Agronomic gain: Definition, approach, and application

Random forest machine learning for maize yield and agronomic efficiency prediction in Ghana

Maize (Zea mays) is an important staple crop for food security in Sub-Saharan Africa. However, there is need to increase production to feed a growing population. In Ghana, this is mainly done by increasing acreage with adverse environmental consequences, rather than yield increment per unit area. Accurate prediction of maize yields and nutrient use efficiency in production is critical to making informed decisions toward economic and ecological sustainability. We trained the random forest machine learning algorithm to predict maize yield and agronomic efficiency in Ghana using soil, climate, environment, and management factors, including fertilizer application. We calibrated and evaluated the performance of the random forest machine learning algorithm using a 5 × 10-fold nested cross-validation approach. Data from 482 maize field trials consisting of 3136 georeferenced treatment plots conducted in Ghana from 1991 to 2020 were used to train the algorithm, identify important predictor variables, and quantify the uncertainties associated with the random forest predictions. The mean error, root mean squared error, model efficiency coefficient and 90 % prediction interval coverage probability were calculated. The results obtained on test data demonstrate good prediction performance for yield (MEC = 0.81) and moderate performance for agronomic efficiency (MEC = 0.63, 0.55 and 0.54 for AE-N, AE-P and AE-K, respectively). We found that climatic variables were less important predictors than soil variables for yield prediction, but temperature was of key importance to yield prediction and rainfall to agronomic efficiency. The developed random forest models provided a better understanding of the drivers of maize yield and agronomic efficiency in a tropical climate and an insight towards improving fertilizer recommendations for sustainable maize production and food security in Sub-Saharan Africa.

Filling the agronomic data gap through a minimum data collection approach

Context

Agronomic data such as applied inputs, management practices, and crop yields are needed for assessing productivity, nutrient balances, resource use efficiency, as well as other aspects of environmental and economic performance of cropping systems. In many instances, however, these data are only available at a coarse level of aggregation or simply do not exist.

Objectives

Here we developed an approach that identifies sites for agronomic data collection for a given crop and country, seeking a balance between minimizing data collection efforts and proper representation of the main crop producing areas.

Methods

The developed approach followed a stratified sampling method based on a spatial framework that delineates major climate zones and crop area distribution maps, which guides selection of sampling areas (SA) until half of the national harvested area is covered. We provided proof of concept about the robustness of the approach using three rich databases including data on fertilizer application rates for maize, wheat, and soybean in Argentina, soybean in the USA, and maize in Kenya, which were collected via local experts (Argentina) and field surveys (USA and Kenya). For validation purposes, fertilizer rates per crop and nutrient derived at (sub-) national level following our approach were compared against those derived using all data collected from the whole country.

Results

Application of the approach in Argentina, USA, and Kenya resulted in selection of 12, 28, and 10 SAs, respectively. For each SA, three experts or 20 fields were sufficient to give a robust estimate of average fertilizer rates applied by farmers. Average rates at national level derived from our approach compared well with those derived using the whole database ( ± 10 kg N, ± 2 kg P, ± 1 kg S, and ± 5 kg K per ha) requiring less than one third of the observations.

Conclusions

The developed minimum crop data collection approach can fill the agronomic data gaps in a cost-effective way for major crop systems both in large- and small-scale systems.

Significance

The proposed approach is generic enough to be applied to any crop-country combination to guide collection of key agricultural data at national and subnational levels with modest investment especially for countries that do not currently collect data.

Special Issue "State-of-the-Art Molecular Plant Sciences in Japan"

Food shortages are one of the most serious problems caused by global warming and population growth in this century [...].

Evaluating responses by ChatGPT to farmers' questions on irrigated lowland rice cultivation in Nigeria

The limited number of agricultural extension agents (EAs) in sub-Saharan Africa limits farmers' access to extension services. Artificial intelligence (AI) assistants could potentially aid in providing answers to farmers' questions. The objective of this study was to evaluate the ability of an AI chatbot assistant (ChatGPT) to provide quality responses to farmers' questions. We compiled a list of 32 questions related to irrigated rice cultivation from farmers in Kano State, Nigeria. Six EAs from the state were randomly selected to answer these questions. Their answers, along with those of ChatGPT, were assessed by four evaluators in terms of quality and local relevancy. Overall, chatbot responses were rated significantly higher quality than EAs' responses. Chatbot responses received the best score nearly six times as often as the EAs' (40% vs. 7%). The evaluators preferred chatbot responses to EAs in 78% of cases. The topics for which the chatbot responses received poorer scores than those by EAs included planting time, seed rate, and fertilizer application rate and timing. In conclusion, while the chatbot could offer an alternative source for providing agricultural advisory services to farmers, incorporating site-specific input rate-and-timing agronomic practices into AI assistants is critical for their direct use by farmers.

Deep Learning Enables Instant and Versatile Estimation of Rice Yield Using Ground-Based RGB Images

Rice (Oryza sativa L.) is one of the most important cereals, which provides 20% of the world's food energy. However, its productivity is poorly assessed especially in the global South. Here, we provide a first study to perform a deep-learning-based approach for instantaneously estimating rice yield using red-green-blue images. During ripening stage and at harvest, over 22,000 digital images were captured vertically downward over the rice canopy from a distance of 0.8 to 0.9 m at 4,820 harvesting plots having the yield of 0.1 to 16.1 t·ha-1 across 6 countries in Africa and Japan. A convolutional neural network applied to these data at harvest predicted 68% variation in yield with a relative root mean square error of 0.22. The developed model successfully detected genotypic difference and impact of agronomic interventions on yield in the independent dataset. The model also demonstrated robustness against the images acquired at different shooting angles up to 30° from right angle, diverse light environments, and shooting date during late ripening stage. Even when the resolution of images was reduced (from 0.2 to 3.2 cm·pixel-1 of ground sampling distance), the model could predict 57% variation in yield, implying that this approach can be scaled by the use of unmanned aerial vehicles. Our work offers low-cost, hands-on, and rapid approach for high-throughput phenotyping and can lead to impact assessment of productivity-enhancing interventions, detection of fields where these are needed to sustainably increase crop production, and yield forecast at several weeks before harvesting.

Elevated CO2 Priming as a Sustainable Approach to Increasing Rice Tiller Number and Yield Potential

Tillering and yield are linked in rice, with significant efforts being invested to understand the genetic basis of this phenomenon. However, in addition to genetic factors, tillering is also influenced by the environment. Exploiting experiments in which seedlings were first grown in elevated CO2 (eCO2) before transfer and further growth under ambient CO2 (aCO2) levels, we found that even moderate exposure times to eCO2 were sufficient to induce tillering in seedlings, which was maintained in plants grown to maturity plants in controlled environment chambers. We then explored whether brief exposure to eCO2 (eCO2 priming) could be implemented to regulate tiller number and yield in the field. We designed a cost-effective growth system, using yeast to increase the CO2 level for the first 24 days of growth, and grew these seedlings to maturity in semi-field conditions in Malaysia. The increased growth caused by eCO2 priming translated into larger mature plants with increased tillering, panicle number, and improved grain filling and 1000 grain weight. In order to make the process more appealing to conventional rice farmers, we then developed a system in which fungal mycelium was used to generate the eCO2 via respiration of sugars derived by growing the fungus on lignocellulosic waste. Not only does this provide a sustainable source of CO2, it also has the added financial benefit to farmers of generating economically valuable oyster mushrooms as an end-product of mycelium growth. Our experiments show that the system is capable of generating sufficient CO2 to induce increased tillering in rice seedlings, leading eventually to 18% more tillers and panicles in mature paddy-grown crop. We discuss the potential of eCO2 priming as a rapidly implementable, broadly applicable and sustainable system to increase tillering, and thus yield potential in rice.

Assessing genetic and agronomic gains in rice yield in sub-Saharan Africa: A meta-analysis

Research for development efforts for increasing rice yield in sub-Saharan Africa (SSA) have largely concentrated on genetic improvement and agronomy for more than 50 years. Here we perform the first meta-analysis to quantify genetic gain - yield increase through use of new variety and calculated by yield difference between new variety and variety popularly grown in the target site, and agronomic gain - difference in yield between improved agronomic practices and the control in SSA using 208 paired observations from 40 studies across 12 countries. Among the studies, 41 %, 34 %, and 25 % were from irrigated lowland, rainfed lowland, and rainfed upland rice, respectively. Seventy percent of the studies reported in this paper were conducted on research stations. In agronomic practices, inorganic fertilizer management practices accounted for 78 % of the studies, of which 48 % were nitrogen (N) management. In each study, we identified four types of varieties: check variety (VC), variety with highest yield in the control (VHC), variety with highest yield under improved agronomic practices (VHT), and variety with largest yield difference between improved agronomic practices and control (VHR). VHT was the same as VHC in 35 % of observations, whereas VHR and VHT were the same in 51 %. These indicate that it is possible to develop varieties adapted to different agronomic practices and high-yielding varieties tend to be responsive to improved agronomic practices. On average, total gain in yield with improved agronomic practices and VHT was 1.6 t/ha. Agronomic practice accounted for 75 % of the total variation in total yield gain with variety and agronomic practice by variety interaction responsible for 19 % and 6 %, respectively. Genetic gains in yield with VHC, VHT, and VHR were 0.7, 0.3, and -0.3 t/ha in control, and 0.4, 0.9, and 0.5 t/ha in improved agronomic practices. Agronomic gain in yield averaged 0.5, 0.8, 1.4, and 1.6 t/ha in VHC, VC, VHT, and VHR, respectively. Agronomic gain in yield of VHT was higher than genetic gain under improved agronomic practices in 54 % of observations. Agronomic gain was highest in irrigated lowland rice, followed by rainfed lowland rice. Higher agronomic gain in yield was also associated with larger difference in N application rate between improved agronomic practices and control. Whereas agronomic practices had larger contribution to total gain in yield than genetic improvement in this study, future assessment of agronomic and genetic gains in yield is warranted. Such assessment should focus more on rainfed rice systems, where agronomic gain was small, take into account genetic improvement rate over time and integrated agronomic practices rather than single intervention like nutrient management practice only, and be conducted in farmers' fields.

Impact of cropping system diversification on productivity and resource use efficiencies of smallholder farmers in south-central Bangladesh: a multi-criteria analysis

Diversification of smallholder rice-based cropping systems has the potential to increase cropping system intensity and boost food security. However, impacts on resource use efficiencies (e.g., nutrients, energy, and labor) remain poorly understood, highlighting the need to quantify synergies and trade-offs among different sustainability indicators under on-farm conditions. In southern coastal Bangladesh, aman season rice is characterized by low inputs and low productivity. We evaluated the farm-level impacts of cropping system intensification (adding irrigated boro season rice) and diversification (adding chili, groundnut, mungbean, or lathyrus) on seven performance indicators (rice equivalent yield, energy efficiency, partial nitrogen productivity, partial potassium productivity, partial greenhouse gas footprint, benefit-cost ratio, and hired labor energy productivity) based on a comprehensive survey of 501 households. Indicators were combined into a multi-criteria performance index, and their scope for improvement was calculated by comparing an individual farmer's performance to top-performing farmers (highest 20%). Results indicate that the baseline system (single-crop aman season rice) was the least productive, while double cropped systems increased rice equivalent yield 72-217%. Despite gains in productivity, higher cropping intensity reduced resource use efficiencies due to higher inputs of fertilizer and energy, which also increased production costs, particularly for boro season rice. However, trade-offs were smaller for diversified systems including legumes, largely owing to lower N fertilizer inputs. Aman season rice had the highest multi-criteria performance index, followed by systems with mungbean and lathyrus, indicating the latter are promising options to boost food production and profitability without compromising sustainability. Large gaps between individual and top-performing farmers existed for each indicator, suggesting significant scope for improvement. By targeting indicators contributing most to the multi-criteria performance index (partial nitrogen productivity, energy efficiency, hired labor energy productivity), results suggest further sustainability gains can be achieved through future field research studies focused on optimizing management within diversified systems.

SUPPLEMENTARY INFORMATION

The online version contains supplementary material available at 10.1007/s13593-022-00795-3.

Progress in research on site-specific nutrient management for smallholder farmers in sub-Saharan Africa

Increasing fertilizer access and use is an essential component for improving crop production and food security in sub-Saharan Africa (SSA). However, given the heterogeneous nature of smallholder farms, fertilizer application needs to be tailored to specific farming conditions to increase yield, profitability, and nutrient use efficiency. The site-specific nutrient management (SSNM) approach initially developed in the 1990 s for generating field-specific fertilizer recommendations for rice in Asia, has also been introduced to rice, maize and cassava cropping systems in SSA. The SSNM approach has been shown to increase yield, profitability, and nutrient use efficiency. Yield gains of rice and maize with SSNM in SSA were on average 24% and 69% when compared to the farmer practice, respectively, or 11% and 4% when compared to local blanket fertilizer recommendations. However, there is need for more extensive field evaluation to quantify the broader benefits of the SSNM approach in diverse farming systems and environments. Especially for rice, the SSNM approach should be expanded to rainfed systems, which are dominant in SSA and further developed to take into account soil texture and soil water availability. Digital decision support tools such as RiceAdvice and Nutrient Expert can enable wider dissemination of locally relevant SSNM recommendations to reach large numbers of farmers at scale. One of the major limitations of the currently available SSNM decision support tools is the requirement of acquiring a significant amount of farm-specific information needed to formulate SSNM recommendations. The scaling potential of SSNM will be greatly enhanced by integration with other agronomic advisory platforms and seamless integration of digital soil, climate and crop information to improve predictions of SSNM recommendations with reduced need for on-farm data collection. Uncertainty should also be included in future solutions, primarily to also better account for varying prices and economic outcomes.

Assessing rice production sustainability performance indicators and their gaps in twelve sub-Saharan African countries

The benchmarking and monitoring of rice production performance indicators are essential for improving rice production self-sufficiency, increasing profitability, reducing labor requirements, optimizing fertilizer inputs, engaging youths in rice production, and increasing the overall sustainability of smallholder rice production systems in countries in sub-Saharan Africa (SSA). In this paper, we quantified five sustainability performance indicators (grain yield, net profit, labor productivity, and nitrogen (N) and phosphorus (P) use efficiencies) to benchmark rice production systems in SSA. Data were collected between 2013-2014 from 2907 farmers from two rice production systems (irrigated and rainfed lowlands) across five agroecological zones (arid, semiarid, humid, subhumid and highlands) in 12 countries (Benin, Cameroon, Cote d'Ivoire, Ghana, Madagascar, Mali, Niger, Nigeria, Senegal, Sierra Leone, Tanzania and Togo). The exploitable gap for each indicator (the difference between the mean of 10 % highest-yielding farms and the mean-yielding farms) was calculated across the countries, the two production systems and agroecological zones. The mean yield varied widely between 2.5 to 5.6 t ha^-1 and 0.6 to 2.3 t ha^-1 in irrigated and rainfed lowlands, respectively, with an average yield of 4.1 and 1.4 t ha^-1, respectively. Across the country-production system combinations, there were yield gaps of 29-69 %, profit gaps of 10-89 %, and labor productivity gaps reaching 71 %. Yield, profit, and labor productivity were positively correlated. They were also positively correlated with N and P fertilizer application rate, but not with N and P use efficiencies. Only between 34-44 % of farmers had desirable ranges in N- or P-use efficiencies in the two production systems. All sites for rainfed lowlands were characterized by low-yield and large gaps in yield, profit, and labor productivity, whereas irrigated lowlands in some countries (Madagascar, Mali, and Togo) have similar characteristics as rainfed ones. We conclude that there is an urgent need to disseminate precision nutrient management practices for optimizing nutrient use efficiency and enhancing rice performance indicators especially in rainfed lowlands as well as low-yielding irrigated lowlands. Furthermore, we propose recommendations for specific categories (i.e. farmer, rice production system, agroecological zone and country) to close performance indicator gaps and to allow the production at scale to achieve rice self-sufficiency in SSA.