Long-Term Epidemiological Surveillance Integrated with Data Science and Multimodel Approaches Advances Understanding of Polerovirus SCYLV Dynamics in Colombian Sugarcane

The Polerovirus Sugarcane yellow leaf virus (SCYLV) causes one of the most economically important viral diseases of sugarcane worldwide, including Colombia, where sugarcane is a strategic crop and the virus has high phytosanitary relevance but limited epidemiological characterization. This study provides the first integrative quantitative and multimodel epidemiological assessment of SCYLV in the country, using a 22-year surveillance dataset (1999 to 2020) from Colombia's main sugarcane-producing region. The analysis integrated data science, classical statistics, machine learning, and spatial epidemiology to identify temporal trends, spatiotemporal aggregation, and environmental and agronomic drivers of disease risk. Data science procedures supported data curation, exploratory analysis, and detection of patterns in SCYLV presence–absence, incidence, and prevalence across the historical series. Risk factor identification used logistic regression modeling of presence–absence data to estimate adjusted associations, and the machine learning framework modeled both presence–absence (classification) and incidence (regression) to capture nonlinear interactions. Spatiotemporal dynamics were assessed using presence–absence data in a space–time permutation model fitted to a Poisson-type distribution and the Knox test. SCYLV presence–absence and incidence were significantly associated with varietal susceptibility, with CC 01-1940 and SP 71-6949 showing the highest case burden climatic variability, with major outbreak clusters coinciding with strong El Niño–Southern Oscillation anomalies of both phases (El Niño in 2007 and 2015/2016 and the strong 2010/2011 La Niña), meteorological conditions (temperature, wind speed, and vapor pressure), vegetation greenness (normalized difference vegetation index), and soil properties (pH, cation exchange capacity, sand content, and nitrogen). The Knox test identified a high-risk spatiotemporal window for SCYLV spread, with relative risk > 10 concentrated within 4.5 km and 4 days after a confirmed positive case, highlighting a critical range for targeted surveillance and rapid preventive action. This integrative workflow provides a reproducible, data-driven foundation for delineating differential risk zones and supports evidence-based strategies for the integrated management of SCYLV in Colombian sugarcane systems, with transferable methodological value for long-term surveillance analyses of other aphid-transmitted viruses.