Monitoring ryegrass-clover grasslands with multi-spectral UAV imagery will improve the sustainability of small-and medium-sized livestock farmers in the northern Peruvian highlands

• Autores: Vallejos Fernández, Luis, Alvarez García, Wuesley Yusmein, Abanto urbina, Maycol, Gutiérrez Arce, Felipe, Tapia Acosta, Eduardo, Pizarro, Samuel, Ciprian, Cesar, Naupari, Javier
• Formato: artículo
• Fecha de Publicación: 2026
• Institución: Instituto Nacional de Innovación Agraria
• Repositorio: INIA-Institucional
• Lenguaje: inglés
• OAI Identifier: oai:repositorio.inia.gob.pe:20.500.12955/3030
• Enlace del recurso: http://hdl.handle.net/20.500.12955/3030; https://doi.org/10.1080/27658511.2026.2623335
• Nivel de acceso: acceso abierto
• Materia: Aboveground biomass; Ryegrass-clover; UAVs; Machine learning; Multispectral imaging; Biomasa aérea; Raigrás-trébol; UAVs (vehículos aéreos no tripulados); Aprendizaje automático; Imágenes multiespectrales; https://purl.org/pe-repo/ocde/ford#4.04.01; Lolium multiflorum; Trifolium repens; Teledetección; Remote sensing; Pastizales; Pastures; Praderas; Grasslands; Forrajaes; Forage; Ganadería; Animal husbandry

The underutilization of remote sensing technology has compromised sustainable forage resource management, impeding the progress of livestock farmers in the northern Peruvian highlands. To accurately predict forage biomass in six high-altitude (2600-2800 m) ryegrass (Lolium multiflorum Lam) -clover (Trifolium repens) paddocks, we applied machine learning models implemented in Google Earth Engine using spectral indices derived from UAV-based multispectral imagery captured by a Micasense RedEdge MX camera mounted on a DJI Matrice 600. A total of 75 forage samples were collected from precisely geo-referenced plots to train and validate machine learning models based on 13 spectral indices. The Random Forest (RF) model, comprising 500 trees for green forage and dry matter, demonstrated high accuracy and efficiency. UAV-based biomass prediction using GEE and ML techniques was validated, achieving R² values of 0.671 and 0.747 and low errors. By integrating UAVs, sensors, and cloud-based ML, we can decision-support potential in the inter-Andean valley. This innovative approach reduces costs, ensures high-resolution snapshot biomass assessment, and empowers producers to make data-driven decisions.