Classifying ecosystems with metaproperties from terrestrial laser scanner data

Abstract

In this study, we introduce metaproperty analysis of terrestrial laser scanner (TLS) data, and demonstrate its application through several ecological classification problems. Metaproperty analysis considers pulse level and spatial metrics derived from the hundreds of thousands to millions of lidar pulses present in a single scan from a typical contemporary instrument. In such large aggregations, properties of the populations of lidar data reflect attributes of the underlying ecological conditions of the ecosystems.In this study, we provide the Metaproperty Classification Model to employ TLS metaproperty analysis for classification problems in ecology. We applied this to a proof-of-concept study, which classified 88 scans from rooms and forests with 100% accuracy, to serve as a template.We then applied the Metaproperty Classification Model in earnest, to separate scans from temperate and tropical forests with 97.09% accuracy (N = 224), and to classify scans from inland and coastal tropical rainforests with 84.07% accuracy (N = 270).The results demonstrate the potential for metaproperty analysis to identify subtle and important ecosystem conditions, including diseases and anthropogenic disturbances. Metaproperty analysis serves as an augmentation to contemporary object reconstruction applications of TLS in ecology, and can characterize regional heterogeneity.

Keywords: bioinformatics; conservation; habitats; software; statistics.

Figure 1

Photographs and compact biomass lidar point clouds of a Room (University of Massachusetts Boston), temperate forest (Harvard Forest) and tropical rainforest (La Selva, Costa Rica)

Figure 2

Diagram of metaproperties (descriptions in Table 4) featuring CBL2 TLS

Figure 3

Flowchart of the steps in the Metaproperty classification model. Orange boxes denote discretionary steps to tackle specific data scenarios

Figure 4

Examples of probability plots and logit transforms for the metaproperty mean Intensity in Temperate vs. Tropical Forests. Transformation can be used, as here, to improve the linearity of the variable, providing the smoothest transition of probability across the range of the variable

Figure 5

Example case where a metaproperty (1st:2nd Returns) completely separates two classification groups (Rooms vs. Forests). The probability is 0% or 100% of declaring a scan as a Forest (1). Such variables cannot be used in a binary logistic regression without employing a penalized likelihood method such as Firth's logistic regression

Figure 6

ROC curve (Left) and Precision and Recall Curve (Right) for Temperate vs. Tropical Forests classification. ROC describes changes in false positive rate as true positive rate improves. Precision & Recall describes changes in positive predictive rate (Precision) as true positive rate (Recall) improves. The wedge/step in the bottom right of the plot for the testing data indicates that a range of Recall rates were observed 0.1 Precision

Figure 7

CBL point clouds showing the metaproperty 1st Returns and 2nd Returns, which was consistently higher in Rooms than in Forests

Figure 8

CBL point clouds for no returns in Rooms vs. Forests. Forests had consistently higher no returns:pulses than Rooms

Figure 9

OPA for Rooms vs. Forests. Forests consistently had higher OPA than Forests. OPA, optical plane area

Figure 10

OPA polygons for Temperate vs. Tropical Forests. Temperate Forests tended towards higher OPA

Figure 11

CBL point clouds showing no returns:pulses, which tended to be higher in Temperate Forests than in Tropical Forests, attributable to the denser sub‐canopy layer of Tropical Forests, which intercepts many pulses

Figure 12

OPA polygons for Inland vs. Costal rainforests, plotted to scale. Coastal Rainforests tended towards higher OPA