The role of spectro-temporal remote sensing in vegetation classification: A comprehensive review integrating machine learning and bibliometric analysis

Abstract

Spectro-temporal remote sensing (STRS) is a reliable source for mapping and monitoring earth’s surface dynamics. This review investigates the role of STRS in Vegetation Classification (VC) by analyzing 159 articles from Web of Science Core Collection (WSCC) database, spanning from 1980 to 2023. By integrating machine learning and bibliometric analysis, it provides comprehensive examination of trends, themes and advancements in the application of STRS in VC. The findings indicate significant growth in the use of STRS for VC, highlighted an exponential increase in publications over time. Recently, commonly used classification methods include machine learning, deep learning and spectral matching techniques, with research themes covering crop types, agricultural land, and forests. Notably, the study underscores the dominance of the USA and China in both publication quantity and collaborative efforts, reflecting their leadership in this field. Frequently utilized STRS data sources include MODIS, Landsat and Sentinel-1/2 sensors. Furthermore, the review emphasizes the need for developing flexible frameworks that integrate spectro-temporal data and accuracy evaluation metrics to build robust, intelligent, and transfer-learning classification methods. Overall, this review sheds light on the role of STRS in VC and provides valuable insights for researchers and decision-makers involved in vegetation monitoring and mapping. It emphasizes the potential of STRS to revolutionize VC and outlines directions for further research to address existing challenges and capitalize on emerging opportunities in this rapidly growing field.