The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Articles | Volume XLVIII-1/W2-2023
14 Dec 2023
 | 14 Dec 2023


R. A. Oliveira, R. Näsi, P. Korhonen, A. Mustonen, O. Niemeläinen, N. Koivumäki, T. Hakala, J. Suomalainen, J. Kaivosoja, and E. Honkavaara

Keywords: Hyperspectral, Unmanned Aerial System (UAS), Grass Biomass, Nitrogen Concentration, Machine Learning

Abstract. Monitoring agricultural grass fields is particularly important for meat and milk production in Northern Europe, where three harvests occur during a growing season to maximize yields. Reliable data on forage, including biomass and nitrogen concentration, are essential for making informed decisions regarding seed mixtures, fertilizer rates, and harvest timing. Miniaturized hyperspectral cameras mounted on unmanned aerial systems (UAS) have become increasingly accessible and efficient. These cameras, operating in the visible to near-infrared (VNIR) range, have shown potential in estimating grass sward quantity and feeding quality. Additional advancements in hyperspectral technology have emerged the short-wave infrared (SWIR) range for UAS applications, previously utilized mainly in laboratory and aircraft-based systems. This study aims to explore the potential of VNIR and SWIR hyperspectral UAS-based remote sensing in biomass and nitrogen estimation during primary and re-growth stages. Grass fresh yield and nitrogen concentration prediction models were built after selecting the most significant features from the cameras to cope with the high dimensionality of the data. Using best features and machine learning, both fresh yield and nitrogen concentration were estimated with normalized root mean square error better than 10%. This work contributes to the development of accurate remote sensing techniques, supporting sustainable agricultural practices and climate change studies.