CONTEXT: Sustainable intensification (SI) is a key strategy for improving the productivity, profitability, and resilience of smallholder farming systems while maintaining environmental sustainability. However, the synergies and trade-offs associated with SI interventions vary across farm typologies and farming contexts, making the design of equitable and effective interventions challenging. OBJECTIVE: This study evaluated the synergies and trade-offs of selected SI interventions for improving soil organic matter (SOM), farm profitability, and labour balance across contrasting farm typologies in Ethiopia, Ghana, and Malawi, and assessed how farm typology and farming context influence intervention performance. METHODS: This study uses the FarmDESIGN whole-farm optimization model to assess current farm performance and generate "clouds" of optimized designs for three countries and three farmer resource-endowment groups (high, medium, low). Five SI interventions, identified with farmers and local stakeholders i.e. crop residue retention, intercropping, improved crop varieties, compost application, and improved livestock breeds were evaluated against three objectives: increasing SOM, enhancing operating profit, and reducing labour imbalance. RESULTS AND CONCLUSIONS: Optimization revealed substantial room for improvement under current management with simultaneous gains in SOM and profit and reductions in labour requirements. In contrast, SI interventions increased SOM and profits but also raised labour demand sharply, depending on context. Medium-resource farms benefited most from intercropping and improved varieties, while low-resource farms gained from residue retention and compost. SOM improvements were largest in integrated crop-livestock systems in Ethiopia, whereas labour constraints, particularly in Malawi, limited profitability despite soil benefits. Findings show that performance gaps arise from structural and typology-specific constraints, requiring interventions tailored to local bottlenecks such as labour shortages, soil degradation, or low profitability. The cross-country comparison further demonstrated that SI outcomes are influenced not only by farm typology but also by broader contextual factors affecting labour, soil management, and system integration. SI options must fit farmers' resources, and decision-makers need targeted support strategies to ensure practical and inclusive outcomes. SIGNIFICANCE: This study demonstrates that SI should be designed as a context-specific strategy rather than promoted as a universal technology package. By explicitly quantifying synergies and trade-offs across farm typologies and countries, the findings provide a scientific basis for designing targeted, equitable, and scalable SI strategies for smallholder farming systems in sub-Saharan Africa.

This study evaluates the impact of climate-smart agricultural (CSA) practices on ecosystem functions (EFs) and ecosystem services (ESs). Four CSA practices, namely, soil bund (SB), crop residue (CR), integrated conservation practice (ICP), and berken plowing (BP), were compared with conventional practices (CP) through field experiments over three rain seasons (2019–2021). The experiment used a randomized complete block design with three replications. Linear mixed-effect and Spearman correlation methods were used for data analysis. The findings reveal that, compared with that under CP, runoff under ICP, SB, and BP was significantly (P < 0.05) lower by 57.5, 55.8, and 27.1%, respectively. Similarly, soil loss was significantly lower by 76, 73, 48, and 15% in the ICP, SB, BP, and CR treatments, respectively, than in the CP plots. Nutrient losses were also significantly (P < 0.05) reduced in sequence under the ICP, SB, BP, and CR treatments compared to the CP treatment. The soil organic carbon stock (SCS) was substantially enhanced under CSA practices. The plant available water content (PAWC) was significantly (P < 0.05) higher under ICP, CR, and BP than under CP. Compared with those under CP, the wheat grain yields under ICP, CR, BP, and SB increased significantly (P < 0.05) on average by 27.4, 23.6, 23.0, and 5.0%, respectively. Overall, these CSA practices improved multiple ecosystem functions and services in the short term, as did sustainable land management practices. The findings suggest that properly designed integrated CSA strategies are essential for resilient and sustainable agriculture without compromising ecosystem services (ESs).

Understanding how spatial and temporal factors influence lake areas is crucial for effective water resource management, particularly in regions with limited hydrological records. We developed a spatially explicit deep learning framework for six major lakes in Ethiopia, using long-term sub-catchment climate and biophysical variables as predictor variables, and remotely sensed lake area as response variables. The framework used an Average Hop Downstream Diffusion graph convolution to represent upstream-to-downstream sub-catchment connectivity, followed by average (AVG) and attention (ATTN) pooling to derive basin-level spatial embeddings at each time step. These embeddings were processed with Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM) networks, yielding four model variants: AVGGRU, ATTNGRU, AVGLSTM, and ATTNLSTM. All four variants showed strong predictive skill, with correlation coefficients and coefficients of determination above 0.85 and absolute percentage errors below 1 percent of long-term mean lake area, but limited ability to capture high-frequency variability highlights. Model differences were more evident in reproducing variability in amplitude and temporal phase, where skill was more moderate. ATTNGRU showed the most stable overall performance. Attribution analysis based on learned attention weights and gradient sensitivity revealed physically consistent patterns. In large inflow-dominated lakes, the most influential sub-catchments were located in upstream tributary sources; in smaller lakes they were concentrated near shorelines. Rainfall and vegetation were more influential in upland catchments, while temperature, evapotranspiration, and runoff were more important near the lakes. Overall, connectivity-aware attention-based models improved both prediction and process interpretation of lake area dynamics in ungauged, data-scarce systems. Future research on data-driven lake-area prediction is recommended to capture high-frequency variability.

Context: Cereals are of strategic importance for food and nutrition security in Ethiopia. Considerable yield gains have been reported in the country, yet farm yields remain well below their potential. Targeting technological solutions to current yield constraints can support efforts aiming to increase cereal productivity in a sustainable manner. Objective: The objectives of this study were to quantify yield gaps for major cereals (tef, wheat and sorghum) and to identify the yield constraints under on-farm conditions across major production environments of Ethiopia. Methodology: A large agronomic diagnostic survey with crop cut yield measurements was conducted for tef (n = 1131), wheat (n = 1721) and sorghum (n = 971) during the 2022 main (Meher) growing season. The survey was expanded with growing-season specific weather data, spatially predicted soil properties, and water-limited yields (Yw) simulated with crop models (or obtained from literature in the case of tef). Yield gaps were quantified as the difference between the water-limited yield and the measured actual yield. Field-specific yield constraints were identified with a machine learning approach, which explained 38-49% on the yield variability in independent test data. Model interpretation was done using SHAP plots and SHAP dependence plots for the most important predictors. Results: Actual yields were on average 1.6 t ha(-1) for tef, 3.4 t ha(-1) for wheat, and 2.2 t ha(-1) for sorghum, corresponding to a yield gap closure (defined as the ratio between actual and water-limited yields) of 47, 39 and 31% of Yw, respectively. Wheat yield gap closure was largest (50-60% of Yw) in North Shoa, which is indicative of increases in wheat productivity associated with increased fertilizer use, and lowest (20% of Yw) in North Wollo. The yield variability of the three crops was largely explained by biophysical conditions (e.g., average seasonal minimum and maximum temperature), except for sorghum for which plant population was one of the top predictors. Agronomic constraints for tef were mostly associated with sowing date, yet other important factors constraining tef yield variability (N fertilizer, seed rate, tillage and weed control) were not fully captured in our analysis. Agronomic constraints for wheat were largely associated with nutrient and disease management, and for sorghum with sub-optimal plant populations at harvest. Conclusion: Variation in tef productivity was attributed to seasonal climatic conditions and agronomic management related to sowing dates. Nutrient and disease management remain important constraints to wheat yield, and there is scope for targeting interventions aiming to increase fertilizer rates and use efficiency across the country. Ensuring high plant populations during the cropping season, increasing fertilizer inputs and adopting practices that preserve soil moisture provide entry points to increase sorghum yields in the future. Significance: Sustainable intensification strategies were identified for three major cereal crops in Ethiopia. Results provide disaggregated evidence of management by environment interactions driving yield variability and provide an entry point for integrated assessments of farm performance.

This study aimed to model the impact of climate-smart agricultural (CSA) practices on soil‒water balance, water use efficiency (WUEET), and wheat yield in the face of climate change. The AquaCrop version 7.1 model was used to estimate the water balance and yield under the baseline (1981-2010) and future (2050s, RCP4.5) climate scenarios. We evaluated five CSA practices, varying in tillage, residue management, and water management, based on experiments conducted in 2020 and 2021. Observed data on wheat (Triticum aestivum L.) grain yield and surface runoff were used for model calibration (2020) and evaluation (2021). The model was evaluated using four performance indicators and found to be robust. The treatments included farmers' conventional practices (CPs), soil bunds (SBs), crop residues (CRs), integrated conservation practices (ICPs), and berken plows (BPs). The results show that climate change is likely to reduce grain yield and WUEET under CP by 1% and 16.3%, respectively, by 2050 compared to the current 2021 period. All CSA practices studied increased grain yield and WUEET over the CP in both periods. Under future climates, ICP showed a greater relative grain yield (Y = 4.51 t/ha), water use efficiency (WUEET = 1.32 kg m3), and other soil water balances, followed by CR, BP, and SB over CP. Overall, ICP has shown tremendous potential for climate change adaptation among the other CSA practices tested. Therefore, adaptation to future climate conditions must integrate different practices, and the novel ICP can be promoted as a climate-smart practice in similar farming systems and agro-ecological settings.

Context Low maize productivity in Sub-Saharan Africa is partly attributed to fertilizer recommendations that inadequately capture spatial variability in soils, climate, and economic conditions. Objectives This study aimed to (i) evaluate machine learning approaches for maize yield prediction; (ii) identify the major drivers of maize yield variability; (iii) quantify nutrient–environment interactions; and (iv) develop a framework for generating site-specific nutrient recommendations for Malawi. Methods Using 14,625 geo-referenced multi-season maize fertilizer trial observations collected across Malawi, we developed an explainable machine learning framework based on LightGBM integrated with recursive feature elimination, SHapley Additive exPlanations (SHAP), and high-resolution soil, climate, hydrological, and topographic covariates. The optimized model was used to predict maize yield responses and derive agronomic and economic optimal nutrient recommendations for nitrogen (N), phosphorus (P), potassium (K), sulfur (S), and zinc (Zn) across 626,022 cropland grid cells (250 m resolution). Results The optimized LightGBM model accurately predicted maize yield (Test R² = 0.71). Nitrogen, phosphorus, soil moisture, and solar radiation emerged as the dominant drivers of yield variability, with strong nutrient–environment interactions revealed through SHAP and response surface analyses. The framework generated 165 agronomic and 241 economic site-specific nutrient recommendations, revealing substantial spatial variability in optimal nutrient requirements, particularly for N (40–120 kg ha⁻¹) and P (5–35 kg ha⁻¹). To facilitate implementation, these site-specific nutrient recommendations were subsequently translated into four representative agronomic and five economic nutrient formulations through clustering, providing a practical balance between recommendation precision and operational feasibility. Conclusions Explainable machine learning provides a transparent and scalable framework for generating spatially explicit nutrient recommendations that better reflect local agronomic and economic conditions than generalized recommendation systems. Expressing recommendations as nutrient requirements rather than fixed fertilizer products provides flexibility for implementation using blended fertilizers, straight fertilizers, organic nutrient sources, or integrated nutrient management strategies. The newly proposed nutrient rates should be validated through on-farm trials prior to any recommendations. Significance The proposed framework bridges the gap between high-resolution digital agronomy and practical fertilizer recommendation systems, providing a scalable foundation for improving fertilizer targeting, nutrient use efficiency, profitability, and climate-resilient maize production in Malawi and other smallholder farming systems in sub-Saharan Africa.

Context: Characterizing crop production environments is essential for targeted interventions, resource allocation, scaling localized findings, and agricultural decision-making. However, existing methods lack the spatial and temporal rigor required to capture spatial and temporal variability in crop production environments. Objective: This study aimed to introduce a data-driven and dynamic spatial framework that integrates crop area mapping with the delineation of agro-ecological spatial units (ASUs) to characterize Ethiopia's rainfed wheat crop production environments. Methods: Annual rainfed wheat areas for the 2021 and 2022 Meher growing seasons were mapped using an ensemble machine-learning approach, leveraging time-series satellite images and environmental data. Dynamic ASUs were delineated using pixel- and object-based clustering methods, considering short-term changes (annual ASUs for 2021 and 2022) and longer-term trends (ASUs developed using data aggregated over the period 2016-2022). Clustering was based on key biophysical variables, including climatic, soil, topographic, and vegetation indices derived from satellite images that capture crop growth and development over space and time. Results and conclusions: The framework captured the spatial and temporal variability of wheat production environments, demonstrating its scalability across space and time. Rainfed wheat area mapping across two growing seasons revealed an expansion in rainfed wheat areas, highlighting the evolving nature of rainfed wheat cultivation in Ethiopia. The integration of rainfed wheat area mapping with dynamic ASU delineation identified five main production environments for wheat in Ethiopia, allowing to better target future research and development activities toward increasing wheat productivity in the country. Significance: The developed framework can facilitate agronomic assessments and inform the targeting of agricultural interventions, with potential applications that extend beyond this case study of rainfed wheat in Ethiopia.

Land degradation in Ethiopia is a pressing issue that demands immediate attention. Although various sustainable land management options have been introduced through top-down approaches, farmers have shown low adoption rates. The objective of this research was to assess the community prioritization of landscape restoration technologies and the appropriation of ecosystem services in the Basona-Worena and Doyo-Gena woredas of Ethiopia. The evaluation of land management option tool was used to survey farmers' preferences and compare different land management options based on input, cost, perceived advantages, and potential drawbacks.Data from 64 participants revealed that farmers were interested in a wide range of benefits. However, their top three preferences were increased food supply, enhanced soil fertility, and improved water supply. The study emphasized the need for site-specific land management measures. Farmers in Basona-Worena favored terrace and bund practices, while farmers in Doyo-Gena preferred exclosure and agroforestry practices. Conversely, the propensity of terracing to attract rodents and pests, the lengthy time takes to see results from bunding, and the cost of gabions were among the shortcomings that farmers identify in conservation techniques. Terracing was the first option for supplying fundamental ecosystem services in both locations, followed by biological measures, water percolation pits, and bunds. All farmers ranked the business-as-usual option as their least preferred option because they perceived it to have limited potential for yielding desired benefits. These findings provide a robust model for informed decision-making on suitable restoration technologies, holding promise for landscape restoration initiatives in Ethiopia and similar locations worldwide.

Climate-smart agricultural (CSA) practices have been adopted in various agroecological zones in Ethiopia to enhance productivity, improve resilience to climate change, and reduce greenhouse gas emissions through carbon sequestration. However, the overall impact of different CSA practices on productivity, adaptation, and mitigation metrics has not been exhaustively evaluated. The study employed a meta-analysis approach based on data from 220 peer-reviewed articles to assess the effects of commonly used CSA practices on these indicators across Ethiopia's diverse agroecological regions. The analysis identified over 20 CSA practices, with most-except soil bunds, level Fanya juu, and deficit irrigation-showing positive effects on productivity, peaking with drip irrigation (effect size of 2.15). Almost all practices also effectively reduced runoff and soil erosion, particularly crop residue mulching, which had a remarkable effect size of 2.95. Additionally, the findings indicated that various CSA practices enhanced soil organic matter and carbon stocks. Water management practices, especially drip and deficit irrigation, demonstrated significantly higher water productivity than traditional flood irrigation, with an effect size of up to 2.6. This water use efficiency suggests that these methods could free up substantial water resources for irrigating additional land, thus boosting crop production in water-scarce areas. However, the analysis revealed a negative effect size of up to -0.74 for income derived from drip irrigation, primarily due to the high costs of the necessary equipment. This highlights the need for reforms in duty and tax exemptions to improve farmers' profitability from drip irrigation systems. Overall, this meta-analysis assesses the impact of various CSA practices on key performance indicators productivity, adaptation, and mitigation-providing insights that can guide the packaging and implementation of the most effective CSA strategies across Ethiopia's agroecological zones.

The strategic use of agricultural organic waste, either directly or processed into high-value organic fertilizers (OFs), is crucial for sustainable agricultural land management, particularly in sub-Saharan Africa, including Ethiopia. This strategy addresses soil organic matter (SOM) depletion and rising chemical fertilizer costs, which pose significant challenges to smallholder farmers. Focus has shifted to underutilized agricultural feedstock such as crop residues and animal manure to counter the decline in soil organic carbon and nutrients caused by competing uses like fuel, construction, and livestock feed. Despite the urgency to recycle organic resources through methods like composting and vermicomposting, challenges persist due to limited information. This review aims to (i) highlight available organic resources for valorization in different agro-ecologies, (ii) showcase technologies for converting organic waste into valuable bio-products and assess their characteristics, focusing on nutrient composition, fertilizer replacement value, and impact on crop yield and soil health, and (iii) discuss current gaps and future prospects tailored to Ethiopian conditions. Recent valorization strategies such as vermicomposting convert underutilized organic matter into high-quality amendments with higher nutrient and microbial composition compared to other methods like thermophilic composting. The combined use of organic and chemical fertilizers consistently leads to significant increases in crop yields in various agro-ecological contexts compared to using organic or chemical fertilizers alone. However, the long-term effects on soil health metrics and agricultural productivity need further investigation in Ethiopian farming. Overall, valorizing organic waste into value-added bio-products (OFs) promotes sustainable agricultural land management practices. Collaborative efforts among governments, research institutes, agricultural experts, and local communities are essential to formulate and implement effective strategies that prioritize soil health through the large-scale adoption of valorization methods for producing OFs.

[This retracts the article DOI: 10.1016/j.heliyon.2025.e42796.].

This study assessed farmers' satisfaction with site-specific fertilizer recommendations (SSFR) and identified key determinants influencing their satisfaction in Ethiopia. Data from 202 households, selected through stratified random sampling, was analyzed using Principal Component Analysis (PCA) and Ordered Probit Model. Results show that 58.4% of farmers were satisfied with the quantity, while 56.9% were satisfied with the timing of fertilizer application in the study areas. Strong satisfaction was reported by 37.6% for recommended fertilizer rates and 39.6% for timing, with minimal (4%) dissatisfaction. Partnerships between the Ministry of Agriculture (MoA), LERSHA, and Digital Green reveals varying satisfaction level had varied satisfaction rates, with MoA leading at 44.6%, compared to LERSHA’s 24.8% and Digital Green’s 30.7%. The study identified key factors that affect satisfaction, including, education level, farm size, availability and affordability of SSF recommendations, the quality of information on planting time, information on land preparation, use of SMS for SSFR dissemination, recommendation of SSFR using cluster approach and livestock size. Higher education levels and larger farms are linked to better SSFR application. Participation in cluster recommendation units fosters collective learning and enhances satisfaction, while access to affordable and timely SSF recommendation improves implementation and crop yields. Short Message Services (SMS) communication has proven effective in engaging farmer and enhancing satisfaction. As Ethiopia continues to work towards agricultural modernization and sustainability, addressing these factors will be vital for promoting the use of site-specific fertilizers (SSF) across Ethiopia's diverse farming landscapes.

Assessing the effects of landscape restoration interventions requires a systematic assessment approach that integrates Sustainable Intensification (SI) indicators. The Sustainable Intensification Assessment Framework (SIAF) is an impact assessment approach that incorporates SI indicators including productivity, economic, environmental, human and social conditions in the assessment of the performance of landscape restoration interventions. This study was conducted in the Geda and Doyogena landscapes of Ethiopia to assess the performance of landscape restoration interventions. The SIAF was used in two landscapes of Ethiopia. This paper employed various data sources including primary data from the Rural Household Multi-Indicator Survey (RHoMIS), secondary data from a literature review and satellite imagery. Remote Sensing and Geographical Information System (GIS) techniques were used to analyze satellite images while SPSS (V. 26) package was used to analyze primary survey data. Results showed that restoration in both landscapes improved the Normalized Difference Vegetation Index (NDVI), increased water productivity and reduced soil erosion. Although landscape restoration reduced soil erosion, it is still by far higher than the tolerable soil loss limit in the Doyogena landscape. This indicates that more land management interventions should be implemented in the coming years. We observed high variability of indicator scores across the five domains in both landscapes. The aggregated indicators showed that landscape management contributed to restoring the environment and improving agricultural productivity and income of households in Ethiopian highlands.

Up-to-date digital soil resource information and its comprehensive understanding are crucial to supporting crop production and sustainable agricultural development. Generating such information through conventional approaches consumes time and resources, and is difficult for developing countries. In Ethiopia, the soil resource map that was in use is qualitative, dated (since 1984), and small scaled (1 : 2 M), which limit its practical applicability. Yet, a large legacy soil profile dataset accumulated over time and the emerging machine-learning modeling approaches can help in generating a high-quality quantitative digital soil map that can provide better soil information. Thus, a group of researchers formed a Coalition of the Willing for soil and agronomy data-sharing and collated about 20 000 soil profile data and stored them in a central database. The data were cleaned and harmonized using the latest soil profile data template and 14 681 profile data were prepared for modeling. Random forest was used to develop a continuous quantitative digital map of 18 World Reference Base (WRB) soil groups at 250 m resolution by integrating environmental covariates representing major soil-forming factors. The map was validated by experts through a rigorous process involving senior soil specialists or pedologists checking the map based on purposely selected district-level geographic windows across Ethiopia. The map is expected to be of tremendous value for soil management and other land-based development planning, given its improved spatial resolution and quantitative digital representation.

Capturing the heterogeneity of farming systems through farm typology is essential for targeting agricultural interventions in any mixed crop-livestock farming system. Therefore, this study aims to construct a farm typology for the Doyogena and Basona districts of Southern and Northern Ethiopia, respectively. A combination of principal component analysis (PCA) and hierarchical clustering (HC) was used to develop a generalized and domain-specific farm typology in the study areas using farm household survey data collected from 503 respondents. A generalized farm typology was constructed considering all the dataset variables whereas the domain-specific farm typology was developed once all the data variables had been categorized into three groups: variables that describe i) the resource endowment, ii) technologies used, and iii) food and nutrition characteristics of the farm. The farm types identified from the domain-specific farm typologies were merged to develop comprehensive, representative, and meaningful farm types. In both districts, the results of the generalized farm typology are more generic, and are not able to fully capture the diversity of farmers’ resource endowment and food and nutrition security status. Compared to the generalized farm typology, the domain-specific farm typology is more useful for targeting tailored agricultural development interventions. The merged typology results show that a combination of medium resource endowment with medium income, medium technology, and low food and nutrition security farm type (34%) is the dominant farm type in the study areas followed by a farm type combining low resource endowment with low income, high technology and marginal food and nutrition secure (21%). The findings of this study provide several insights into targeting and scaling domain-specific agricultural development interventions that can be applicable for sustainable intensification of mixed farming systems. For example, growing multiple crops in crop rotations and as intercrops; implementing conservation tillage, and introducing improved seed varieties, and livestock breeds offer possible pathways for sustainable agricultural intensification for medium resource endowment, medium technology, and low food and nutrition security farm types.

Various climate-smart agricultural (CSA) practices are being advocated in different agroecological zones of Ethiopia to enhance the sustainability, resilience, and productivity of the agricultural sector in response to climate change. Prioritizing and packaging these CSA practices are essential to amplify the impact of climate change mitigation efforts. By strategically selecting and prioritizing these practices and technologies, resources can be allocated effectively to activities with the highest potential for reducing greenhouse gas emissions, bolstering resilience, and fostering sustainable development. However, identifying and prioritizing climate-smart practices that cater to the needs of vulnerable farmers and are tailored to specific local contexts remains challenging, often hindered by subjective assessments and limited awareness. The objective of this paper was to enhance the precision and objectivity of prioritizing CSA practices by leveraging a combination of research findings and expert knowledge. The steps included the following: i) a CSA prioritization assessment framework was used to identify and prioritize CSA practices across various agro-ecologies based on the CSA pillars (productivity, adaptation, and mitigation); ii), a meta-analysis approach was employed to determine the effect size of various CSA practices on the three pillars of CSA practices; iii), the effect size values were rescaled and ranked based on effect size categories; and iv), correlation was performed to assess the relationship between the two approaches, and finally, average values were taken to integrate and determine the final rank of CSA practices. Overall, we found out that there were weak correlations between the ranks of the two approaches, resulted in a mismatch between the ranks of CSA practices by experts and meta-analysis results. Using the meta-analysis approach, only 35% of the CSA practices were equally ranked by both approaches, 40% of the CSA practices were more likely ranked by experts, while 25% of the CSA practices were more likely ranked by the meta-analysis approach. This implies that experts overestimated the effect of various CSA practices on various indicators of productivity, soil loss, and run-off and soil organic matter. Integrating the ranks of the two approaches helped to target CSA practices across various agro-ecological zones. According to the combined ranking, several CSA practices were targeted to six major agro-ecological zones in the country. These various CSA practices increase productivity, enhance adaptation, and sequester carbon dioxide from the atmosphere. Based on the availability of these CSA practices, it is possible to package various combinations of these practices.

Rainfall variability coupled with poor land and water management is contributing to food insecurity in many sub-Saharan African countries such as Ethiopia. To address such challenges, various efforts have been implemented in Ethiopia. The objective of this study was to evaluate the long term impacts of different soil and water conservation and water harvesting interventions on groundwater and drought resilience of the Gule watershed, northern Ethiopia. The study involved: (i) documentation of the approaches followed and the technologies implemented in Gule since the 1990s, (ii) monitoring the hydrological effects of the interventions for ten years, and (iii) evaluation of the effects of the interventions on groundwater (level and quality), spring discharge and suspended sediment concentration (SSC) in runoff. Results showed that interventions were implemented at different stages and scales. As a result of the interventions, the watershed was transformed into a landscape resilient to rainfall variability: (a) dry shallow groundwater wells have become productive and the level of water in wells has raised, (b) the groundwater quality has improved, (c) SSC in high floods has reduced by up to 65%, (d) discharge of existing springs has increased by up to 73% and new springs have started to emerge. Due to improved water availability, irrigated land has increased from less than 3.5 ha before 2002 to 166 ha in 2019. Communities have remained water-secure during an extreme drought in 2015/2016. Implementation of watershed management practices has transformed the landscape to be resilient to rainfall variability in a semi-arid environment: a lesson for adaptation to climate variability and change in similar environments.

Abstract1. CONTEXTAddressing the limitations of scaling agronomic recommendations, which are usually confined to small areas, requires a spatial framework for characterizing production environments in a timely and cost-effective manner.2.OBJECTIVEThis study aimed to introduce a data-driven framework to characterize rainfed wheat crop production environments in Ethiopia. The framework entails mapping of the annual rainfed wheat area and the delineation of crop-specific and dynamic agro-ecological spatial units (ASUs).3. METHODSAn ensemble machine learning approach built upon time-series satellite images and environmental data was used for crop type mapping while pixel- and object-based clustering algorithms were used to delineate dynamic ASUs from two temporal perspectives: annual ASUs for the 2021 and 2022 growing seasons to assess short-term dynamism, and ASUs from aggregated data (2016 – 2022) to capture long-term variations in the production environment.4. RESULTS AND CONCLUSIONSModel evaluation showed that the ensemble of random forest, gradient boosting, and classification and regression trees predicted wheat cropland in the 2021 and 2022 growing seasons with 88-90% accuracy. A concordance in defining ASUs between pixel- and object-based approaches was observed with consistency and dynamism in ASUs from 2021 to 2022 and between single-year and aggregated ASUs across approaches. This consistency and dynamism in ASUs highlight the spatial scalability and temporal flexibility of the framework, which allows for characterizing production environments across scales and analyzing trends and fluctuations, providing valuable insights for addressing food security and environmental challenges.5.SIGNIFICANCEThe developed spatial framework could facilitate future yield gap analysis and agronomic assessments for rainfed wheat in Ethiopia and be transfered to other crops and production environments.

Context: Fertilizer use efficiency and profitability are very low due to blanket fertilizer recommendations in sub-Saharan Africa. It is crucial to establish tailored recommendations that account for local conditions. Countries like Ethiopia are moving towards adopting site-specific fertilizer recommendations (SSFR) that are developed using machine learning (ML) and designed to enhance yields, profitability, and environmental benefits. Objective: The objective of this study was to evaluate the performance of ML generated SSFR for wheat in improving resource use efficiency, yields, and profitability compared to local (LBFR) and national (NBFR) blanket fertilizer recommendations. Methods: ML was used to develop SSFR for wheat in Ethiopia. Farmer replicated on-farm validation trials were conducted across 277 sites using a randomized complete block design. Data on farm management history, yields, and prices for fertilizer and grain were gathered using Open Data Kit (ODK) tools. Key performance metrics, including site-specific yield gains or losses, profitability, and resource use efficiencies were computed for each site. Data analysis and the presentation of results were conducted using R software packages. Results: The study indicated variability in resource use efficiency, yields, and profits within and across the testing districts. Performance of SSFR was superior in 75% and 72%, lower in 14% and 21%, and comparable in 10% and 7% of the testing sites compared to NBFR and LBFR, respectively. SSFR led to average grain yield increase of 16% and 25% over NBFR and LBFR, respectively. P and S use efficiency were low with SSFR compared to the blanket recommendations. SSFR increased nitrogen use efficiency by 30% and water use efficiency by 0.58 kg and 0.83 kg of grain per ha per mm of water over NBFR and LBFR, respectively. Furthermore, SSFR yielded profit gain of USD580 per hectare per season over LBFR and USD412 over NBFR. Conclusions: SSFR using ML was effective at enhancing wheat productivity, profitability, and resource use efficiency. The yield loss at a few locations and reductions in P and S use efficiency with the SSFR underscore the importance of improving the predictive ability of the ML algorithm by incorporating a broader array of variables and data from diverse wheat farming contexts. Significance: This study underlines the innovative use of data-driven ML approach to optimize fertilizer use in developing countries. The findings support the pilot expansion of SSFR under diverse conditions to optimize fertilizer efficiency and increase crop productivity and profitability for smallholder farmers.

In most developing countries, smallholder farms are the ultimate source of income and produce a significant portion of overall crop production for the major crops. Accurate crop distribution mapping and acreage estimation play a major role in optimizing crop production and resource allocation. In this study, we aim to develop a spatio–temporal, multi-spectral, and multi-polarimetric LULC mapping approach to assess crop distribution mapping and acreage estimation for the Oromia Region in Ethiopia. The study was conducted by integrating data from the optical and radar sensors of sentinel products. Supervised machine learning algorithms such as Support Vector Machine, Random Forest, Classification and Regression Trees, and Gradient Boost were used to classify the study area into five first-class common land use types (built-up, agriculture, vegetation, bare land, and water). Training and validation data were collected from ground and high-resolution images and split in a 70:30 ratio. The accuracy of the classification was evaluated using different metrics such as overall accuracy, kappa coefficient, figure of metric, and F-score. The results indicate that the SVM classifier demonstrates higher accuracy compared to other algorithms, with an overall accuracy for Sentinel-2-only data and the integration of optical with microwave data of 90% and 94% and a kappa value of 0.85 and 0.91, respectively. Accordingly, the integration of Sentinel-1 and Sentinel-2 data resulted in higher overall accuracy compared to the use of Sentinel-2 data alone. The findings demonstrate the remarkable potential of multi-source remotely sensed data in agricultural acreage estimation in small farm holdings. These preliminary findings highlight the potential of using multi-source active and passive remote sensing data for agricultural area mapping and acreage estimation.