INTRODUCTION
Mixed crop-livestock systems are important in terms of area and contribution to people's livelihood in sub-Saharan Africa. Apart from constraints such as low crop and livestock productivity, as in Ethiopia (Benin et al., Reference Benin, Ehui, Pender, Pender, Place and Ehui2006), increasing water scarcity and water competition between different uses hamper meeting increasing food demands in these systems. This is further aggravated by severe land degradation and climate change. Therefore, there is an urgent need to improve agricultural productivity, and water productivity in particular, in order to secure both people's livelihoods and the environment (Faurès and Santini, Reference Faurès and Santini2008; Molden et al., Reference Molden, Oweis, Steduto, Bindrban, Hanjra and Kijne2010).
Water productivity measures the ability of agricultural systems to convert water into food and feed and is defined as the ratio of agricultural outputs to the volume of water depleted for production (Molden et al., Reference Molden, Oweis, Steduto, Bindrban, Hanjra and Kijne2010). In the crop sector, crop water productivity (CWP) is a well-known concept (Bouman, Reference Bouman2007) and interventions to improve CWP are well established (Rockström and Barron, Reference Rockström and Barron2007). However, in mixed crop-livestock systems the interactions between livestock and water are also important, as livestock production depletes large amounts of water for feed production and is partly responsible for environmental degradation due to overgrazing (Amede et al., Reference Amede, Descheemaeker, Peden and van Rooyen2009a). The concept of livestock water productivity (LWP) was introduced by Peden et al. (Reference Peden, Tadesse and Misra2007) to investigate these interactions and find ways to increase livestock production without depleting more water or causing further environmental degradation. Although the LWP concept is still relatively new, recent attention to it (e.g. Amede et al., Reference Amede, Descheemaeker, Peden and van Rooyen2009a; b; Haileslassie et al., Reference Haileslassie, Peden, Gebreselassie, Amede and Descheemaeker2009) revealed that improvements in LWP help to meet increased demands for animal products without further depleting and degrading already scarce water resources.
Based on a literature review, Descheemaeker et al. (Reference Descheemaeker, Amede and Haileslassie2010) showed that interventions to improve LWP in crop-livestock systems could be grouped into three biophysical categories, namely interventions targeting feed management, water management and animal management. Technical interventions have to be integrated with institutional and policy measures to ensure adoption (Amede et al., Reference Amede, Geheb and Douthwaite2009b). Although this is true in general and applicable in most circumstances, concrete information on the potential effects of various interventions on LWP is lacking.
Therefore, the purpose of this study was to conduct an ex-ante evaluation of the effect of promising interventions on LWP in mixed crop-livestock systems in a water scarce area in Ethiopia. This theoretical assessment of potential improvements could help to inform policy-makers, development actors and investors about where to focus and invest to achieve positive impacts on water productivity. The objectives of this study consisted of: (i) the characterization of the baseline situation, (ii) the assessment of productivity gaps in the baseline situation, (iii) the identification of a set of promising interventions to alleviate the productivity gaps and (iv) the assessment of the effects of the interventions on livestock outputs and water depletion, and therefore LWP.
MATERIALS AND METHODS
For the simulation of the effects of different interventions, this study used the farm characteristics of a typical household in a water scarce region in the highlands of northern Ethiopia.
The ex-ante evaluation of interventions consisted of four steps, including the baseline characterization, the productivity gap analysis, the identification of promising interventions and the simulation of the resulting changes in water productivity. Only the potential biophysical effects of the interventions were assessed, and it was assumed that money, labour and other inputs were sufficiently available for implementation. As such, this study did not include an evaluation of the achievability of the interventions in terms of resource availability, labour requirements and social acceptance. Although the proposed interventions may not be immediately viable for smallholder farmers, the analysis may point to relevant investment options for sustainable intensification of farming systems.
Baseline characterization
A typical mixed crop-livestock system of the northern Ethiopian highlands in Lenche Dima watershed (see Mekonnen et al., (Reference Mekonnen, Descheemaeker, Tolera and Amede2010) for a detailed description of the study area) was taken as the current practice (baseline). The study site is representative of the highlands of Ethiopia where subsistence agriculture dominates and water scarcity is a constraint to agricultural production. The farming system and biophysical environment in the study site was characterized based on field observations, farmer group discussions and semi-structured interviews with key informants. In addition, a detailed questionnaire was used to conduct a survey of 54 households. The questionnaire covered issues of land and livestock holding, crop and livestock production, farming practices, water use and socio-economics, and covered the production year 2007/2008. The mean land and livestock holding as well as different crop and livestock production components were determined for an average household based on the survey results. Water flows for the different crops, land uses and feed types were determined with the soil water balance model BUDGET (Raes et al., Reference Raes, Geerts, Kipkorir, Wellens and Sahli2006). More details on the methodology for the baseline characterization in the study area can be found in Mekonnen et al. (Reference Mekonnen, Descheemaeker, Tolera and Amede2010).
Productivity gap assessment
In this study, a productivity gap refers to one of two concepts: (i) the actual difference between the current and the common or potential level of productivity, and (ii) a reason for a low level of productivity. For the first concept, current crop and livestock productivity levels were compared with common and potential productivity values reported in the literature. For the second concept, reasons for low productivity were sought by analysing water flows and energy expenditure by animals for different activities. For the water flows, productive (transpiration) and unproductive (evaporation, runoff, deep percolation) water flows were determined and compared to water input (rainfall). By investigating the energy consumption of animals for different activities, a better understanding was gained on how much energy was lost for unproductive purposes.
Ex-ante evaluation of the effects of interventions on LWP
A broad set of promising technical interventions was determined based on the productivity gap analysis and a literature review. From that set, the locally available and viable options were identified after an assessment of the available inputs and through talks with farmers and local extension agents.
A spreadsheet model was developed to conduct the ex-ante evaluation of the effect of interventions on LWP. The model operates in Excel and is a simple tool to monitor changes in the system components when certain factors or variables are modified. The building blocks of the model include different interlinked components of a production unit, their interactions and influencing factors (Figure 1). Crop and livestock interventions exert their effect through influencing certain factors, components and/or interactions, thus affecting LWP. In what follows, the different building blocks of the model are explained in more detail.
Structure of the spreadsheet model with production components (boxes), interactions (arrows) and influencing factors (bullet points).
LWP: livestock water productivity.
Depending on the scale of the analysis, the production unit can be a household, community or a larger scale unit like a basin. In this study, the model was used to investigate effects of interventions at the household level. The production unit contains land from which the feed is produced and from which the water is depleted. The production unit also manages a livestock herd, producing livestock outputs. The two sides are linked by energy production from feed and energy requirements for livestock maintenance and production, and they combine to determine LWP.
The landholding of the production unit comprises different land uses. Depending on the cropping pattern, the crop and feed productivity, and the feed management, certain amounts of different feed types are produced. In producing that feed, water is evaporated and transpired. Besides climate, land use and soil type, the volume of water depleted also depends on agronomic practices, grazing management, land management and water management. The produced feed supplies energy to the livestock to perform activities and produce outputs. The metabolizable energy (ME) supply depends on the energy content of the feed, its digestibility and feed management (e.g. residue treatment, feed processing, feed storage), and is defined by ARC (1980) as the gross energy of the feed less that of the faeces, urine and combustible gases, expressed as megajoules (MJ) of ME per day for a diet or MJ per kilogram of feed or diet dry matter (MJ kg−1 DM). It represents that portion of the feed energy that can be used by the animal. The ME supplied by the feed is used by the livestock for maintenance, different physical activities and production. The required energy depends on the size and composition of the livestock herd, the production level and the activities. The energy requirements were calculated based on the equations from King (Reference King1983). The livestock outputs are determined by the livestock herd, animal productivity (including mortality, milk productivity, off-take rates, etc.), and the livestock preferential uses by the production unit. Market values of the different livestock outputs determine the total value of the livestock output, constituting the nominator of the LWP equation (see Mekonnen et al. (Reference Mekonnen, Descheemaeker, Tolera and Amede2010)).
The following assumptions and steps were taken in the analysis:
– It was assumed that in the baseline scenario, the energy balance was neutral, i.e. the available energy from the feed matched the energy required to attain the current level of production.
– The change in livestock production caused by the application of an intervention was determined based on the change in available energy. This change could be caused by an increase in the available ME because of more or better quality feed, or by a decrease in energy requirements. If the available energy increased, it was assumed that it was entirely used for milk production and live weight gain, which was converted into meat production using the dressing percentage.
– The change in water flows was determined based on changes in land use or land management due to the interventions, and soil water balance modelling with BUDGET (Raes et al., Reference Raes, Geerts, Kipkorir, Wellens and Sahli2006).
More details on the methodology to calculate LWP can be found in Haileslassie et al. (Reference Haileslassie, Peden, Gebreselassie, Amede and Descheemaeker2009) and in Mekonnen et al. (Reference Mekonnen, Descheemaeker, Tolera and Amede2010). Besides LWP, other productivity variables were calculated for the different interventions. These included feed productivity indicators such as feed biomass and energy productivity per ha and per cubic meter of water. Livestock productivity indicators included milk production per ha and per cubic meter of water as well as compound financial livestock outputs per tropical livestock unit (TLU, equivalent to 250 kg live animal weight), ha and cubic meter of water (i.e. LWP).
For each of the identified interventions specific assumptions and calculations are discussed below. The interventions are categorized in three groups, related to feed management, water management and animal management.
Feed management
Fertilizer application. For this intervention, it was assumed that farmers applied 100 kg of urea (46% N) and 100 kg of diammonium phosphate (46% P205 and 18% N) to a hectare of teff fields. As this equalled the recommended rate, the teff grain yield was assumed to increase to the potential yield of 2 t ha−1. In addition, it was assumed that the harvest index did not change and that transpiration increased by 20% because of the increased biomass production. As a result of better ground cover by the increased biomass, it was assumed that evaporation was decreased by the same amount, so that ET stayed constant.
Introduction of fodder trees. The effect of this intervention was simulated by assuming that half of the homestead area, normally planted with maize, was dedicated to growing 300 sesbania trees (Sesbania sesban). The feed production from the leaves and shoots of sesbania was 8 t DM ha−1 (Gutteridge and Shelton, Reference Gutteridge and Shelton1994) and its ME content 10.3 MJ kg−1 DM (SLP, 2008). For determining the water flows from the fodder trees, a simplified approach was used. Runoff was set at 10 % of the rainfall and evapotranspiration was assumed to equal the effective rainfall (total rainfall − runoff). It was further assumed that evapotranspiration in the growing season (May to December) accounted for 80% of the annual evapotranspiration. Of this, 25% was lost through evaporation and 75% through transpiration.
Urea treatment of crop residues. It was assumed that one third of the teff straw was treated with urea for silage production. The ME content of the silage was 9.2 MJ kg−1 DM (Dejene et al., Reference Dejene, Bediye, Kehaliw, Kitaw and Nesha2009). There was no change in the water flows for this intervention.
Feeding concentrates. Cottonseed cake and wheat bran were assumed to be fed preferentially to the oxen and cows at a total rate of 1 kg day−1. The ME content was 10.5 MJ ME kg−1 for cottonseed cake (SLP, 2008) and 10.1 MJ ME kg−1 for wheat bran (MAFF, 1975). As neither cotton nor wheat was produced locally, their water consumption was added to the calculation as virtual water. However, as these are by-products, the water and area required for their production was derived by applying a biomass-based partitioning of the total crop water consumption (see Mekonnen et al. (Reference Mekonnen, Descheemaeker, Tolera and Amede2010) for more details on this approach). A country-wide average cottonseed yield of 0.9 t ha−1 (FAOSTAT, 2009) was combined with an oil extraction rate of 70% and an oilcake conversion factor of 0.05 (Ramachandra et al., Reference Ramachandra, Taneja, Sampath, Anandan and Angadi2007) to calculate the required production area for the amount of cottonseed cake given to the livestock. For wheat, the same procedure was followed based on a wheat grain yield of 1.6 t ha−1 (CSA, 2008) and a wheat bran conversion factor of 0.08 (Ramachandra et al. Reference Ramachandra, Taneja, Sampath, Anandan and Angadi2007). The evapotranspiration from cotton was determined based on its crop water requirements and the assumption that 75% of the water requirements was actually evaporated and transpired in the growing season. The crop water requirements were calculated using the crop coefficients reported in Allen et al. (Reference Allen, Pereira, Raes and Smith1998) and climate factors for cotton-growing areas in Ethiopia. For wheat, 450 mm of evapotranspiration was adopted from Haileslassie et al. (Reference Haileslassie, Peden, Gebreselassie, Amede and Descheemaeker2009) for a wheat growing area in Ethiopia.
Improved food-feed crops. In this intervention, chickpea was replaced by an improved cowpea variety, producing 1.5 t grain ha−1 and 2.5 t haulms ha−1, which could be used for fodder (Singh et al., Reference Singh, Ajeigbe, Tarawali, Fernandez-Rivera and Abubakar2003). The ME content of the haulms was 8.5 MJ kg−1 DM (SLP, 2008). The evapotranspiration for cowpea was assumed to be the same as that for chickpea, because of the similarity in length of the growth stages and crop coefficients (Allen et al., Reference Allen, Pereira, Raes and Smith1998). However, the water depleted for feed production slightly increased due to the higher ratio of residue biomass over grain biomass for cowpea (1.67) (Singh et al., Reference Singh, Ajeigbe, Tarawali, Fernandez-Rivera and Abubakar2003), as compared to chickpea (1.2).
Water management
Surface water harvesting structures. The effect of the water harvesting structures was manifested through a reduction in the energy required for walking to drinking points in the dry season. Instead of walking 9 km, animals were assumed to walk 2 km, as in the wet season (see Descheemaeker et al. (Reference Descheemaeker, Amede and Haileslassie2010) for more details on the assumptions).
Field level soil and water conservation measures. As a result of soil and water conservation, the grain yield increase was assumed to be 10% for all crops (with no change in harvest index). This is a cautious increase, given the range of 5–296% grain yield increase reported in the literature for Ethiopian conditions (Alemayehu et al., Reference Alemayehu, Yohannes and Dubale2006; Vancampenhout et al., Reference Vancampenhout, Nyssen, Gebremichael, Deckers, Poesen, Haile and Moeyersons2006). The conservation measures resulted in a decrease in the runoff coefficient of about 10–20% (McHugh, Reference McHugh2006). This was simulated in BUDGET by lowering the runoff curve number. The effect on other water flows was manifested as an increase in deep percolation, whereas evapotranspiration changed only slightly.
Animal management
Improved health services. It was assumed that improved health services reduced the current mortality rates from 10% to 5%. The effect on LWP of lowering the mortality rate was simulated by increasing the livestock financial outputs by an overall 5%. This was based on the assumption that the baseline production level was attained under the burden of a 10% mortality rate. The production level increased by 0.95/0.90, or ≈5%, if the mortality rate was reduced to 5%. In addition, it was assumed that this change may not affect the water flows, as the consumed feed remained almost the same, because animals that died also consumed feed before dying.
Crossbred cows. A crossbred cow was introduced in addition to the existing herd. The crossbred cow was assumed to weigh 400 kg and produce an average 8 litres of milk per day during a nine-month lactation period (Ahmed et al., Reference Ahmed, Ehui and Assefa2004). The extra energy required for the crossbred cow was calculated taking into account energy for maintenance, feeding, lactation, pregnancy, walking (0.5 km every day) and growth. This energy was converted into water requirements by assuming that the energy intake was derived from a 4:1:1:1 ratio of untreated teff straw: urea-treated teff straw: wheat bran: oilseed cake, and by taking into account the water evapotranspired in producing these feeds.
Reducing animal numbers. Two different interventions were evaluated, namely (i) removing the small ruminants and (ii) removing one of the two oxen from the herd of the baseline situation.
In the first case, energy for maintenance, feeding, lactating, pregnancy and walking for small ruminants was saved. As such, more energy became available for the other animals, which was converted into milk and meat production. On the other hand, no benefits were gained from selling goats. The evapotranspiration decreased slightly because of the smaller TLU and therefore the smaller grazing area. In the case of the reduced number of oxen, the energy required for maintenance, feeding and walking was reduced. As the farmer still had to plough his fields with a pair of oxen by hiring an ox from another farmer, the energy required for ploughing did not change. The overall saved energy was invested in milk and meat and the benefits from draught power were halved. The water used for feed production decreased because of the smaller TLU and therefore the smaller grazing area.
RESULTS AND DISCUSSION
Characteristic of the farming system
In the baseline scenario, the average household's landholding was comprised of 2 ha of cropland and 1.2 ha of communal grazing land (Table 1). From that landholding, 6.7 t of feed was produced, supplying a total of about 54 GJ ME (Table 1). The depleted water per ha for the production of feed varied from 212 to 289 mm for crop residues to 415 mm for grazing land (Table 2). The livestock holding of an average household comprised five cattle heads, three goats and one donkey (Table 3). The off-take rates varied from 5% for cattle to 10% for goats and mortality rates were quite high at around 10% for both cattle and small ruminants. Milk production was determined by a lactation period of 200 days and a daily milk yield ranging from a maximum of 2.1 litres to a minimum of 1.3 litres (Table 3). Other annual livestock outputs included 6 t manure, 88 animal days for ploughing and threshing, and 220 transport events (e.g. transporting water and crop produce) by donkeys. Based on the local market values of 2008 for the different livestock outputs and services, a financial output of US$ 591 was obtained, and LWP was US$ 0.052 m−3.
Landholding and feed production from different crops and communal grazing land for an average household in the baseline scenario
†Ratio of residue biomass over grain biomass, harvest index = 1/(1 + conversion factor).
‡Fraction of the total residue biomass used as feed.
§ME: metabolizable energy.
Evapotranspiration† (mm ha−1) for the production of grain and feed from major crops and land use types in the baseline scenario.
†Evapotranspiration during the growing season, after partitioning based on the harvest index and the feed use factor.
Livestock holding and some livestock production variables for the baseline production system.
Gap analysis
Based on the water flows from the main land uses (Figure 2), it was found that in the baseline scenario, productive water flows (i.e. transpiration) accounted for 61% and 48% of the total water flows in the growing season in cropland and grazing land, respectively. This meant that in the baseline scenario the productivity gap consisted of a loss of roughly half of the rainfall through unproductive evaporation, runoff and deep percolation. If some of these unproductive water flows could be shifted to productive transpiration, more biomass would be produced with the same amount of water, which would boost the system's water productivity.
Water flows per ha of cropland and grazing land in the growing season, including evaporation (E), transpiration (T), runoff (RF) and deep percolation (DP) for the baseline scenario.
The breakdown of the energy expended by the livestock for maintenance, different activities and production provided insight into gaps in animal productivity. For the livestock herd of an average household in the baseline scenario, about 70% of the annual energy was spent on maintenance, and 12% was lost by walking (Figure 3). Contrastingly, only 7% and 3% of the annual energy was used for live weight gain and milk production, respectively (Figure 3). The main reason for this was the inadequate quantity and quality of the feed. Improving the feed availability, its digestibility and energy content would allow the livestock to produce more, so that the share of maintenance to the total annual energy budget would decrease. The reason for the high energy loss by walking was the inadequate availability of livestock drinking points. In the dry season, when water shortage was severe, animals had to walk three to four hours per day to access water.
Annual energy budget for the livestock herd (see Table 3) of an average household in the baseline scenario.
Productivity gaps were also diagnosed by comparing prevailing levels of crop and livestock productivity in the study area with common levels. With regard to cattle milk production, all parameters (lactation period, daily milk yield, milk production per lactation period, age at first calving and calving interval) fell within the reported range for indigenous cattle in Ethiopia, but below the average of the reported values (Table 4). The combined slaughter and commercial off-take rates of cattle and goats were well below the national average of about 10–15% for cattle and 25–35% for small ruminants (Negassa and Jabbar, Reference Negassa and Jabbar2008). Based on the above, and the relatively high mortality rates of 10%, it was concluded that animal productivity in the baseline scenario was low.
Milk production parameters for the baseline scenario, as compared to the range and the average of reported values in the literature† for indigenous cattle in Ethiopia.
†Reported values are extracted from the literature, including Debrah and Anteneh (Reference Debrah and Anteneh1991); Desta (Reference Desta2002); Hussen et al. (Reference Hussen, Tegegne, Yousuf and Gebremedhin2008); Tessema et al. (Reference Tessema, Gebrewold, Tegegne, Diediou and Hedge2003).
With respect to crops, the attained crop yields per hectare were similar to the yield values reported by CSA (2008) for the same region (North Wollo) (Table 5). However, given the higher potential yields under recommended fertilizer rates (Table 5), improvements would be possible.
Current yield levels (t ha−1) of major crops for the baseline scenario and in North Wollo (CSA, 2008) compared to potential yields under recommended fertilizer rates.
†For chickpea, there are no fertilizer recommendations in Ethiopia.
Effects of different interventions
Promising interventions were identified in three categories of technical strategies, including feed management, water management and animal management. The interventions focusing on feed management all resulted in increased feed availability and feed ME as compared to the baseline (Table 6). Except for feeding concentrates and introducing a crossbred cow, which required extra (virtual) land and water, the land area and the volume of water depleted changed only slightly (Table 6). The two interventions directed at water management exerted their effect in different ways. The increase in milk and meat production due to water harvesting resulted from saving energy for walking to drinking points. On the other hand, the increased livestock production due to soil and water conservation measures was explained by increased biomass yield and, therefore, feed production, because of better soil water availability. Also the effects of the animal management interventions were manifested differently. Improved livestock health, leading to lower mortality rates, led to higher animal outputs from the same feed and water consumption. Bringing in a crossbred cow increased the livestock holding, the (virtual) land in use to produce the required feed and the water depleted from that, but also the milk production from 172 to 1252 litres year−1 (Table 6). Reducing animal numbers led to small reductions in depleted water, land in use and feed production, but resulted in higher milk and meat production (Table 6), because of savings in energy for non-productive activities and maintenance.
Characteristics and effects on production per annum of the different interventions in comparison with the baseline scenario.
†Includes virtual land and water for the production of purchased concentrates.
‡Combination 1 includes fertilizer application, urea treatment, concentrates, water harvesting, animal health, crossbred cow.
§Combination 2 includes all interventions under combination 1, plus the reduction of the number of oxen to only one.
TLU: tropical livestock unit.
The land and water productivity of feed and ME increased with most interventions in varying degrees (Table 7). Land productivity decreased with feeding concentrates and introducing a crossbred cow, which was partly fed on concentrates, because of the virtual land and water needed to produce the concentrates (Table 7). Feed and energy water productivity increased particularly with interventions like fertilizer application, and the introduction of fodder trees, concentrates, improved food-feed crops, and soil and water conservation measures (Table 7).
Land and water productivity of feed and metabolizable energy (ME) of the system per annum after introduction of different interventions as compared to the baseline scenario.
†Combination 1 includes fertilizer application, urea treatment, concentrates, water harvesting, animal health, crossbred cow.
§Combination 2 includes all interventions under combination 1, plus the reduction of the number of oxen to only one.
Livestock land and water productivity values are summarized in Table 8. Land and water productivity of milk increased most strongly with the introduction of a crossbred cow, reducing the number of oxen by one and applying fertilizer to the teff fields (Table 8). With respect to the total financial livestock outputs per hectare, the top three interventions were applying fertilizers, reducing the number of oxen and providing adequate animal drinking points through water harvesting. For the total financial livestock outputs per TLU, the introduction of a crossbred cow entered the top three of most successful interventions. The highest monetary gains in LWP were made by reducing the number of oxen and applying fertilizers (Table 8).
Productivity of livestock (LS) products and outputs after introduction of different interventions as compared to the baseline scenario.
†Combination 1 includes fertilizer application, urea treatment, concentrates, water harvesting, animal health, crossbred cow.
§Combination 2 includes all interventions under combination 1, plus the reduction of the number of oxen to only one.
In order to take away the main productivity gaps all at once, interventions improving feed quality and quantity were combined with providing adequate drinking points, improved health and improved animal productivity. The combination of several interventions like fertilizer application, urea treatment, feeding concentrates, water harvesting, animal health, introducing a crossbred cow and reducing the number of oxen did not result in considerable improvements in feed biomass productivity (Table 7). However, in terms of milk productivity, overall financial animal productivity (per TLU) and livestock water productivity, the combinations of interventions showed considerably higher returns than the single interventions (Table 8).
Context and applicability of the ex-ante evaluation
The low animal productivity found in the smallholder mixed crop-livestock systems of Ethiopia is common throughout Africa (Otte and Chilonda, Reference Otte and Chilonda2002). Together with inadequate water resources and low animal potential, under-nutrition due to low quantity and quality of feed are the main explanatory factors (Kebreab et al., Reference Kebreab, Smith, Tanner, Osuji, Ayantunde, Fernández-Rivera and McCrabb2005). One of the reasons for the bad animal nutrition is the high dependence on crop residues, which are deficient in ME and important nutrients, and characterized by low intake, low protein content, poor digestibility and rough texture (Blümmel et al. Reference Blümmel, Samad, Singh and Amede2009). Low cost strategies to overcome the inadequate energy supply from a diet based on crop residues include supplementation and treatment of crop residues (Lenné et al., Reference Lenné, Fernandez-Rivera and Blümmel2003), which can increase both the intake and the digestibility of the feed (Blümmel et al., Reference Blümmel, Samad, Singh and Amede2009). Feeding trials in Ethiopia indicated that as a result of supplementation and/or residue treatment, live weight gain and milk yield were improved to a varying degree (Dejene et al., Reference Dejene, Bediye, Kehaliw, Kitaw and Nesha2009; Koralagama et al., Reference Koralagama, Mould, Fernandez-Rivera and Hanson2008; Melaku et al. Reference Melaku, Peters and Tegegne2004). In accordance with these findings, our spreadsheet model simulated that the feed management interventions led to potential improvements in daily milk production to about 3.2 litres and daily liveweight gain for a 250 kg animal to about 0.8 kg.
With respect to improving inadequate water resources, a study in the Ethiopian Rift Valley demonstrated that water harvesting structures enabled farmers to combine vegetable production with small-scale dairying, which significantly increased milk production (Puskur, personal communication). Also Tesfay (Reference Tesfay2007) showed that providing on-site drinking water reduced animal stress and energy costs and consequently increased milk production. This is in line with the simulation in this study of increased milk production due to water harvesting structures in the homestead.
Considering animal productivity in terms of growth, meat and milk production, and reproduction, Zebu cattle performance is inferior to improved Western breeds, as they have been selected for adaptive rather than productive traits, which enables them to survive in harsh conditions. The potential of crossbred cows in smallholder farming systems in the tropics to improve household nutrition, family income and food security, and foster sustainable system intensification is well documented (Bebe et al., Reference Bebe, Udo, Rowlands and Thorpe2003). In this study, the introduction of a crossbred cow was one of the more effective interventions in terms of improving livestock water productivity. The effect of reducing animal numbers is less studied, although its potential to reduce the proportion of available feed used for maintenance was mentioned by Kebreab et al. (Reference Kebreab, Smith, Tanner, Osuji, Ayantunde, Fernández-Rivera and McCrabb2005). The work of Zerbini et al. (Reference Zerbini, Wold, Shapiro, Starkey and Kaumbutho1999) on multipurpose cows (milk and draught) explored ways to intensify mixed farming systems by using fewer, but more productive animals. The present study provided additional evidence that reducing animal numbers resulted in more productive use of feed and water resources.
Assumptions and limitations of the study
The simulation of the effects of interventions on livestock water productivity was based on a number of simplifications and assumptions, as described under Materials and Methods. The theoretical assessment of potential effects did not take into account the economic, market, institutional or bio-physical constraints limiting the adoption of the interventions and, therefore, did not include an analysis of the achievability of the interventions. For example, in many cases farmers might prefer growing eucalyptus trees in their homestead instead of sesbania trees, as the former will produce more cash income in the short term. In addition, we only considered the average situation, ignoring conditions leading to suboptimal effects of the interventions in bad years. For example, high livestock mortality due to drought or crop failure due to pest infestations might wipe out the positive effect of certain interventions, thus increasing the risk for farmers and reducing the likelihood of adoption. Besides that, the simulations were not calibrated and validated against actual productivity, so that the results have to be interpreted cautiously. Keeping these limitations in mind, the present study had its value in informing development actors on which interventions are potentially effective in improving livestock water productivity, thus guiding investment options.
CONCLUSIONS
This paper presented a theoretical ex-ante analysis of the potential effects of interventions in smallholder mixed crop-livestock farming systems. Although presented for the specific case of the Ethiopian highlands, the approach and spreadsheet model are also applicable to analyse systems and identify best bets for improving water productivity in other locations and circumstances. The findings of this study have to be interpreted cautiously because of a number of simplifications and assumptions but are valuable in informing development actors about the effectiveness of interventions.
The identification of major productivity gaps related to energy budgets, water flows and animal productivity pointed to the importance of interventions targeting these gaps. The specific interventions focused at improving feed management, water management and livestock management had a positive effect on improving livestock water productivity, ranging from a potential 4% to 94% improvement of LWP. Combining the different interventions led to a stronger improvement than any single interventions.
This analysis also indicated how far the livestock water productivity of the existing system could be increased within reasonable limits of herd size and landholding. Assuming that farmers had the financial and labour resources to purchase fertilizers and concentrates, build and maintain a water harvesting structure, buy and manage a crossbred cow, and have access to veterinary services, LWP could be tripled on the same land holding (excluding the virtual land needed to produce feed supplements). This confirmed the notion that intensification, consisting of keeping fewer, but more productive animals, in combination with more and higher quality feeds and better animal husbandry, could lead to production increases without further depletion and degradation of natural resources.
Acknowledgements
The authors would like to acknowledge the Bundesministerium für wirtschaftliche Zusammenarbeit und Entwicklung (BMZ) for funding the project on Improving Water Productivity of Crop-Livestock Systems of Sub-Saharan Africa. We would like to thank the district Bureau of Agriculture and the development agents in the site for sharing information with us and facilitating our work. Many thanks go to the farmers of Lenche Dima watershed for their co-operation.
