Participant Workshop Models


An introduction was presented by Dork Sahagian on the STELLA modelling package, a convenient graphical box model developer. Using STELLA, it is possible to conceptually construct a simple system model with the relevant stocks (reservoirs) and fluxes (exchanges between reservoirs) of various entities such as carbon, trees, elephants, people, money, etc. The fluxes are specified by equations relating the various stocks, constants, and various functions determined by the models which then graphically display the changes in the stocks and values of fluxes as a function of time during a model run. While it is exceedingly simple to construct a model in this way, a stumbling block that appeared throughout the workshop was that the participants did not know the appropriate constants and functions to specify in formulating the equations for the fluxes. This directly highlighted the gaps in scientific understanding of the systems they were modelling. While STELLA and other models can describe a system conceptually, in order to accurately apply the model, it is necessary to quantify the relationships between system parameters. This caused workshop participants to think back to their systems and consider the types of field campaigns and other research that would be necessary to quantify these relationships. This was probably the most important result of the workshop because it clearly demonstrated the utility and limitations of modelling techniques, as demonstrated with models developed by the participants.

The workshop participants formed groups of three to four to construct models using the STELLA modelling package. In forming groups, the rule was to include at least one participant who specialized in hydrologic problems and one who specialized in ecological problems. The groups were multinational and multidisciplinary and resulted in the simple models shown below.


R. Boukchina, S. Jain, W. Goma

A STELLA model was developed to simulate Nitrate concentration in runoff to the natural reservoir (lake) from agriculture watershed. The flow chart of STELLA linked water budget process to nitrogen budget in soil to simulate runoff nitrate exportation to the lake. The hydrological, micro biological, agronomical and hydrolic processes were the basis for the logical and mathematical relationships among the input variables which were rain, evapotranspiration, organic and inorganic fertilizers, field capacity, rate of nitrification and plant uptake. The model was run with two fertilizer scenarios. Results showed that nitrate output can be controlled by the fertilizer application management.

View Figure: Stella Model


A. Adewale, A. Amissah - Arthur, G. Chavula, N. E. Ngwa, K. Hailemariam


Reservoir sedimentation remains a serious problem in Africa. The situation has been exacerbated by the intensification of agricultural production and the encroachment upon marginal lands by settlements as a result of population pressure. This in turn has led to the accelerated rates of soil erosion with subsequent siltation of river channels and reservoirs. Experience has shown that although it may cost taxpayers several million dollars to construct dams for the purpose of retaining runoff, it may take only a few years to fill reservoirs with sediments if up stream catchment areas are poorly managed. It is in this respect that computer models, can be used in catchment management by exploring attentive management schemes.

Since many governments in Africa spend large sums of money to construct dams for various purposes (among which are water supply, irrigation hydroelectric power generation), it is imperative that accurate predictions are made of erosion in the catchment area (under varying conditions) and of the eventual deposition of these mobilized sediments in reservoirs. Such information is valuable in the implementation of sustainable catchment area management for water resources.

In this study, Stella was used to develop a program to simulate reservoir sedimentation under different catchment management practices and climate change scenarios.


The general objective of this study was to develop a preliminary model which could be used as a basis for writing a more comprehensive simulation of reservoir sedimentation.

Specific Objectives


In order to accomplish the above objectives, the following procedure was adopted:


Two scenarios were attempted : With 1% & 3% population growth rate.

Three graphs were obtained from the two model runs :

These show the sensitivity of sediment erosion and reservoir siltation to population growth.

With other parameters kept constant, a population density of 10 inhabitants / km2 and a growth rate of 1% per annum gave a reservoir sedimentation period of 65 years - see Figure 1.

With an annual population growth rate of 3% and population density still maintained at 10 people per square kilometer, the reservoir siltation period dropped to 55 years - see Figure 2.

Finally, the model showed that the capacity and residence time of the reservoir decreases with further progression of sedimentation in the catchment area resulting in active sediments being transported over the reservoir site ( which has at this particular stage filled up with sediments ) into the ocean.


1. Test the model using real data. We plan to identify sites in Nigeria, Ethiopia, Cameroon, Ghana and Malawi with sufficient underlying data set to perform the analysis.

2. To modify the model structure as appropriate as possible in order to carry out simulations under the different scenarios

3. To apply model results to the sustainable management of catchment areas for more effective water resources management.


We would like to express our heartfelt gratitude to Charlie Vorosmarty, Dork Sahagian and Berrien Moore for their valuable contributions to our model building exercise using STELLA.

View Figure: Effect of Land Use and Climate Change on Reservoir Siltation


Presented by P. C. OKE & A. ABDELLAOUI

The objectives of this project are :

The predictive model dealt with one case of disease which correlates with climatic parameters. We have introduced an input Harmattan index, microbe opportunity, immunization condition, nutritional condition, random seed factor. Thus, this rate has permitted us to calculate the population morbidity . In the second trial, we introduced the parameters (discomfort index, treatment, delays, random seed 2) which permitted us to determine the lethality. This model permitted us to simulate the effects of climate parameters on the population submitted to a virulence of microbe.

This model is constituted by three blocks :

Each block has several components which are traduced by this diagram.

In this model, there are several sensitivity selectors which permit us to simulate different conditions.

These graphics traduce the results of simulation :

This model must be completed by other blocks to integrate other variables. The equations must be corrected to permit the best calibration of the model.

View Predictive Model


(C. Gumbie, M.B.M. Sekhwela, F.K. Mensah, K.V. Rabah)


The development of natural woodlands or forests is influenced by many factors, some of which include fire, drought, harvesting, animal browsing and damage, just to mention a few. The growth of the forest depends on precipitation and natural or artificial regeneration. Human influence can be expressed in terms of the clearing of forest for arable use, cutting for energy, and logging for construction and industrial timber.

Thus interactive relationships can be better depicted or expressed with an ecological model which details inputs to the system, outputs from the system, and flows within the system. Where such relationships can be defined or expressed mathematically, simulation models can be developed to further explore and increase the understanding of such interactive ecosystem processes. This can also help to provide possibilities for predicting the likely outcomes of various management options for sustainable development. An example of this is a simulation model of a community managed natural woodland in a dry Savanna adjacent to a Game Park in Zimbabwe.

Model Conceptualization

The model assumes an existing natural woodland of approximately 5000 hectare, which is exposed to pressure from human-related activities, climatic, and elephant disturbance.

Stage 1: Woodland

An initial stand of 15000 trees is assumed, and managed for a 50 year period. The dynamics of the woodland is dependent on human activities, climatic, regeneration, mortality, and elephant disturbance.

Stage 2: Human Activity

The human component plays a significant role whereby some proportion of the woodland is cleared by the neighboring community for agricultural purpose (15%), harvested for fuel(19%), and for construction and industrial timber(15%). All these activities contribute to tree cutting which lead to tree removal (death).

Stage 3: Tree Mortality

Factors contributing to tree death in this type of woodland are known to be fires and drought. In the model it is assumed that the contribution due to fire is 15% and that to drought is 25%. The contribution of elephant damage also is quite significant, due to the proximity of the woodland to a Game Park.

Stage 4: Elephant Population

The dynamics of elephant population has a direct effect on the woodland system. It is therefore important to control the elephant population through culling, assuming the elephant population grows at a rate of 1% per annum.

Stage 5: Climatic Factor

In the dry Savanna woodland rainfall plays a significant role on vegetation, particularly on seedling development. It is assumed that 25% of the seedlings will survive and be recruited into the woodland system. It is also assumed that some of the trees in the woodland system die through natural thinning.


The seedling and the recruitment levels fluctuated in line with rainfall pattern as expected, as both germination and seedling establishment depend on soil moisture levels. Initially the woodland shows fluctuating disturbances due to elephant damage until the elephant population is reduced by culling to sustainable levels. At this sustainable level both seedling and recruitment are replenishing the woodland stock which appears to stabilize for the next twenty years.

View Model for a Community Managed Woodland



Alec Joubert

Climatology Research Group, University of the Witwatersrand


Estimates of climate change based on general circulation models (GCMs) are available for Africa south of the equator (defined below as southern Africa). Point estimates for two climatically-distinct sites in South Africa have been used as inputs to the ACRU model in order to investigate the potential hydrological impacts of regional climate change. The two sites selected are near Cape Town (a winter rainfall region) and north of Pretoria (a summer rainfall region).

The aims of the project are to model the potential hydrological impacts of expected climate changes on two climatically-distinct sites in South Africa. In particular, the objective of the study is to identify the relative sensitivities of simulated total stream flow (runoff) and soil moisture in the first soil layer (A-horizon) to imposed climate change. In addition, we aim to develop familiarity with the ACRU modeling system, and particularly to improve the understanding of the sensitivity of the hydrological cycle to exceed on ACRU by means of adjustments to mean monthly maximum and minimum temperatures. Regardless of the sign of change in DTR, the magnitude of the change was limited to 1°C (in keeping with GCM estimates).

For the summer rainfall site, summer and autumn DTRs (September-May) were increased by 1°C by imposing a 2°C increase in maximum temperature and a 1°C increase in minimum temperature. During winter (June-August), the DTR was decreased by 1°C by imposing a 2°C increase in minimum temperature and a 1°C increase in maximum temperature. For the winter rainfall site, the DTR was decreased by 1°C throughout the year (as described for winter above).


Two climatically-distinct sites have been selected: Elsenburg, situated in a winter rainfall region near Cape Town; and Roodeplaat, situated in a summer rainfall region north of Pretoria. Daily rainfall series for both sites were estimated using the stochastic rainfall generator within the ACRU model. The use of stochastic rainfall series implies that inter-annual and intra-annual variability of rainfall at the actual sites is slightly under-estimated. A correction was applied to these data using the PPTCOR multiplication factor within ACRU to adjust the median monthly rainfall of the stochastic rainfall series to that of the actual station.

Mean monthly maximum and minimum temperatures were estimated by the author to represent the actual observed time series. No actual data were available at the workshop.

Imposed Climate Changes

The GCM-derived changes in regional climate are based on those models which provide the most reliable estimates of regional climate change. In all cases, the imposed changes probably represent a conservative estimate of the changes which are most likely to occur regionally. For each site, two simulations were run. The first represents a control simulation, in which no climate changes were imposed. Climate changes as described below were imposed for a second simulation at each site. The imposed changes include:


Austral (Southern Hemisphere) seasons are used below. Mean monthly rainfall for the both the early and late summer seasons (September-February) as well as the late summer and autumn (March-May) were decreased by 10 percent. During the winter months (June-August), mean monthly rainfall was increased by 10 percent. These changes represent the consensus of a wide range of GCM projections and were applied at both stations. The ACRU model provides a decomposition of mean monthly rainfall to daily time series internally, by means of Fourier analysis.

Diurnal Temperature Ranges

Changes in diurnal temperature range (DTR) were imposinear response of the hydrological system to relatively small changes in mean monthly rainfall. The changes in total stream flow at the winter rainfall site are smaller (on a percentage basis) than those at the summer rainfall site possibly indicating a different hydrological response on the basis of rainfall intensity. Rainfall resulting from convective thunderstorms may be expected to be more intense, than rainfall resulting from frontal systems, and this may be expected to cause the difference in runoff response at the two sites. Soil moisture changes in the A-soil horizon are less than 1 mm/month throughout the year, suggesting again that changes in soil moisture are more conservative than changes in runoff.

Preliminary Results

Winter Rainfall Region:

The annual rainfall cycle at Elsenburg exhibits are marked winter-season peak . The perturbed rainfall used to simulated anthropogenically-induced climate change does not alter the shape of the annual cycle, although its amplitude is affected. The primary mode of rainfall production at this site is frontal. The 10% decrease in winter rainfall results in a reduction of approximately 10 mm/month . During summer, the 10% increase in rainfall results in an increase of approximately 2 mm/month.

Percentage changes in total stream flow (runoff) are similar to those for rainfall (approximately 10% in winter). However, these changes represent only a 3-5 mm/month change in the magnitude of runoff . Runoff changes in during summer are much smaller than during winter, given the generally lower rainfall. Changes in soil moisture in the A-soil horizon are very small (less than 1 mm/month) throughout the year, representing a change of less than 5% in all months . Therefore in all cases the changes in soil moisture are smaller than the changes in rainfall imposed in the perturbed simulation.

Summer Rainfall Region:

The summer rainfall site (Roodeplaat) also exhibits a strong annual rainfall cycle, with a pronounced summer rainfall maximum. Mean monthly rainfall throughout the winter months is less than 10 mm/month. The primary mode of rainfall production over this region is convective.

The change in mean monthly rainfall induced in the perturbed simulation resulted in a decrease of summer rainfall by approximately 10 mm/month, but negligible changes in mean rainfall during winter. The annual cycle in total stream flow (runoff) also reflects a strong annual cycle, with very little runoff during the winter months.

The changes in total stream flow simulated in the perturbed experiment result in a 2-3 mm/month decrease in runoff during the summer months, and a change of less than 1 mm/month during winter. The changes in total stream flow during summer represent a decrease of approximately 30% over the control simulation, and are indicative of the highly non-expected climate change.

Summary and Conclusions

This study has indicated the sensitivity of hydrological systems to relatively conservative (small) changes in mean climate. At both the summer and winter rainfall sites used in the present study, the changes in mean rainfall are unlikely to have important climatological implications. However, the highly non-linear response of total stream flow at Roodeplaat during summer is indicative of the potential for considerable hydrological implications resulting from those small changes in rainfall. Importantly, the hydrological response of a system is (among others) a function of soil type, mode of rainfall production (as a control on rainfall intensity), and antecedent conditions (what is the existing soil moisture content?). The potential exists in future to extend this preliminary study to other regions of South and southern Africa, and to examine the potent hydrological implications of other climate changes, for example those related to expected changes in the frequency and magnitude of extreme rainfall events.

Prediction of Soil Loss in the Sub-humid Tropics Using the STELLA Model

O. Totolo, P. Yanda, D. Olago and E. Sambo


Soil erosion is a problem in most countries of the world. The process seriously impacts on the productivity and usefulness of the land. The aerial coverage of the land is seriously reduced by gullies and other erosion features. Soil erosion also impacts negatively on the aesthetic beauty of the land. It is therefore of paramount importance to try to understand this insidious process.

One way to understand the process of soil erosion is to try to understand the interactions and interrelationships of the input parameters and their ultimate influence on the output parameters. However, input parameters (data) are limited in most of African countries. Such relationships between variables can therefore be achieved through extrapolation and interpolation of limited data. This is made possible by creating models or using the available models. In this exercise, STELLA was used to create a model to explain the interactions and interrelationships of the input parameters and to also predict the output parameters affected by one or more parameters. This is very useful because given a set of parameters, the model enables you to forecast a number of scenarios which could be helpful in land management and policy decisions. Modeling enables scientists to experiment with different situations and in this way, policy makers and land managers can make well informed decisions on the basis of modeling results.


1. estimate soil loss in the sub-humid area in the tropics

2. establish levels of influence on soil loss for each of the variables

3 develop different scenarios on measures to mitigate soil loss

View Conceptual Framework for Soil Erosion Model

View Figure: Relationship bewtween soil loss, vegetation degradation, vegetation growth and rainfall

Future Plans

It is envisaged that this project will be extended to other parts of Africa to cover as many climatic zones as possible. Initially the project will select sites in Kenya, Tanzania, Malawi and Botswana. A comparative study will be made to determine the parameters that should be incorporated within the STELLA model. Cross-model comparisons will be made with, for example, ACRU, HYDRO and CENTURY. Apart from the vegetation modeling that has been initiated at the workshop, there is also interest in erosion aspects.

Land Cover Changes on Zomba Mountain and Impacts on Water Yields in the Mulunguzi Catchment

Paul V. Desanker, Eston Y. Sambo, Geoffrey Chavula


The Zomba Mountain Catchment area of the Mulunguzi River and Dam has provided drinking water to the town of Zomba for many years. Good data, covering a period of 40 years or more, for runoff exists from a gauging station just before the Mulunguzi River flows into the Mulunguzi Dam at the edge of the Zomba Mountain. From around 1908-1920, the Department of Forestry has converted indigenous vegetation over most of the mountain to highly productive pine, cypress and Mulanje cedar plantations (Pinus patula, Cupressus lustanica, Widringtonia nodiflora). The forests are managed on a clear-cutting cycle regardless of slope. Fires and wind have also taken their toll recently in creating drastic cover changes. Siltation is also a major problem and the dam had to be dredged in the late 1980's when it almost dried out. Water yields from the mountain have diminished to crisis levels (University of Malawi and schools have changed schedules, water is rationed as a matter of routine, etc.). Other possible reasons advanced for water shortages in the Zomba Municipality have included the presence of a crack in the base of the dam, leading to loss of reservoir water, or changes in climatic patterns that have resulted in increased drought incidences. In this study, we hypothesize that the aforestation program has modified the water balance to a point that water yields have been adversely affected, compounded by poor land management leading to siltation problems.

Mulunguzi dam lies in a catchment of around 18 km2 on Zomba plateau (one of four main catchments for Lake Chilwa, an inland drainage lake). The dam was originally constructed for electricity generation for Zomba town in the 1940's when the population was less than 10,000 but it has now grown to over 50,000. There are plans to start construction of a new dam upstream beginning 1997/98. This is expected to further change the hydrological characteristics.

Overall Goal

Quantify water balance components in relation to land cover types over the Mulunguzi River catchment and apply results to forest scheduling and collection dam dynamics.

Specific Objectives

1. Calculate water balance of the Mulunguzi basin and compare with gauge data

2. Reconstruct a hypothetical time series of forest cover (forest cover types, etc.) for the Zomba Mountain Forest Reserve to provide inputs of the vegetation component in the hydrological model.

3. Relate forest cutting patterns to water yields and attempt to optimize spatial patterns to maximum water yields.

Related Studies

Design of the study

The catchment will be divided into land cover types or compartments which will form units of 5-10 hectare in size:

The historical data will characterize land cover changes in the periods:


Expected Results

The models are expected to provide information to answer questions on how the hydrological cycle has been impacted by the aforestation program. An important outcome will be to see how the present forest cutting patterns relate to water yields compared with other management regimes. This may provide ideas on how to optimize spatial patterns to maximize water yields.

View Effects of Fire on Population Dynamics of Teak Woodlands

View Figure: Ajavon

Go back to African Workshop Report