Characterizing forest structural types and shelterwood dynamics from Lorenz-based indicators predicted by airborne laser scanning

Abstract

In this study, Lorenz curve descriptors of tree diameter inequality were used to characterize the dynamics of forest development in a shelterwood-managed Pinus sylvestris (L.) dominated area. The purpose was to stratify the forest area into forest structural types (FST) from airborne laser scanning (ALS)-based wall-to-wall predictions of the chosen indicators: Gini coefficient (GC) and Lorenz asymmetry (LA). A clear boundary at GC = 0.5 was found, which separated even-sized (below) and uneven-sized (above) areas. Furthermore, a need for including LA in the characterization of the uneven-sized areas was detected, to distinguish bimodal from reverse J-shaped stands. Beta regression was used for the ALS predictions, yielding RMSEs of 19.67% for GC and 11.01% for LA. Based on our results, we concluded that forest disturbance decreases GC, whereas seed regeneration increases GC and, therefore, gap dynamics are characterized by shifts between either side of the GC = 0.5 threshold. In even-sized stands, GC decreases toward maturity owing to self-thinning occurring at the stem exclusion stage. In uneven-sized stands, the skewness of the Lorenz curve indicates understory development, as ingrowth decreases LA. The possible applications of the resulting FST map are discussed; for instance, in identifying areas needing silvicultural treatments or evaluating forest recovery from disturbances.

A Comparison of Airborne Laser Scanning and Image Point Cloud Derived Tree Size Class Distribution Models in Boreal Ontario

“…The stratification is similar to that used by Bollandså s and Naesset [23] who used the Gini coefficient to group by distribution types, ranging from normal, to uniform, to reverse-J. Alternatives to VCI that better characterize forest structural types [24] are available. VCI was used here to ensure the validation data covered a range of conditions, not necessarily to identify forest structural types, and was felt to be adequate for this use.…”

Comparison of airborne laser scanning methods for estimating forest structure indicators based on Lorenz curves

“…The second indicator was NSLM, the proportion of number of stems which are larger than the weighted mean (QMD). In other words, NSLM is the share of stem density stocked above the QMD, which is the value of the Lorenz plot's xaxis at the inflexion point of the curve x QMD (Valbuena et al, 2013a). The third indicator was the corresponding value for the y-axis M(x QMD ), which is the proportion basal area larger than the QMD (BALM; Gove, 2004).…”

“…Because of its complete three-dimensional characterization of vegetation, airborne laser scanning (ALS) remote sensing allows for evaluating properties related to forest structure in broad forest areas (Lefsky et al, 2005;Maltamo et al, 2005). These properties can be exploited to study forest successional stages Valbuena et al, 2013a), the risk of wild fire propagation (Andersen et al, 2005;Hall et al, 2005)o r wind-throw damage (Suárez et al, 2008), or characteristics related to habitat quality (Lefsky et al, 2002;Martinuzzi et al, 2009). Lexerød and Eid (2006) pointed out a number of motivations for using indicators derived from Lorenz curves to describe forest structure.…”

“…The attention has recently been turned to studying tree size inequality by L-moments, especially with regards to their relations with ALS datasets (Ozdemir and Donoghue, 2013;Valbuena et al, 2013b). Furthermore, the coefficient of LA developed by Damgaard and Weiner (2000) was employed by Valbuena et al (2013a) for characterizing the relation between dominant and subdominant cohorts in multi-layered forests. LA is a join description of BALM and NSLM, two important structural characteristics of forests closely related to one another.…”

“…Previous research aiming at estimating these indicators from ALS were mainly based on parametric modelling, including best subset (Valbuena et al, 2013b) and beta regression (Valbuena et al, 2013a). Most difficulties were found in the variance structure observed on the prediction, and also on the complexity of the relation of LA with ALS metrics.…”