Increasing barley production in the face of declining soil fertility on smallholder farms is a critical challenge in Ethiopia. The objectives of this study were to (1) determine the optimum rates of nitrogen (N), phosphorus (P), potassium (K), and sulfur (S) fertilizers under a balanced fertilization of other macro- and micronutrients (zinc and boron) and (2) quantify the agronomic efficiency of N and P on Cambisols and Vertisols. The trials were conducted for three consecutive years (2014–2016) on farmers’ fields on 83 sites under rain-fed conditions. On each farm, six rates of N, P, K, and S with a balanced application of the other nutrients were laid out separately in a randomized block design with three replicates. Balanced fertilization resulted in yield increments of 26–59% on Cambisols and 18–74% on Vertisols, as compared with the control (0 N). Application of P increased grain yield from 11% to 30% and from 7% to 14% due to the interaction of P by soil type as compared with the control. The application of K and S, however, had no significant effect on barley yields on either soil type. The highest agronomic efficiency of N (AEN) was obtained from 46 kg N ha−1, with AEN increasing by 117% on Cambisols and 33% on Vertisols, as compared with the highest N rate of 230 kg N ha−1. The maximum agronomic efficiency of P (AEP) was recorded with 10 kg P ha−1 on Vertisols, with AEP increasing by 73% on Cambisols and 452% on Vertisols, as compared with the application of 50 kg P ha−1. With balanced application of other nutrients, use of 230 kg N and 20 kg P ha−1 was recommended for tepid sub-moist mid-highlands and Vertisols. For tepid sub-humid mid-highlands, application of 46 kg N and 20 kg P ha−1 with balanced application of other nutrients was recommended. For Cambisols, application of 230 kg N and 40 kg P ha−1 was recommended with balanced application of the other nutrients. Similarly, application of 230 kg N, 40 kg P, and 10 kg S ha−1 was tentatively recommended for cool sub-most mid highlands. However, response to K was inconsistent and therefore a firm recommendation could not be proposed for the study sites. Further research needs to be conducted on more soil types and locations.

Optimisation of fertiliser use and site-specific nutrient management are increasingly becoming critical because of the growing need to balance agricultural productivity with the growing demand for food and environmental concerns. Trials to determine responses of crops to fertilisers have been widely conducted in sub-Saharan Africa (SSA) with increasing emphasis on the development of economically optimum rates (EORs). Computation of EORs depends on accurate estimation of both the optimum nutrient rate and the agronomic maximum yield response; however, estimation of nutrient-response parameters and EORs is beset by a number of problems. Therefore, the objectives of this paper were to (1) point out common problems in the development and use of nutrient dose-response models and (2) provide corrective measures to facilitate future trial design and data analysis. This review outlines the underlying assumptions, strengths and limitations of the various response functions in order to facilitate informed choices by practitioners. Using specific examples, it also shows that (1) the commonly used trial designs do not allow examination of interactions between two or more nutrients and (2) trial designs with ≤5 nutrient levels and wide spacing between the levels result in large uncertainty in dose-response parameters. The key recommendations emerging from the review are as follows: (1) factorial designs and response surface models should be used more widely to address interactions between nutrients; (2) a minimum of six carefully spaced nutrient levels should be used to correctly estimate dose-response parameters; and (3) when locating field trials, Reference Soil Groups and cropping history should be carefully considered to produce site-specific EORs.

Although soybean is emerging as an important commercial crop in sub-Saharan Africa (SSA), its productivity on smallholder farms is very low. Soybean requires application of phosphorus (P) fertilizer and inoculation with the right rhizobium strains to achieve optimum biological nitrogen fixation and higher yields. However, subsistence farmers in SSA rarely invest in P fertilizers and inoculants due to lack of knowledge of their use and benefits. Most of the early reports on soybean in SSA have been based on work on research stations; hence, information is lacking on the profitability of fertilizer and inoculant use on smallholder farms in SSA. The main hypothesis of the present study was that the combined application of P and inoculants significantly reduces yield risks and increases P use efficiency and profitability compared with P fertilizer alone under smallholder farm conditions. We analyzed a data set of over 2,800 observations from on-farm demonstrations across Ghana, Kenya, Malawi, Rwanda, Tanzania, Uganda, and Zambia. Soybean yields, the partial factor productivity of P (PFPP), agronomic efficiency of P (AEP), and the value cost ratio (VCR) were significantly improved by the combined application of P fertilizer with inoculants than with P fertilizer alone. Combining P and inoculants increased yields over P alone by 17.3% in Kenya, 21.4% in Zambia, 25.7% in Ghana, 56.4% in Tanzania, and 57.1% in Malawi. However, soil organic matter was an important determinant of yield response and P use efficiency. The VCR increased linearly with increasing AEP in P + inoculant (R2 = 0.829) and less so with P fertilizer alone (R2 = 0.672). Net present values were positive in all countries, indicating that investments in P fertilizer and inoculants will generate profits over time. In order to increase uptake of fertilizers and inoculants among subsistence farmers and make soybean production more profitable, appropriate policies and market incentives need to be created.

Meeting the burgeoning global demand for both food and energy requires substantial yield increases through the efficient use of inputs like fertilizers. Prompted by the result of a soil survey expedition in the late 1950s, which signaled a widespread deficiency of nitrogen (N) and phosphorus (P), plant nutrition research in Ethiopia began in the 1960s, focusing on the response of prioritized cereals – tef (Eragrostis tef), wheat (Triticum aestivum), and maize (Zea mays) – to the application of N and P fertilizers. Nationwide on-farm trials conducted in the early 1970s led to a blanket recommendation of 64 kg N ha−1 and 20 kg P ha−1, irrespective of the crop and soil types, which were applied in the form of di-ammonium phosphate (18-46-0) and urea (46-0-0), respectively. Research conducted in the 1980s across agro-ecological and edaphic spectrum recommended 30–138 kg N ha−1 and 0–50 kg P ha−1, respectively. However, studies show that only 30–40% of the smallholder farmers use fertilizers at a rate less than recommended (on average at 37–40 kg ha−1). This rate reflects limited supply, high prices, and the low and declining crop response to fertilizers. As a result, cereal yields increased only 10% despite a fivefold increase in fertilizer application since the 1980s. Owing to the limited and declining crop response and the increased price of fertilizer in the 1990s, research on the integrated application of inorganic and organic sources of fertilizers was initiated. Although the integrated use resulted in increased yield and better economic benefits, it was not mainstreamed into the national agricultural extension system. The soil survey expedition that began in 2011 culminated in the mapping of the soil nutrient status using literature-based critical limits. The maps have persistently revealed the deficiency of N, P, potassium, sulfur, zinc, and boron across the surveyed areas. Despite the above efforts, the data sets generated through the soil surveys conducted at different times during the last half-century and the agronomic research during the same period have never been fully exploited. It is believed that the recent development in data mining and machine-learning approaches creates the opportunities to use the data sets in conjunction with other covariates in order to generate evidence that helps to make better decisions both at strategic and operational levels. The development of decision support tools based on such large datasets and analytical capacity is believed to facilitate better-informed decisions that lead to increased resource use efficiency and sustainability.

The potential yield of improved maize varieties usually cannot be fully realised mainly due to inappropriate soil nutrient management practices in most parts of Ethiopia. Site-specific fertiliser recommendations are rarely used in the farming systems of Ethiopia. There is also a lack of data to develop or validate decision support tools for targeting specific crop production. A study was conducted for three consecutive rainy seasons (2016–2018) in the maize belt of the north-western parts of the Amhara National Regional State of Ethiopia. The objectives were to obtain the maximum achievable yield potential of maize, determine the most yield-limiting nutrients and create a database of maize responses to applied nutrients so that decision support tools could be developed for the study areas. Treatments were individual nutrients (nitrogen (N), phosphorus (P) and potassium (K)) and combinations of the three. In some treatments, NPK was also combined with sulphur, zinc, lime and compost. Two hybrid maize varieties (BH-540 and BH-660) adaptable to the study areas were used. BH-540 was used for the Mecha district, while BH-660 was used for the south Achefer, Jabitahnan–Burrie–Womberma districts. Maize yield increased by more than 50% due to fertiliser applications compared to without fertiliser. The study showed that the possibility of increasing maize productivity to more than 12 t ha-1 for the study sites. The most yield-limiting nutrient in the study sites was N, followed by P; K was not a yield limiting. Without N the yield of both varieties was non-significant from the control (without added nutrients). Maize grain yield did not respond to application of lime, compost, zinc and sulphur. The result also showed very high variability across sites, indicating that it is important for policymakers, farmers and investors to consider site-specific fertiliser recommendations. Finally, a database containing intensive plant response to NPK for maize was generated and could be used as input in site-specific decision support tools development.

Increasing yields in farmer fields is a priority to address increasing food demands. The study was conducted within four wheat-growing areas in Ethiopia: Debre Birhan, Hosaina, Sinana and Maychew. The objectives were to identify (1) best-bet soil fertility management options based on agronomic performance and economic evaluation and (2) key yield-reducing factors in farmer fields based on an agronomic survey among 55 participating farmers. Two types of on-farm experiments were conducted: researcher-managed trials that tested combinations of nutrients, including micronutrients, organic resources or both over two cropping seasons and farmer-managed trials comparing ‘improved practice’ against ‘farmer’s practice’. Fertilizer treatment affected wheat productivity in Debre Birhan (p < 0.01), a site limited in sulphur. Here, full NPK increased yields over the control (p < 0.05), whereas a combination of NPK and manure was better than the application of manure as the only source of added nutrients (p < 0.05). Applying half the recommended NPK with micronutrients and manure achieved similar yields as the full fertilizer treatment. In Hosaina, treatment had no significant effect on wheat productivity, although a combination of NPK and zinc resulted in an additional 26–57% yield relative to the other treatments. In Maychew, a significant treatment effect (p < 0.05) was observed. Here, the treatment with lower rates of nitrogen and phosphorous had lower yields than the full NPK treatment. A significant effect of plant densities on on-farm productivity was also observed. We conclude that although nutrient management including use of micronutrients is important in specific cases, investments to optimize plant densities have a huge potential to increase food productivity.

A number of studies across sub-Saharan Africa have recently reported poor crop responses and low agronomic use efficiencies of applied nitrogen (AEN), phosphorus (AEP) and potassium (AEK). However, the conditions under which non-responsiveness occurs, its underlying causes and its probability of occurrence on different soil types are not well understood. Using data from 542 sites and 14 soil types in 23 African countries, we provide novel insights into the linkage between lack of response to applied N, P and K, the mineralogy of soils and their resilience to erosion. We estimated mean responses as well as the probabilities (ϕ) of no response in terms of response ratio (RR), yield gain (YG) and agronomic efficiency. Here we defined ‘no response’ as zero agronomic response to fertilizer inputs in a given site and year indexed by either RR ≤ 1, AEN ≤ 0, AEP ≤ 0 or AEK ≤ 0. The highest risks of no response were recorded on the iron-rich Plinthosols (ϕ = 0.26) followed by the aluminium-rich Alisols (ϕ = 0.16) and the erosion-prone Lixisols (ϕ = 0.16) and Leptosols (ϕ = 0.13). In terms of yield gains, the highest risk of low response (i.e., YG ≤ 0.5) was recorded on Alisols (ϕ = 0.47) and the lowest on Fluvisols (ϕ = 0.05). Cambisols, Fluvisols, Luvisols and Nitisols were deemed highly responsive to NPK fertilizer. The risks of no response were significantly higher on soils derived from siliceous than mafic parent materials, soil types with low resilience to erosion, soils with low-activity clays and high P fixation capacity. It is concluded that maize grain yields can exceed 3 t ha-1 with high probability (ϕ > 0.80) on Andosols, Nitisols and Vertisols, but with very low probability (ϕ < 0.30) on Alisols and Arenosols. It is also concluded that across soil types and agroecological zones, the risk of no response is up to two times more on farmers’ fields than on research stations. Here, we discuss the implications of these finding for the design and location of future agronomic trials. We also provide insights to guide the targeting of fertilizer subsidies where nutrients can be more efficiently used.

The study at hand measures the value of improving data governance and access in the Supporting Soil Health Interventions (SSHI) project in Ethiopia. We applied two separate but interlinked models, one qualitative and one quantitative, to create a new framework enhancing the traditional cost–benefit analysis. The qualitative analysis provided novel insights into the specific types of value and the mechanisms through which they are generated. These results underpinned the development of an innovative framework to measure this perceived value quantitatively. By combining the quantitative and qualitative framework, the study demonstrated that it is possible to generate plausible and credible quantitative estimates of both costs and benefits of data governance and access. While acknowledging that the estimates are only illustrative, the case study results suggested on a direct cost measure, at a particular point in time, the SSHI data governance activities yielded a negative return. However, indirect social and public benefits are rarely quantified, but this paper shows that relatively few “indirect” benefits (current but unmeasured, or measurable but in the future) are necessary to reverse that view, at least from the point of the economy more generally.

The use of chemical fertilizers is among the main innovations brought by the 1960s Green Revolution. In Ethiopia, fertilizer application during the last four decades has led to significant yield gains, yet yield remains below its potential across much of the country. One of the main challenges responsible for low yield response to fertilizer application has been the use of ‘blanket’ recommendations, whereby no tailoring of fertilizer amount and frequency is done based on soil requirements. As a result, the amount of fertilizer applied ranges widely, and can be either sub- or supra-optimal. There is thus an increasing need for site-specific fertilizer recommendations which take into account site characteristics such as climate variables (temperature, rainfall, and solar radiation); soil factors (soil organic carbon, moisture, pH, texture, cation exchange capacity, and level of macro- and micronutrients); and topographic position indices. This article reports on a data-mining approach we developed on a large dataset of 6585 wheat (Triticum aestivum) field trials. The dataset includes detailed, site-specific biophysical variables to create nutrient response functions that can guide optimal site-specific fertilizer application. The approach used a machine-learning model (random forest) to capture the relationship between nutrients – nitrogen (N), phosphorous (P), potassium (K), and sulfur (S) – and wheat yield. The model explained about 83, 82, 47, and 69% of variances of yield for N, P, K, and S omission, respectively, with consistent performance across training and testing datasets. Expectedly, for N and P omission data, the most important explanatory variables are nutrient rate, followed by soil organic carbon and soil pH. For K and S, however, climatic variables played an important role alongside nutrient rates. The site-specific yield–fertilizer response curves derived from our model are highly variable from location to location, as they are affected by the climatic, soil, or topographic conditions of the site. Importantly, using principal component analysis, we showed that the shape of the fertilizer response curves is a result of the multiple environmental factors (including soil, topography, and climate) that are at play at a given site, rather than of a specific dominant one. The research output is expected to respond to the national policy demands for a sound method to identify the optimal fertilizer rate to increase economic returns of fertilizer investments and take fertilizer utilization research one step further.

Grain sorghum [Sorghum bicolor (L.) Moench] is the major cereal crop used as staple crop in the arid and semi-arid regions of Ethiopia. Low sorghum yields are attributed to soil, climate and topographic factors. We investigated sorghum yield response to factorial combination of nitrogen and phosphorous (NP) as well as potassium (K), sulphur (S) and zinc (Zn), and how the position of farmers’ fields belonging to different landscape positions (i.e., upslope, mid-slope, and foot slope) could explain fertilizer response and yield variability. The analysis in this study made use of dataset from two sets of on-farm experiments where trials were set at two farmers’ fields for NPKS and three farmers’ fields for NPZn experiments in each landscape position. The experiments were implemented at two sorghum-growing locations (i.e., Hayk and Sirinka) in parts of the north-eastern Amhara region in Ethiopia. Sorghum yield response to fertilizer application was strongly linked to the spatial variation along landscape positions and varied over locations. Fertilizer response was significantly higher at foot slopes compared to mid-slopes and upslope positions, where fields at foot slopes exhibited relatively homogeneous responses. Application of combined nitrogen (N) and phosphorus (P) fertilizers, landscape position and the interaction of fertilizer application and landscape positions strongly affected sorghum yield. There was a linear and significant increase in sorghum yield with the increase in the NP rates. The combined application of NP with different levels of KS as well as NP with Zn fertilizer rates did not result in significant yield difference. The results indicated that local factors were much more influential when accounting for the heterogeneity in sorghum yield response to fertilizer. This further acknowledges the importance of a landscape-based fertilizer management approach to respond yield potential variability related with the farmers’ fields and landscape environment. Further investigation is needed to develop homogeneous fertilizer response units based on spatial variability of soil and topographic attributes along the landscape.

The use of fertilizers in balanced and adequate amounts is a prerequisite for increasing crop productivity and production. Unbalanced plant nutrient management continues to be a major factor contributing to low maize (Zea mays L.) yields due to lack of information on the dose–responses to macronutrients on different soil types in Ethiopia. This study was carried out to quantify maize yield response and agronomic efficiency of varying application rates of nitrogen (N), phosphorus (P), potassium (K) and sulphur (S) under balanced application of other nutrients across two soil types in Ethiopia. Field trials were set up on 29 farmers’ fields in four districts of Oromia and Southern Nations, Nationalities and Peoples Region (SNNPR) for three consecutive cropping seasons (2014–2017). The treatments consisted of six rates of N, P and S each and eight rates of K combined with balanced application of the remaining macronutrients, zinc (Zn) and boron (B). The treatments were laid out in randomised complete blocks design with three replicates per farm. Using nutrient dose–response modelling, the agronomic optimum rates of N, P, K and S were estimated at 46, 40, 17 and 10 kg ha−1 on Nitisols, with balanced application of the other nutrients. On Andosols, the optimum rates of N, P and S were estimated at 184, 20 and 30 kg ha−1, respectively, but the optimum K rate could not be estimated. The predicted maximum yields obtained with balanced nutrient application were lower on Andosols (3397–3640 kg ha−1) than on Nitisols (4630–6094 kg ha−1). Using the Mitscherlich dose–response model, the percentage deficiencies of N, P, K and S were estimated to be 1.3–3.3 times more on Nitisols than Andosols. Consequently, agronomic efficiencies of N, P, K and S were significantly lower on Andosols than on Nitisols. It is concluded that balanced application of 46 kg N ha−1, 40 kg P ha−1, 17 kg K ha−1, 10 kg ha−1 S, 2 kg Zn ha−1 and 0.5 kg B ha−1 could be recommended for maize on Nitisols in the study area. Although this recommendation may also apply to Andosol, further research is needed as the productivity of Andosols appears to be limited by constrains other than N, P, K, S, Zn and B. We also recommend a shift from the blanket fertilizer recommendations to site-specific nutrient management based on good understanding of the variations in crop response with soil type and agroecology and appropriate soil and plant analyses.

Nutrient deficiency is a major constraint in tef (Eragrostis tef (Zucc.) Trotter) production in Ethiopia. In the past, a blanket recommendation of nitrogen (N) and phosphorus (P) fertilizers has been applied regardless of the diversity of agroecological zones and soil types. As crop responses declined with widespread deficiencies of nutrients, farmers have lost interest in applying the recommended fertilizer rates. The present study was conducted with the objective of quantifying the response of tef to different rates of N, P, potassium (K) and sulphur (S) fertilizers under balanced application of the nutrients other than the one under investigation. In each region, the same set of trials was implemented on farmers’ fields for three years. All trials were implemented on Vertisols across four agroecological zones (AEZs). The treatments were seven rates of N, six rates each of P and S and eight rates of K with a basal application of zinc (Zn) and boron (B). The results showed that the most limiting nutrient is N followed by P in each agroecological zone on the Vertisols. There was clear evidence of increases in grain yield with increasing rates of N and P, but the responses to K and S rates did not follow clear trends across AEZs. With balanced application of the other nutrients, 23–92 kg N ha−1 increased grain yield by 11–92%, while 10–40 kg P ha−1 increased yields by 12–33% over the control in the different AEZs. Based on results from dose–response models, it is concluded that the agronomic optimum rates for tef production are 69 kg N, 20 kg P, 30 kg K and 10 kg S ha−1 in the cool sub-moist mid-highlands and the tepid sub-humid mid highlands; 92 kg N, 30 kg P, 15 kg K and 10 kg S ha−1 in tepid moist mid highlands but 69 kg N, 10 kg P, 15 kg K and 10 kg S ha−1 in the tepid sub-moist mid highlands. On Vertisols not covered by this study, these recommendations should be subject to appropriate soil and plant analysis. We also recommend further studies on the nutrient requirement of tef on other soil types and AEZs.

The productivity of wheat is low on smallholder farms in Rwanda. Although mineral fertiliser use is being promoted as a sustainable intensification (SI) pathway, little is known about the nutrient use efficiency and profitability of various fertiliser inputs in Burera, Musanze and Nyamagabe districts of Rwanda. The objective of this study was to assess the use of combinations of nitrogen (N), phosphorus (P), potassium (K), magnesium (Mg), zinc (Zn) and boron (B) in wheat production in terms of nutrients management specifically, crop yield, production risk, input use efficiency and economic returns on smallholder farms. The study was conducted in three wheat-growing regions of Rwanda (i.e., Nyamagabe, Musanze and Burera districts) with contrasting soil conditions. The treatments included combinations of different levels of N (0, 30, 60, 90 and 120 kg ha−1) with P (0, 7.5, 15 and 22.5 kg ha−1) and K (10, 20 and 30 kg ha−1) and the control with no applied nutrients. A diagnostic treatment composed of 90 kg N, 15 kg P, 20 kg K, 10 kg Mg, 2.5 kg Zn and 0.5 kg B ha−1 was also included. Mean grain yield and its variability, rainfall use efficiency (RUE), agronomic use efficiency (AE) of N and P and the value cost ratios (VCRs) were calculated to assess the sustainability of the nutrient rates. Across all sites, wheat grain yield and RUE increased with increase in N rates up to 90 kg N ha−1, beyond which no further increase was observed. The highest wheat yield (5.5 t ha−1) and RUE (6.6 kg ha−1 mm−1) with the lowest production risk (coefficient of variation [CV] = 20%) were recorded in the diagnostic treatment. Although the highest AEN and AEP were recorded at lower N and P levels, the CVs of VCR were high (>64%), indicating higher production risk to wheat farmers. In all cases, an optimum VCR (5.6), with the lowest CV (44.4%), was recorded in the diagnostic treatment. We conclude that application of 90 kg N, 15 kg P, 20 kg K, 10 kg Mg, 2.5 kg Zn and 0.1 kg B can guarantee a more SI of wheat production in Burera, Musanze and Nyamagabe districts of Rwanda.

The use of fertilisers in maize production has been the focus for many years of agronomic studies on research stations in East Africa. However, information on production risks and profitability of fertiliser use on smallholder farms is generally lacking because most of the early studies have focused on mean yields and yield components on research stations. This study applied rigorous analyses to determine conditions under which (1) production risks are low; (2) the recommended nitrogen (N) and phosphorus (P) fertiliser rates achieve a yield target of ≥3 t ha−1 believed to be a necessary condition to kick start a smallholder-led ‘green revolution’ in Africa and (3) N and P fertiliser use is profitable on smallholder farms in East Africa. Analysis of data from 464 on-farm trials in Kenya, Rwanda, Tanzania and Uganda revealed significant variations in production risks and nutrient use efficiency with season and soil type. On most sites, except in Uganda, production risks were lower with the recommended N and P fertilisers than the control during both the short and long rains. Production risks were three to four times higher with N and P fertiliser relative to the control on Lixisols and Ferralsols, but such risks were much lower on Nitisols, Leptosols, Vertisols, Plinthosols and Cambisols. The probability of exceeding grain yields of 3 t ha−1 with the recommended N and P rates was over 0.60 on Nitisols and Leptosols, but less than 0.20 on Lixisols and Plinthosols. The agronomic use efficiencies of N (AEN), P (AEP) and value cost ratios (VCR) were highest on Cambisols and lowest on Plinthosols. The VCR increased linearly with increase in AEN (R2 = 0.92) and AEP (R2 = 0.87) and less so with increase in grain yields (R2 = 0.47–0.60). Net present values indicated profitably of N and P fertiliser over the long term in only 30% of the site by season combinations in Uganda compared with 69% in Kenya, 81% in Rwanda and 84% in Tanzania. Patterns of N use efficiencies were different from P use efficiencies across soil types. Therefore, we recommend that N and P fertilisers should be appropriately targeted to soils where applied nutrients are used efficiently by maize crops.

The response of wheat to the application of different rates of nitrogen (N), phosphorus (P), potassium (K), and sulfur (S) under balanced fertilization on different soil types and agroecologies has not been well studied in Ethiopia. Therefore, the objectives of this study were to (1) determine soil-specific responses of wheat to N, P, K, and S under balanced fertilization; (2) quantify agroecology-specific N, P, K, and S response of wheat under balanced fertilization; and (3) determine nutrient use efficiency of wheat on different soil types under balanced fertilization. Trials were conducted on farmers’ fields across 24 locations covering 4 soil types and 5 agroecological zones (AEZs) from 2013 to 2017. The mean grain yields of wheat significantly varied with applied N and P fertilizer rates with soil types and AEZs. With balanced application of other nutrients, the optimum N rates for wheat were 138 kg N ha−1 on Cambisols and Luvisols, 92 kg N ha−1 on Vertisols, and 176 kg N ha−1 on Nitisols, while the optimum P rate was 20 kg P ha−1 on Cambisols and Vertisols. The nutrient dose–response curve did not reveal consistent pattern for K and S applications on all soil types. The agronomic efficiency of wheat decreased with increasing rates N and P on all soil types. The highest agronomic efficiency of N (15.8 kg grain kg−1 applied N) was recorded with application of 92 kg N ha−1 on Vertisols, while the highest agronomic efficiency of P (49 kg grain kg−1 applied P) was achieved with application of 10 kg P ha−1 on Cambisols. We conclude that applications of 92–138 kg N ha−1, 20 kg P ha−1, 18 kg K ha−1, and 10 kg S ha−1 under balanced application of zinc and boron could be recommended depending on soil type for wheat production in the study areas.

Ethiopia has heterogeneous topographic, climatic and socio-ecological systems. Recommendations of agricultural inputs and management practices based on coarse domains such as agro-ecological zones (AEZ) may not lead to accurate targeting, mainly due to large intra-zone variations. The lack of well-targeted recommendations may contribute to the underperformance of promising technologies. Therefore, there is a need to define units where similar environmental and biophysical features prevail, based on which specific recommendations can be made for similar response units (SRUs). We used unsupervised machine learning algorithms to identify areas of high similarity or homogeneous zones called ‘SRUs’ that can guide the targeting of agricultural technologies. SRUs are landscape entities defined by integrating relevant environmental covariates with the intention to identify areas of similar responses. Using environmental spatial data layers such as edaphic and ecological variables for delineation of the SRUs, we applied K- and X-means clustering techniques to generate various granular levels of zonation and define areas of high similarity. The results of the clustering were validated through expert consultation and by comparison with an existing operational AEZ map of Ethiopia. We also augmented validation of the heterogeneity of the SRUs by using field-based crop response to fertiliser application experimental data. The expert consultation highlighted that the SRUs can provide improved clustering of areas of high similarity for targeting interventions. Comparison with the AEZ map indicated that SRUs with the same number of AEZ units captured heterogeneity better with less within-cluster variability of the former. In addition, SRUs show lower within-cluster variability to optimal crop response to fertiliser application compared with AEZs with the same number of classes. This implies that the SRUs can be used for refined agricultural input and technology targeting. The work in this study also developed an operational framework that users can deploy to fetch data from the cloud and generate SRUs for their areas of interest.

In sub-Saharan Africa (SSA), farmers intercrop common beans with maize but apply inorganic or organic fertilisers targeting only maize. Effects of this practice on bush bean yield have not been fully evaluated with respect to input use and compatibility when intercropped with maize. An on-farm trial managed by smallholder community members was conducted to assess the influence of various soil fertility management options and cropping systems on the yield of two bush bean genotypes (SER45 and SER83) in two agro-ecological zones of Malawi. The farmer-managed trials were laid out in split-plot design, with the bean genotypes as main plots and a combination of the soil fertility management options (i.e., no input, manure, fertiliser and fertiliser + manure) and cropping systems (i.e., sole crop and intercrop) as subplots. The trials were affected by terminal drought and dry spells, but results show that manure and fertiliser application enhanced the resilience of the drought-tolerant bean genotypes. The genotype SER45 was responsive to manure application in the sole crop, giving a 44.4% yield increase over no-manure application. In sole cropping with fertiliser plus manure, bean yields improved by 40.1% for SER45 and 78.3% for SER83 relative to the no-input control. Although sole cropping had higher bean yields, the treatment with manure and fertiliser had a higher land equivalence ratio for intercrop of 1.54 for SER45 and 1.32 for SER83 over sole cropping. These results show that, under smallholder farmer management, the climate adaptability of bush bean genotypes could be enhanced by the combined application of organic and inorganic fertilisers in maize–bean intercrop. The combined application also enhances whole-farm productivity of the common maize–bean intercrop practice than monocrop, hence is of benefit to most low-input smallholder farmers of SSA.