< Back to previous page

Project

Urban treescape analysis using ground-based and airborne remote sensing to support conservation of tree-provided ecosystem services

Urban trees are the most important green infrastructure in cities to mitigate climate change, improve the urban environment, and promote human health and well-being by providing numerous critical ecosystem services. However, they experience various pressures, such as the urban heat island (UHI) effect, soil sealing, and air pollution, potentially affecting the tree health condition and thereby the quantity and quality of the ecosystem services they offer to humanity. Urban tree introduction and conservation initiatives are thus important, and various management strategies and policies have been formulated. In support of such efforts, it is essential to monitor the dynamics of important tree properties, in particular those related to tree functioning, across space. Moreover, the adaption capability of trees to environmental changes and how tree-provided ecosystem services influence human health and well-being need to be well understood. Based on these scientific outputs and insights, appropriate and targeted management activities can be implemented, and their effectiveness can be evaluated. However, conventional tree inventory is spatially and temporally constrained by a range of difficulties. Therefore, further study on promising solutions is required. In this dissertation, we explored and improved the potential of ground-based and airborne (hyperspectral and light detection and ranging (LiDAR)) remote sensing technology in achieving the abovementioned requirements of urban treescape analysis. Our study area was the Brussels Capital Region in Belgium.

Driven by a lack of knowledge on the capability of trees to adapt to the urban environment, we investigated the within-species variation in leaf functional and optical traits (i.e., water- and pigment-related traits) and their phenology induced by the UHI effect and soil sealing using generalized additive models (Chapter 2). We focused on Tilia × euchlora trees and found that the intra-species trait variations among different environmental conditions ranged from 8% to 38%. The trees growing in sealed soils were observed to start the autumn downregulation of photosynthetic pigments earlier (up to 13 days) than those at unpaved sites. These intra-species variations indicate the leaf plasticity of Tilia × euchlora trees, enabling them to adapt to the fast-changing urban environment. We demonstrated that using leaf optical traits to act as a proxy of leaf functional traits is promising, and may allow for the examination of environmental impacts on trees at broader spatial scales using spectral sensors onboard airplanes or satellite platforms.

Despite the adaptive capability, tree health conditions can to different extents be impaired by severe or chronic environmental stress, leading to degraded tree-provided ecosystem services. In Chapter 3, we therefore assessed the potential of airborne hyperspectral and LiDAR data with a Random Forest classifier to detect urban tree defoliation, discoloration, and a combination thereof at the individual tree crown level. We found that the fusion of hyperspectral and LiDAR features achieved the highest accuracies, with overall accuracies ranging from 0.81 to 0.89. The overall better performance of LiDAR features over hyperspectral features might be case-specific, needing further examination. Importantly, we demonstrated that a species-specific modelling approach should be adopted in mapping urban tree health.

Tree species is an important tree property, determining the ability of trees to provide ecosystem services, and can serve as basic information to enhance the mapping of other tree properties (e.g., tree health) using remote sensing. Chapter 4 was therefore dedicated to assessing and improving the potential of airborne hyperspectral and bi-temporal LiDAR (leaf-on and leaf-off) data in urban tree species classification at the individual tree crown level. Additionally, we looked at how planting locations (i.e., streets vs parks) drive intra-species spectral and structural variations and in consequence affect classification accuracies. We found that the importance of hyperspectral and LiDAR features for species discrimination varied within species between street and park trees. The proportions of intra-species variation in spectral reflectance, leaf-on and leaf-off LiDAR features explained by planting locations were up to 40.6%, 63.9%, and 64.6% respectively. These results supported our finding that a planting location-specific modelling approach significantly improved urban tree species mapping, with the highest classification accuracies (85.1%) achieved by using the combined hyperspectral and leaf-on and leaf-off LiDAR data. Built upon these findings, we suggest integrating a step of semantic classification of trees into urban tree species discrimination.

Central to the tree-provided ecosystem services is to improve human health and well-being. In Chapter 5, we explored how airborne LiDAR technology can be used to improve the quantification of exposure to trees and enhance our understanding of the associations with cardiovascular and mental health. We developed a complete workflow, including individual tree delineation, screening of incorrect trees, and estimation of tree traits, to map three-dimensional tree traits at the city level. We showed that medication sales for cardiovascular disease and mental disorders were negatively associated with crown volumes but positively associated with tree density in models including both exposure indicators. We thereby hypothesize that living in areas with larger crown volumes and lower tree densities may be more beneficial to cardiovascular and mental health, compared to living in areas with higher tree densities and smaller crown volumes. These findings underscore the need to conserve large trees in cities.

The research conducted in this dissertation has confirmed the potential of ground-based and airborne (hyperspectral and LiDAR) remote sensing for a comprehensive and advanced analysis of urban treescape to support the assessment, conservation, and improvement of ecosystem services that trees provide to human society. Despite the operationality of the developed methodologies, future research should further address various uncertainties in urban tree monitoring by coupling a well-designed urban laboratory with multisource data (e.g., remote sensing, sensor networks, citizen science) and state-of-the-art techniques in order to better inform management and policy.

Date:25 Sep 2018 →  22 Aug 2022
Keywords:Hyperspectral, LiDAR, Sentinel-2, functional/optical traits, urban tree health, environmental stressors, leaf phenology, public health, bird species diversity
Disciplines:Landscape architecture, Art studies and sciences, Physical geography and environmental geoscience, Communications technology, Geomatic engineering, Forestry sciences, Ecology, Environmental science and management, Other environmental sciences
Project type:PhD project