File Download
Links for fulltext
(May Require Subscription)
- Publisher Website: 10.1016/j.rse.2018.09.018
- Scopus: eid_2-s2.0-85054034567
- WOS: WOS:000449449800009
- Find via
Supplementary
- Citations:
- Appears in Collections:
Article: Invasive tree species detection in the Eastern Arc Mountains biodiversity hotspot using one class classification
Title | Invasive tree species detection in the Eastern Arc Mountains biodiversity hotspot using one class classification |
---|---|
Authors | |
Keywords | Africa Biased support vector machine Imaging spectroscopy Invasive tree species One class classification |
Issue Date | 2018 |
Citation | Remote Sensing of Environment, 2018, v. 218, p. 119-131 How to Cite? |
Abstract | Eucalyptus spp. and Acacia mearnsii are common exotic tree species in eastern Africa that have shown (strong) invasive behavior in some regions. Acacia mearnsii is considered a highly invasive species that is replacing native species and Eucalyptus spp. are known to consume high amounts of groundwater with suspected effects on native flora. Mapping the occurrence of these species in the Taita Hills, Kenya (part of the Eastern Arc Mountains Biodiversity Hotspot) is important as there is lack of knowledge on their occurrence and ecological impact in the area. Mapping methods that require a lot of fieldwork are impractical in areas like the Taita Hills, where the terrain is rugged and the infrastructure is poor. Our aim was hence to map the occurrence of these tree species in a 100 km2 area using airborne imaging spectroscopy and laser scanning. We used a one class biased support vector machine (BSVM) classifier as it needs labeled training data only for the positive classes (A. mearnsii and Eucalyptus spp.), which potentially reduces the amount of required fieldwork. We also introduce a new approach for parameterizing and setting the threshold level simultaneously for the BSVM classifier. The second aim was to link the occurrence of these species to selected environmental variables. The results showed that the BSVM classifier is suitable for mapping Acacia mearnsii and Eucalyptus spp., holding the potential to improve the efficiency of field data collection. The introduced parametrization/threshold selection method performed better than other commonly used approaches. The crown level F1-score was 0.76 for Eucalyptus spp. and 0.78 for A. mearnsii. We show that Eucalyptus spp. and A. mearnsii trees cover 0.8% and 1.6% of the study area, respectively. Both species are particularly located on steeper slopes and higher altitudes. Both species have significant occurrences in areas close to the biggest remaining native forest patch (Ngangao) in the study area. Nonetheless, follow-up studies are needed to evaluate their impact on the native flora and fauna, as well as their impact on the water resources. The maps created in this study in combination with such follow-up studies could serve as base data to generate guidelines that authorities can use to take action in handling the problems these species are causing. |
Persistent Identifier | http://hdl.handle.net/10722/309247 |
ISSN | 2023 Impact Factor: 11.1 2023 SCImago Journal Rankings: 4.310 |
ISI Accession Number ID |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Piiroinen, Rami | - |
dc.contributor.author | Fassnacht, Fabian Ewald | - |
dc.contributor.author | Heiskanen, Janne | - |
dc.contributor.author | Maeda, Eduardo | - |
dc.contributor.author | Mack, Benjamin | - |
dc.contributor.author | Pellikka, Petri | - |
dc.date.accessioned | 2021-12-15T03:59:49Z | - |
dc.date.available | 2021-12-15T03:59:49Z | - |
dc.date.issued | 2018 | - |
dc.identifier.citation | Remote Sensing of Environment, 2018, v. 218, p. 119-131 | - |
dc.identifier.issn | 0034-4257 | - |
dc.identifier.uri | http://hdl.handle.net/10722/309247 | - |
dc.description.abstract | Eucalyptus spp. and Acacia mearnsii are common exotic tree species in eastern Africa that have shown (strong) invasive behavior in some regions. Acacia mearnsii is considered a highly invasive species that is replacing native species and Eucalyptus spp. are known to consume high amounts of groundwater with suspected effects on native flora. Mapping the occurrence of these species in the Taita Hills, Kenya (part of the Eastern Arc Mountains Biodiversity Hotspot) is important as there is lack of knowledge on their occurrence and ecological impact in the area. Mapping methods that require a lot of fieldwork are impractical in areas like the Taita Hills, where the terrain is rugged and the infrastructure is poor. Our aim was hence to map the occurrence of these tree species in a 100 km2 area using airborne imaging spectroscopy and laser scanning. We used a one class biased support vector machine (BSVM) classifier as it needs labeled training data only for the positive classes (A. mearnsii and Eucalyptus spp.), which potentially reduces the amount of required fieldwork. We also introduce a new approach for parameterizing and setting the threshold level simultaneously for the BSVM classifier. The second aim was to link the occurrence of these species to selected environmental variables. The results showed that the BSVM classifier is suitable for mapping Acacia mearnsii and Eucalyptus spp., holding the potential to improve the efficiency of field data collection. The introduced parametrization/threshold selection method performed better than other commonly used approaches. The crown level F1-score was 0.76 for Eucalyptus spp. and 0.78 for A. mearnsii. We show that Eucalyptus spp. and A. mearnsii trees cover 0.8% and 1.6% of the study area, respectively. Both species are particularly located on steeper slopes and higher altitudes. Both species have significant occurrences in areas close to the biggest remaining native forest patch (Ngangao) in the study area. Nonetheless, follow-up studies are needed to evaluate their impact on the native flora and fauna, as well as their impact on the water resources. The maps created in this study in combination with such follow-up studies could serve as base data to generate guidelines that authorities can use to take action in handling the problems these species are causing. | - |
dc.language | eng | - |
dc.relation.ispartof | Remote Sensing of Environment | - |
dc.rights | This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. | - |
dc.subject | Africa | - |
dc.subject | Biased support vector machine | - |
dc.subject | Imaging spectroscopy | - |
dc.subject | Invasive tree species | - |
dc.subject | One class classification | - |
dc.title | Invasive tree species detection in the Eastern Arc Mountains biodiversity hotspot using one class classification | - |
dc.type | Article | - |
dc.description.nature | published_or_final_version | - |
dc.identifier.doi | 10.1016/j.rse.2018.09.018 | - |
dc.identifier.scopus | eid_2-s2.0-85054034567 | - |
dc.identifier.volume | 218 | - |
dc.identifier.spage | 119 | - |
dc.identifier.epage | 131 | - |
dc.identifier.isi | WOS:000449449800009 | - |