input.
‘dy iid at iW | |
HANAN AT A
Acai AAA
;
fe a a le A a i ee ae
‘Time (Month) ‘Time (Month)
Running Average Running Average
: :
iT Ahi hh Jatt Ai
Ta A a Cu
AOE LEE EE
eee a see OR te a a all
‘Time (Month) ‘Time (Month)
c
Figure 20: Tests with different time steps
The time steps used from right to left, top to bottom: 1; 0,5; 0,25; 0,125; 0,0625; 0,03125; 0,015625 and 0,0078125.
25
