Alwin A. Hardenbol, Hans Ole Ørka, Terje Gobakken, Joel Kostensalo. (2026). Optimising rare tree species detection for large-area mapping: a case study on European aspen (Populus tremula). Silva Fennica vol. 60 no. 3 article id 26012. https://doi.org/10.14214/sf.26012

Keywords: Bayesian statistics; biodiversity conservation; base rate problem; broadleaves; individual tree map; model transferability

Highlights: Balancing false discovery and false negative rates improves rare tree species detection; Preferring volume- over count-based approaches makes ecological sense for rare tree species; We detected 70% of aspen volume and classified this volume with 31% false discovery rate and 54% false negative rate; Base rates of rare tree species have been overlooked in remote sensing literature.

Abstract | Full text in HTML | Full text in PDF | Author Info

Some locally rare tree species – like European aspen (Populus tremula L.) – contribute disproportionately to forest biodiversity, making their accurate monitoring important for conservation. However, most remote sensing studies of rare tree species ignore their low prevalence/base rate, necessary for model transferability. We aimed to assess European aspen classification under realistic landscape prevalence and find methods to improve classification performance. Across 253 plots from field data in Våler, Norway we detected and segmented 5455 living trees, including 55 European aspens (ca. 1% prevalence), from high-density airborne laser scanning (ALS) data merged with multispectral aerial imagery. Individual segments were then classified as European aspen or non-aspen using random forest and synthetic minority oversampling (SMOTE). We evaluated classification based on tree counts and summed tree volumes. To underpin the importance of prevalence consideration and compare our results, we also re-analysed published Fennoscandian European aspen classification studies. Contrary to a perfect balance, we found an optimal SMOTE-ratio at ca. 0.07 (European aspen–non-aspen ratio). Our count-based classification achieved 32% (16–52) false discovery rate (FDR) and 73% (60–83) false negative rate (FNR), whereas volume-based classification achieved 31% (11–51) FDR and 54% (38–71) FNR. Re-analysed studies averaged 66% FDR and 33% FNR at a 1% prevalence. Our results demonstrate that rare tree species classification benefits from optimising majority–minority ratios and, given the ecological value of larger trees, volume-based approaches are warranted. Finally, we urge reporting landscape prevalence and adjust classification metrics to reflect these to enable transferability to real-world applications in rare tree species classification.

Hardenbol, Natural Resources Institute Finland (Luke), Natural Resources Unit, Yliopistokatu 6B, FI-80100 Joensuu, Finland https://orcid.org/0000-0002-0615-505X E-mail: alwin.hardenbol@luke.fi
Ørka, Norwegian University of Life Sciences (NMBU), Faculty of Environmental Sciences and Natural Resource Management, P.O. Box 5003, NO-1432 Ås, Norway https://orcid.org/0000-0002-7492-8608 E-mail: hans-ole.orka@nmbu.no
Gobakken, Norwegian University of Life Sciences (NMBU), Faculty of Environmental Sciences and Natural Resource Management, P.O. Box 5003, NO-1432 Ås, Norway https://orcid.org/0000-0001-5534-049X E-mail: terje.gobakken@nmbu.no
Kostensalo, Natural Resources Institute Finland (Luke), Natural Resources Unit, Yliopistokatu 6B, FI-80100 Joensuu, Finland https://orcid.org/0000-0001-9883-1256 E-mail: joel.kostensalo@luke.fi

article id 10515, category Research article

Alwin A. Hardenbol, Anton Kuzmin, Lauri Korhonen, Pasi Korpelainen, Timo Kumpula, Matti Maltamo, Jari Kouki. (2021). Detection of aspen in conifer-dominated boreal forests with seasonal multispectral drone image point clouds. Silva Fennica vol. 55 no. 4 article id 10515. https://doi.org/10.14214/sf.10515

Keywords: Populus tremula; deciduous trees; mixed forest; protected areas; tree species classification; unmanned aerial vehicles

Highlights: Four boreal tree species (Scots pine, Norway spruce, birches and European aspen) classified with an overall accuracy of 95%; Presence of European aspen detected with excellent accuracy (UA: 97%, PA: 96%); Late spring is the best time for species classification by remote sensing; Best time to separate aspen from birch was when birch had leaves, but aspen did not.

Abstract | Full text in HTML | Full text in PDF | Author Info

Current remote sensing methods can provide detailed tree species classification in boreal forests. However, classification studies have so far focused on the dominant tree species, with few studies on less frequent but ecologically important species. We aimed to separate European aspen (Populus tremula L.), a biodiversity-supporting tree species, from the more common species in European boreal forests (Pinus sylvestris L., Picea abies [L.] Karst., Betula spp.). Using multispectral drone images collected on five dates throughout one thermal growing season (May–September), we tested the optimal season for the acquisition of mono-temporal data. These images were collected from a mature, unmanaged forest. After conversion into photogrammetric point clouds, we segmented crowns manually and automatically and classified the species by linear discriminant analysis. The highest overall classification accuracy (95%) for the four species as well as the highest classification accuracy for aspen specifically (user’s accuracy of 97% and a producer’s accuracy of 96%) were obtained at the beginning of the thermal growing season (13 May) by manual segmentation. On 13 May, aspen had no leaves yet, unlike birches. In contrast, the lowest classification accuracy was achieved on 27 September during the autumn senescence period. This is potentially caused by high intraspecific variation in aspen autumn coloration but may also be related to our date of acquisition. Our findings indicate that multispectral drone images collected in spring can be used to locate and classify less frequent tree species highly accurately. The temporal variation in leaf and canopy appearance can alter the detection accuracy considerably.

Hardenbol, University of Eastern Finland, School of Forest Sciences, P.O. Box 111, FI-80101 Joensuu, Finland https://orcid.org/0000-0002-0615-505X E-mail: alwin.hardenbol@uef.fi
Kuzmin, University of Eastern Finland, School of Forest Sciences, P.O. Box 111, FI-80101 Joensuu, Finland; University of Eastern Finland, Department of Geographical and Historical Studies, P.O. Box 111, FI-80101 Joensuu, Finland E-mail: anton.kuzmin@uef.fi
Korhonen, University of Eastern Finland, School of Forest Sciences, P.O. Box 111, FI-80101 Joensuu, Finland E-mail: lauri.korhonen@uef.fi
Korpelainen, University of Eastern Finland, Department of Geographical and Historical Studies, P.O. Box 111, FI-80101 Joensuu, Finland E-mail: pasi.korpelainen@uef.fi
Kumpula, University of Eastern Finland, Department of Geographical and Historical Studies, P.O. Box 111, FI-80101 Joensuu, Finland E-mail: timo.kumpula@uef.fi
Maltamo, University of Eastern Finland, School of Forest Sciences, P.O. Box 111, FI-80101 Joensuu, Finland E-mail: matti.maltamo@uef.fi
Kouki, University of Eastern Finland, School of Forest Sciences, P.O. Box 111, FI-80101 Joensuu, Finland E-mail: jari.kouki@uef.fi

Click this link to register to Silva Fennica.

If you are a registered user, log in to save your selected articles for later access.

Contents alert

Your selected articles

Follow @SilvaFennica