Integration: the Road to Food Security Tigray region, north Ethiopia

(1)

Integration: the Road to Food Security

Tigray region, north Ethiopia

Interdisciplinary Project 2015

Terracing in the Ethiopian highlands

By: Ivo de Klerk 10442626 (Political Science), Chris van Diemen 10372652 (Earth Sciences), Jolien van der Krogt 10447458 (Earth Sciences), and Quiri Passchier 10445188 (Artificial Intelligence).

Coordinators: Jaap Rothuizen and Crelis Rammelt Date: 22-5-2015

(2)

Abstract

The Tigray region in Ethiopia is characterized by a poor population that is largely dependent on low-yield, rain fed sustenance agriculture and extremely variable precipitation, resulting in very vulnerable livelihoods. Applying the new technique of road water harvesting (RWH) could enhance livelihood security in the area. In this research, an interdisciplinary approach is used to calculate the impacts that RWH would have. Its effects on erosion and water availability are modelled and then analysed in a sustainable livelihood framework. Additionally, using phone data to analyse poverty incidence in multiple dimensions is a promising method to monitor the results of RWH. It is found that RWH reduces erosion only slightly, but the irrigation it enables can increase yields significantly. This benefits not only the farmers close to the roads but the wider population, although the resulting inequality can cause problems on the long-term. Experimenting with RWH in the region is recommended as this can strengthen livelihood sustainability.

(3)

1. Introduction

Global food security has steadily risen throughout the last decades (Pinstrup-Andersen, Pandya-Lorch & Rosegrant, 2001), but it is estimated that famine will still stay a worldwide problem for at least another 50 years (Godfray et al., 2010). However, this prediction is based on current development and a question to be asked is if that development can be extrapolated into the future. There are numerous factors that should be taken into account when it comes to food security, ranging from the proliferation of infrastructure (Godfray et al., 2010; Rosegrant, & Cline, 2003) to the speed at which fertile cropland is lost due to erosion (Hanjra & Qureshi, 2010).

A region that is known for its low food security is the Tigray region in Northern Ethiopia (Figure 1.1). Ethiopia has known a structural food deficit since at least 1980 (Devereux, 2000). Tigray is subject to considerable erosion and irrigation is inadequate to ensure a stable crop yield for local farmers (Taddese, 2001; Sonneveld, 2002), which is a major problem in a country largely dependent on low-productivity agriculture (Devereux, 2000). Recently parts of Tigray have been indicated as a famine crisis area for April - June 2015 by the Famine Early Warning System (Fews.net, 2015) (Figure 1.1 Right).

Figure 1.1. Left: overview maps indicating Ethiopia and the Tigray region, Right: Tigray area map with famine risk areas April - June 2015 superimposed (Fews.net, 2015)

One often mentioned way to increase the food security in an area is road building: better infrastructure creates more opportunities for trade and specialisation and thus enhances the prosperity of an area. But beside the connectivity function of roads, the construction also has environmental effects. One of these is the impact roads have on local hydrology, which can, for better or for worse, be substantial (Garcia-Landarte Puertas et al., 2014). Road water harvesting (RWH), the collecting of rainwater from roads, has

(5)

in already two case-studies in the Tigray region been shown to be effective (Teweldebrihan, 2014; Woldearegay, 2014).

Consequences of RWH for local livelihoods are still uncertain. Analyzing these requires an interdisciplinary approach, which combines different perspectives on infrastructure and knowledge from multiple disciplines, because “each research perspective offers partial insights towards infrastructure development knowledge” (Gartner, 2014:33). Modelling the effects of RWH requires an integration of knowledge from earth science, insights from the social sciences and data-collection techniques from artificial intelligence. In doing so, this research aims to contribute to the literature investigating whether these so called ‘multifunctional roads’ could become a major boost for food security in the region (Demenge et al., 2014; Teweldebrihan, 2014).

Research question

How can water management practices integrated in roads contribute to strengthening livelihood sustainability in the Tigray region, Ethiopia?

(Sub-)questions

● How does the location of RWH systems influence erosion? ● How does RWH affect irrigation and thus yield?

● How does RWH impact the livelihoods of both direct beneficiaries and others?

In the next chapters, first the situation in the Tigray region will be briefly sketched. The focus will then shift to some different perspectives on infrastructure and how these can be integrated. Building on this, an integrated approach to analysing the effects of RWH will be proposed. The methods, involving modelling the effects on erosion and water availability and the inference of poverty rates from call data records, will then be discussed. After presenting the results, the paper will wrap up with a conclusion about the viability of RWH.

2. The Tigray region

The Tigray region is situated in the northern part of Ethiopia. The area has a large highland area at about 2000 - 3000 meter above sea level but also extensive lowland plains at about 500 - 1500 meter. The climate is regarded to be generally sub-tropical with an extensive dry period that can last up to 10 months. More than 90% of the area is regarded to be semi-arid (Taffere, 2003). The area is largely dependent on small scale agriculture as a main source of economical activity adding up to about 43% of the GDP in 2008/2009 and most of these agricultural practices are rainfed (Bues, 2011). The area has a lack of measures countering the influences of weather variability (Araya & Stroosnijder, 2011) and most local farmers produce just one crop a year. These factors make them vulnerable to droughts and result in widespread crop failings (Bues, 2011). Hence, 41% of the population was undernourished in the years 2005 to 2007 (Bues, 2011).

Currently only 25% of the population has access to roads and ambitious road building plans have been initiated by the government in recent years. There are plans to construct about 5177 km of roads in the area during 2014/15 (Woldearegay et al., 2014). These roads can be a blessing for the area but can also cause problems on many levels. In this report, the focus will be on enhancing positive effects as well

(6)

3. Theoretical perspectives on RWH

Infrastructural projects like RWH are a topic relevant for many academic disciplines, which has resulted in different perspectives on how to research it. Gartner (2014) distinguishes three: technocratic research emphasizes the physical properties and effects of infrastructure, but neglects the social complexity surrounding it; interventionist research stresses the instrumental value of infrastructure for development goals, but neglects power relations and the heterogeneity of societies; critical research, finally, draws attention to the socio-political environment and inequality of access, but neglects the adaptability and positive impacts of many infrastructure projects. Of course these are ideal types and many theories will combine them to some extent, but the distinction is useful to ensure research covers all angles. Each of these perspectives offers its own particular insights and in theory they are complementary, but in practice they are characterized by differing and sometimes opposing assumptions. Next, the three perspectives will be elaborated upon for RWH. In the subsequent section the relevance of integrating the three perspectives will be explained, along with ontological and epistemological challenges in doing so.

3.1 The technocratic perspective: the physical impacts of RWH

Technocratic research into RWH would look at the physical impacts of roads on their environment. The impact that is looked at in this research is the influence of roads on water flows. Roads can be a significant obstacle to water, blocking and canalizing water flows. On the one hand this is known to result in enhanced erosion (DeGrassi, 2005), which is detrimental to the quality of the surrounding soil. On the other hand it increases runoff, meaning valuable water is lost for agriculture. Rethinking the hydrological properties of roads and consequently changing their location and integrating RWH has the potential to turn both these effects around (Teweldebrihan, 2014).

3.2 The interventionist perspective: the social benefits of RWH

An interventionist perspective on RWH would emphasize the social benefits of irrigation and the reduction in soil loss. Both increase the agricultural potential, which is very relevant in a country in which economic growth strongly covariates with agricultural yields (Devereux, 2000). Increased productivity also creates employment opportunities, thus creating additional demand for goods and laying the foundation for a more developed market (Smith, 2004). Finally, it increases food security, an important limiting factor on investments and productivity. This creates a virtuous circle, as a higher income provides a stronger buffer against food shortages (Timmer, 2004).

3.3 The critical perspective: the differential consequences of RWH

Critical research into RWH would emphasize the differential effects it can have and the role of the socio-political environment in determining these. RWH can create economic inequalities because it only benefits some farmers, but who profits is not necessarily neutrally determined. Access to any sort of infrastructure is determined by an actor’s agency and by mediating institutions, which are often biased towards certain interests (Gartner, 2014). In areas where land and water rights have not been formalized, more powerful actors can often benefit at the expense of others (see e.g. Boelens, Guevara-Gil & Panfichi, 2010) and investment in agricultural land is often accompanied by land grabbing (De Schutters, 2011). Additionally, irrigation systems can be a source of power as “[c]anals are essentially devices for rationing a valuable commodity among competing claimants” (Wade, 1975:1743), meaning they are prone to corruption.

(7)

3.4 Integrating the perspectives

As said before, these perspectives are ideal types. This is especially true for the interventionist perspective, which Gartner (2014) uses to juxtapose her own (more moderate) critical approach with. Nonetheless, they are valuable for developing an integrated, interdisciplinary approach. This approach is necessary as the effects of RWH do not fit neatly in one of the perspectives, let alone in a single academic discipline. For this research, the technocratic perspective offers detailed information on the physical effects of RWH; the interventionist perspective offers nomothetic data for generalization and comparison according to outside norms; and the critical perspective offers idiographic knowledge about local particularities and effects.

There is another reason for integrating these perspectives, specifically the critical one. As Gartner reflects, “it is worth questioning whether the certitude of technocratic and interventionist perspectives is derived from scientific rigour or is reflective of political asymmetries in the process of knowledge production” (p. 28). This echoes earlier criticism by Chambers (1995) that the development discourse is biased towards the perspective of western professionals, whose realities are “universal, reductionist, standardized and stable”, thus further marginalizing the poor, whose realities are “local, complex, diverse and dynamic” (p. 173). In this view, integrating the perspectives of the marginalized is not just a form of thoroughness but of emancipation.

The technocratic and interventionist perspective on the one hand and the critical perspective on the other have some fundamental differences however. The first two are based on a foundationalist ontology, assuming there is a real world independent of our meaning, and a positivist epistemology, which assumes objective knowledge about the world is possible; the latter has an anti-foundationalist ontology1, viewing the world as socially constructed, and an interpretivist epistemology, which assumes knowledge is always influenced by the discourse within which it is constructed (Furlong & Marsh, 2010). In their extreme form, these two positions are incompatible. Taken more moderately, however, they can be reconciled in a critical realist position. This position has a foundationalist ontology but is closer to an interpretivist epistemology (Furlong & Marsh, 2010). It accepts the existence of phenomena which are not directly observable, which positivism does not, assumes that knowledge is fallible and theory-laden and states that both the physical world and our interpretation of it affects outcomes. It often gives rise to mixed-methods research that combines quantitative analyses of causal relationships with qualitative, more detailed analyses of causal mechanisms.

(8)

4. Approach

In the previous chapter, Gartner’s (2014) three perspectives on infrastructure were discussed along with their integration on a philosophical level. In this chapter, this integration will be made more concrete by proposing an integrated, interdisciplinary approach on the impacts of RWH. This approach will be based on the Sustainable Livelihoods Approach (SLA), which combines elements of the interventionist and critical perspectives. It will be extended in two ways. First, poverty, an important element of the SLA, will be further specified using the Multidimensional Poverty Index (MPI). This substantiates poverty as well as quantifying it, thus incorporating more interventionist elements into the SLA. Secondly, the effects on natural capital in the SLA will be conceptualized in more detail for RWH by integrating the effects on erosion and water availability, thus accounting better for the technocratic perspective. This is an integration technique Rutting et al. (2014) call adding, in which a concept from one discipline (natural capital) is connected to insights from another discipline (earth science). After discussing these elements in order, the integrated approach will be presented.

4.1 The Sustainable Livelihoods Approach

The SLA rose as a response to the dominance of simplistic, often income-based definitions of poverty. Chambers (1995) notes that this dominance is understandable, but reflects the reality of western professionals instead of that of the poor. The SLA does this by accounting for the diverse and variable ways in which poor (and non-western) people often make a living. It is wide and holistic, but pays little attention to the underlying causes of poverty (Fisher et al., 2013).

Chambers (1995) defines a livelihood as “the means of gaining a living, including livelihood capabilities, tangible assets and intangible assets” (p. 174). A sustainable livelihood is “adequate for the satisfaction of basic needs, and secure against anticipated shocks and stresses” (p. 175). Scoones (1998) adds that it should maintain or enhance its capabilities and assets, while not undermining its natural resource base. The SLA conceptualises livelihoods as livelihood resources being used in various livelihood strategies, constrained (and facilitated) by institutions and organisations (Scoones, 1998). Livelihood resources involve various forms of capital, such as natural, economic and human capital. Seen in this framework, the goal of RWH is to enhance the natural capital of farmers; the other factors do not change (directly at least).

Scoones (1998) describes five key assessment criteria relating to sustainable rural livelihoods. The first is the creation of working days, as sufficient work is necessary to sustain a livelihood. Second is poverty reduction, which has an obvious link to livelihoods. The third criterion is the enhancement of well-being and capabilities, as a key element of this view is that the poor should be able to pursue goals they themselves value. The fourth criterion is livelihood adaptability, vulnerability and resilience, which is especially important in the context of Ethiopia’s droughts. The final criterion is natural resource base sustainability.

In this research, not all criteria will be used. Poverty reduction is an important criterion as this relates directly to local living conditions; it will be linked to the MPI below. Reduction of vulnerability and natural resource base sustainability relate strongly to irrigation and erosion reduction respectively. Creation of working days and enhancement of well-being and capabilities will not be used though, the first as its relation to actual improvement of living conditions is unclear and the second as this research lacks the means to tackle it. RWH can theoretically be expected to improve both though.

(9)

4.2 The Multidimensional Poverty Index

One of the SLA assessment criteria is poverty. The MPI, developed by Oxford Poverty & Human Development Initiative (OPHI) with the UN Development Programme, quantifies poverty through its underlying indicators (Figure 4.1). As stated by SLA and implemented by the MPI, income alone is insufficient to quantitatively characterise poverty.

Figure 4.1. MPI dimensions and indicators (OPHI, 2013)

4.3 The environmental impacts of roads

The Tigray region is subject to considerable erosion and irrigation alone is inadequate to ensure a stable crop yield for local farmers (Taddese, 2001; Sonneveld, 2002). It is estimated that the Ethiopian region can expect a 20 - 40% loss in yield over the next 20 years only due to soil erosion (Pimentel et al., 1995; Lal, 1995; Shiferaw & Holden, 1999). It is therefore key to understand the influence of roads on these systems and how roads can be enhanced to improve on both problems.

Erosion

The concept of RWH directs to the possibility of a road to not only provide transportation but also provide a barrier against erosion and a chance to enhance irrigation (Teweldebrihan, 2014). Roads by default are known to often enhance erosion in surrounding areas (DeGrassi, 2005). Erosive properties could be turned around when RWH is implemented. To predict erosion levels, the revised universal soil loss equation (RUSLE) is used to model the mean annual soil loss. The theory has been validated with a wide range of empirical evidence and has been used in similar environments (Foster et al., 2000). This makes the RUSLE the most comprehensive tool for modelling soil loss in the Tigray region. The equation consists of six factors as seen in equation 4.1.

𝐴 = 𝑅𝐾𝐿𝑆𝐶𝑃 (4.1) [𝑡𝑜𝑛/ℎ𝑎/𝑦𝑒𝑎𝑟]

The result of the equation is the long-term average annual soil loss (A) given in ton per hectare per year. The equation takes account of the rainfall erosivity factor (R), the soil erodibility factor (K), topographic factors (L and S) and cropping management factors (C and P).

(10)

Evaluation of irrigation

RWH could potentially be beneficial for areas such as the Tigray region. When transported to areas in need, water can be used for irrigation. Since the precipitation is poorly distributed over the cropping season in this arid area, storing rainwater could be a way to secure yields (Teweldebrihan, 2014). To quantify the potential benefits of irrigation for food security Araya et al. (2010) use the Aqua Crop Model of FAO to calculate yield response of teff on irrigation.

Teff is a grass, mostly used for bread, that accounts for two-thirds of human nutrition in Ethiopia (Stallknecht et al., 1993). The grains of teff are high in calcium and the protein level is comparable with other cereal grains. The grain is also high in iron, which lowers the occurrence of anemia in Ethiopia (Stallknecht et al., 1993). The crop can be cultivated in several different circumstances, for example wet and dry areas. Because of this, it can also grow in the dry Tigray region. Moreover, the straws can be used as cattle feed in the dry season. Another advantage of production of this crop is that prices are relatively high. Farmers can earn substantial money with it (Stallknecht et al., 1993). Finally, the crop is disease resistant.

Since teff is not a demanding crop, it is interesting to see how it responds to additional water input. This is analyzed by Araya et al. (2010). The results are showed in the table 4.1.

Table 4.1: Influence of irrigation water on teff yields. From: Araya et. al., 2010.

Year Rain + Irrigation

[m3 h-1] Grain yield [kg ha-1] Biomass yield [kg ha-1] Irrigation [mm] 2008 2696 1950 7620 95 2008 2556 1170 6220 69 2008 2316 970 5400 45 2008 2046 730 4630 20 2008* 1846 640 3930 0 2009 1280 1400 5800 90 2009 945 610 4530 40 2009* 696 370 3480 0

From this table it can be concluded that the use of irrigation water has a positive effect on the growth of the grain yield and the aboveground biomass. When water is harvested from roads and is applied to the fields adequately, it can indeed increase teff yields.

4.4 The integrated approach

The integrated approach based on these elements is shown in figure 4.3. RWH has two main physical impacts: on water availability for irrigation and on erosion. Of these, irrigation affects vulnerability to droughts and poverty and erosion affects natural resource base sustainability and poverty. In these ways, RWH affects livelihood sustainability in the region.

(11)

(12)

5. Methods

For the research this approach requires, a variety of methods is used, which will be discussed here. For determining the physical impacts of a road an area of interest is used to model the erosion with and without RWH and for roads high and low on a hillside using the RUSLE. The amount of water that could be harvested will be modelled for the same road and will subsequently be related to teff yields. To establish the social effects induced by the physical results of RWH, on-site research would be desirable; however, this is beyond the scope of this paper. While it is not possible to conduct in this research, methods for deriving the MPI for an area from mobile phone data will be discussed. Additionally, a literature study will be done into the social impacts.

5.1 Erosion modelling

Area of Interest

Figure 5.1 Map visualizing area of interest with A) the location of figure 5.2

To be able to model for erosion a small area that is representative for agriculture in the Tigray region is chosen (Figure 5.1). This area of interest (AOI) is a ± 4 km2 area around a small village at 38°17'0.06"E

(13)

13°34'7.30"N close to Socota del Semien. This location was chosen considering a number of factors, including available DEM, size, steepness of slopes, inhabitants, and the occurrence of a road. Gullies are recognizable on satellite images (figure 5.2B), indicating problems with erosion.

Figure 5.2 Satellite picture of part of the AOI, A) gully formed in agricultural field B) gully formed where water accumulates downslope of a road.

The equation

When modelling one should always keep in mind that reality is simplified and results should be considered as indications only. The RUSLE equation can be used to model soil erosion in the AOI. This equation has shown to hold up to many empirical studies. How the different RUSLE factors are derived is given in table 5.1, detailed calculations can be found in appendix C. Justification of factors is beyond the scope of this paper and is provided by Renard et al. (1991). It is important to note that a higher value for either one of the factors results in higher soil loss.

Table 5.1 Overview of methods for different factors.

Factors Method

Slope length (L) Calculated from 30x30m DEM for 4 different scenarios

Slope angle (S) Calculated from 30x30m DEM Land cover (C),

Management practices (P)

Manually mapped in ArcMap

Soil (K), Rainfall (R),

(14)

The slope length factor (L) is higher when flow accumulation is higher. This factor can be largely influenced by roads that cut through the area and possibly change flow accumulation and will be modelled for different scenarios.

Modelling

The modelling is divided in two parts:

1. Calculation of stable factors - K, C, P, S, and R 2. Calculation of variable factor L

Consequently the results will be combined in ArcGIS to calculate Erosion change and distribution for different scenarios.

Part 1 The control variables

The K, C, P, S, and R factors are important for erosion. None of these factors is expected to significantly change thanks to RWH and are therefore kept constant in all scenarios in the model. The calculation of these factors can be found in appendix C.

Part 2 Modelling of dependent variable

The L factor is a positive function of water accumulation. Water accumulation is defined as amount of water that potentially flows over a particular area. Figure 5.3 shows a situation that is not influenced by any roads. Areas with high rates of water accumulation (figure 5.3A, blue pixels) can be interpreted as streams of water in the landscape and are mainly influenced by topography. The corresponding L factor can be seen in figure 5.3B. Streams correspond to the areas with a high L factor indicated by red pixels. Without roads the streams cut through the locations where roads are now. In the actual situation roads and the placements of bridges influence the location of streams. Water will be blocked and stream alongside roads until a bridge or conduit is reached. This leads to water accumulation, stream forming, and enhanced erosion.

Water can be harvested in many ways at points where streams and roads intersect as will be discussed in the next chapter. The different scenarios will be modelled with a situation where all water that reaches the roads is harvested. This will cause the water accumulation downslope of the roads to be lower and this will cause erosion to drop in these areas.

(15)

Figure 5.3 Oblique view of AOI, different road scenarios indicated by red, orange, and green lines vertical dimension 2x amplified, A) visualization of elevation in AOI, water accumulation superimposed on area. B)

(16)

Scenarios

Figure 5.4 Oblique view of the AOI, agricultural land highlighted in green.

Figure 5.5 Schematic representation of slope and road locations.

In order to model different potential road locations three scenarios are devised. All scenarios are modelled with a perfect water harvesting mechanism in place where water is being tapped from the accommodation space between slope and road. The first scenario considers the current road location. The other two scenarios are chosen considering road locations higher and lower relative to the current road, slope, and the agricultural areas. It is expected that more can be gained when the downslope area of the road is an agricultural area because water is kept from streaming over agricultural land. A road lower on the slope will influence less agricultural land and is expected to have less influence on erosion in this area.

(17)

Table 5.2. Overview of scenarios.

Scenario Description Expected results

No Road The situation as is modelled without any roads, the result of this scenario is available and described below

Results are modelled and shown in figure. This will be regarded as

base level.

Current road The situation with roads in current position and RWH system installed. Indicated by orange line in figures above. The modelling of this scenario needs an enhancement of the GIS model.

It is expected that erosion will be lower. Water harvesting halfway along the slope will lower water accumulation and thus the L factor. Road High The situation with the road higher up on the slopes with

RWH system installed, indicated by the red line in figures above. Note the extra part of road to the West.

It is expected that erosion will be lower. A road higher up on the hill will lower water accumulation and thus the L factor.

Road Low The situation with the road lower down on the slopes with RWH system installed, indicated by the green line in figures above. Note the extra part of road to the West.

It is expected that erosion will be similar. The lower road is expected to have a neglectable effect on slope length since road is on low part in agricultural area.

Base level erosion map (A)

Figure 5.6 Oblique visualization of base level annual soil loss (road only for orientation), ranges from 0 to 134 ton ha-1 year-1, A) Steep areas with high soil loss B) low areas with high water accumulation (steams) and high soil loss

(18)

The result of the base level erosion model is visualized in figure 5.6. The red areas indicate areas with high annual soil loss and correspond to agricultural areas with high slope gradients (figure 5.3 A) or areas with high rates of water accumulation/streams (figure 5.3 B). How this will be influenced by the installation of water harvesting and different road locations will be discussed below.

5.2 Modelling RWH potential

To discover which amount of water can be used for irrigation, a few calculations need to be made. First, the amount of water that enters the system should be clear. To be able to compare the amount of collected rainwater with other studies (Araya et. al., 2010; Carter and Miller, 1991), the amount of rainfall in the rainy season between July till October (kiremt) is used for calculations. From a 30 meter resolution DEM, the direction of flow is derived. This direction is based on the elevation of a surface (Appendix B). The direction of flow map gives an overview of flow direction of runoff. This is useful knowledge when locating a road with a water harvesting function. With the use of the flow accumulation map (figure 5.7), it can be calculated how much water will be captured after construction of this road.

Figure 5.7 Flow accumulation and RWH location in area of interest.

412 mm of rain water falls during the Kiremt rainy season. This amount equals 0.412 m/m2 water. According to Yimer et al. (2008), infiltration rate of rainwater is dependent on land use and the slope of an area. Because of this dependence, another study in runoff harvesting in a similar area is used for the calculations, which state that 1% of the precipitation can be harvested (Carter and Miller, 1991). The catchment area consists of 4199251 m2 and the study area includes 173 hectare of agricultural fields.

5.3 Learning poverty maps from call detail records

An alternative and addition to on-site research, which as mentioned earlier would be preferable, is the extraction of social statistics from call detail records (CDR). Those statistics can be used to estimate

(19)

RWH impacts on poverty as quantified by the Multidimensional Poverty Index (MPI). Unfortunately this data is not readily available. Therefore the methodological possibilities and possible results will be discussed. The outcomes and implementation of such a data study should always be understood in the regional context.

Quinn et al. (2014) suggest intelligence gathering as one of the “research topics in which ... artificial intelligence can be applied specifically to the developing world” (p. 2). Data gathering in a developing country like Ethiopia is time consuming and expensive, leaving official statistics scarce, incomplete and above all outdated. Carrying out local surveys to measure success of projects on for example RWH, on the other hand, is difficult in Ethiopia as well. This is due to cultural reservedness to complain, in general and to outsiders in particular (C. Rammelt, personal communication, 8 April 2015). Telephone data can be a good alternative as upcoming or established use of mobile phones leads to vast amounts of data being produced. This data is, if analysed with the right techniques, rich in behavioral and social information for policy use. Pokhriyal, Dong and Govindaraju (2015) have developed two techniques to predict the MPI from CDR data in Senegal on different spatial scales. This data was provided by the Data4Development Orange challenge Senegal 2015. It will be discussed if those techniques can be applied to Ethiopia, after which a time component will be added to be able to compare the impact of the implementation of water management practises integrated in roads.

Using CDR learning techniques in to local context of Ethiopia

The state-run telecommunications company Ethio Telecom is the only provider of mobile phone services in Ethiopia. This could be an advantage because in this way the government has direct access to the produced data to study poverty. On the other hand this state-owned monopoly can aggravate privacy questions.

Figure 5.8. Ethiopia had 27,25 mobile phones per 100 people as of 2013. The change over the last year for which data was available was a 21,81% increase. (Quandl adapted from The World Bank, 2015)

Ethiopia had 27.25 mobile phones per 100 people in 2013 (Figure 5.8; appendix C contains comparative numbers). The World Bank (2015) states that as of 2013, 81% or 76.2 out of 94.1 million Ethiopians live in rural areas. For the upper bound of rural mobile phones, phone subscriptions would be equally likely in

(20)

lower bound on the other hand, assuming all urban Ethiopians have a phone, the number of rural inhabitants with a phone would come to 7.8 million. In comparison, Senegal has 9.0 million out of 13.6 million inhabitants with a mobile phone. So there is no question as to whether there are enough phones in Ethiopia, but there is as if to whether it is possible to get a representative sample of the population out of the phone subscribers and whether a representative sample is actually needed to get a reliable correlation with the MPI dimensions.

Using CDR learning techniques to infer the MPI

When using CDR to give an indication of the MPI in another time period or on another spatial scale, different data sets can be available and those sets can be used to achieve different results. The data is grouped in different sets because the total of raw detailed data in one set would pose the highest privacy risk. Privacy sensitivity of CDR dataset is due the possibility to trace back individual users. In each set a different component is taken out or aggregated to a higher scale to reduce the privacy sensitivity of the data.

One of the sets available to Pokhriyal, Dong and Govindaraju (2015) was one year of hourly antenna to antenna data (Dataset 1). This data is aggregated over the users. When further aggregating this data in the time and space dimension to a yearly region to region level, different measures of the data can be correlated with the regional MPI components to learn a model with which MPI components at a finer time and or spatial scale can be predicted. Pokhriyal, Dong and Govindaraju conclude the PageRank measure works best to predict the H (percentage of poor) and A (average intensity of poverty among H) components of the MPI in Senegal with a correlation -0.8 and a p value of less than 0.0005. Other measures and learning methods can be tested to best indicate the other dimensions of the MPI in Ethiopia. The MPI report on Ethiopia (OPHI 2013) MPI values for the regions of Ethiopia in 2011 can be seen in figure 5.9 (appendix D contains tabulated and detailed data). Those are the dependent variables that will be correlated with the independent CDR variables.

(21)

Figure 5.10 and 5.11 give an overview of the finer grained spatial levels for which the MPI could be predicted and followed over time by using CDR data.

(22)

The other dataset used by Pokhriyal, Dong and Govindaraju (2015) contains data aggregated on “arrondissement” level in Senegal, which would be woreda level in Ethiopia (which can be seen in figure 5.10 and 5.11), and on a monthly basis, to calculate Bandicoot2 behavioural indicators for individual users (Dataset 3). Because Bandicoot indicators are calculated for each user individually they are more suitable for making predictions on a finer spatial scale and include: number of incoming and outgoing calls, number of active days, number of people in contact with etc. With less pre-aggregation of the data, finer grained results can be obtained.

This data first has to be aggregated to the level of known MPI scores, being yearly and on the regional level, before the model learned in this way can be used to make small-scale predictions. The research done so far indicates 11 out of 33 individual indicators provided in dataset 3 have a good correlation with MPI. Combinations of Bandicoot behavioural indicators and data measures and their predictive value for the multiple dimensions of poverty should be studied in the context of Ethiopia. Also, while making MPI predictions on time levels finer than one year seasonality has to be taken into account. This can be done either by taking out this variance, but it would take multiple years of study to find seasonal patterns, or by only comparing MPI between the same times of year.

What can be achieved with CDR data

Resulting from reliably deriving the MPI from CDR data more often than official statistics and on smaller scales, is the possibility to study the impact of RWH on all poverty dimensions and on for this paper of specific relevance the food security, which is one of the MPI indicators. With a finer space scale of the MPI it would be possible to compare small regions such as Tabias, with and without new RWH techniques in place. And when comparing dry years before and after RWH implementation potential mitigating effects on drought vulnerability can be studied.

(23)

6. Results

6.1 Erosion

Fig 6.1 (left) Total annual soil loss for the agricultural area (± 174 ha) in ton/year, (right) Difference in percent between base level and scenarios for total area (± 400 ha) and agricultural area.

The total erosion of the AOI is modelled to be around 1350 ton/year. The implementation of a perfect RWH system can lead to a reduction of soil loss of about 75 ton/year and leads to a reduction of erosion in all scenarios (figure 6.1, left). However these number will only be used to get an indication of magnitude rather than an absolute number. The focus will be on the relative change in soil loss (figure 6.1, right). The modelled situation without a road is used as a base level. A difference of about -4 % for the high road to a -5.5% change for the low road is achieved in the modelled scenarios. These numbers do not correspond to the expected results given in table 3. And the differences between high and low road scenarios are insignificant, at maximum 1.5 %. The spatial distribution of soil loss change and the roads in different scenarios can explain this small difference.

The differences are mainly due to the location of the roads relative to the agricultural areas and the slopes these agricultural fields are built upon. The base level soil loss map in chapter 5 indicates that steep slopes and water accumulation create high soil erosion (figure 5.3, A & B). The spatial results indicate that erosion is lower downslope of roads (figure 6.2A). The roads increase upslope erosion where water accumulates at the sides of the roads before it reaches the open water basins (figure 6.2B). The absolute difference is difficult to visually recognize on the scale of the model at only 1.5 %.

(24)

Figure 6.2 (left) Difference with base level when low road is modelled green indicates lower soil loss, red indicates higher soil loss, A) indicates areas where soil loss is less due to reduced water accumulation, (right) difference with

base level when high road is modelled, B indicates areas where water flows along road and increases soil loss (vertical dimension 2x amplified).

It is estimated that the Ethiopian region can expect a 20 - 40% loss in yield over the next 20 years due to soil erosion alone (Pimentel et al., 1995; Lal, 1995; Shiferaw, & Holden, 1999). RWH could reduce the soil losses by 6.2 - 8.8 ton ha-1 yr-1 and save 1 - 2 % of the yield losses over the next 20 years. Spatial distribution of potential yield gain is concentrated downslope of roads, which could have consequences for the distribution of benefits as will be discussed in section 6.3. As a single solution the road location change would have a neglectable effect, but in combination with irrigation systems this measure could have significant benefits.

(25)

6.2 Water harvesting

Figure 6.3 Types of runoff harvesting (from Teweldebrihan, 2014).

The amount of water available for irrigation is measured and will be described below. Ways to collect this water are displayed in figure 6.3. The kind of water harvesting suitable for the Tigray region is micro-catchment harvesting (Teweldebrihan, 2014). Water from the roads and ditches alongside the roads is harvested and stored for future use. Moreover, the road also collects water because of its barrier function. This collected water can then be transported to agricultural fields or be stored.

In the ideal situation for water harvesting, which would be when the road is built lower than the current road, 98 m3 water per hectare can be harvested. In this situation, this water can either be captured at the hillside of the road, with risk of flooding, or at the downside. When collected at the lower side of the road, the water can be lead to a basin or directly to agricultural fields. With the use of a culvert, the water can flow underneath the road and can in that way be controlled more easily (Woldearegay, 2014).

Compared to table 4.1, an additional amount of approximately 100 m3 water will not increase yield significantly. Since the agricultural fields are distributed over 4 km2, it would probably not be possible to irrigate all fields with this collected water. Therefore, it would be more beneficial if only a few agricultural field make use of this collected water. These fields should be located near the road, to minimize water loss due to transport. When the water would be used to irrigate only 80 hectares of agricultural fields, almost 200 m3 water can be applied as irrigation. When this amount is compared to the values in table 4.1, one can say that it should result in higher agricultural yields. What the effect of this distribution of irrigation water will be described in paragraph 6.3.

(26)

Figure 6.4 Teff response to irrigation.

An additional amount of 200 m3 water per hectare of agricultural field can have different effects (Figure 6.4). An increase in both grain and biomass yield is to be expected, but the factor in which this happens can vary from 5 to 100%. In dry years, the additional water will increase yields more significantly than in years with more rainfall. According to Araya and Stroosnijder (2011), as dry spells are often periodic and irrigation provides buffering capacity for these periods.

6.3 Livelihood impacts

That irrigation is generally good for farmers is an obvious conclusion. How exactly rural livelihoods are affected is a more complex question though. The average farm size in Tigray is 0,96 hectare (Heady et al., 2013), meaning that in the AOI 80 farmers could properly have their fields irrigated using RWH. This number should not at all be seen as representative for its potential in the region, but does provide an indication of the number of people who directly benefit. In the first paragraph, livelihood impacts for these farmers will be assessed according to Scoones’ (1998) criteria. In the next section then, impacts on others will be assessed. Because of the importance of inequality both in the effects of RWH and in determining these effects, a final section will pay attention to this.

Impacts for direct beneficiaries

While much has been written about the benefits of irrigation for poverty reduction, there is actually no consensus on this (Smith, 2004). According to Gebregziabher et al. (2009) macro-studies often find no relationships between use of irrigation and poverty rates, while micro-studies of households do. A study into the MPI of an area can be classified as a macro-study. Unfortunately, either underpinning or refuting this statement through conducting such a study proved impossible at this point. In their own study in Tigray, Gebregziabher et al. (2009) find that non-irrigator income is almost 50% lower than that of

(27)

irrigators (although the exclusive focus on income creates a bias). Namara et al. (2008) confirm the findings about poverty, reporting a “significant difference in incidence, depth and severity of poverty” (p. 114), although income inequality is higher for irrigators.

Better choosing the locations of roads also benefits natural resource base sustainability somewhat, as erosion slightly decreases. As the modelling shows, RWH can reduce erosion by 5 %, mainly for the land downslope. Irrigation could also have an adverse effect, as it can result in overexploitation of the soil (Smith, 2004); this indirect effect is hard to take into account however.

Finally, RWH also decreases vulnerability to droughts. Araya and Stroosnijder (2011) name three major causes of crop failure in Tigray: dry spells, a short growing period (due to a late onset of the kiremt rainy season) and a total lack of rain. Irrigation could reduce the risk of the first two of these. However, a trade-off needs to be made between buffer capacity and the area benefiting from irrigation, and thus between total yield and yield security. Mitigating the vulnerability to droughts can have further positive effects for poverty reduction however. As Devereux (2001b) discusses, hazards like droughts can lead to reduced productivity because of risk mitigation strategies, like growing low-risk, low-productivity crops, and coping strategies, trading off long-term productivity for short-term consumption. Dercon (2004) offers some evidence of the second effect in Ethiopia, showing that droughts cause welfare losses persisting for several years. Reducing the risk that droughts pose thus has wider benefits for poverty alleviation.

Impacts for others

Apart from the direct effects discussed previously, RWH is also bound to have indirect effects for others in the area. First, providing access to irrigation to some farmers is bound to create unequal outcomes. Irrigating farmers gain a comparative advantage and can thus outcompete non-irrigating farmers. Additionally, irrigation benefits resource-rich farmers most as these have better access to the means and skills needed to profit from irrigation (Smith, 2004; Namara, 2008). Finally, droughts often decrease inequality. Thiede (2014) shows, in a study that includes Tigray, that droughts increase equality within communities. Even though this is most likely because the rich ‘liquidize’ their livestock to keep up consumption while the poor go hungry, it is important to take into account. Inequality can thus be expected to rise.

RWH would not only benefit the people who profit directly however. Smith (2004) examines ways in which irrigation can reduce poverty in a literature review and comes up with four mechanisms. First, it increases the level and security of productivity and employment, something benefiting just the irrigating farmers and their farm workers. Secondly, though, it affects other sectors through linkages, for example because it creates economic growth (and thus demand for local goods) and lowers food prices. A 1.0 % growth of agricultural output is thus reported to lead to a 0.3-1.0 % growth of non-agricultural output. Thirdly, irrigation can lead to greater livelihood diversity for both the irrigating farmers (who can diversify into more risky crops) and other rural households (because of the demand it creates). Finally, irrigation water can be used for other purposes, like drinking water, too.

It can benefit non-irrigators during droughts too. Devereux (2007) argues that famines caused by droughts can be seen as a series of entitlement failures and that “effective intervention to address any one of these entitlement failures can prevent the drought or flood event from evolving into a food crisis” (p. 47). This suggests that RWH, even though it directly benefits only a part of the community, benefits all

(28)

and 4) improves possibilities for inter-community aid during droughts. These last two might prove to be problematic though, as they very directly create economic and social inequality.

Social-political impacts

While the picture so far seems quite positive, there is another side irrigation construction. Smith (2004) warns that it often mainly benefits the rich, thus increasing inequality, and can even worsen the situation of the poor, for example if it leads to mechanisation and local labour thus is not needed anymore. While it may be true that absolute poverty is more important for Ethiopia than economic inequality (Namara et al., 2008), the latter can never be separated from socio-political inequality. It can be especially problematic in conjunction with droughts, when food stocks become a strong source of power. The results could be increased economic inequality, as irrigating farmers could increase prices while others would have to sell assets, and increased social inequality, as food-aid from the irrigating farmers could create patronage networks.

According to Smith (2004), one important risk is that the poor are displaced of their lands, either because their land is appropriated by others or because canals and reservoirs are constructed on it. Land rights are thus a critical issue. In Ethiopia all land was until recently owned by the state, reflecting the country’s communist past. Though it was (and is) de facto owned by the people who work it, series land reforms made land tenure quite insecure (Dercon & Alayew, 2007). Starting in the late 1990s in Tigray, land registration programmes have been set up to privatise ownership in a process that has been described as low-cost and fair (Deininger et al., 2008). This is important as these enhance tenure security and guarantee transfer-rights, both of which increase land investments (Deininger & Jin, 2006). Holden et al. (2009) report land registration in Tigray to have resulted in less land conflicts, higher investments and soil conservation and a 45% increase in productivity. The low barrier for people to register the land they work means they are unlikely to be displaced because of irrigation construction and this risk is thus limited.

(29)

7. Conclusion and recommendations

Ethiopia in general and Tigray in specific are characterized by a poor population that is largely dependent on low-yield, rain fed sustenance agriculture. This in combination with the variable precipitation patterns has resulted in stunted economic growth and insecure livelihoods for the farmers. At the same time, it is a region in which the road network is rapidly expanding. Roads are known to cause enhanced erosion due to runoff. By integrating water management measures into the road through its location and through RWH, however, they can also positively impact local livelihoods.

The effects of RWH for livelihood sustainability have been analysed in this research using an interdisciplinary approach combining the different perspectives that are usually used in infrastructure analysis. This research has some limitations though. First, modelling is intrinsically limited in accuracy due to simplifications in the modelling environment and results should be considered as indications only. Second, no actual on-site research has been done, despite the fact that this is very relevant for predicting the impacts of an infrastructural intervention. A third limitation is that the costs of implementing RWH have not been taken into account. And finally, RWH has not actually been applied yet. This makes predicting its consequences a very much theoretical exercise and limited in reliability.

That being said, the results are positive for RWH. Smart choice of road locations could save 1-2 % of the yield losses over the next 20 years. As a single solution, the road location would thus have a neglectable effect. As calculations show however, the amount of water that can be harvested from roads is substantial. If this water could all be used for irrigation instead of recharging ground water, the teff yields and total biomass yields will increase, as is displayed in table 4.1. In this way, RWH improves livelihood sustainability for both its direct beneficiaries and other locals, but also increases economic inequality. This can be problematic as it also creates socio-political inequality, which can threaten the livelihoods of the less powerful. While this is by no means certain, the risk should be taken into account in development plans. When experimenting with RWH, following MPI impacts over time through CDR data is seen as a promising method to quantify poverty impact of RWH projects, although yet of limited applicability to Ethiopia as one of the countries with the lowest mobile phone ownership levels.

One thing that this research shows is the need for an interdisciplinary approach. Road water management is a physical aspect of a subject that is mostly analysed by social scientists. Connecting the various elements of this physical base to its various social impacts is difficult and leads to internal tensions, but is a necessary step. In practice this integration is done anyway, but usually with a simplified view of either one or the other. Doing it consciously, while trying to combine the different possible perspectives, is important for research to be relevant not just on paper but in practice as well.

In conclusion, implementing RWH can contribute to more sustainable livelihoods in Tigray. It is a relatively new concept, which first of all needs experimentation, but the research so far has been positive about its potential. While the benefits are limited, so are the likely costs. All in all, RWH is a promising approach for strengthening livelihood sustainability and experimenting with it in Tigray is recommendable.

(30)

References

- Alkire, S., Conconi, A., Robles, G. and Seth, S. (2015). Multidimensional Poverty Index, Winter 2014/2015: Brief

Methodological Note and Results. OPHI Briefing 27, University of Oxford, January.

- Araya, A., Keesstra, S. D., Stroosnijder, L. (2010). Simulating yield response to water of Teff with FAO’s aquaCrop model. Field crop research, 116, 196-204.

- Araya, A., & Stroosnijder, L. (2011). Assessing drought risk and irrigation need in northern Ethiopia. Agricultural

and Forest Meteorology, 151, 425–436

- Bizuwerk, A., Peden, D., Taddese, G., & Getahun, Y. (2005). GIS Application for analysis of Land Suitability and Determination of Grazing Pressure in Upland of the Awash River Basin, Ethiopia. International Livestock Research

Institute (ILRI), Addis Ababa, Ethiopia.

- Boelee, E., Yohannes, M., Poda, J., McCartney, M., Cecchi, P., Kibret, S., Hagos, F., Laamrani, H. (2012) Optios for water storage and rainwater harvesting to improve health and resilience against climate change in Africa.

Regional environmental change, 13, 509-519.

- Boelens, R., Guevara-Gil, A., & Panfichi, A. (2010). Indigenous water rights in the Andes: struggles over resources and legitimacy. Water Law, 20, 268-277

- Bues, A. (2011, April). Agricultural foreign direct investment and water rights: An institutional analysis from Ethiopia. In International Conference on Global Land Grabbing (pp. 6-8).

- Carter, D. C., & Miller, S. (1991). Three years experience with an on-farm macro-catchment water harvesting system in Botswana. Agricultural water management, 19(3), 191-203.

- Chambers, R. (1995). Poverty and livelihoods: whose reality counts? Environment and Urbanization, 7(1), 173-204

- DeGrassi, A. (2005). Transport, Poverty and Agrarian Change in Africa: Models, Mechanisms and New Ways Forward. IDS Bulletin, 36(2), 52-57

- Deininger, K., Ayalew Ali, D., Holden, S., & Zevenbergen, J. (2008). Rural Land Certification in Ethiopia: Process, Initial Impact, and Implications for Other African Countries. World Development, 36(10), 1786–1812 - Deininger, K., and S. Jin. (2006). Tenure Security and Land-Related Investment: Evidence from Ethiopia.

European Economic Review, 50, 1245–77.

- Dercon, S. (2004). Growth and shocks: evidence from rural Ethiopia. Journal of Development Economics, 74, 309–329

- Dercon, S., & Ayalew, D. (2007). Land Rights, Power and Trees in Rural Ethiopia. Working Paper. CSAE (University of Oxford)

- Devereux, S. (2000). Food insecurity in Ethiopia. Discussion paper for the Department for International Development, United Kingdom.

- Devereux, S (2001b). Livelihood Insecurity and Social Protection: A Re-emerging Issue in Rural Development.

Development Policy Review, 19(4), 507-519

- Devereux, S. (2007). The impact of droughts and floods on food security and policy options to alleviate negative effects. Agricultural Economics, 37, 47-58

- DRMFSS, Disaster Risk Management and Food Security Sector. (N.D.) Tigray Livelihood Zone Reports. retrieved May 2015 from www.dppc.gov.et/Livelihoods/Tigray/Pages/Tigray_Livelihood%20_Zone%20_Report.htm -Ethiopian demography and health. (N.D.). Tigray. Retrieved May 2015 from http://www.ethiodemographyandhealth.org/Tigray.html

- Fews.net,. (2015). Ethiopia | Famine Early Warning Systems Network. Retrieved 25 April 2015, from http://www.fews.net/east-africa/ethiopia

- Fisher, J. A., Patenaude, G., Meir, P., Nightingale, A. J., Rounsevell, M. D., Williams, M., & Woodhouse, I. H. (2013). Strengthening conceptual foundations: analysing frameworks for ecosystem services and poverty alleviation research. Global Environmental Change, 23(5), 1098-1111.

- Furlong, P., & Marsh, D. (2010). Ontology and Epistemology in Political Science. In: Marsh, D., & Stoker, G. (eds.). Theory and Methods in Political Science (3rd ed.). Houndmills: Palgrave Macmillan

(31)

- Gangnon, B. (2012). File:Simien Mountains National Park 12.jpg - Wikimedia Commons.

Commons.wikimedia.org. Retrieved 19 April 2015, from

http://commons.wikimedia.org/wiki/File:Simien_Mountains_National_Park_12.jpg

- Garcia-Landarte Puertas, D., Woldearegay, K., Mehta, L., van Beusekom, M., Agujetas Peréz, M., & van Steenbergen, F. (2014). Roads for water: the unused potential. Waterlines, 33(2), 120-138

- Gartner, C.M. (2014). The Agency of Infrastructure: A Critical Acquisition Framework for Understanding

Infrastructure Development within Inequitable Societies. Doctoral dissertation, University of Waterloo, Ontario,

Canada.

- Gebregziabher, G., Namara, R.E., & Holden, S. (2009). Poverty reduction with irrigation investment: An empirical case study from Tigray, Ethiopia. Agricultural Water Management, 96, 1837–1843

- Godfray, H. C. J., Beddington, J. R., Crute, I. R., Haddad, L., Lawrence, D., Muir, J. F., ... & Toulmin, C. (2010). Food security: the challenge of feeding 9 billion people. science, 327(5967), 812-818.

- Hanjra, M. A., & Qureshi, M. E. (2010). Global water crisis and future food security in an era of climate change.

Food Policy, 35(5), 365-377.

- Heady, D., Dereje, M., Ricker-Gilbert, J., Josephson, A., & Tafesse, A.S. (2013). Land Constraints and

Agricultural Intensification in Ethiopia: A Village-Level Analysis of High-Potential Areas. Ethiopia Strategy

Support Program working paper 58

- Heath, J., & Beeby, J. (2015). NR505 :: Concepts In GIS. Gisedu.colostate.edu. Retrieved 3 March 2015, from http://gisedu.colostate.edu/webcontent/nr505/ethiopia/group4/GIS%20Analyses.html#Crop

- Help.arcgis.com,. (2015). ArcGIS Desktop. Retrieved 26 April 2015, from http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//009z00000062000000.htm

- Holden, S.T., Deininger, K., & Ghebru, H. (2009). Impacts of low-cost land certification on investment and productivity. American Journal of Agricultural Economics, 91(2), 359–373

- Hurni, H. (1985). Erosion–productivity–conservation systems in Ethiopia. In: Proceedings 4th international

conference on soil conservation, Maracay, Venezuela (pp. 654-674).

- Lal, R. (1995). Erosion-crop productivity relationships for soils of Africa. Soil Science Society of America Journal,

59(3), 661-667.

- Namara, R., Makombe, G., Hagos, F., & Awulachew, S. B. (2008). Rural poverty and inequality in Ethiopia: does access to small-scale irrigation make a difference?. In: Bekele, S., Loulseged, M., & Denekew, A. (eds.). Impact of

irrigation on poverty and environment in Ethiopia.

- Oxford Poverty and Human Development Initiative (OPHI). (2013). Country Briefing: Ethiopia.

- Pimentel, D., Harvey, C., Resosudarmo, P., Sinclair, K., Kurz, D., McNair, M., ... & Blair, R. (1995). Environmental and economic costs of soil erosion and conservation benefits. Science-AAAS-Weekly Paper Edition,

267(5201), 1117-1122.

- Pinstrup-Andersen, P., Pandya-Lorch, R., & Rosegrant, M. W. (2001). Global food security. The Unfinished

Agenda. IFPRI, Washington, 7-17.

- Pokhriyal, N., Dong, W. and Govindaraju, V. (2015). Virtual Networks and Poverty Analysis in Senegal. Book of Abstracts: Scientific Papers, Orange Data for Development Challenge Senegal.

- Quandl (2015) Ethiopia: Mobile cellular subscriptions (per 100 people). retrieved May 2015 from https://www.quandl.com/data/WORLDBANK/ETH_IT_CEL_SETS_P2 and

https://www.quandl.com/collections/society/cell-phone-subscriptions-by-country

- Quinn, J., Frias-Martinez, V., & Subramanian, L. (2014). Computational Sustainability and Artificial Intelligence in the Developing World. AI Magazine Special Issue on Computational Sustainability.

- Rosegrant, M. W., & Cline, S. A. (2003). Global food security: challenges and policies. Science, 302(5652), 1917-1919.

(32)

- Scoones, I. (1998). Sustainable rural livelihoods: A framework for analysis. Institute of Development Studies, IDS Working Paper 72

- Shiferaw, B., & Holden, S. (1999). Soil erosion and smallholders' conservation decisions in the highlands of Ethiopia. World development, 27(4), 739-752.

- Shiferaw, A. (2011). Estimating soil loss rates for soil conservation planning in the Borena Woreda of South Wollo Highlands, Ethiopia. Journal of Sustainable Development in Africa, 13(3), 87-106.

-Smith, E.D.L. (2004). Assessment of the contribution of irrigation to poverty reduction and sustainable livelihoods.

International Journal of Water Resources Development, 20(2), 243-257

- Sonneveld, B. G. J. S. (2002). Land under pressure: the impact of water erosion on food production in Ethiopia. Shaker.

- Stallknecht, G. F., Gilbertson, K. M., Eckhoff, J. L. (1993) Teff: food crop for humans and animals. In: Janick, J., & Simon, J. E. (eds.). New crops. New York: Wiley

- Taddese, G. (2001). Land degradation: a challenge to Ethiopia. Environmental management, 27(6), 815-824. - Taffere, B. (2003). Efforts for Sustainable Land Management in Tigray: The Role of Extension. In Policies for

Sustainable Land Management in the Highlands of Tigray, Northern Ethiopia: Summary of Papers and Proceedings of a Workshop Held at Axum Hotel, Mekelle, Ethiopia, 28–29 March 2002.

- Teweldebrihan, M.D. (2014). Optimizing Intensified Runoff from Roads for Supplemental Irrigation: Tigray

Region, Ethiopia. Master of Science thesis, Technical University Delft, Delft, Netherlands

-The World Bank (2015). Urban population (% of total) retreved May 2015 from http://data.worldbank.org/indicator/SP.URB.TOTL.IN.ZS

Ethiopia. retreved May 2015 from http://data.worldbank.org/country/ethiopia

- Thiede, B.C. (2014). Rainfall Shocks and Within-Community Wealth Inequality: Evidence from Rural Ethiopia.

World Development, 64, 181–193

- Tibebe, D., & Bewket, W. (2011). Surface runoff and soil erosion estimation using the SWAT model in the Keleta watershed, Ethiopia. Land Degradation & Development, 22(6), 551-564.

- Timmer, C. P. (2004). Food Security and Economic Growth: an Asian perspective. Asian-Pacific Economic

Literature, 19(1), 1–17

- Wade, R. (1975). Administration and the Distribution of Irrigation Benefits. Economic and Political Weekly, 10, 1743-1747

- Wischmeier, W. H., & Smith, D. D. (1978). Predicting rainfall erosion losses-A guide to conservation planning.

Predicting rainfall erosion losses-A guide to conservation planning.

- Woldearegay, K., Garcia-Landarte Puertas, D., Van Steenbergen, F., Van Beusekom, M., & Agujetas, M. (2014).

Water harvesting from roads in Tigray, Northern Ethiopia: Practices, Opportunities and Design Considerations.

roadsforwater.org.

- Yimer, F., Messing, I., Ledin, S., Abdelkadir, A., (2008) Effects of different land use types on infiltration capacity in a catchment in the highland of Ethiopia. Soil use and management, 24, 344-349.

(33)

Appendix

A: Slope length map

(34)

B: Factor calculations

Area considerations: The area should not be too large to allow for the modelling of the effects of roads and be able to recognize the slope and stream scale influences. Also a digital elevation model (DEM) with a resolution of 30 x 30 meters is available for the area. This allows for an approximate modelling of roads. The area is inhabited and already contains a road. This makes the different scenarios and a comparison with a current situation possible. The area also harbours relatively steep slopes and on aerial pictures gully forming can be recognized (figure x, A). This indicates that erosion is a problem.

Rainfall Erosivity Factor (R)

Precipitation is a significant factor in soil erosion and is represented by the R factor

𝑅 = 0.0483 ∗ 𝑃1.610 ₍₂₎

Calculated from mean annual precipitation using the GIS raster calculator.

Only data with large grid cells (± 1 km cells) available, one mean is used for entire area. 𝑃 = 788.4 𝑚𝑚 < mean annual precipitation over area.

𝑅𝑚𝑒𝑎𝑛= 2227.4

Soil Erodibility Factor (K)

Erosion of soils is partly dependent on the composition of the soils. The composition determines the resistance particles have against erosion and is represented by the K factor.

𝐾 = 0.0293(.65 − 𝐷𝑔 + .24𝐷𝑔2_{) ∗. ..} 𝑒𝑥𝑝[−.0021(_𝑓𝑂𝑀 𝑐𝑙𝑎𝑦) − .00037𝑓𝑐𝑙𝑎𝑦( 𝑂𝑀 𝑓_{𝑐𝑙𝑎𝑦})2 − 4.02𝑓𝑐𝑙𝑎𝑦 + 1.72𝑓𝑐𝑙𝑎𝑦 2_] (3) 𝐷𝑔 = −3.5𝑓𝑐𝑙𝑎𝑦− 2.0𝑓𝑠𝑖𝑙𝑡 − 0.5𝑓𝑠𝑎𝑛𝑑 (4) 𝑓𝑠𝑎𝑛𝑑 = sand fraction 𝑓𝑠𝑖𝑙𝑡 = silt fraction 𝑓_{𝑐𝑙𝑎𝑦} = clay fraction

𝐷𝑔 = mean geometric particle size 𝑂𝑀 = percentage organic matter

Table B.1: Soil characteristics (Tibebe, & Bewket, 2011)

Soil type Bulk density (g cm−3 ) AWC (mmH2O/mm soil) Hydraulic conductivity (mm h−1 )

Textural composition (per cent by weight)

Organic carbon (per cent by weight)

Clay Silt Sand

Eutric Cambisol

Integration: the Road to Food Security Tigray region, north Ethiopia