Estimating aboveground carbon density and its uncertainty in Borneo’s structurally complex tropical forests using airborne laser scanning

Abstract

<p><strong>Abstract.</strong> Borneo contains some of the world’s most biodiverse and carbon dense tropical forest, but this 750 000-km2 island has lost 62 % of its old-growth forests within the last 40 years. Efforts to protect and restore the remaining forests of Borneo hinge on recognising the ecosystem services they provide, including their ability to store and sequester carbon. Airborne Laser Scanning (ALS) is a remote sensing technology that allows forest structural properties to be captured in great detail across vast geographic areas. In recent years ALS has been integrated into state-wide assessment of forest carbon in Neotropical and African regions, but not yet in Asia. For this to happen, new regional models, need to be developed for estimating carbon stocks from ALS in tropical Asia, as the forests of this region are structurally and compositionally distinct from those found elsewhere in the tropics. By combining ALS imagery with data from 173 permanent forest plots spanning the lowland rain forests of Sabah, on the island of Borneo, we develop a simple-yet-general model for estimating forest carbon stocks using ALS-derived canopy height and canopy cover as input metrics. An advanced feature of this new model is the propagation of uncertainty in both ALS- and ground-based data, allowing uncertainty in hectare-scale estimates of carbon stocks to be quantified robustly. We show that the model effectively captures variation in aboveground carbons stocks across extreme disturbance gradients spanning tall dipterocarp forests and heavily logged regions, and clearly outperforms existing ALS-based models calibrated for the tropics, as well as currently available satellite-derived products. Our model provides a simple, generalised and effective approach for mapping forest carbon stocks in Borneo, and underpins ongoing efforts to safeguard and facilitate the restoration of its unique tropical forests.</p>

Topography shapes the structure, composition and function of tropical forest landscapes

“…At Sepilok, the optimal balance is struck at mid elevations and on moderate-to-steep slopes that characterise sandstone hill forests, where low turnover rates allow for tightly packed and carbon-rich forests to form (Nilus 2004). This underscores why TCH alone cannot always be used to reliably predict ACD in mature forests (Duncanson et al 2015;Coomes et al 2017), highlighting the need for new general models for estimating ACD from ALS data (Jucker et al 2018). An integral part of this process will involve refining current approaches to estimating WD from remotely sensed data.…”

“…For each 1-ha plot, we used the field data to calculate three plot-level attributes: ACD (Mg C ha À1 ), the communityweighted mean wood density (WD; g cm À3 ) and tree species richness. ACD was estimated by first calculating the aboveground biomass (AGB; kg) of individual trees using Chave et al's (2014) pantropical biomass equationwhere AGB ¼ 0:067 Â D 2 Â H Â WD À Á 0:976and then summing the AGB of all trees within a plot after applying a carbon content conversion factor of 0.47 (see Jucker et al (2018) for details). Plot-level WD was quantified as…”

“…The reflectance profiles of the three forest types obtained from the airborne hyperspectral imagery are shown below. of ACD at Sepilok and across Sabah more widely (Coomes et al 2017;Jucker et al 2018): mean top-of-canopy height (TCH; m) and gap fraction at 20 m aboveground. TCH is the mean height of the pixels which make up the surface of the CHM, while gap fraction is the proportional area with no crown cover at a specific height aboveground (20 m in this case), and captures heterogeneity in canopy height and density.…”

“…Accurately estimating plot-level WD from remotely sensed data has proven challenging, and remains a major source of uncertainty in efforts to map forest carbon stocks (Mitchard et al 2014). Based on the observation that community-mean WD tends to increase during forest succession, previous work has attempted to estimate WD from ALS-derived metrics such as TCH, but with mixed results (Asner & Mascaro 2014;Jucker et al 2018). Here, we take a different approach which assumes that because leaf and stem traits are generally coordinated in trees (M endez-Alonzo et al 2012;Nolf et al 2015;Zeballos et al 2017; although see Baraloto et al 2010), we can use hyperspectral imagerywhich has been shown to reliably capture leaf properties (e.g.…”

“…In this regard, airborne remote sensing promises to revolutionise the way we study forests by providing large-scale coverage of key biophysical parameters at high resolution. For instance, airborne laser scanning (ALS, or LiDAR) can be used to simultaneously capture the 3D structure of both the canopy and the underlying terrain (Lefsky et al 2002;Wulder et al 2012;Detto et al 2013), and is now routinely used as a tool to map ACD in tropical forests (Asner & Mascaro 2014;R ejou-M echain et al 2015;Jucker et al 2018). By coupling ALS with novel spectranomic approaches that are able to resolve the biochemical makeup of forest canopies based on how they reflect light (Townsend et al 2008;Asner et al 2015;Schneider et al 2017), we are now in a position to comprehensively assess the relative importance of different topographic features in driving heterogeneity in the structure, composition and function of tropical forest landscapes.…”