Pasture monitoring using remote sensing and machine learning: A review of methods and applications

Abstract

Pastures are important feed sources for livestock and require an optimal management strategy to boost the productivity and sustainability of grassland. Remote sensing (RS) has been explored for grassland monitoring and estimating pasture biophysical characteristics. The array of sensors, including hyperspectral, multispectral, and RGB, integrated with sensing platforms such as satellites, drones, and ground-based vehicles, yields massive amounts of data. This heterogeneous RS data necessitates machine learning (ML) methods for precisely estimating pasture quality and quantity. This survey aims to provide a systematic review and meta-analysis of the progress in pasture monitoring using RS with ML. First, we propose a taxonomy that assimilates and categorises the existing works based on the various approaches used in the RS data processing pipeline. Second, we analyse and synthesise the performance of ML methods on the RS data for pasture monitoring tasks such as pasture identification and classification, biomass estimation, and pasture quality estimation. Finally, we report the survey findings and underscore the challenges and future avenues of research in pasture modelling using hybrid RS with ML approaches. The survey highlights that integrating RS data into ML models has demonstrated considerable success in pasture monitoring, particularly in biomass estimation, whereas pasture quality estimation warrants elevated focus in future research.