A multi-plot assessment of vegetation structure using a micro-unmanned aerial system (UAS) in a semi-arid savanna environment

Abstract

Unmanned Aerial Systems (UAS) represent an important niche platform for measuring vegetation health, structure, and productivity; metrics that directly inform sustainable conservation and development initiatives in rural African savannas. Products derived from UAS imagery have much finer spatial resolutions than traditional satellite or aircraft imagery, allowing the spectral and structural heterogeneity of vegetation to be mapped and monitored with more detail, an advantage especially useful for challenging environments such as dryland savannas. This study uses UAS-captured imagery to assess the efficacy of UAS for monitoring structural characteristics of vegetation in a mixed savanna woodland. The main objective was to compare multiple approaches for extracting woody vegetation structure from UAS imagery and assess correlations between in situ field measurements and UAS estimates. We compare different sensor types to determine whether multispectral data improve estimates of vegetation structure at the expense of spatial resolution. Results indicate that leveraging multispectral reflectance information, particularly in the near-infrared portion of the spectrum, aids in crown delineation, areal estimates, and fractional cover of woody and non-woody vegetation within the study area. We also compare two image segmentation techniques for crown delineation and found that all techniques perform best in grassy savanna sites where trees and shrubs are easily distinguishable. Overall, a region-growing technique consistently exhibits highest levels of agreement with in situ height and crown area measurements, while a simple height threshold is best for determining fractional coverage of structural classes present. Findings from this work contribute to the advancement for applying high spatial resolution, UAS-derived methods in remote sensing analyses with specific consideration towards autonomous crown delineation and resource management initiatives in dryland systems. Lastly, data-informed analyses, as presented here, provide robust scientific evidence that contribute to informing environmental management decisions when considering the use of UAS technology in conservation and wildlife management across Africa.

Introduction

Land functions represent the goods and services provided by the landscape and are dictated by land use and land cover dynamics (Verburg et al., 2009). A thorough understanding of land functions is integral to natural resource management, and advancements in remote sensing techniques and increasing access to imagery makes the use of such data in monitoring and managing land cover and land use dynamics increasingly tractable. Specifically, remote sensing across a range of environments and scales provides a fundamental tool towards quantifying and understanding the roles of various species and surficial features present (Verburg et al., 2004, Olson et al., 2008). In savannas, heterogeneous landscapes make it challenging to work with low (>500 m) or medium resolution (~5 m–30 m) data. For fine-scale processes typical in these dynamic landscapes, unmanned aerial systems (UAS) provide a viable toolset for offering site specific, flexible, high-resolution options for monitoring and management efforts (Anderson and Gaston, 2013).

Savannas represent an important type of dryland system, covering one fifth of the Earth’s land mass and supporting large wildlife and human populations (Herrero et al., 2016). In savanna systems worldwide, altered precipitation regimes and increases in population of humans, livestock, and wildlife can potentially affect land functions, and may exhibit positive feedbacks leading to degradation of the savanna landscapes (Van Langevelde et al., 2003, Ward, 2005, Hyvärinen et al., 2019, Nkosi et al., 2019). This perception and quantification is often tied back to fire and grazing disturbance regimes and closely related to the positive feedback phenomenon known as bush or shrub encroachment (Roques et al., 2001). While degradation of savanna environments may be variously defined (Eldridge et al., 2011), a shift towards a shrub-dominated equilibrium has implications for land functions in terms of changing the type or composition of biodiversity and resource availability (Roques et al., 2001, Van Auken, 2009). Traditional remote sensing analyses of vegetation structural characteristics tend to fall short, as spatial resolutions of data from satellite-based platforms are inherently coarse and contain pixels with mixed cover (Herrmann and Tappan, 2013). These implications are important while considering the type of fine-scale metrics needed to address United Nations Sustainability Development Goals (SDGs) that aim to halt and reverse land degradation and biodiversity loss (SDG 15) and promote sustained economic growth (SDG 8) in the savanna context. One current area of need directly relates to the underlying rigor of scientifically informed analyses that provide information used for more effective decision making regarding interacting SDG initiatives (ICSU, 2017).

Challenges and inconsistencies associated with using imagery collected via satellite, such as atmospheric effects, cloud cover, temporal constraints, and seasonality, can be addressed by UAS-sourced data with flexible timing, in desirable conditions, and at low altitudes (Zhang and Kovacs, 2012). Furthermore, low altitude flights produce centimetric ground sampling distances that are much finer than data collected via satellite platforms. Researchers tailor their data collection needs based on specific studies through field-controlled flight parameters, delivering fine-scale imagery without many of the limitations typical of data collected by sensors aboard satellite platforms (Anderson and Gaston, 2013). Most importantly, UASs are effective for quantifying vegetation structure as well as estimating fractional vegetation coverage (Cunliffe et al., 2016, Mayr et al., 2017, Yan et al., 2019), as differences in woody species establishment can be difficult to resolve, but important when assessing functional properties of vegetation cover (Brown et al., 1997, Mayr et al., 2017, Ludwig et al., 2019), and are often related to biodiversity, livelihoods, and economic health.

Photogrammetric techniques allow for the derivation of three-dimensional estimates from sets of overlapping two dimensional photos. Advances in computational efficiency in recent years enable for efficient and realistic representations of surficial features through the production of three-dimensional point clouds analogous to the output of Light Detection And Ranging (LiDAR) surveys, which represent the current standard in three-dimensional surface estimation (Smith et al., 2015, Zhu et al., 2019). This process of establishing keypoints in multiple overlapping photos to produce a sparse point cloud (Structure from Motion (SfM)) and subsequently densifying the point cloud (Multi-View Stereo (MVS)) has been applied in many fields and provides safe, relatively inexpensive opportunities for extraction of detailed surface and structural information.

Three-dimensional remotely-sensed vegetation structure has shifted scientific understanding of landscape phenomena and conditions, particularly in savannas (Fisher et al., 2014). Information regarding vegetation structure derived from UAS imagery via photogrammetric techniques such as SfM-MVS can provide grounds upon which to estimate relative value of vegetation present in terms of structure for in systems of interest. Although it is not uncommon for studies to utilize data from UAS platforms in this manner (Dandois and Ellis, 2010, Mayr et al., 2017, Guan et al., 2020), incorporation of data in the NIR is less explored, and a gap remains in understanding the utility of the spectral range of sensor payloads extended into the red edge and near infrared wavelengths (700–900 nm). Combined with the spatial grain and temporal fidelity of UAS data increased spectral detail could further enable the ability to extract structural information for a given landscape, important for the potential of leveraging derived UAS data with other scales of remotely sensed information in monitoring and supporting SDG metric development (Anderson et al., 2017).

This study asks three questions, stratified across savanna sites with different classes of dominant vegetation, to determine how structural composition relates to estimation of vegetation characteristics using SfM-MVS. (RQ1) How does accuracy of SfM-MVS point clouds vary when derived from data collected within tree-, shrub-, and grass-dominated savannas? (RQ2) Are data captured in the NIR portion of the electromagnetic spectrum better suited for estimates of canopy height and crown area? (RQ3) Which of several, tested methods used for LiDAR analyses are most appropriate for SfM-MVS approaches? These questions directly focus on vegetation structure since it affects and informs land functions, as degradation in savannas can be related directly to structure rather than productivity in many contexts (Roques et al., 2001, Van Langevelde et al., 2003, Pricope et al., 2015).