Volume: 5, 2026
5th International PhD Students’ Conference at the University of Life Sciences in Lublin, Poland:
ENVIRONMENT – PLANT – ANIMAL – PRODUCT
Abstract number: E020
DOI: https://doi.org/10.24326/ICDSUPL5.E020
Published online: 22 April 2026
UAV RGB reference data for national Solidago spp. mapping
Bożena Omeliańska*1, Ewa Kołaczkowska1, Anna Kowalska1, Martyna Zarzycka1, Edyta Regulska1, Anna Jarocińska2, Marlena Kycko2, Anna Foks-Ryznar3, Anna Wawrzaszek3, Emil Wrzosek3, Edyta Woźniak3, Marek Ruciński3, Szymon Sala3, Małgorzata Jenerowicz-Sanikowska3 and Andrzej Affek1
1 Institute of Geography and Spatial Organization, Polish Academy of Sciences, 51/55 Twarda St., Warsaw, Poland
2 Faculty of Geography and Regional Studies, University of Warsaw, Warsaw, Poland
3 Space Research Centre, Polish Academy of Sciences, Warsaw, Poland
* Corresponding author: bomelianska@twarda.pan.pl
Reliable large-scale mapping of invasive plant species from satellite imagery depends on high-quality reference data, yet the performance and limitations of unmanned aerial vehicle (UAV)-derived red–green–blue (RGB) imagery in heterogeneous landscapes remain insufficiently understood. This study presents a nationally distributed UAV RGB reference dataset and an operational workflow designed to support wall-to-wall satellite mapping of invasive goldenrods (Solidago spp.) in Poland.
During the peak flowering period (August–September 2024), 79 UAV orthomosaics were acquired across environmentally diverse habitats. Reference polygons were delineated and independently validated by a botanist. A feature set combining RGB-based spectral indices and Grey Level Co-occurrence Matrix (GLCM) texture metrics was used in Random Forest classification. Classification performance was evaluated using user’s accuracy (UA), producer’s accuracy (PA), and F1-score, and complemented by expert visual validation and site-level modelling of accuracy drivers. Mean detection accuracy for Solidago spp. reached an F1-score of 0.873 ±0.070, with strong variability across sites. Errors were consistently dominated by false positives, indicating systematic overestimation of goldenrod cover. Regression models revealed repeatable drivers of reduced precision, including strong direct illumination, disturbance such as mowing, and structurally complex vegetation contexts, which increase spectral and textural ambiguity in RGB imagery. The visual inspection of classified orthomosaics, combined with standardized ground observations conducted by a botanist, demonstrates that classification accuracy based solely on standard quantitative metrics (UA, PA, F1) is insufficient for assessing the quality of UAV-derived reference data, as these metrics fail to capture systematic error patterns and the ecological plausibility of mapped outputs.
Integrating quantitative metrics, expert validation, and modelling of error drivers provides a more robust and ecologically meaningful evaluation framework. The proposed workflow supports efficient and standardised reference-data collection and improves the reliability of subsequent satellite-based mapping of invasive species.
This study was funded by the Ministry of Science and Higher Education of Poland under the Science for Society Programme (grant No. NdS-II/SP/0216/2024/01).
Keywords: GLCM texture; invasive species; Solidago spp.; UAV mapping
How to cite
Omeliańska B., Kołaczkowska E., Kowalska A., Zarzycka M., Regulska E., Jarocińska A., Kycko M., Foks-Ryznar A., Wawrzaszek A., Wrzosek E., Woźniak E., Ruciński M., Sala S., Jenerowicz-Sanikowska M., Affek A., 2026. VUAV RGB reference data for national Solidago spp. mapping. In: 5th International PhD Students’ Conference at the University of Life Sciences in Lublin, Poland: Environment – Plant – Animal – Product. https://doi.org/10.24326/ICDSUPL5.E020