Introduction

Soil fertility is a key production factor for most farmers in Sub-Sahara Africa. Low intrinsic soil fertility, limited replenishment of removed nutrients and high erosion rates in mountainous areas cause soil fertility decline to become a major threat to current and future food production (Gachimbi et al. 2005; AGRA 2014). Many farming systems still rely on natural release of nutrients from the soil through mineralization and de Jager et al. (2001) estimated that approximately 60–80 % of the farm income was obtained at the expense of soil nutrient depletion, i.e. through un-replenished nutrient uptake in marketable crops. Soil nutrient mining cannot be visualized easily and, therefore, indicators are often used to facilitate discussions on soil fertility management. Soil nutrient balances are a commonly used indicator and are defined as the difference between the sum of nutrient input flows and the sum of nutrient output flows within a specific system (field, farm, nation) over a certain period (season or year). Soil nutrient balances reflect the net change in soil fertility and indicate trends in time, but do not necessarily determine the current state of soil fertility. To determine the severity of the depletion rate, nutrient balances can be related to the soil nutrient stocks, where typically a depletion rate of more than 2 % per year of the soil nutrient stock is considered unsustainable (Elias 2002).

Ethiopia is considered as one of the most vulnerable countries in sub-Sahara Africa (SSA) with regard to soil fertility depletion because of its mountainous topography and intensive farming systems based on small cereals. Indeed, national averages of nutrient balances were estimated at −41 kg N, −6 kg P and −26 kg K per ha per year, which is among the highest nutrient depletion rates for sub-Saharan Africa (Stoorvogel and Smaling 1993). Yet, soil nutrient balances can differ considerably between different crops, farming systems and agro-ecological zones (Sommer et al. 2014). Restoring soil fertility is, therefore, a key priority of the Government of Ethiopia (GoE). The Agricultural Growth Plan (AGP) is the flagship programme of the Ethiopian government designed to achieve the Growth and Transformation Plan (GTP) which was running from 2011 to 2015 and is currently extended with a second phase until 2019. One of the objectives of the GTP is to double agricultural production by scaling up of existing best practices and introducing new promising innovations (World Bank 2010). It is increasingly realized that it could be difficult to achieve GTP/AGP objectives with the current levels of nutrient mining. To this effect, more balanced and site-specific fertilizer recommendations and improved management practices have been proposed and are being implemented by different projects including Ethiopian Soil Information System (EthioSIS) and Capacity building for Scaling up of Evidence-based best Practices for increased Production in Ethiopia (CASCAPE).

In this paper, soil nutrient balances and nutrient flows were analysed at field level in six areas of Ethiopia. Soil nutrient depletion rates were related to soil nutrient stocks to evaluate the longer term sustainability of the agro-ecosystem.

Materials and methods

Selected sites

The results described in this paper are part of the results of the broader study of the CASCAPE project (www.cascape.info). This project operates in 30 AGP woredas (districts) of five main agricultural production areas. The woredas represent different farming systems, agro-ecological zones and soil types. Table 1 summarizes the general features of the selected woredas and Fig. 1 presents the geographical distribution of the woredas included in this study. The woredas are located in the eastern part of Tigray National Regional State, western part of Amhara National Regional State and some parts of Southern Nations Nationalities and Peoples Regional State. Given the spread of the Oromiya National Regional State, three sites were selected in its central, eastern and western highlands. In each area local teams collected and analysed data from selected woredas. Woredas managed by one team were termed clusters and covered 4 or 5 woredas. In each cluster, soil nutrient balances were determined for approximately 60 farms, comprising more than 500 fields for 3 years (2012–2014).

Table 1 Balance entries for soil nutrient balances at field level as used in the MonQI toolbox
Full size table
Fig. 1
figure 1

Locations of the intervention woredas

Full size image

Soil characterization

In each kebele (a kebele being the lowest administrative unit) of the woredas, soil profile studies were performed. A total of 706 soil samples were taken for soil analysis to the Soil Fertility Laboratory of the Federal Water Works Design and Supervision Enterprise of Ethiopia. Amongst others, samples were analyzed for pH-H2O, in 1:2.5 soil:water solution, organic carbon content (%), total N content (%), available P content (mg/kg), cation exchange capacity [CEC cmolc (+)/kg] and exchangeable K (cmolc/kg). The analyses were carried out following the soil laboratory procedures as outlined in Van Reeuwijk (2002). Soil organic carbon was analyzed using the Walkley and Black method (Nelson and Sommers 1996) and total nitrogen according to the Macro-Kjeldahal method that involves digestion of the sample and a wet-oxidation procedure (Bremner and Mulvaney 1982). Available phosphorus content was determined using the Olsen sodium bicarbonate extraction solution (pH 8.5) method as described by Van Reeuwijk (2002) whereby the amount of phosphate was determined by spectrophotometer at 882 nm. Exchangeable basic cations and the cation exchange capacity (CEC) of the soils were determined by using the 1 M ammonium acetate (pH 7) method according to the percolation tube procedure (Van Reeuwijk 2002) whereby exchangeable K was determined by flame photometer. The soil data were used in the background database of the MonQI software to estimate hard to quantify flows through pedo-transfer functions (Lesschen et al. 2007) and to calculate soil nutrient depletion rates, defined as the nutrient balance divided by the relevant soil nutrient pool.

Nutrient balances

Soil nutrient balances were calculated as the net differences of nutrient inputs and outputs flows taking the soil as “black box” (Smaling and Stoorvogel 1990). Input flows include farmer managed flows like applications of fertilizer and seeds and natural processes like atmospheric deposition and biological N fixation. Output flows include removal of nutrients from soil by harvests, leaching, volatilization and soil erosion by water. The different balance entries were assessed through farm surveys (farmer managed nutrient flows) and transfer functions (naturally occurring nutrient flows) as shown in Table 2. Nutrient balances were estimated using the monitoring for quality improvement (MonQI) toolbox. The MonQI toolbox is a method for monitoring management and performance of small scale farming systems world-wide. MonQI has been used in the Tropics for over 30 years (www.monqi.org; Smaling et al. 2013; Vlaming et al. 2012). Basically, farmers provide information about their farm management and farm activities using standardized questionnaires. The questionnaire consists of different sections related to the main farm activities (livestock activities, crop activities, household activities, redistribution activities and storage activities). Data analysis is automated and combined with background data on e.g. nutrient contents of products, soil data and conversion factors from farmer used units (e.g. headloads) to SI units.

Table 2 Characterization of the farming systems of the different woreda
Full size table

The software produces a wide range of farm management and farm performance indicators like NPK balances and gross margins at plot, activity, and farm level. Data collection started in 2012 for all woredas except those in the Addis Ababa cluster (Table 3) who joined the project only after 2012. Table 3 gives an overview of the number of surveyed farms per cluster and per year. Participating farmers were beneficiaries of the CASCAPE project and were selected on the basis of representativeness for their region. The soil data were used in the background database of the MonQI software to estimate hard to quantify flows through pedo-transfer functions (Lesschen et al. 2007) and to calculate soil nutrient depletion rates, defined as the nutrient balance divided by the relevant soil nutrient pool. Hard to quantify flows included atmospheric deposition, biological N fixation, leaching, gaseous losses and erosion. Hard to quantify flows, as the term suggests, are extremely difficult to quantify, both by physical measurements as by simulation modelling. This is especially true for soil erosion which is the results of many, highly variable factors (Hessel et al. 2006). In MonQI soil erosion is estimated using the modified USLE equation which was considered as the best compromise between accuracy and data requirements, though the results should be interpreted with care (Sonneveld et al. 2011). Briefly, soil loss by erosion was estimated by multiplying site specific parameters on rainfall, soil erodibility, slope, slope length, protection by crops and protection by management factors (i.e. anti-erosion measures). The estimated soil loss was subsequently corrected for enrichment of nutrients in the topsoil and soil formation (i.e. weathering of nutrients in newly accessible soil layers).

Table 3 Number of farms surveyed using the MonQI toolbox per cluster per year
Full size table

Data analysis

The current state of the soil fertility was assessed using soil fertility thresholds of 0.15 %, 10 ppm and 1.0 cmol kg−1 for N, P and K respectively (Hazelton and Murphy 2007). Soil nutrient depletion rates were calculated by dividing soil nutrient balances by relevant woreda average soil nutrient pools, using soil properties of the upper 30 cm. We used average woreda soil data because the soil sampling was part of a different project and selected sites were in close proximity, but did not exactly coincide with MonQI survey sites.

Data was subject to an extensive data cleaning process. Data was checked for consistency, completeness and correctness. Consistency checking included systematic reviewing of similarities of sources and destinations of flows. E.g. maize harvest could only originate from a maize plot or a plot with maize intercropping, but not from plots with a different crop. Completeness checking was the systematic reviewing of input and output flows. E.g. plots with only output data (harvests) were double checked when inputs were missing. In some cases, especially for leguminous crops used for home consumption, this turned out to be correct, and no inputs were provided. Outlier checking was done for income and nutrient balances. Outliers were defined as fields with N balances smaller than −500 kg N/ha and higher than 500 kg N/ha. Typically these outliers were found on small fields (less than 100 m2) where small flaws in original data were exacerbated when data was normalized to hectares. These outliers were often due to difficulties of farmers to estimate inputs (especially organic inputs), land sizes and, sometimes, typing errors. Several data cleaning sessions were organized and suspicious data were corrected or removed from the dataset.

Results

Soil data

Table 4 presents the chemical soil fertility data per woreda. In general, soil fertility was low to medium for most sites; soil N contents were below the thresholds for sufficient soil fertility in 68 % of the sites. For P and K these numbers were respectively 11 % and 89 %. In 2 woredas (out of 28, namely Girar Jarso and Munessa) all macro nutrients were below the threshold values. For P, relatively high values were found for the woredas in the Mekelle cluster. For N and K, no regional differences were observed. Variabilities were high for soil P and soil K contents with standard errors of 0.43 and 0.51, respectively. For soil N contents the standard error was 0.21.

Table 4 Average soil data of upper 30 cm and standard deviations (n = 4–10), nutrient balances and depletion rates per woreda
Full size table

Nutrient balances

Balance entries per woreda are shown in Table 5 for N. For P and K comparable patterns were found, but with less balance entries (because biological fixation, atmospheric deposition and gaseous losses mainly refer to N). The differences between the inputs and the outputs are the soil nutrient balances, which are shown in Fig. 2. Soil nutrient balances varied considerably between woredas and on average equalled −23 ± 73, 9 ± 29 and −7 ± 64 kg ha−1 year−1 for N, P and K, respectively. Whereas some sites showed negative P balances, most sites showed P-accumulation (Fig. 2). Average soil N depletion rates were 0.2 % per year when calculated as percentage of the total soil N stock. Typically, only 5 % of the total soil N pool is available for crop uptake (Havlin et al. 2014). When N depletion was expressed as percentage of available N, depletion rates were on average 4.2 % which was considered as severe. For K, average depletion rate was 0.1 % year−1, with higher values in South Achefer and Ofla (Fig. 3). On average, soil nutrient balances decreased with 20, 3 and 12 kg ha−1 year−1 for N, P and K, respectively during the monitoring time (2012–2014) (Fig. 4). These values equal accelerated soil nutrient depletion of 33, 14 and 41 % for N, P and K, respectively each year. The soil nutrient balances are the net result of various balance entries as listed in Table 1. Mineral inputs were the main nutrient inputs and predominantly consisted of mineral fertilizers. Erosion, leaching and harvests were main nutrient output processes (Fig. 5). Figure 6 shows the inputs of N per woreda, which predominantly consisted of urea. For P more than 95 % of the total inputs originated from DAP and for K hardly any inputs were given. Soil erosion was a key loss pathway of nutrients (Fig. 5). Estimated erosion rates were most often moderate or moderate to high according to the classification of Beskow et al. (2009), but especially in the Addis Abeba cluster erosion rates could be very high (Fig. 7).

Table 5 Input and outputs of N per balance entry per woreda (kg/ha/year) and standard deviations (n = 51–570)
Full size table
Fig. 2
figure 2

Average nutrient balances (kg ha−1 year−1) per woreda for N (top), P (middle) and K (bottom). Error bars show standard deviations

Full size image
Fig. 3
figure 3

Nitrogen depletion rates per woreda (% year−1)

Full size image
Fig. 4
figure 4

Average changes in soil N (above), P (middle) and K (below) balances (kg ha−1 year−1) for main crops between 2012 and 2014

Full size image
Fig. 5
figure 5

Average soil N balance entries per cluster (kg ha−1 year−1)

Full size image
Fig. 6
figure 6

Sources of N inputs per woreda

Full size image
Fig. 7
figure 7

Relative share of land within erosion classes: slight (<2.5 tonne/year), slight/moderate (2.5–5.0 tonne/year), moderate (5.0–10 tonne/year), moderate/high (10–15 tonne/year), high (15–25 tonne/year) and very high (>25 tonne/year)

Full size image

Standard deviations of the nutrient balance entries were in the order of 100–300 % (Fig. 2; Table 5) and were due to the high diversity of agro-ecological conditions and farming systems. For reasons of readability of the graph, no standard errors were added in Fig. 5. Soil nutrient balances found in this study compared reasonably well to results found by others as shown in Table 6.

Table 6 Nutrient balances of CASCAPE and previously reported values for Ethiopia (kg ha−1)
Full size table

Nutrient balances and crop yields

Figure 4 shows soil N balances and yield levels for the major crops, namely maize, wheat, barley and teff. On average, soil N balances were negatively correlated to yield level, indicating that for every kg of yield increment, soil N depletion increased, i.e. became more negative (Fig. 8). On average, i.e. averaged for all crops, years and sites, yield induced depletion was about 18 g N per kg of yield. For P and K, no relations between yields and soil nutrient balances were found.

Fig. 8
figure 8

Relation between yields and soil N balances for barley (upper left), wheat (upper right), maize (below left) and teff (below right). Each symbol represents a unique field-season combination. Solid line shows linear regression

Full size image

Discussion

Soil fertility and agricultural development in Ethiopia

Ethiopia’s agriculture is on the rise. Projects like CASCAPE demonstrate that yields can be doubled or tripled by the combination of good seeds, sound nutrient supply and proper management practices. Indeed production levels in Ethiopia have been rising for the last decade (Simons et al. 2014) mainly because of an effective extension mechanism in combination with market reforms. Yet, on average fertilizer strategies, to compensate for the increased withdrawal of nutrient in the harvested products and crop residue removals, did keep pace with the yield increments (e.g. Fig. 8). As a result, soils of Ethiopia are now, more than ever, under threat for soil nutrient depletion. Soil nutrient depletion is often regarded as a precursor of soil fertility loss which in its turn can eventually lead to abandonment of land (Sanchez 2002). This is confirmed by the current study which shows that nutrient depletion has increased over time as well as with yield levels (Fig. 8). The Ethiopian government is aware of this and responded with massive soil sampling campaigns to detect deficient nutrients though the EthioSIS project. This effort culminated in the establishment of fertilizer blending plants in the country to produce specific fertilizer blends targeted at the observed deficient nutrients. Notwithstanding the success of the EthioSIS project, this study showed that soil erosion accounted for about 50 % of the nutrient losses (Fig. 5). Hence, it is possible to argue that an effective soil restoration campaign should include soil and water conservation measures to combat erosion, improve on-farm nutrient use efficiency by promoting composting and mulching (and reduce competing claims for crop residues) in combination with targeted fertilizer recommendations that at least compensate for nutrient removal. Without such measures, soil nutrient depletion will continue and eventually soils will become exhausted.

Interpretation of soil nutrient balances and erosion rates

The nutrient depletion rates found in this study are commonly interpreted as ‘mild’. However, even ‘mild’ depletion may become serious when prolonged over time. The classification of ‘mild’ is based on the comparison with previous studies (Table 6). Moreover, the results of this study showed that nutrient depletion increased over time and worsened with increasing crop yields. Consequently, the results are alarming and call for structural improvement of soil nutrient management practices in the highlands of Ethiopia. In general, previous studies performed at national level showed more negative soil nutrient balances compared to the results found in this study. To some extend these differences can be related to differences in spatial scales, i.e. system boundaries. Nutrient balances increase with increasing spatial scales, because losses remain within system boundaries when assessed at higher spatial scale. For instance, erosion is a major loss item at field level, but not at regional scale when the sediments remain within system boundaries.

Soil erosion equalled 8.3 ± 16.1 tonne ha−1 year−1. This average can be interpreted as ‘moderate’ according to the classification of Beskow et al. (2009). Figure 7 shows the distribution of the estimated erosion rates over the different classes for each woreda. The vast majority (59 %) of the woreda average erosion rates was classified as ‘moderate’ with an estimated erosion rate between 5 and 10 tonnes per hectare per year. However, variability was high and severe erosion rates of more than 25 tonnes year−1 occurred, especially in the woredas within the Addis Ababa cluster (Fig. 7).

Data variability

Variabilities within the dataset were high, with standard errors often exceeding 100 % (Fig. 2). Partly, this is caused by the nature of topic; diverse agro-ecological conditions, socio-economic conditions and personalities of farmers cause high variabilities in farm management, including nutrient management. High variability between soil nutrient balances within a specific region was confirmed by Elias (2002) and Aticho et al. (2011) who found considerably variabilities in soil nutrient balances across socio-economic groups and agro-ecological zones in Southern and Western Ethiopia, respectively. For another part, the variability is caused by biases in yield estimations by local enumerators and farmers. In this study, flows of nutrient inputs and outputs were estimated by the farmers and local enumerators. Although this approach may have contributed to biases in our dataset, it also gives a genuine reflection of farmers’ appreciation of nutrient flows. Moreover, the high number of data records (6944 records in total, each record representing a unique plot-season combination) compensates for the relatively high variability.

We tried to relate soil nutrient balances to farm and farmer characteristics like size of the farm, age and education level of the household head etc., but could not find convincing relations. Contrarily, Haileslassie et al. (2007) found that environmental conditions, farming systems (e.g. crop selection), access to resources (e.g. land, livestock and fertilizer) and smallholders’ source of off farm income determined the magnitude of nutrient fluxes.

Context specific interventions

This study shows that, at present, soil fertility decline in the highlands of Ethiopia is continuing and, on average, is worsening. Continuing soil degradation is a serious threat to future food security. Bindraban et al. (2012) emphasised that a comprehensive approach is needed to assess both extent and impact of soil degradation interlinking various scales. To effectively combat soil nutrient depletion, various spatial scales should interact to link depletion (rural) areas with accumulation (urban) areas. Notably, depletion and accumulation of nutrients (the polarization of the nutrient system) expresses at several spatial scales: organic wastes pile up at the homestead fields, but also at the city borders and at the global level where developed countries experience the environmental hazards of superfluous nutrients through eutrophication while developing countries are struggling to maintain and, preferably, increase food production. In fact, these are two sides of the same coin caused by unbalanced nutrient management at various spatial scales. According to Dietz et al. (2012), urbanization goes hand in hand with globalization and population growth. Although current urbanization level in Africa is only 36 %, this number is rapidly increasing because of population growth and rural–urban migration. Many Africans combine a rural and an urban existence, others move from rural areas to small towns and then to Africa’s booming cities. These trends may have serious consequences for nutrient polarization and may lead to increased soil nutrient depletion in the production sites when nutrient removal is not sufficiently addressed. Recycling of organic waste streams is a potential way out, but requires willingness, legislation and investment, all being limited in many developing countries. At the same time, Nigussie et al. (2015) found that less than about 10 % of manure and crop residues available to smallholder farmers in Ethiopia was applied to soils, indicating that re-use efficiency can be largely improved at the farm level.

Erosion was a key determinant of the negative nutrient balances emphasizing the need for improved soil and water conservation measures at farm and at watershed level. Current efforts on increased inputs like blended mineral fertilizers are a step in the good direction, but to really halt and reverse soil fertility decline, organic fertilizer application and soil and water conservation (SWC) measures should be an integral part of the intervention strategy.

Halting the current trend of soil nutrient depletion will remain a major challenge and needs context specific solutions. Farmers (and donors) look for immediate, or at least short term effects. Yirga and Hassan (2010) found that smallholder farmers tend to discount the future at higher rates leading to overexploitation of soil nutrients, whereas soil conservation or restoration efforts become effective only at the medium or long term. Tenure security is a crucial element to overcome the costs of time dimensions related to soil management.

Conclusions

Nutrient balances in the highlands of Ethiopia, typically the high potential areas for agricultural production are currently exposed to severe nutrient depletion. Yet, agricultural production in this area is increasing which benefits farmers livelihoods and contributes to food security of the country as a whole, but at the expense of the natural resource base. To maintain the current increase in food production, the trade-offs towards soil nutrient depletion need to be counterbalanced by improved soil management. Erosion is a key determinant of the negative nutrient balances emphasizing the need for improved soil and water conservation measures at farm level and at catchment level. It was, therefore, recommended that an effective strategy to increase the productive capacity of land in the studied woredas include targeted soil and water conservation measures and improved integration of organic matter management with mineral fertilizer application.