Potentials and Limitations of WorldView-3 Data for the Detection of Invasive <i>Lupinus polyphyllus</i> Lindl. in Semi-Natural Grasslands

Semi-natural grasslands contribute highly to biodiversity and other ecosystem services, but they are at risk by the spread of invasive plant species, which alter their habitat structure. Large area grassland monitoring can be a powerful tool to manage invaded ecosystems. Therefore, WorldView-3 multi...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Damian Schulze-Brüninghoff, Michael Wachendorf, Thomas Astor
Formato:	article
Lenguaje:	EN
Publicado:	MDPI AG 2021
Materias:	invasive species WorldView-3 grassland machine learning feature selection Science Q
Acceso en línea:	https://doaj.org/article/81ca9fb6ec984a8ca7bc93899bd41e5b
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:81ca9fb6ec984a8ca7bc93899bd41e5b
record_format	dspace
spelling	oai:doaj.org-article:81ca9fb6ec984a8ca7bc93899bd41e5b2021-11-11T18:54:17ZPotentials and Limitations of WorldView-3 Data for the Detection of Invasive <i>Lupinus polyphyllus</i> Lindl. in Semi-Natural Grasslands10.3390/rs132143332072-4292https://doaj.org/article/81ca9fb6ec984a8ca7bc93899bd41e5b2021-10-01T00:00:00Zhttps://www.mdpi.com/2072-4292/13/21/4333https://doaj.org/toc/2072-4292Semi-natural grasslands contribute highly to biodiversity and other ecosystem services, but they are at risk by the spread of invasive plant species, which alter their habitat structure. Large area grassland monitoring can be a powerful tool to manage invaded ecosystems. Therefore, WorldView-3 multispectral sensor data was utilized to train multiple machine learning algorithms in an automatic machine learning workflow called ‘H2O AutoML’ to detect <i>L. polyphyllus</i> in a nature protection grassland ecosystem. Different degree of <i>L. polyphyllus</i> cover was collected on 3 × 3 m<sup>2</sup> reference plots, and multispectral bands, indices, and texture features were used in a feature selection process to identify the most promising classification model and machine learning algorithm based on mean per class error, log loss, and AUC metrics. The best performance was achieved with a binary classification of lupin-free vs. fully invaded 3 × 3 m<sup>2</sup> plot classification with a set of 7 features out of 763. The findings reveal that <i>L. polyphyllus</i> detection from WorldView-3 sensor data is limited to large dominant spots and not recommendable for lower plant coverage, especially single plant detection. Further research is needed to clarify if different phenological stages of <i>L. polyphyllus</i> as well as time series increase classification performance.Damian Schulze-BrüninghoffMichael WachendorfThomas AstorMDPI AGarticleinvasive speciesWorldView-3grasslandmachine learningfeature selectionScienceQENRemote Sensing, Vol 13, Iss 4333, p 4333 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	invasive species WorldView-3 grassland machine learning feature selection Science Q
spellingShingle	invasive species WorldView-3 grassland machine learning feature selection Science Q Damian Schulze-Brüninghoff Michael Wachendorf Thomas Astor Potentials and Limitations of WorldView-3 Data for the Detection of Invasive <i>Lupinus polyphyllus</i> Lindl. in Semi-Natural Grasslands
description	Semi-natural grasslands contribute highly to biodiversity and other ecosystem services, but they are at risk by the spread of invasive plant species, which alter their habitat structure. Large area grassland monitoring can be a powerful tool to manage invaded ecosystems. Therefore, WorldView-3 multispectral sensor data was utilized to train multiple machine learning algorithms in an automatic machine learning workflow called ‘H2O AutoML’ to detect <i>L. polyphyllus</i> in a nature protection grassland ecosystem. Different degree of <i>L. polyphyllus</i> cover was collected on 3 × 3 m<sup>2</sup> reference plots, and multispectral bands, indices, and texture features were used in a feature selection process to identify the most promising classification model and machine learning algorithm based on mean per class error, log loss, and AUC metrics. The best performance was achieved with a binary classification of lupin-free vs. fully invaded 3 × 3 m<sup>2</sup> plot classification with a set of 7 features out of 763. The findings reveal that <i>L. polyphyllus</i> detection from WorldView-3 sensor data is limited to large dominant spots and not recommendable for lower plant coverage, especially single plant detection. Further research is needed to clarify if different phenological stages of <i>L. polyphyllus</i> as well as time series increase classification performance.
format	article
author	Damian Schulze-Brüninghoff Michael Wachendorf Thomas Astor
author_facet	Damian Schulze-Brüninghoff Michael Wachendorf Thomas Astor
author_sort	Damian Schulze-Brüninghoff
title	Potentials and Limitations of WorldView-3 Data for the Detection of Invasive <i>Lupinus polyphyllus</i> Lindl. in Semi-Natural Grasslands
title_short	Potentials and Limitations of WorldView-3 Data for the Detection of Invasive <i>Lupinus polyphyllus</i> Lindl. in Semi-Natural Grasslands
title_full	Potentials and Limitations of WorldView-3 Data for the Detection of Invasive <i>Lupinus polyphyllus</i> Lindl. in Semi-Natural Grasslands
title_fullStr	Potentials and Limitations of WorldView-3 Data for the Detection of Invasive <i>Lupinus polyphyllus</i> Lindl. in Semi-Natural Grasslands
title_full_unstemmed	Potentials and Limitations of WorldView-3 Data for the Detection of Invasive <i>Lupinus polyphyllus</i> Lindl. in Semi-Natural Grasslands
title_sort	potentials and limitations of worldview-3 data for the detection of invasive <i>lupinus polyphyllus</i> lindl. in semi-natural grasslands
publisher	MDPI AG
publishDate	2021
url	https://doaj.org/article/81ca9fb6ec984a8ca7bc93899bd41e5b
work_keys_str_mv	AT damianschulzebruninghoff potentialsandlimitationsofworldview3dataforthedetectionofinvasiveilupinuspolyphyllusilindlinseminaturalgrasslands AT michaelwachendorf potentialsandlimitationsofworldview3dataforthedetectionofinvasiveilupinuspolyphyllusilindlinseminaturalgrasslands AT thomasastor potentialsandlimitationsofworldview3dataforthedetectionofinvasiveilupinuspolyphyllusilindlinseminaturalgrasslands
_version_	1718431655825244160

Potentials and Limitations of WorldView-3 Data for the Detection of Invasive <i>Lupinus polyphyllus</i> Lindl. in Semi-Natural Grasslands

Ejemplares similares