Introduction

There is a strong and complex relationship between natural resources and rural livelihoods. Rural people in low income countries depend on the availability and access to natural resources for supporting their livelihoods (Ellis and Allison 2004). A livelihood comprises the assets (natural, physical, human, financial and social capital), the activities and the access to these (mediated by institutions and social relationships) that together determine the living gained by the individual or household (Ellis 2000).

The increasing global concerns about the sustainable management of natural resources which followed the UN summit in Rio de Janeiro in 1992 have not visibly reduced the pace of deforestation in the tropics, which is caused by a complex mixture of demographic, economic, technological, cultural and institutional factors (Hartemink and others 2008; Lambin and others 2003). In the international discourse on natural resources conservation, there are diverse relationships between conservation strategy and poverty reduction that reflect conflicting paradigm positions in the current debate conservation (Adams and others 2004). The proponents of the so-called “zero conservation” or ‘‘fortress conservation’’ approach advocate protection measures that seeks to exclude local people from natural resources (Hutton and others 2005; Sanderson and Redford 2003). However, others still equally have different views, community based management projects can make both development and conservation economically viable and attractive for the local communities to maintain biodiversity and integrity of nature (Singh 2008; Sunderland and others 2008). Furthermore, Robbins and others (2006) and Romero and Andrade (2004) suggested that the exclusion of communities from conservation ultimately leads to social conflict and noncompliance with conservation-related regulations (Chan and others 2007). The combining of participatory modeling and livelihood studies could contribute to sustainable natural resource management and livelihood improvement by building shared understanding of critical issues and helping to focus on conservation and development interventions (Campbell and others 2008).

Africa is already a region under pressure from climate stresses and is highly vulnerable to the impacts of climate change (UNFCCC 2007). Thus, African countries working in the conservation-development nexus need to take active part in the current global and regional processes on climate change adaptation.

Ethiopian subsistence agriculture is heavily dependent on rain-fed production. The erratic nature of rainfall leads to reduced crop production. The main reason is the daily, seasonal and inter-annual rainfall variability (Segele and Lamb 2005). In Ethiopia, widespread land degradation has led to severe challenges for the people (Amsalu and de Graaff 2006; Argaw 2005; Mahmud and others 2005; Taddese 2001). Socio-economic and institutional factors, such as population pressure, poverty and land tenure arrangements are the main contributors to land degradation. Population growth raises the demand for subsistence cropland and for biomass (fuel and fodder). Both are leading to deforestation. People’s lack of access to alternative sources for livelihood exacerbates this, and, therefore increases the problems of erosion and nutrient depletion (Haile 2004; Nyssen and others 2009).

In Ethiopia, land tenure is a disputed issue. All rural land is owned by the state and part of this land is allocated to farmers on a use-right basis (Bogale and others 2006). The rural land reform policy strictly prohibits the transfer of land by sale or mortgage. However, it does allow transfer of use-right in the form of gift, inheritance, restricted leasing and sharecropping (Crewett and others 2008). Most debates and studies on land ownership of the state mainly revolve around questions of insecurity (redistributions) of landholdings, degradation of soil quality, unsuitable land use practice and fragmentation of farms. The federal government and the regional states have started a process of land registration and certification to address farmers rural land insecurity (Deininger and others 2008).

The rapidly growing population in Ethiopia adds approximately two million people per year and population is predicted to be more than double current levels by the year 2050 (United Nations Population Division 2009). Internal migration in Ethiopia is high and associated with education, demographic, economic and environmental and security reasons (Mberu 2006). Historically rural-to-urban, urban-to-rural and rural-to-rural migration has varied dramatically with famines and political reforms (Ezra 2001; Tegenu 2003). Households living in the rural areas such as the Rift Valley are faced with a number of constraints, including erratic rainfall, recurrent droughts, rapid population growth, deforestation, soil degradation, food insecurity and low education (Garedew and others 2009). People are the main agents of environmental degradation and they are also the victims. Villagers who were involved in previous research with the same authors acknowledged the importance of a continuous discussion on the sustainable use of forests and other natural resources, which enabled them to move towards more sustainable practices. Farmers realized that they over used the local woodland. Recurrent drought and the constrained subsistence agriculture drove people to overuse the forest as safety net strategy, because other livelihood options than dry-land agriculture are very limited in the study area. High population increase would have a major impact on the remaining forests. Further, we experienced that farmer’s still perceive land tenure as insecure. The establishment and successful implementation of a forest restoration site by the Forestry and Natural Resource College and an action research project on the community land has had a positive impact on the behavior of the surrounding farmers towards forest increase. Accordingly, people are increasingly aware of the importance of woodland forests as a safety net during recurrent droughts and of the need to manage them sustainably. During discussions, villagers have also expressed interest in increasing the forest cover and forest area on the landscape. Some of them already have woodlots around their homesteads which is an encouraging drive to others.

Further depletion of the environment, low agricultural production and worsening socio-economic conditions, including rapid population growth could be foreseen if current trends remain. The objective of this study is to explore a participatory dynamic simulation modeling approach based on a dialogue with farmers in order to test different development strategies (scenarios). The approach would include the generation of forward projections (from 2006–2036) of land-use, population and income under various assumptions discussed with the farmers and should contribute to the debate on how to address social-economical and environmental changes.

Material and Methods

Study Area

The study area is “Keraru” Kebelle (Kebelle is the lowest administration unit in the government structure) in the “Arsi-Negele” district of Oromiya National Regional State located between 205 and 210 km south of Addis Ababa (Fig. 1). It covers 2932 ha and represents a semi-arid flat land of the district’s lowland climatic zone and is situated below 1800 m ASL (Garedew and others 2009). The nearby climate data from the National Meteorological Services Agency for the years’ 1972–2005 shows that the annual rainfall ranges between 264 mm and 968 (the mean is 710 mm), while the mean annual minimum and maximum temperatures are 13.5 and 27.7°C, respectively. The study area had a population of 3647 in 2004, the population density was 124 persons per km2, the annual population growth rate was 2.5% ± 0.2 and the rural-urban migration was low (Garedew and others 2009). The people in the area are farmers, most of them from the Oromo ethnic group, who practice Islam and live in polygamous families. The mainstay of their livelihoods is agriculture including mixed livestock raising and rain-fed crop production. Major crop types grown are maize, wheat and tef (Eragrostis tef). The government extension service is minimal. The natural woody vegetation was dominated by woodland’s and wooded-grassland of Acacia trees, but the area has experienced a rapid deforestation at a rate of 1% per year with cropland successively replacing woodland and wooded-grassland Acacia forests (Garedew and others 2009).

Fig. 1
figure 1

Map of Ethiopia, shows location of the study area

Full size image

Methods

We based the participatory modeling on an approach described by the Center for International Forestry Research (CIFOR) (http://www.cifor.cgiar.org/conservation/_ref/research/index.htm). A system dynamics model was built using the stock- and- flow model software (STELLA v.8) with an icon based interface and availability of array functions (Costanza and Voinov 2001; High Performance Systems Inc. 1996). System dynamics is a concept that considers the dynamic interaction between the elements of the studied system and can help to understand their behavior over time, build models, identify how information feedback governs the behavior of the system and develop a strategy for better management of the studied system (Doerr 1996).

The study was conducted in 2009 using data inputs and assumptions from a previous study (Garedew and others 2009), farmers and experts from the district Agricultural & Rural Development Bureau were involved, unpublished data and other sources (Tables 1, 2 3). The present study involved a process of model building with active participation of 20 informants (focus group) representing purposively selected households from diverse categories of wealth, age and gender within the community to obtain diverse information. The selection was made with the help of the Kebelle council by picking up those individuals who had formal education and considered reasonably able to understand the topics, express feelings, opinions and perspective on the situations. Some members of the Kebelle council were also involved in the study. Repeated meetings and discussions were also made with the entire community to triangulate the data obtained from the focus group. The purpose was to obtain good understanding of their objectives in resource management and building on their knowledge about the trends of the local environment and livelihood (Sayer and Campbell 2004). Wherever data was lacking, information was provided through the focus group dialogue and consensus. This helped to improve the input data of the different sectors of the model for exploring reasonable socio-economical and environmental pathways.

Table 1 Data inputs and assumptions for ‘land-use model sector’ in studying the trends of land-use using various scenarios
Full size table
Table 2 Data inputs and assumptions for ‘human population model sector’ in studying the trends of population using various scenarios
Full size table
Table 3 Data inputs and assumptions for ‘income’ and ‘rainfall’ sector models in studying the trends of income and rainfall using various scenarios
Full size table

Three main scenarios were elaborated. The first one was named “business as usual” and did not assume any significant change in the future conditions or stakeholders’ behavior. In the second scenario, “strategies for socio-economic development”, a number of assumptions reflecting government (MoFED 2007) and local efforts for socio-economic change, including micro-finance, better family planning, better health and better education services, were made. The third scenario, “forest increase” was put in focus and modeled as a pathway for restoring the woody vegetation in the landscape through an area closure strategy (e.g., by excluding cattle). Woodland forest is a source of firewood, charcoal, construction material for the local farmers’ consumption, and also fodder for livestock. This scenario was initiated by the farmers themselves in order to express the availability of wood for households’ consumption, improve livestock productivity and reduce soil erosion (water and wind) and resuming of additional forest cash income for livelihood. Currently, the woodland forest is almost disappearing and the important forest income and biomass collection (firewood and charcoal for sale and consumption and livestock fodder) are shrinking rapidly.

The model structure included several sub-models or sectors representing components of the socio-economical and environmental systems. These are land-use, human population, rainfall and a variety of incomes from crop and livestock production and non-farm activities (Fig. 2). The model simulated all variables over a period of 30 years. In the model the land-use stock is described as a function of changes in different categories of land-uses, human population dynamics and forest increase scenario. Land-use data inputs and assumptions are presented in Table 1. Those assumptions were based on historically observed trends and a discussion with the local farmers on what would be reasonable in the dynamics of land-use. The human population size is described as a function of growth rate, death rate and emigration. The population growth is influenced by the proposed family planning, health and education scenarios. The livestock number and livestock income are modeled as a function of the estimated losses and increases in livestock number, the carrying capacity (feed resources) of the area in terms of tropical livestock unit (TLU), human population dynamics and rainfall. Livestock carrying capacity was calculated based on the total animal feed available from different sources: grassland, crop residues and forest land (Table 3). Crop production is based on farm size of the households, human population dynamics, the variability of rainfall and the availability of micro-finance. Furthermore, non-farm income is based on the human population and land-use dynamics, the availability of micro-finance and educational conditions.

Fig. 2
figure 2

The general structure of the model

Full size image

The model was built for an average household whose farm size is 1.5 ha, with a cropping area of maize (65%), wheat (25%) and tef (10%). The estimated average annual crop productivity of maize was 1.25 ton/ha (varying between 0.7 and 2.2) while wheat was 1.1 ton/ha (varying between 0.5 and 1.4) and tef 0.5 ton/ha (varying between 0.2 and 0.7). On the farmland, food crops are grown for subsistence and cash needs of the farming households. Crop net income (both consumption and cash) was calculated by subtracting the estimated crop cost and loss (30% of total crop income) from the total household crop income. In the study area, an average household owns five cattle, three goat/sheep, one donkey and two chickens which generate household livestock income, including sale of livestock products (mainly milk and eggs), sale of livestock, plough oxen rent, transport rent and consumption uses. All farmers do not have all kind of livestock goods throughout the year but buy and exchange internally for their own use while they also supply to the market. The economic contribution of the livestock sector is considerable and accounts for 15% of the total household income. In the study, non-farm income comprises wage labor, forest-based activities, small scale fishing, sale of salt-rich soil for cattle feed, petty trading, sale of sand for construction, sale of traditional drink, government safety net transfer and remittance (little was reported). A household averagely enrolled in at least the three of these non-farm income generation sources. All monetary values are reported in Ethiopian Birr, where USD $1 = ~11.50 in 2009.

Model testing was an essential part of the model development process. If the model is to be used, it should provide relatively accurate information about the system being modeled. In this study, the model could be validated by using land-use data from 1973 and the actual population data of the Kebelle from 1975 as input variables (Garedew and others 2009) and modeling of the period 1973/75–2006. The resulting simulated land-use values for three occasions (1986, 2000 and 2006) and simulated population values for four occasions (1984, 1994, 2004 and 2006) could then be compared to observed conditions and values derived from Garedew and others (2009).

Result

Dynamics in Population

The population sector model simulates natural population growth annually. Table 4 shows the simulations of population growth based on various intervention strategies, including “business as usual” and “better family planning”, “better health”, “better education” and a combination of the three latter. Over the simulation period (2006–2036), the total population growth varies between 68% and 136% among the simulated strategies. However, when compared to “business as usual”, the scenario “better health” actually rise population growth (through reduced mortality), while “better family planning” (implying reduced birth rate), “better education” (meaning increased emigration) and the combined scenario significantly reduce population growth, Apparently, better family planning and the combined scenarios would be the best pathways for a balanced population growth compared to other strategies if considering the carrying capacity and the sustainable use of natural resources.

Table 4 Simulation of human population growth based on the different strategies
Full size table

Dynamics in Land-Use

The simulations of the two land-use scenarios “with” and “without” the forest increase strategy were based on assumptions of land-use change as specified in Table 1. The simulation outcome as presented in Table 5 illustrate that small modifications in the assumptions of annual land transfers in the scenario “with forest increase strategy” (as compared to the scenario “without”) gave as an outcome that the area of woodland increased quite considerably (203%) at the expense of other land-use types over a 30 years period. For the villagers who initially defined what they want to achieve, (e.g., increased woodland) the interesting part (“the result”) would be what input data generate that output (e.g., more woodland) and how to go about to harmonize the input data in their daily life situation.

Table 5 Simulation of land-use types (ha) based on without (A) and with (B) forest increase strategies
Full size table

The different rates of population growth in different scenarios affect the settlement area and the farm size per household (see Figs. 3, 4). Overall, the total area of settlements is increasing throughout the simulation period while the increments follow different pattern of pathways for different intervention strategies (Fig. 3). For instance, area of settlement dramatically increases with better health scenario compared to other intervention strategies. While the farm size per household tends to decrease throughout the simulation years irrespective of intervention strategy (Fig. 4). Here also better family planning and the combined scenario options are the best alternative pathways to slowing down the trends of decreasing farm size of households.

Fig. 3
figure 3

Simulation of settlement areas under five different integrated strategies of scenarios: 1 = business as usual, 2 = better family planning, 3 = better health, 4 = better education, 5 = combined scenarios (2, 3 & 4)

Full size image
Fig. 4
figure 4

Simulation of per household farm size under five different integrated strategies of scenarios: 1 = business as usual, 2 = better family planning, 3 = better health, 4 = better education, 5 = combined scenarios (2, 3 & 4)

Full size image

Livelihood Strategies and Income Dynamics in the Households

Over the last three decades the households have mainly followed an increasingly extensive mixed agricultural livelihood strategy (crop and livestock). Farmers have recognized that recurrent droughts, erratic rainfall and soil degradation have influenced the agricultural productivity and food security. As those droughts occurred and the population increased, the forest cover has decreased when farmers tried to compensate the declining crop productivity by opening new croplands for subsistence agriculture. At this stage no more suitable land is left for cropland expansion. The demand for land by new households has also increased and as a result farm size per household is diminishing. During normal rainfall seasons high costs for agricultural inputs (chemical fertilizer and improved seeds) and lack of plow oxen exacerbate the challenges for the crop production sector. Households’ efforts to diversify incomes through non-farm economic activities in order to buy food for the dry season can only provide marginal opportunities to fill the food gap.

The simulation of the average household income from crops and livestock (Table 6) followed a range of patterns between different intervention strategies. For agricultural income, all of the intervention strategies, both the micro-finance and the combined scenarios, considerably improved household incomes in the long-term but they had no regular patterns over the separate years of simulation. The reason is that households’ income is regulated by the amount of income generated from agricultural production, which is largely dependent on the amount of rainfall and its distribution within the growing season, since agriculture is mostly rain-fed in the study area. A rainfall model was produced by a random generator providing annual rainfall values between 250 mm and 950 mm. The simulated output shows that the magnitude of agricultural income (in particular income from crops) per household varies with the amount of rainfall in the area (Fig. 5). On the other hand, nonfarm income (Table 7) was constant throughout the simulation period at a level specific for each of the strategies. Informants reported that in the past many households had been involved in forest activities and generate substantial non-farm income from sale of firewood, charcoal and other forest-based products. As an example, 69% of the households of the study are extracting some income from the remnant Acacia forest. Hence, the modeling output showed that there would be an increasing non-farm income through the “forest increase” strategy and this was simulated to be doubled when compared to the “business as usual” strategy.

Table 6 Simulation of farm household incomes (Birr) based on different strategies
Full size table
Fig. 5
figure 5

Relationships between the simulated rainfall (mm) and household income under the micro-finance strategy

Full size image
Table 7 Simulation of non-farm household income (Birr) based on different strategies
Full size table

Model Validation

Figures 6, 7, 8 and 9 show the comparisons between the historical development and simulated model for changes in population size and areas of woodland, wooded-grassland and farmland. Generally, the simulated curves approximately match the historical development of all studied variables.

Fig. 6
figure 6

Comparison of simulated and actual population size

Full size image
Fig. 7
figure 7

Comparison of simulated and actual woodland size in hectare

Full size image
Fig. 8
figure 8

Comparison of simulated and actual wooded-grassland size in hectare

Full size image
Fig. 9
figure 9

Comparison of simulated and actual farm size in hectare

Full size image

Discussion

The tested model, STELLA, provides a basis for better understanding of socio-economic and environmental interactions. The model was built based on assumed relationships between different variables. The outcome of a simulation is entirely dependent on those relationships and the input data. Therefore, any output always needs to be analyzed in relation to those input assumptions. The use of a simulation model to predict the future development of the dynamic system under various conditions (or to study what input data generate a certain desired output) is important in developing effective strategies. There are many examples of similar system models that could contribute to the environmental management practices (Helldén 2008; Kassa and others 2009; Sandewall and Nilsson 2001; Sayer and others 2007; Stéphenne and Lambin 2001). A participatory approach in scenario modeling is also an excellent platform for discussing strategies among different concerned stakeholders. If research data on historical trends are available it adds quality to the discussion on future developments.

We emphasize that a simulation model is not a forecasting instrument but a planning and analysis tool. It generates questions to be asked rather than direct answers. If a scenario suggests that farmers need to convert a certain cropland to woodland the question would be what efforts, resource inputs or strategies are required to achieve that. If that is not possible the question would be what other strategies could achieve an acceptable result. A more technical type of questions would be if the scenario or even the model accurately responds to or describes to the real world changes or if the model needs to be adjusted. One simple example of the later could be changes in birth rate as a result of “family planning” which may not happen instantly but change gradually over time.

In our model testing, it was not possible to undertake a strict statistical evaluation of the model because of the nature of the input data sets that encompasses a few separate years only. Therefore, instead of calculating error variance we used a graphical approach. The simulation outcomes are rough indications rather than very precise predictions. Validation with the historical development of some of the variables indicates that the model responds to key input variables more or less in a correct way.

The relationship between population growth and environmental changes is still an area of active debate (Alexandratos 2005; Carr and others 2005; Grau and others 2008; Jha and Bawa 2006; Nyssen and others 2004). In Ethiopia, population growth increases the demand for arable land and encourages the conversion of forests to agriculture. It also increases the demand for wood. The link between population growth and land degradation are thought to be very strong (Bishaw 2001; Dessie and Christiansson 2008; Feoli and others 2002; Hans and others 2005; Taddese 2001; Teketay 2001). A previous study of land-use dynamics in the study area documented rapid population growth, declining crop productivity and rapid deforestation (Garedew and others 2009). In the present study, the output of the simulation indicates a further rapid population growth, declining farm size and worsening environmental degradation and socio-economic conditions if “business as usual” continues. However, through strategies such as those indicated in the other scenarios, there could be an opportunity to reverse environmental degradation and reduce population growth. It requires, however among other things, that farmers are motivated to participate in increasing the forest in the landscape and that the government actively promotes family planning, health, education, micro-finance, securing of land property rights and sustainable natural resource management. There are encouraging experiences of natural resource restoration (flora, fauna and soil) through local people participation in different degraded dry-land regions of Ethiopia and other developing countries (Lamb and others 2005; Mengistu and others 2005; Verdoodt and others 2009). A scenario based study in a forested part of Ethiopia suggested that participatory forest management (PFM) could provide higher forest cover and more sustainable household incomes for the local community (Kassa and others 2009).

In southern and eastern African countries, farm sizes have been declining over time and a quarter of the agricultural households are controlling less than 0.10 hectares per capita (Jayne and others 2003). In Ethiopia, the availability of land suitable for agriculture is shrinking due to land degradation, while the amount of land required to feed the growing population is steadily increasing (Haile 2004; Teketay 2001). Food security continues to deteriorate, the country has not been food self-sufficient for the last 3 decades and the gap has been filled by food-aid (Kirwan and McMillan 2007). In the present study, household farm size could decline due to population growth. As a result, low per capita income in the households is a major hindrance in providing adequate food to the members in the household. Household food security is likely deteriorating severely if crop productivity per unit area is unable to improve simultaneously with the rapidly increasing population. Informants have been mentioned repeatedly that erratic rainfall and shortage of land for crop production contributes to the challenge faced by the people living in the study area. In this respect, improving agriculture and diversifying livelihood options can help to reduce people’s economic difficulties.

Conclusions

The model predicts an extensive land-use change, largely based on both the decisions of the community and natural population growth. The study simulates rapid population growth, declining household farm size, declining household income, further deterioration of forest cover and worsening land degradation if current practices continue.

The outlined “forest increase scenario” suggests a pathway that might possibly improve the restoration of forest cover in the landscape and subsequently raising household income. It addresses a critical issue but is not an easy way to go, which in practice requires the right decisions, confidence and interplay among farmers as well as government in order to bring back the forest.

The scenarios suggest that the level of population growth could be reduced with various strategies of family planning and education. This has an implication on the land-use patterns, the per capita household income and thereby on household food security. The amount of household income is largely dependent on the amount and distribution of rainfall and use of micro-finance. There was a strong relationship between rainfall variability and agricultural production.

Although, the simulation outcomes are predicted values, the study illustrates that the model can be used as a valuable supporting tool which can aid in the decision making processes in natural resource management. Local or regional planners can easily adapt the model and change variables following additional knowledge and discussions with interested stakeholders in the local area.