Summary
This study demonstrates the first successful application of remote sensing to assess liana (woody vine) infestation at landscape scale in an aseasonal tropical forest in Sabah, Malaysia. Using hyperspectral and LiDAR data with neural network classification, the authors compared pixel-based and object-based approaches, finding pixel-based methods more accurate for continuous infestation prediction but equivalent performance when data were categorised into three classes. The work addresses a methodological gap in extending remote sensing liana assessment beyond seasonal forests and highlights the importance of aligning spatial units of remotely sensed and field observations.
Regional applicability
This study was conducted in Malaysian tropical forest and is primarily applicable to aseasonal tropical forest systems globally rather than United Kingdom conditions. The methodology and findings may inform forest monitoring and conservation approaches in tropical regions with similar closed-canopy forest structures, but direct transfer to temperate UK forestry would require further methodological adaptation.
Key measures
Root mean square deviation (RMSD) of predicted versus observed liana infestation; pixel-based and object-based classification accuracy; McNemar's χ² test comparing classification methods; liana infestation categories (Low 0–30%, Medium 31–69%, High 70–100%)
Outcomes reported
The study compared pixel-based and object-based remote sensing classifications for detecting liana infestation across landscape scales in aseasonal tropical forest. Pixel-based classification achieved lower root mean square deviation (27.0% ± 0.80) compared to object-based classification (32.6% ± 4.84), with no significant difference when data were grouped into three infestation categories.
Topic tags
Dig deeper with Pulse AI.
Pulse AI has read the whole catalogue. Ask about this record, its theme, or how the findings apply to UK farming and policy — every answer cites the underlying studies.