remote sensing Article Detection of Spatial and Temporal Patterns of Liana Infestation Using Satellite-Derived Imagery Chris J. Chandler *, Geertje M. F. van der Heijden, Doreen S. Boyd and Giles M. Foody   Citation: Chandler, C.J.; van der Heijden, G.M.F.; Boyd, D.S.; Foody, G.M. Detection of Spatial and Temporal Patterns of Liana Infestation Using Satellite-Derived Imagery. Remote Sens. 2021, 13, 2774. https://doi.org/10.3390/rs13142774 Academic Editor: Damiano Gianelle Received: 11 June 2021 Accepted: 12 July 2021 Published: 14 July 2021 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affil- iations. Copyright: © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). School of Geography, University of Nottingham, University Park, Nottingham NG7 2RD, UK; Geertje.Vanderheijden@nottingham.ac.uk (G.M.F.v.d.H.); doreen.boyd@nottingham.ac.uk (D.S.B.); giles.foody@nottingham.ac.uk (G.M.F.) * Correspondence: Christopher.chandler1@nottingham.ac.uk Abstract: Lianas (woody vines) play a key role in tropical forest dynamics because of their strong influence on tree growth, mortality and regeneration. Assessing liana infestation over large areas is critical to understand the factors that drive their spatial distribution and to monitor change over time. However, it currently remains unclear whether satellite-based imagery can be used to detect liana infestation across closed-canopy forests and therefore if satellite-observed changes in liana infestation can be detected over time and in response to climatic conditions. Here, we aim to determine the efficacy of satellite-based remote sensing for the detection of spatial and temporal patterns of liana infestation across a primary and selectively logged aseasonal forest in Sabah, Borneo. We used predicted liana infestation derived from airborne hyperspectral data to train a neural network classification for prediction across four Sentinel-2 satellite-based images from 2016 to 2019. Our results showed that liana infestation was positively related to an increase in Greenness Index (GI), a simple metric relating to the amount of photosynthetically active green leaves. Furthermore, this relationship was observed in different forest types and during (2016), as well as after (2017–2019), an El Niño-induced drought. Using a neural network classification, we assessed liana infestation over time and showed an increase in the percentage of severely (>75%) liana infested pixels from 12.9% ± 0.63 (95% CI) in 2016 to 17.3% ± 2 in 2019. This implies that reports of increasing liana abundance may be more wide-spread than currently assumed. This is the first study to show that liana infestation can be accurately detected across closed-canopy tropical forests using satellite-based imagery. Furthermore, the detection of liana infestation during both dry and wet years and across forest types suggests this method should be broadly applicable across tropical forests. This work therefore advances our ability to explore the drivers responsible for patterns of liana infestation at multiple spatial and temporal scales and to quantify liana-induced impacts on carbon dynamics in tropical forests globally. Keywords: airborne hyperspectral and LiDAR; aseasonal forest; Greenness Index; liana infestation; Sentinel-2 imagery 1. Introduction Lianas (woody vines) are a pervasive component of tropical forests [1,2]. They are non-self-supporting structural parasites that use the architecture of trees to extend their leaves to the forest canopy [3]. As competition between lianas and trees is stronger than tree-tree competition [4], lianas can negatively impact the growth [5,6] and survival of their host [7,8] and therefore suppress the ability of tropical forests to sequester and store carbon [9]. Lianas have been proliferating in some tropical forests [10,11], which may lead to a stronger negative impact on carbon storage and sequestration in these areas. Several putative mechanisms have been suggested for this increase, such as elevated atmospheric CO2, an increase in forest disturbance and an increase in the frequency and severity of Remote Sens. 2021, 13, 2774. https://doi.org/10.3390/rs13142774 https://www.mdpi.com/journal/remotesensing