Reducing uncertainty in nitrogen budgets for African livestock systems

These models aggregate data from soil or landscape units using a set of rules to calculate N budgets for countries, regions, continents, and the world. Stoorvogel and Smaling (1990) developed a method to calculate nutrient (NPK) budgets for croplands, including livestock-related flows of manure applied to fields and excretion during grazing of crop residues. Budgets were calculated for systems aggregated into land and water classes. Since Stoorvogel and Smaling (1990) others groups have implemented modifications to the method to make it applicable at other levels (Sheldrick et al 2002, 2003, Sheldrick and Lingard 2004), or to make the calculations spatially-explicit (Henao and Baanante 1999). Lesschen et al (2007) used process-based simulation models to estimate some of the flows; Liu et al (2010) used finer spatial resolution, and Potter et al (2010) used refined land cover layers.

Bouwman et al (2005) developed the livestock module of the land use model IMAGE (Bouwman et al 2006), which calculates N budgets for two livestock systems. IMAGE has been used for scenarios of change such as the Millennium Ecosystem Assessment (2005), and the International Assessment of Agricultural Knowledge, Science and Technology for Development (IAASTD) (McIntyre et al 2009), for which its resolution and main assumptions seem appropriate. However, for analyses of nutrient management in Africa, resolution appears too coarse, and the systems definition (pastoral, and mixed-landless) does not cover the diversity of livestock systems found in Africa. The FAO livestock's long shadow (Steinfeld et al 2006) assessed the role of livestock in the global N cycle estimating N₂O emissions at sub-continental level for grazing, mixed and industrial systems. Steinfeld et al (2006) followed IPCC (1997) to estimate emissions from manure management using excretion rates by livestock type, fraction of manure managed, and emission factors for each manure management system. The authors produced tables and maps depicting emissions of N₂O and N deposition following volatilization. Unfortunately, they did not present the data used for the calculations, neither do they refer to the sources used. Liu et al (2010) estimated global N flows for croplands using coarse assumptions on the management of feeds and manure and their spatial allocation, which may have influenced largely the N budgets calculated and conclusions drawn for Africa.

More than two decades after Stoorvogel and Smaling (1990), there has been progress towards reducing the uncertainty of large scale N budgets. However, the inclusion of livestock systems appears still rudimentary. Major short comes are: (a) livestock densities are usually modelled based on environmental factors because of the lack of recent livestock censuses, (b) land cover maps lack the resolution needed to capture the diversity observed in Africa, (c) constant factors to estimate livestock excretion rates are used, which also lack spatial distribution, (d) livestock excreta are usually assumed to be uniformly distributed onto the land, and (e) nutrient losses and emission factors are often guessed using estimates from other continents and climates. Moreover, all studies use coarse assumptions on production, availability, and the use of crop residues in livestock systems and mostly feedbacks between croplands and grazing lands are not considered, with some assessments assuming that rangelands are endless nutrient sources. All these pitfalls reflect not only that there are few data for livestock systems for Africa, but also that systems thinking is needed to address complexity and to reduce uncertainty.

We evaluated studies including livestock in N budgets at plot, farm and village and community level (table 1). In the 90s, N budgets for croplands focused on plot and farm level (e.g. Baijukya and De Steenhuijsen Piters 1998, Defoer et al 1998, Wortmann and Kaizzi 1998), with some extrapolations to regional or landscape level (e.g. Van der Pol and Traore 1993, Folmer et al 1998). These studies included livestock as consumer of crop residues and/or as a producer of manure for cropland. Some studies estimated N losses through erosion, leaching, and volatilization with field measurements (e.g. Baijukya and De Steenhuijsen Piters 1998), or with simulation models (e.g. Wortmann and Kaizzi 1998), concluding that soil mining through cropping must be compensated with fertilizers.

Partial N budgets estimated using information collected through farm surveys are common in the literature (e.g. Defoer et al 1998, Esilaba et al 2005). These studies do not include estimates of N losses, but are useful to compare farms and management practices. Livestock-related flows usually include on-farm and off-farm feeds, and manure production as reported by farmers (e.g. Wortmann and Kaizzi 1998, Elias et al 1998, Nkonya et al 2005), where losses are sometimes estimated using the transfer functions of Stoorvogel and Smaling (1990). Shepherd et al (1996) built a static model to estimate nutrient balances at farm level. The authors recognized that estimated nutrient losses were uncertain, and that the missing feedbacks between farm components limited the evaluation of the impact of alternative practices. Later on, Shepherd and Soule (1998) developed a simulation model to include biophysical and economic flows between farm components.

Mohamed Saleem (1998) discusses the potential contribution of livestock management to balance N in different livestock system. Some other studies quantified in more detail on crop–livestock interactions. For example, Dugue (1998) studied livestock-related nutrient transfers at community level in Northern Cameroon. The analyses included temporal characterization of manuring in cropland and grazing land, with records of livestock population, and intake of crop residues but no losses from manure management. Harris (1998) quantified N balances following Stoorvogel and Smaling (1990) measuring manure application rates and manure quality, and quantified excretion during grazing. Ramisch (2005) studied nutrient flows at household and community levels in Southern Mali using resource flow maps, which allowed householders to describe spatially management strategies.

The studies by Achard and Banoin (2003), Schlecht et al (2004, 2006), and of Manlay et al (2002, 2004) in West Africa showed that only 10–20% of the excreta can be used for cropping, manure which tends to concentrate in relatively small areas of cropland usually owned by the wealthiest communities members. This has also been shown in the work of Tittonell et al (2005, 2009b), and of Rufino et al (2011), where the potential for using manures to support crop production was explored. These studies concluded that livestock can play a role in increasing crop yields, but remains insufficient if manures are not combined with inorganic fertilizers and biological N₂ fixation. Mapping landscapes, livestock grazing itineraries and excretion behaviour in livestock systems is time consuming and expensive, but crucial to understand nutrient flows, determine hot-spots for emissions, and to improve nutrient management. Manlay et al (2002) and Schlecht et al (2004) showed that livestock management creates areas of high concentration of nutrients, which are not efficiently used to produce crops and might contribute largely to nitrogen losses.

Brouwer and Powell (1998) measured N leaching in micro plots following corralling in Niger. These measurements showed the heterogeneity of soil fertility and how this could be managed with livestock density of the corralling practices. Nyamangara et al (2003) measured in Zimbabwe N leaching losses for maize plots that increased largely when manure or fertilizers were applied. Turner et al (2005) showed that grazing management affects largely feed intake and excretion. Research in Kenyan savannas indicated that corraling increases largely soil organic matter, an effect that persists more than 30 years after corrals were abandoned (Augustine 2003). The large concentrations of N in surface soil resulted in 5–10 times larger N leaching rates in abandoned corrals than in adjacent soils. Losses of N through surface run-off are directly related to the management of the land. In Ethiopia, Girmay et al (2009) estimated losses larger than 30 kg of N ha⁻¹ yr⁻¹ for cultivated land, 8–10 kg N ha⁻¹ for grazing land, and less than 1 kg N ha⁻¹ for exclosures or plantations.

Few studies have quantified NH₃ and N₂O emissions in livestock systems. Murwira (1995) measured NH₃ losses from manure which did not exceed 4% of total N for a 30 day incubation period. Nzuma and Murwira (2000) found that addition of straw to faeces reduced ammonia losses by up to 85%, and by 50% when added to combined faeces and urine. Few studies have estimated N₂O emissions from manure or manured soils. In Zimbabwe, N₂O emissions from sandy soils ranged between 0.1 and 0.3 kg N ha⁻¹ depending on the organic fertilizer added, while N leaching ranged between 29 and 40 kg N ha⁻¹ (Chikowo et al 2004). Adding manure to maize resulted in N₂O emissions ranging from 0.06 to 0.4 kg N ha⁻¹ (Mapanda et al 2011). In Mali, experiments showed that applications of manure and urea depressed N₂O emissions and resulted in lower NH₃ concentrations in the soils than in the crops that received only urea. Manure applications alone had no effect on the N₂O emissions (Dick et al 2008). The onset of rains caused an increase (45–80%) in N₂O emissions, whereas tillage increased N₂O emissions by 34–86%.

Process and materials added determine the quantity and quality of the final compost and the crop response to the manure N (Delve et al 2001). Manure decomposition is affected by macro-organism's activity. In Niger, Esse et al (2001) estimated that termites were responsible for up to 75% manure loss within ten weeks of exposure onto the soil. While it is possible to collect nearly all of the excreta, in practice the efficiency of collection of excreta N is much lower. In a study in Kenya manure stored in a covered concrete storage, lost up to 40% of the N before composting (Lekasi et al 2001). Manure is commonly left in corrals for many weeks before collection, and losses of urinary-N and labile faeces N through leaching and volatilization are usually very large. During 150 days storage of manure in humid tropical central Uganda 20–50% of the initial N was lost (Rufino et al 2007). Experiments in Kenya showed that pit storage conserved 20% more carbon than heap storage during a 180 days period. However, heaps conserved more N and K than pits (Tittonell et al 2009a). Predotova et al (2010) in Niger showed that roofing manure heaps had no effect on N₂O emissions, and reduced NH₃ losses only during the dry season. Practices that favour C conservation in manures can be detrimental for N losses (Tittonell et al 2009a), and justify the measurement of the main pathways for nutrient losses for different systems. Measurements of N flows in fields, rangelands and livestock facilities are needed to assess nutrient use efficiency for different management practices.

Livestock populations at country levels are collected by FAO. This data is disaggregated using various procedures (Kruska et al 2003, Robinson et al 2011). The gridded livestock of the world (GLIS) (Wint and Robinson 2007) uses environmental variables to spatially distribute national data. The authors of GLIS recognized that the main sources of uncertainty are the variability in spatial and temporal resolution of the data used, and therefore generated recently an improved database using sub-national data (Robinson et al 2014). Household surveys provide livestock numbers at household level, so the uncertainty around these numbers is usually small. Scaling household livestock numbers to community or regional levels is problematic without a farm typology and measures of variability among types (Tittonell et al 2010). The work of Cecchi et al (2010) using livelihood data collected by non-governmental organizations to define livestock production systems is a great step towards making spatial data useful for research. More detail could be added by splitting households into groups including the use of natural resources, and connecting spatially-explicit livestock populations to the land using process-based modelling.

Nutrient budgets are snap-shots, and do not account for temporal variability driven by grazing management (Scoones and Toolmin 1998, Augustine 2003, Turner et al 2005, Schlecht et al 2006). Although livestock graze extensive rangelands, excreta concentrate in areas where livestock overnight, where feeds are more abundant or where feed quality is better. That modifies vegetation in the long term creating and maintaining emission hotspots in the landscape (Augustine et al 2003). Long-term monitoring is needed to capture peaks in N₂O emissions and leaching. Wolf et al (2010) showed that missing short-lived pulses of emissions during spring thaw in the Mongolian steppe leads to large underestimations of the emissions for those systems. Unfortunately, there are no such data describing spatial and temporal variability of N flows for African systems, even though we can assume that re-wetting events at the onset of the wet season drives emission pulses (Brümmer et al 2008). Hickman et al (2011) questions the validity of IPCC emission factors to quantify emissions from African agriculture arguing that there are very few studies that looked at cropping practices, and none estimated all N flows on-farm. The same applies for livestock systems: there are no studies that look at the full balance using systems types that allow scaling to continental level. N budgets for African systems have mostly considered cropping and livestock being independent and having hardly any interaction. The inaccuracy of the estimates is partly a result of that flawed assumption. A multilevel systems typology that includes livestock (from fields, to farm and landscapes) can be used to encompass spatial variability.

Uncertainty in N budgets is also associated with the inaccuracy of the land cover maps. N flows estimated at field level or for individual animals are scaled to farm, region and continent using land areas. Efforts to improve quality of the N measurements, their frequency, and spatial spread, can be impaired by large disagreement in land cover maps. The choice of land cover database has a large effect on N budgets as it introduces a systematic error that propagates in the estimations (Nol et al 2008). The sort of information needed can be collected using comprehensive field surveys. There are initiatives aiming at improving the quality of spatial data. Geo-wiki (www.geo-wiki.org) (Fritz et al 2009) aims at using crowdsourcing to improve global land cover. Internet users are asked to recognize land cover features and report them online. The input is saved in a database with photographs to create an improved global land cover map.

In most large scale N budgets for Africa there is no interaction between livestock and crops, or these are extremely simplified. Crops are used to feed livestock and manures are applied to the land. However, both flows appear not to be constrained by the existence and availability of land or by the existence of livestock and their location. Amounts of crop residues fed to livestock are largely unknown, in spite of the wealth of research on crop residues and livestock (e.g. Renard 1997, Valbuena et al 2012). Besides, there has been much research on the use of organic residues as soil inputs, although with no good estimates of regional scope for manure utilization. Some research showed the importance of livestock for cropping, but not enough research quantified the importance of biomass production for livestock production (Rufino et al 2011, Herrero et al 2013).

Feed intake, livestock excretion and manure use are the main flows connecting crop and livestock systems. These flows can be estimated using calibrated models, given there is input data. Herrero et al (2013) developed a global livestock systems characterization that uses a nested system definition to include feeding and manure management using the livestock systems of Robinson et al (2011). The nested-systems definition allows using ranges for difficult-to-estimate flows, reflecting results obtained experimentally. The livestock systems characterization includes livestock diets that have been described for ten livestock systems: Livestock Grazing Temperate/tropical highlands (LGT), Livestock Grazing Humid-Subhumid Tropics and Subtropics (LGH), Livestock Grazing Arid/semiarid (LGA), Mixed Farming Temperate /tropical highlands (MRT), Mixed Farming Humid-Subhumid Tropics and Subtropics (MRH), Mixed Farming Arid/semiarid (MRA), and two landless systems (smallholder and industrialized). For each of these systems, the diets of cattle (dairy and non-dairy), small ruminants, poultry and pigs are described. For each combination of systems and livestock type, manure management and associated N losses are described (supplementary material, tables 1 and 2, available at stacks.iop.org/ERL/9/105008/mmedia).

The diets described for different systems in Herrero et al (2013), and livestock types and a process-based model were used to estimate excretion rates. Pathways for N losses in livestock systems differ from those of only crops systems. Herrero et al (2013) followed IPCC (2006) to estimate N₂O emissions, though emission factors are mainly based on measurements in temperate regions. One of the most uncertain parameters in the regional and continental N budgets is the fraction of manure managed. This varies largely across and within systems, with systems recycling most manure excreted and others almost nothing. Our assessment including the most recent livestock population data of Robinson et al (2014) shows the geographic spread of available manure and its uncertainty (figure 1). Comprehensive surveys could be used to estimate proportion of manure used in croplands, or for other purposes by recording 'time spent' grazing in different land units, which can be cross-checked and geo-referenced. This data collected for all the range of systems and complemented with experimental data to estimate full losses for selected systems will improve the accuracy and usefulness of N budgets.

Zoom In Zoom Out Reset image size