Big data, small explanatory and predictive power: Lessons from random forest modeling of on-farm yield variability and implications for data-driven agronomy.

Context: Collection and analysis of large volumes of on-farm production data are widely seen as key to understanding yield variability among farmers and improving resource-use efficiency. Objective: The aim of this study was to assess the performance of statistical and machine learning methods to explain and predict crop yield across thousands of farmers' fields in contrasting farming systems worldwide. Methods: A large database of 10,940 field-year combinations from three countries in different stages of agricultural intensification was analyzed. Random effects models were used to partition crop yield variability and random forest models were used to explain and predict crop yield within a cross-validation scheme with data re-sampling over space and time. Results: Yield variability in relative terms was smallest for wheat and barley in the Netherlands and for wheat in Ethiopia, intermediate for rice in the Philippines, and greatest for maize in Ethiopia. Random forest models comprising a total of 87 variables explained a maximum of 65 % of cereal yield variability in the Netherlands and less than 45 % of cereal yield variability in Ethiopia and in the Philippines. Crop management related variables were important to explain and predict cereal yields in Ethiopia, while predictive (i.e., known before the growing season) climatic variables and explanatory (i.e., known during or after the growing season) climatic variables were most important to explain and predict cereal yield variability in the Philippines and in the Netherlands, respectively. Finally, model cross-validation for regions or years not seen during model training reduced the R2 considerably for most crop x country combinations, while for wheat in the Netherlands this was model dependent. Conclusion: Big data from farmers' fields is useful to explain on-farm yield variability to some extent, but not to predict it across time and space. Significance: The results call for moderate expectations towards big data and machine learning in agronomic studies, particularly for smallholder farms in the tropics where model performance was poorest independently of the variables considered and the cross-validation scheme used.

Rice quality: How is it defined by consumers, industry, food scientists, and geneticists?

BACKGROUND: Quality is a powerful engine in rice value chain upgrading. However, there is no consensus on how "rice quality" should be defined and measured in the rice sector. SCOPE AND APPROACH: We adopt a Lancasterian definition of rice quality as a bundle of intrinsic and extrinsic attributes. We then review how rice quality is (i) perceived and defined by consumers and industry stakeholders in rice value chains in Southeast and South Asia; (ii) measured and defined by food technologists; and (iii) predicted through genetics. KEY FINDINGS AND CONCLUSIONS: Consumers are heterogeneous with respect to their perceived differentiation of rice quality among regions, countries, cities, and urbanization levels. Premium quality is defined by nutritional benefits, softness and aroma in Southeast Asia, and by the physical appearance of the grains (uniformity, whiteness, slenderness), satiety, and aroma in South Asia. These trends are found to be consistent with industry perceptions and have important implications for regional and national breeding programs in terms of tailoring germplasm to regions and rice varieties to specific local market segments. Because rice is traded internationally, there is a need to standardize definitions of rice quality. However, food technologists have not reached unanimity on quality classes and measurement; routine indicators need to be complemented by descriptive profiles elicited through sensory evaluation panels. Finally, because rice quality is controlled by multiple interacting genes expressed through environmental conditions, predicting grain quality requires associating genetic information with grain quality phenotypes in different environments.

Helping feed the world with rice innovations: CGIAR research adoption and socioeconomic impact on farmers.

Rice production has increased significantly with the efforts of international research centers and national governments in the past five decades. Nonetheless, productivity improvement still needs to accelerate in the coming years to feed the growing population that depends on rice for calories and nutrients. This challenge is compounded by the increasing scarcity of natural resources such as water and farmland. This article reviews 17 ex-post impact assessment studies published from 2016 to 2021 on rice varieties, agronomic practices, institutional arrangements, information and communication technologies, and post-harvest technologies used by rice farmers. From the review of these selected studies, we found that stress-tolerant varieties in Asia and Africa significantly increased rice yield and income. Additionally, institutional innovations, training, and natural resource management practices, such as direct-seeded rice, rodent control, and iron-toxicity removal, have had a considerable positive effect on smallholder rice farmers' economic well-being (income and rice yield). Additional positive impacts are expected from the important uptake of stress-tolerant varieties documented in several Asian, Latin American, and African countries.

Southeast Asia must narrow down the yield gap to continue to be a major rice bowl.

Southeast Asia is a major rice-producing region with a high level of internal consumption and accounting for 40% of global rice exports. Limited land resources, climate change and yield stagnation during recent years have once again raised concerns about the capacity of the region to remain as a large net exporter. Here we use a modelling approach to map rice yield gaps and assess production potential and net exports by 2040. We find that the average yield gap represents 48% of the yield potential estimate for the region, but there are substantial differences among countries. Exploitable yield gaps are relatively large in Cambodia, Myanmar, Philippines and Thailand but comparably smaller in Indonesia and Vietnam. Continuation of current yield trends will not allow Indonesia and Philippines to meet their domestic rice demand. In contrast, closing the exploitable yield gap by half would drastically reduce the need for rice imports with an aggregated annual rice surplus of 54 million tons available for export. Our study provides insights for increasing regional production on existing cropland by narrowing existing yield gaps.

RiceAtlas, a spatial database of global rice calendars and production.

Knowing where, when, and how much rice is planted and harvested is crucial information for understanding the effects of policy, trade, and global and technological change on food security. We developed RiceAtlas, a spatial database on the seasonal distribution of the world's rice production. It consists of data on rice planting and harvesting dates by growing season and estimates of monthly production for all rice-producing countries. Sources used for planting and harvesting dates include global and regional databases, national publications, online reports, and expert knowledge. Monthly production data were estimated based on annual or seasonal production statistics, and planting and harvesting dates. RiceAtlas has 2,725 spatial units. Compared with available global crop calendars, RiceAtlas is nearly ten times more spatially detailed and has nearly seven times more spatial units, with at least two seasons of calendar data, making RiceAtlas the most comprehensive and detailed spatial database on rice calendar and production.

Farmers' Preference for Rice Traits: Insights from Farm Surveys in Central Luzon, Philippines, 1966-2012.

Many modern rice varieties (MVs) have been released but only a few have been widely adopted by farmers. To understand farmers' preferences, we characterized MVs released in the Philippines from 1966 to 2013 and identified important characteristics of the varieties that were widely adopted in Central Luzon using farm surveys conducted in 1966-2012. We found that farmers adopt MVs that are high yielding, mature faster, and have long and slender grains, high milling recovery, and intermediate amylose content. The amylose content of adopted varieties has been declining, suggesting value in developing softer rice. To have a high potential for adoption, new MVs should have characteristics within the ranges of values observed for the adopted MVs. In addition, new MVs should have higher head rice recovery, less chalky grains, and better resistance to pests and diseases. Most MVs released in 2005-2013 compared poorly in these three traits. To reduce the risk of severe outbreaks, broad spectrum resistance should be incorporated into new MVs. This analysis of five decades of farm surveys provides insights into the varietal characteristics preferred by farmers which could contribute to the establishment of a product profile for developing improved MVs that are more targeted and, hence, would have high potential for adoption by farmers in Central Luzon and similar areas. We recommend a similar analysis be done in other major rice growing regions to aid the development of MVs that are more responsive to farmers' needs and preferences.

Spectral signature generalization and expansion can improve the accuracy of satellite image classification.

Conventional supervised classification of satellite images uses a single multi-band image and coincident ground observations to construct spectral signatures of land cover classes. We compared this approach with three alternatives that derive signatures from multiple images and time periods: (1) signature generalization: spectral signatures are derived from multiple images within one season, but perhaps from different years; (2) signature expansion: spectral signatures are created with data from images acquired during different seasons of the same year; and (3) combinations of expansion and generalization. Using data for northern Laos, we assessed the quality of these different signatures to (a) classify the images used to derive the signature, and (b) for use in temporal signature extension, i.e., applying a signature obtained from data of one or several years to images from other years. When applying signatures to the images they were derived from, signature expansion improved accuracy relative to the conventional method, and variability in accuracy declined markedly. In contrast, signature generalization did not improve classification. When applying signatures to images of other years (temporal extension), the conventional method, using a signature derived from a single image, resulted in very low classification accuracy. Signature expansion also performed poorly but multi-year signature generalization performed much better and this appears to be a promising approach in the temporal extension of spectral signatures for satellite image classification.