Search

Rationale and Objectives To date, no clinically useful classification system has been developed for reliably differentiating mucinous cystic neoplasm (MCN) from a benign hepatic cyst (BHC) in the liver. The objective was to use machine learning and a multi-center study design to develop and assess the performance of a novel classification system for predicting whether a hepatic cystic lesion represents MCN or BHC. Materials and Methods A multi-center cohort study identified 154 surgically resected hepatic cystic lesions in 154 subjects which were pathologic confirmed as MCN (43) or BHC (111). Readers at each institution recorded seven pre-determined imaging features previously identified as potential differentiating features from prior publications. The contribution of each of these features to differentiating MCN from BHC was assessed by machine learning to develop an optimal classification system. Results Although several of the assessed imaging features demonstrated statistical significance, only 3 imaging features were found by machine learning to significantly contribute to a potential classification system: (1) solid enhancing nodule (2) all septations arising from an external macro-lobulation (3) whether the lesion was solitary or one of multiple cystic liver lesions. The optimal classification system had only four categories and correctly identified 144/154 lesion (93.5%). Conclusion This multi-center follow-up study was able to use machine learning to develop a highly accurate classification system for differentiation of hepatic MCN from BHC, which could be readily applied to clinical practice.

Background: Non-contrast chest CTs (NCCT) are performed routinely for coronary artery calcium (CAC) scoring and lung cancer screening. However, a large amount of noncoronary and nonpulmonary data from these scans remain unassessed. We aimed to validate a deep learning model to automatically segment and measure left atrial (LA) volumes from routine NCCT and evaluate prediction for cardiovascular outcomes. Methods: We retrospectively evaluated 273 patients (median age 69 years, 55.5% male) who underwent a routine non-ECG gated NCCT for lung cancer screening. LA volumes were quantified by three expert cardiothoracic radiologists and a prototype AI algorithm. LA volumes were then indexed to the body surface area (BSA). Expert and AI LA volume index (LAVi) were compared and used to predict cardiovascular outcomes within five years. Logistic regression with appropriate univariate statistics were used for modelling outcomes. Findings: There was excellent correlation between AI and expert results with an LAV intraclass correlation of 0.950 (0.936-0.960). Bland-Altman plot demonstrated the AI underestimated LAVi by a mean 5.86 mL/m 2 . AI-LAVi was associated with new-onset atrial fibrillation (AUC 0.86; OR 1.12, 95% CI 1.08-1.18, p

To evaluate the effectiveness of a novel artificial intelligence (AI) algorithm for fully automated measurement of left atrial (LA) volumes and function using cardiac CT in patients with atrial fibrillation.We included 79 patients (mean age 63 ± 12 years; 35 with atrial fibrillation (AF) and 44 controls) between 2017 and 2020 in this retrospective study. Images were analyzed by a trained AI algorithm and an expert radiologist. Left atrial volumes were obtained at cardiac end-systole, end-diastole, and pre-atrial contraction, which were then used to obtain LA function indices. Intraclass correlation coefficient (ICC) analysis of the LA volumes and function parameters was performed and receiver operating characteristic (ROC) curve analysis was used to compare the ability to detect AF patients.The AI was significantly faster than manual measurement of LA volumes (4 s vs 10.8 min, respectively). Agreement between the manual and automated methods was good to excellent overall, and there was stronger agreement in AF patients (all ICCs ≥ 0.877; p0.001) than controls (all ICCs ≥ 0.799; p0.001). The AI comparably estimated LA volumes in AF patients (all within 1.3 mL of the manual measurement), but overestimated volumes by clinically negligible amounts in controls (all by ≤ 4.2 mL). The AI's ability to distinguish AF patients from controls using the LA volume index was similar to the expert's (AUC 0.81 vs 0.82, respectively; p = 0.62).The novel AI algorithm efficiently performed fully automated multiphasic CT-based quantification of left atrial volume and function with similar accuracy as compared to manual quantification. Novel CT-based AI algorithm efficiently quantifies left atrial volumes and function with similar accuracy as manual quantification in controls and atrial fibrillation patients.• There was good-to-excellent agreement between manual and automated methods for left atrial volume quantification. • The AI comparably estimated LA volumes in AF patients, but overestimated volumes by clinically negligible amounts in controls. • The AI's ability to distinguish AF patients from controls was similar to the manual methods.

Vulnerabilities of the global fuel-fertilizer-food nexus have been revealed by a regional geopolitical conflict causing sudden and massive supply disruptions. Across over- and under-fertilized agricultural systems, nitrogen (N) fertilizer price spikes will have very different effects and require differentiated responses. For staple cereal production in India, Ethiopia, and Malawi, our estimates of N-fertilizer savings show the value of integrated organic and inorganic N management. N-deficient systems benefit from shifting to more cost-effective, high-N fertilizer (such as urea), combined with compost and legumes. N-surplus systems achieve N savings through better targeted and more efficient N-fertilizer use. Globally, there is a need to re-balance access to N-fertilizers, while steering the right fertilizer to the right place, and managing N in combination with carbon through near-term interventions, while striving for longer-term sustainable management. Nationally, governments can invest in extension and re-align subsidies to enable and incentivize improved N management at the farm level.

Crop yields across sub-Saharan Africa are much lower than what is attainable given the environmental conditions and available technologies. Closing this ‘ecological yield gap’ is considered an important food security and rural welfare goal. It is not clear, however, whether it is economically sensible for farmers to substantially increase crop yields. Here we estimate the local yield response of maize to fertilizer across sub-Saharan Africa with an empirical machine-learning model based on 12,081 trial observations and with a mechanistic model. We show that the average ‘economic yield gap’—the difference between current yield and profit-maximizing yield—is about one-quarter of the ecological yield gap. Furthermore, although maize yields could be profitably doubled, the economic incentives to do so may be weak. Our findings suggest that agricultural intensification in sub-Saharan Africa could be supported by complementary agronomic approaches to improve soil fertility, lowering the fertilizer cost, and by spatial targeting of fertilizer recommendations. To increase crop yields in many regions of sub-Saharan Africa, farmers would need to use more fertilizers. However, estimating the local maize yield responses to fertilizers shows that the fertilizer and maize price ratio presents a strong barrier to achieving higher yields in most regions.

This is the accepted manuscript version of the work published in its final form as Ricker-Gilbert, Jacob;Jayne, T. S.;Chamberlin, Jordan. JDevelopment in Practice, page(s): 1–19.https://www.tandfonline.com/10.1080/09614524.2021.1937567 Deposited byshareyourpaper.organdopenaccessbutton.org. We've taken reasonable steps to ensure this content doesn't violate copyright. However, if you think it does you can request a takedown by emailinghelp@openaccessbutton.org.

Background Evaluation for gastrointestinal leak is a frequent imaging indication, and dual-energy computed tomography (DECT) with oral or rectally administered contrast can be used to improve efficiency and diagnostic confidence. Purpose To assess the value of the DECT iodine overlay (IO) reconstruction as a stand-alone image set compared to routine CT in assessing oral or rectal contrast leak from the gastrointestinal system. Material and Methods A blinded, retrospective audit study was performed by three readers who each interpreted 50 studies performed for assessment of oral or rectal contrast leak that were acquired using DECT. Each reader independently assessed both the routine CT images and the images of the reconstructed IO for contrast leak in random order with a six-week “wash-out period” between readings. Clinical follow-up provided the reference standard. Readers recorded the presence/absence of a leak, diagnostic confidence, image quality score, and interpretation time for each image set. Results Pooled data for overall accuracy in identification of a leak increased from 0.81 (95% confidence interval [CI]=0.74–0.87) for routine CT to 0.91 (95% CI=0.85–0.95) with IO, and the area under the curve (AUC) was significantly higher for IO than routine CT ( P = 0.015). Readers required significantly less time to interpret IO than routine CT (median improvement of 12.5 s per image using pooled data; P Conclusion Use of DECT IO reconstructions for identification of oral or rectal contrast leak requires less time to interpret than routine CT with improved accuracy and maintained diagnostic confidence and perceived image quality.

The expansion of agro-dealers into remote areas can be seen as conducive to more smallholders adopting new technologies and inputs, to include improved seed and fertilizer. However, lower travel costs may be offset by agro-dealer decisions on stocking and pricing, reflecting both travel time from wholesale markets as well as the level of competition in localized areas. This paper investigates the geographical distribution of agro-dealers and related patterns of local market competition on the availability and prices of maize seed and fertilizer. We use a unique census of agro-dealers in eight districts of Tanzania (n = 299) which maps distribution points for agricultural inputs in these areas. Results suggested that despite a high number of agro-dealers, almost 30% of farmers lived more than an hour travel time from at least one agro-dealer. Instead of wide geographical coverage, agro-dealers tended to be found in clusters, with strong variation in cluster sizes between different districts. Overall, more remote agro-dealers faced less competition, resulting in fewer stocked product choices and charging higher prices to customers, even after controlling for travel time from district headquarters. Remote farmers are disadvantaged in their uptake of new technologies and critical production inputs due to lack of competition among agro-dealers. Our results suggest that highly aggregated and/or simplified measures of market access fail to reflect important heterogeneity in the market access conditions faced by farmers; a better understanding of distribution networks and competition is needed.

We use nationally representative survey data from Zambia to evaluate whether rural youth out-migration is associated with a lack of rural economic vibrancy, as often asserted in policy discussion. After controlling for distance from markets, we do find some evidence that areas with more productive and market-oriented agriculture are associated with lower rates of out-migration, particularly for those who leave explicitly to find work. However, measures of non-farm orientation of the local economy are positively associated with out-migration, indicating that the sectoral composition of the local economy has some bearing on the relationship between vibrancy and migration. When we examine more disaggregated patterns, we find that the out-migration decisions of young males are particularly responsive to rural vibrancy conditions. However, we find that the migration decisions of those who have completed secondary school are influenced less by vibrancy measures than the decisions of less educated individuals, possibly indicating that individuals who aspire to leave rural areas, and who invest in educational attainment to achieve such aspirations, may be less susceptible to influence by rural productivity changes. Our results suggest that the narrative of a stagnation-driven exodus of educated rural youth is oversimplified and, as such, is of limited policy relevance.

Credit markets are key instruments by which liquidity-constrained smallholder farmers may finance productive investments. However, the documented low demand and uptake of agricultural credit by smallholder farmers in sub-Saharan Africa pose challenges for energizing rural transformation in the region. In this paper, we investigate the impact of rainfall uncertainty—a major source of production risk—on the uptake of credit by rural farm households in Ethiopia. We further examine whether rainfall uncertainty explains credit rationing among those households not participating in rural credit markets. We find that rainfall variability discourages the uptake of agricultural credit. We also find that rainfall variability is associated with credit risk rationing, expressed as low demand for agricultural credit. We show that our findings are robust to alternative ways of constructing rainfall variability (inter-annual or inter-seasonal) and a battery of robustness checks. For instance, we show that rainfall variability is a strong predictor of credit uptake in rural areas while less relevant in urban areas. We also document heterogeneous responses to rainfall variability; those households living in the arid and semi-arid lands of Ethiopia, which are believed to be more vulnerable to recurrent weather shocks, are more responsive to rainfall variability in terms of reduced uptake of agricultural credit. Our results highlight the impacts of uninsured production risk on the demand for agricultural credit and hence smallholder agricultural investments. Our findings suggest the importance of interventions aimed at relaxing smallholders’ credit rationing while also reducing their production risk.

While much has been said in recent years about the importance of engaging rural youth in sub-Saharan Africa’s development, the factual data about how African youth currently engage in rural economi...

While a considerable body of literature has developed in recent years around the drivers and consequences of rural out‐migration in sub‐Saharan Africa, relatively little work has been done to understand the impacts of migration into rural areas. We use nationally representative household survey data from Zambia to explore the relationship between rural in‐migration and agricultural productivity outcomes in receiving communities. We document high levels of rural in‐migration throughout Zambia—12% of rural household heads having moved from elsewhere within the previous 10 years—with two‐thirds of rural in‐migrants originating from other rural areas. Migrants are, on average, better endowed with capital resources than their nonmigrant neighbors and are more engaged with input and output markets. After controlling for other factors, we find that higher rates of rural in‐migration are associated with greater agricultural productivity outcomes in receiving communities. These positive associations are particularly pronounced in more remote rural areas, and where in‐migration originates from other rural areas. Taken together, our results suggest that rural in‐migrants play an important role in the rural transformation processes underway in Zambia.

Ethiopia has achieved the second highest maize yield in sub-Saharan Africa. Yet, farmers’ maize yields are still much lower than on-farm and on-station trial yields, and only ca. 20% of the estimated water-limited potential yield. This article provides a comprehensive national level analysis of the drivers of maize yields in Ethiopia, by decomposing yield gaps into efficiency, resource and technology components, and accounting for a broad set of detailed input and crop management choices. Stochastic frontier analysis was combined with concepts of production ecology to estimate and explain technically efficient yields, the efficiency yield gap and the resource yield gap. The technology yield gap was estimated based on water-limited potential yields from the Global Yield Gap Atlas. The relative magnitudes of the efficiency, resource and technology yield gaps differed across farming systems; they ranged from 15% (1.6 t/ha) to 21% (1.9 t/ha), 12% (1.3 t/ha) to 25% (2.3 t/ha) and 54% (4.8 t/ha) to 73% (7.8 t/ha), respectively. Factors that reduce the efficiency yield gap include: income from non-farm sources, value of productive assets, education and plot distance from home. The resource yield gap can be explained by sub-optimal input use, from a yield perspective. The technology yield gap comprised the largest share of the total yield gap, partly due to limited use of fertilizer and improved seeds. We conclude that targeted but integrated policy design and implementation is required to narrow the overall maize yield gap and improve food security.

Introduction/Hypothesis: To develop and validate a deep learning model to automatically segment and measure left atrial (LA) volumes from routine noncontrast chest CTs and evaluate for outcomes prognosis. Methods: A prototype AI algorithm, with a conditional variational auto-encoder architecture, was initially developed and trained using training data from 1313 patients (paired CTA and non-contrasted exams; “training group”). Left atrium contours from contrasted cases were used to train the model on corresponding non-contrasted dataset. We then retrospectively evaluated 274 patients (mean age 67.9 years, 54.6% male, 57.2% caucasian) who underwent a routine non ECG-gated, noncontrast chest CT for coronary artery calcium or lung cancer screening for validation of the AI model (“validation group”). Outcomes included heart failure (HF) hospitalization and new-onset atrial fibrillation within 5 years. Simple logistic regressions with appropriate univariate statistics were used for modelling outcomes. Results: In the validation group, there was excellent correlation between AI and expert results with a LA volume index (LAVi) intraclass correlation coefficient of 0.950 (0.936-0.960). Bland-Altman plot demonstrated that the AI underestimated LAVi by a mean 5.86 mL/m2. AI-LAVi independently predicted new-onset atrial fibrillation (AUC = 0.855, OR = 1.12, 95% CI 1.08 - 1.18; p Conclusion: In this large validation cohort, AI for automated measurement of LA volume on routine noncontrast chest CTs had excellent agreement with manual quantification. AI-LAVi is independently predictive for new-onset atrial fibrillation and HF hospitalization within 5 years.

Introduction/Hypothesis: Myocardial native T2 quantification using cardiac magnetic resonance (CMR) may be helpful to diagnose active cardiac sarcoidosis and predict prognosis without the need for IV gadolinium. Methods: At a large academic institution, 66 consecutive patients with biopsy-proven pulmonary sarcoidosis underwent CMR for evaluation of active cardiac sarcoidosis (CS) from January 2017 to January 2020. Of these, 56 patients were identified who fit inclusion criteria. Active clinical CS was defined according to the Japanese Ministry of Health and Welfare criteria for diagnosis of CS. Quantitative T2 mapping of all AHA myocardial segments was evaluated and incident heart failure (HF) was recorded. A power calculation was performed beforehand which suggested n > 50 was adequate. Results: The best fit model to identify active clinical CS was the mean T2 value of the basal segments (AUC 0.902, 95% CI 0.821-0.983; see Figure 1). Mean basal T2 >51 ms was the optimal cutoff with accuracy of 91.1%, sensitivity of 71.4%, and specificity of 97.6%. T2 >49 ms was the most sensitive cutoff (92.9%) with a negative predictive value of 96.2%. The best fit model to predict incident HF was the average T2 value of the basal segments (AUC 0.830, 95% 0.702-0.959). For every 1 ms increase in T2 value, there was a 41% increased risk for incident HF (odds ratio 1.41 95% CI 1.12-1.92, p=0.013). Conclusion: Measurement of mean basal myocardial T2 values efficiently identified patients with active clinical cardiac sarcoidosis and predicted incident heart failure. Quantitative T2 mapping is highly sensitive and specific for the diagnosis.

Introduction/Hypothesis: Transcatheter aortic valve replacement (TAVR) requires pre-surgical planning via CTA. Our purpose is to compare pre-TAVR CTA expert measurements versus an AI in a pilot study. Methods: An AI model was created using Unett++ with a Resnet18 encoder, and was trained using transfer learning. A total of 52 randomized patients with a preTAVR CTA were identified for AI validation. The average diameter from 3 aortic segments was measured by AI: annulus, sinuses of Valsalva, and sinotubular junction. Of the original 52 subjects, 39 patients had expert measurements for comparison with AI results. Standard statistical analysis was performed. Results: Patients were 65% male and 88% Caucasian. The median age and BMI were 82 years and 28 kg/m2, respectively. A Sorensen-Dice coefficient of 0.93 was attained between the expert and AI measurements. Annulus (Pearson r = 0.911), sinuses of Valsalva (r = 0.919), and sinotubular junction (r = 0.874) average diameter measurements correlated with the experts’ (p Conclusion: In this validation study, AI measurements of the aortic annulus, sinuses of Valsalva, and sinotubular junction correlate with expert measurements. There was high inter-relater reliability for all sites. Further study will include expanded patient cohorts and accurately streamlined preoperative assessment for TAVR. Figure. Pre TAVR CT angiogram images in orthogonal projection for aortic annulus (A), sinuses of Valsalva (B), and sinotubular junction (C). “Ground Truth” is expert segmentation and “Threshold 0.5” is AI segmentation.

Introduction: This study was performed to evaluate a novel AI algorithm for fully automated measurement of LA volumes and function using cardiac CTA in patients with atrial fibrillation (AF). Methods: A prototype AI algorithm (AI-LAV), with a conditional variational auto-encoder architecture, was developed and trained using 555 manually annotated ECG-gated, contrast-enhanced cardiac CTs. A total of 79 patients (mean age 63 ± 12 years; 35 with AF and 44 controls) evaluated between 2017 and 2020 were included in this retrospective study to validate the AI model. Images were analyzed by the trained AI algorithm and experienced radiologists. Left atrial volumes were obtained at cardiac end-systole (LAVmax), end-diastole (LAVmin), and pre-atrial contraction (LAVpc), which were then used to obtain LA function indices (total, passive, and active emptying fractions). Statistical analysis was performed. Results: AI was significantly faster than manual measurement of LA volumes and function (4 s vs 10.8 min, respectively). Agreement between the manual and automated methods was good to excellent overall, and there was stronger agreement in AF patients (all ICCs ≥ 0.877; p Conclusion: In a validation cohort, the novel AI algorithm efficiently performed fully automated multiphasic CTA-based quantification of left atrial volume and function with similar accuracy to expert manual quantification. Figure. Scatter and Bland-Altman plots for LAVmax and LAVmin. Segmentation comparison between expert (manual) and AI on cardiac CTA.

Highlights • We examine inter- and intra-plot heterogeneity of productivity in maize plots. • We find high inter- and intra-plot variabilities in plant population and yield. • Intra-plot heterogeneity overrides inter-plot heterogeneity in smallholder maize. • The magnitude and distribution of yield estimation errors are method-dependent. • More intensively managed plots show lower intra-plot heterogeneity and lower yield., Intra-plot heterogeneity in yield is often observed in smallholder farming systems, although its implications for yield measurement remain under-investigated. Using a unique dataset on smallholder maize production in Ethiopia, we quantify the magnitude of inter- and intra-plot heterogeneity, describe the relationship between intra-plot heterogeneity and maize productivity, and document the implications of intra-field heterogeneity on the accuracy of alternative yield estimation protocols. Our data include five common yield estimation protocols, as well as full plot harvests of 230 smallholder maize fields. We surveyed agronomic decisions, biophysical variables, and accessibility characteristics of the surveyed fields. We quantify intra-plot heterogeneity using the coefficient of variation (CV) of stand density, cob weight, and maize grain yield. A generalized linear mixed model is used to explore the relationship between these variables and the method- and heterogeneity-dependence of yield estimation accuracy. We find inter-plot CV values ranging from 32 to 56 %, 22 to 73 % and 39 to 49 % in population density, cob weight and grain yield, respectively. Intra-plot heterogeneity constituted most of this variation, with across-method mean CV values of 41 %, 82 % and 63 %, respectively, of the total variability in population density, cob weight and grain yield. A rise in intra-plot heterogeneity of 0.5 % to 0.8 % is associated with a significant increase in yield estimation error under alternative yield estimation protocols. Regression analysis shows that interactions in agronomic decisions, input intensity and plot accessibility factors dictate intra-plot heterogeneity and method accuracy in smallholder systems. Intra-plot heterogeneity is larger than inter-plot heterogeneity in the current study area. Our analysis shows that the effect of intra-plot heterogeneity on yield estimation accuracies is method-dependent and yield estimation methods that fail to capture true intra-plot heterogeneity are more error-prone. Results of such estimations should be considered with caution when used as the basis of decision-making.

ispartof: AMERICAN JOURNAL OF AGRICULTURAL ECONOMICS vol:104 issue:2 pages:831-852 status: published

Rural economic transformations in Africa are generating new opportunities to engage with agricultural value chains. However, many young people are said to be locked out of such opportunities because of limited access to farmland, which pushes them out of agriculture and rural areas, and/or hinders their autonomy. This framing of the ‘land problem’ imperfectly reflects rural young people’s livelihoods in much of sub-Saharan Africa, and therefore does not provide a solid basis for policy. Policy-relevant discussions must consider the diversity of rural contexts, broader land dynamics and more nuanced depictions of youth engagement with the rural economy.

We estimate the efficiency and equity returns to farmland rental markets in Malawi using a matched tenant–landlord survey of smallholder farm households in four districts. Our sample allows us to more fully observe the landlord side of the rental market, which is almost always missing in previous studies. Our results suggest that land rental markets promote efficiency by facilitating a net transfer of land to more productive farmers. We also find that land rental markets promote equity as conventionally defined in the land markets literature, that is, by transferring land from land‐rich households to land‐poor households, and from labor‐poor to labor‐rich households. However, our study identifies some important challenges for land rental markets in this context. First, we find that tenants in our sample are wealthier than their landlord counterpart on average in all dimensions other than landholding. In addition, most landlords report the motive for renting out their land as either the need for immediate cash, or the lack of labor and/or capital to cultivate the plot that was rented out. These findings align with concerns about potential “stress renting” by poor landlords and suggest the value of defining equity along a broader set of dimensions other than simply equalizing the distribution of farmland and labor.

Agricultural extension to improve yields of staple food crops and close the yield gap in Sub-Saharan Africa often entails general recommendations on soil fertility management that are distributed to farmers in a large growing area. Site-specific extension recommendations that are better tailored to the needs of individual farmers and fields, and enabled by digital technologies, could potentially bring about yield and productivity improvements. In this paper, we analyze farmers' preferences for high-input maize production supported by site-specific nutrient management recommendations provided by an ICT-based extension tool that is being developed for extension services in the maize belt of Nigeria. We use a choice experiment to provide ex-ante insights on the adoption potentials of site-specific extension services from the perspective of farmers. We control for attribute non-attendance and account for class as well as scale heterogeneity in preferences using different models, and find robust results. We find that farmers have strong preferences to switch from general to ICT-enabled site-specific soil fertility management recommendations which lend credence to the inclusion of digital technologies in agricultural extension. We find heterogeneity in preferences that is correlated with farmers' resource endowments and access to services. A first group of farmers are strong potential adopters; they are better-off, less sensitive to risk, and are more willing to invest in a high-input maize production system. A second group of farmers are weak potential adopters; they have lower incomes and fewer productive assets, are more sensitive to yield variability, and prefer less capital and labor intensive production techniques. Our empirical findings imply that improving the design of extension tools to enable provision of information on the riskiness of expected outcomes and flexibility in switching between low-risk and high-risk recommendations will help farmers to make better informed decisions, and thereby improve the uptake of extension advice and the efficiency of extension programs., Highlights • Choice experiement on farmer preferences for site-specific extension recommendations • High willingness to accept ICT-enabled and site-specific extension recommendations • Observed heterogeneity in farmers' willingness to intensify production systems • Information on riskiness of return crucial for improved design of extension tools

Smallholder farmers in Africa typically only have access to blanket fertilizer recommendations which are defined over very broad areas and may not be optimal for local production conditions. The response to such recommendations has generally been poor. Using a randomized control trial in Ethiopia, we explore whether targeted extension advice leads farmers to align fertilizer usage to the recommended levels and whether this impacts productivity. We also consider whether coupling the targeted information with agricultural insurance encourages fertilizer investment. Results show that targeted recommendations closed the gap between the amount of fertilizer used and the recommended amounts and this in turn increased productivity and profits. We found no differential effect of the targeted recommendation when coupled with agricultural insurance, suggesting that the risk of crop failure is not a binding constraint to fertilizer adoption in this context, or that farmers do not consider agricultural insurance a useful risk-mitigating mechanism.

Soil property and class maps for the continent of Africa were so far only available at very generalised scales, with many countries not mapped at all. Thanks to an increasing quantity and availability of soil samples collected at field point locations by various government and/or NGO funded projects, it is now possible to produce detailed pan-African maps of soil nutrients, including micro-nutrients at fine spatial resolutions. In this paper we describe production of a 30 m resolution Soil Information System of the African continent using, to date, the most comprehensive compilation of soil samples ($$N \approx 150,000$$ N ≈ 150 , 000 ) and Earth Observation data. We produced predictions for soil pH, organic carbon (C) and total nitrogen (N), total carbon, effective Cation Exchange Capacity (eCEC), extractable—phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), sulfur (S), sodium (Na), iron (Fe), zinc (Zn)—silt, clay and sand, stone content, bulk density and depth to bedrock, at three depths (0, 20 and 50 cm) and using 2-scale 3D Ensemble Machine Learning framework implemented in the (Machine Learning in ) package. As covariate layers we used 250 m resolution (MODIS, PROBA-V and SM2RAIN products), and 30 m resolution (Sentinel-2, Landsat and DTM derivatives) images. Our fivefold spatial Cross-Validation results showed varying accuracy levels ranging from the best performing soil pH (CCC = 0.900) to more poorly predictable extractable phosphorus (CCC = 0.654) and sulphur (CCC = 0.708) and depth to bedrock. Sentinel-2 bands SWIR (B11, B12), NIR (B09, B8A), Landsat SWIR bands, and vertical depth derived from 30 m resolution DTM, were the overall most important 30 m resolution covariates. Climatic data images—SM2RAIN, bioclimatic variables and MODIS Land Surface Temperature—however, remained as the overall most important variables for predicting soil chemical variables at continental scale. This publicly available 30-m Soil Information System of Africa aims at supporting numerous applications, including soil and fertilizer policies and investments, agronomic advice to close yield gaps, environmental programs, or targeting of nutrition interventions.