Evidence for Resilient Agriculture in Latin America

Funded by IDRC's project 'Agroecology and the resilience of small-scale farmers to climate change: Evidence to transform food systems in the Dry Corridor of Central America', this dataset represents a major expansion of the Evidence for Resilient Agriculture (ERA) system into Latin America and the Caribbean (LAC), with a dedicated focus on three cornerstone crops for the region’s food systems: maize, common bean, and coffee. Building on ERA’s established methodology, the LAC expansion systematically compiles, harmonizes, and structures agronomic data from experimental studies conducted across diverse environments, ranging from humid tropical highlands to dry corridor systems.The dataset captures the effects of a wide array of agricultural management practices and technologies, including soil amendments, nutrient management, tillage, cultivar choice, intercropping, agroforestry, pest and disease control, climate-smart practices, and water management interventions. Each observation is standardized following ERA protocols for management classification, outcome reporting, and contextual metadata tagging.All data points are georeferenced and temporally annotated, enabling robust linkages to external datasets such as historical and projected climate conditions, soil properties, elevation, and market access. This contextual integration allows users to assess not only the average effects of management options, but also how these effects vary across biophysical and climatic gradients. Methodology: Overview of Methodological Approach: The expansion of the Evidence for Resilient Agriculture (ERA) system into Latin America and the Caribbean (LAC) followed standard procedures for systematic evidence synthesis, while integrating AI-assisted search and screening tools to improve efficiency and transparency. Methods build on the ERA global protocol and were adapted to the regional crop focus on maize, common bean, and coffee, as well as the unique availability of literature in English, Spanish, and Portuguese. For more information: https://eragriculture.github.io/ERA_Agronomy/ERA-Agroecology.html Literature Search: Search Platform and Query Structure: Unlike the original ERA protocol, which relied on Web of Science and Scopus, the ERA–LAC expansion used OpenAlex, an open-access, AI-enabled bibliographic index. OpenAlex provides semantic search capability and assigns an internal relevance score to each returned article based on natural-language processing (NLP) models. Search queries combined three conceptual elements: Management practice Outcome domain (e.g., yield, resilience indicators, environmental performance, mitigation) Regional context (LAC countries and subregions) Search strings were constructed using Boolean operators (AND, OR) and translated into OpenAlex’s semantic query format. Separate queries were run for major outcome domains (productivity, resilience, environmental indicators, barriers to adoption). The final search yielded 7615 unique records published between 1950 and 2024. AI-Assisted Relevance Scoring and Pre-Screening: OpenAlex assigns each article a numeric relevance score based on machine-learning models that evaluate title, abstract, topic clusters, and metadata. We used this score to streamline the screening process such that: High-relevance articles were screened first; mid-relevance articles were screened subsequently; low-relevance articles were reviewed only if they contained regional keywords or crop-specific terms. This hybrid approach allowed us to preserve comprehensiveness while reducing manual screening volume. For more information: https://eragriculture.github.io/AI-Powered-Meta-Analysis-Automation/docs/OA-vignette.html Screening and Study Selection: Inclusion Criteria: Studies were considered eligible if they: Reported at least one management practice relevant to maize, common bean, or coffee systems; Reported at least one agronomic, environmental, or resilience outcome of interest; Included a control treatment; Presented primary data from field trials; Were conducted in LAC countries Data Extraction: Data extraction followed a structured and standardized workflow designed to harmonize agronomic evidence across diverse studies, consistent with ERA protocols. Standardized Excel Template: All eligible articles were extracted using a standardized Excel data template, adapted from the ERA global protocol and tailored for LAC crop systems. The template included predefined fields for: Study identifiers (authors, year, country, site) Experimental design characteristics Management practices tested Control descriptions Outcome measurements (yields, biophysical indicators, environmental metrics) To ensure consistency in coding across reviewers and across studies, the template integrated a controlled vocabulary. Extraction Process: Each article was reviewed in full by a trained extractor who entered data into the structured template following standardized coding rules. A...