Towards reliable monitoring of grass species in nature conservation: Evaluation of the potential of UAV and PlanetScope multi-temporal data in the Central European tundra

0577091 - BÚ 2024 RIV NL eng J - Článek v odborném periodiku
Kupková, L. - Červená, L. - Potůčková, M. - Lysák, J. - Roubalová, M. - Hrázský, Z. - Březina, S. - Epstein, H. E. - Müllerová, Jana
Towards reliable monitoring of grass species in nature conservation: Evaluation of the potential of UAV and PlanetScope multi-temporal data in the Central European tundra.
Remote Sensing of Environment. Roč. 294, August (2023), č. článku 113645. ISSN 0034-4257. E-ISSN 1879-0704
Institucionální podpora: RVO:67985939
Klíčová slova: multi-temporal species classification * relict arcto-alpine tundra * grass vegetation * UAV hyperspectral data
Obor OECD: Plant sciences, botany
Impakt faktor: 13.5, rok: 2022
Způsob publikování: Omezený přístup
https://doi.org/10.1016/j.rse.2023.113645

The study considered three different data types with various spatial, temporal, and spectral resolutions, and therefore, the results varied. Despite these differences, general trends were discerned for both the data type and the classifiers. The accuracy increased with increasing spatial resolution, and the ultra-high spatial resolution data of the UAV brought very good OA > 90%. The coarser satellite data of PS performed with maximum OA = 82.1%. As for the temporal resolution, the multi-temporal composites outperformed the mono-temporal data in all cases. We also identified the best time window during the peak of the growing season - the best OAs were recorded in most cases for images from July.The results show high accuracies for most species, >95% for the UAV data, and above 80% for the coarser PS data. The UAV data analysis exhibited excellent results. In particular, the values of >99% were reported for UA or PA and > 96% for F1-score, for the following species: CV (both PA and UA of composites July/August for both MS and HS UAV data, June/July/August composite of MS data), DC (PA of June/July/August composite for HS data), MC (UA of mono-temporal and both multi-temporal composites for MS data, and June/July/August composite for HS data). Slightly lower, but still very good, values were found for NS.
Trvalý link: https://hdl.handle.net/11104/0348978