Journal
REMOTE SENSING
Volume 9, Issue 1, Pages -Publisher
MDPI
DOI: 10.3390/rs9010081
Keywords
field spectroscopy; classification; random forests; vegetation index; grassland; multitemporal; ground-based sensors
Categories
Funding
- German Research Foundation (DFG) [SCHE 549/2-1]
- Faculty of Agriculture of the University of Bonn, Germany
Ask authors/readers for more resources
Monitoring the reaction of grassland canopies on fertilizer application is of major importance to enable a well-adjusted management supporting a sustainable production of the grass crop. Up to date, grassland managers estimate the nutrient status and growth dynamics of grasslands by costly and time-consuming field surveys, which only provide low temporal and spatial data density. Grassland mapping using remotely-sensed Vegetation Indices (VIs) has the potential to contribute to solving these problems. In this study, we explored the potential of VIs for distinguishing five differently-fertilized grassland communities. Therefore, we collected spectral signatures of these communities in a long-term fertilization experiment (since 1941) in Germany throughout the growing seasons 2012-2014. Fifteen VIs were calculated and their seasonal developments investigated. Welch tests revealed that the accuracy of VIs for distinguishing these grassland communities varies throughout the growing season. Thus, the selection of the most promising single VI for grassland mapping was dependent on the date of the spectra acquisition. A random forests classification using all calculated VIs reduced variations in classification accuracy within the growing season and provided a higher overall precision of classification. Thus, we recommend a careful selection of VIs for grassland mapping or the utilization of temporally-stable methods, i.e., including a set of VIs in the random forests algorithm.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available