Exploring use of remotely sensed data for capturing biomass accumulation in silage
1H. Huitu, 1O. Niemeläinen, 2R. Näsi, 2N. Viljanen, 2T. Hakala, 2L. Markelin, 2E. Honkavaara, 1H. Ojanen, 3J. Kaivosoja
1. Natural Resources Institute Finland (Luke)
2. Finnish Geospatial Research Institute
3. Natural Resources Institute Finland
Accurate and up-to-date spatial information is fundamental for precision farming, and improves decision making in silage production. Multiple sources of spatial information can be utilized to monitor biomass accumulation of the growing crop stand. Recently, remotely sensed imagery from drones and satellites has become widely available, while their cost has dropped drastically. Also, crop growth models can extend the usability of old canopy information when new data or measurements cannot be captured.
In this study we explore the performance of 1) consumer level UAS data acquisition and processing, 2) professional level Fabry–Pérot interferometer based hyperspectral UAS, and 3) Sentinel-2 MSI imagery in capturing the relative biomass of silage grass. Our main study area consisted of one 8 ha grass field, where spatial variation in biomass was monitored during the growing seasons 2016 and 2017. In 2017, vegetation samples were collected twice from 10 geo-located plots, and analysis was supported by three 20 m * 30 m field trial plots measured at 10 – 15 days interval. Vegetation samples were analyzed for dry and fresh biomass and digestibility. UAS data was captured twice per season, and for the respective dates, best availlable Sentinel-2 MSI images were processed.
We found that all the studied sources of remote sensing information (consumer level UAS data, professional level hyperspectral UAS data and Sentinel-2 MSI imagery) had explanatory value to the spatial variation in biomass. As long as the geographical scale in the variation can be captured at the image resolution of 10 m, the NDVI calculated from the Sentinel-2 imagery is valuable. The consumer level UAS RGB-data was subject to rapid weather changes that were not accurately corrected from the image mosaics with the standard commercial data processing chain. However, also consumer level UAS data produced biomass estimates that were usable in the context of precision farming. Use of accurate hyperspectral mapping purely for relative biomass estimations was not justified, since the changes in biomass were detectable also in wider spectral ranges.