Research project SR/00/307 (Research action SR)
CONTEXT AND OBJECTIVES
The prospected growth in urban population and progressive global warming will put huge pressure on the quality of the environment in densely populated areas. Sustainable development and management of urban areas, especially urban green, is hence crucial to guard the living quality in our future cities.
Unfortunately, current policy support tools such as environmental models are not well adapted to the high level of heterogeneity of urban landscapes and would greatly benefit from detailed, multi-temporal, spatially distributed input data provided by remote sensing.
This project, therefore, aims at exploring the potential of the combined use of recent multi- and
hyperspectral sensors in combination with structural information derived from LiDAR, for detailed,
spatially explicit characterization of morphological and (bio)physical properties of the urban environment.
Remote sensing derived information on the characteristics of green and built-up areas will be used to improve the parameterisation of urban biophysical models. As such, we strive at improving the operational value of urban ecosystem services related to temperature and water regulation.
Objectives: (i) Explore the potential of innovative approaches for spectral unmixing for improved land cover parameterisation of urban biophysical models, using airborne hyperspectral data, and assess the transferability of the approaches proposed to imagery of lower spatial and/or spectral resolution acquired by current and future spaceborne multispectral and hyperspectral sensors; (ii) Examine the use of spectral and LiDAR remote sensing data for characterising chemical and structural properties of urban vegetation for improved land cover parameterisation of urban biophysical models; (iii) consolidate chemical and structural properties of the urban environment, derived by remote sensing, by defining a local climate zone (LCZ) typology that is far better suited to characterize urban climatic conditions than traditional land cover datasets typically employed for urban climate modelling; (iv) Develop a quantitative ecosystem service mapping tool on urban water regulation making optimal use of the detailed, high-resolution remote sensing based characterisation of the urban ecosystem; (v) Develop a simulation framework for assessing impacts of urban growth and alternative urban planning scenarios on urban heat and water regulation, based on an integration of agent-based modelling of human activities at neighbourhood level, and grid-based, remote sensing supported modelling of biophysical processes.
METHODOLOGY
Urban land cover mapping - Starting from a hierarchical scheme of functional land cover classes and properties, relevant for modelling of urban hydrology and climate the potential of different types of sensors for land cover parameterisation will be evaluated. Emphasis will be put on evaluating the transferability of the approaches developed on high-resolution hyperspectral imagery to imagery of lower spatial and/or spectral resolution acquired by current and future spaceborne multispectral and hyperspectral sensors. The monitoring and modelling methods will therefore systematically be evaluated for both real (airborne hyperspectral and Landsat) and simulated (EnMAP, HyspIRI, Sentinel2) data sets and will be tested on different study sites (Brussels, Berlin, Santa Barbara), in order to support the development of generic algorithms. Herefore, we will follow the following methodological scheme: (WP2) definition and optimization of spectral libraries, (WP3) comparative analysis of the performance of different unmixing approaches, both physically based and machine learning based, and modification of these approaches to reduce endmember variability issues.
Characterizing urban vegetation – in WP4 we will evaluate the potential of spectral and LiDAR remote sensing data for characterising biochemical (e.g. canopy chlorophyll and water) and structural (e.g. LAI, biomass, height) properties of urban vegetation for improved parameterisation of urban biophysical models.
Urban heat regulation modeling – In WP5 a remote sensing based framework for urban and rural landscape characterisation adapted to urban heat regulation will be develop through: (1) identifying urban surface parameters relevant for urban heat regulation; (2) characterising local climate zones (LCZ) based on the remote sensing land cover parameterisation provided in WP3 and WP4; mapping LCZ as being functional units from an ecosystem service perspective; (4) exploring and comparing regression/correlation models to investigate the association between LCZ and typical urban climatic parameters like land surface temperature (LST) and air temperature (Tair).
Urban hydrological modelling - In WP6 the urban land cover products provided in WP3 and WP4 through remote sensing analysis will be used for parameterisation and validation of hydrological responses of different land cover types and vegetation properties. The strong relation between specific surface cover (urban green and built-up) and hydrological response at the high resolution (2m) forms the basis for a consistent upscaling of the hydrological response to a coarser scale, more specific for functional entities.
Urban dynamics modelling - Outputs of the remote sensing supported biophysical modelling work in WP5 (heat) and WP6 (water) should assist urban planners in developing smart urban growth management strategies, resulting in land use arrangements that minimize pressure on the environment. To that end, in WP7, a simulation-based modelling approach will be developed, transforming expected changes in the density of human activities into changes in urban form. Alternative scenarios for urban planning will constrain the model in terms of the structural and biophysical properties of new and transformed urban areas. This will enable us to assess the consequences of different planning decisions on urban heat and water regulation.
EXPECTED SCIENTIFIC RESULTS
Peer-reviewed publication on and new algorithms for:
- the comparison of alternative training approaches for unmixing of Landsat/Sentinel OLI data
- the comparison of different endmember library optimization methodologies/workflows
- relevant cross-sensor spectral features for urban mapping applications
- a comparative analysis of the performance of different physical and machine-learning based unmixing approaches for mapping urban land cover from high resolution hyperspectral data
- methodologies and its accuracy by which vegetation properties can be assessed through spectral and LiDAR analysis at different spectral and spatial resolutions and for different unmixing methodologies.
- the transferability of alternative approaches for the estimation of urban land cover fractions, developed on high-resolution hyperspectral data, to data of lower spatial and/or spectral resolution
- relevance of RS-based local climate zone mapping for local policy makers
- the applicability of the LCZ mapping system at different local scale levels
- LCZ characterisation in terms of surface and air temperature
- High resolution modelling and validation of the hydrological response of the urban ecosystem
- Spatial metric based upscaling for the assessment of the hydrological response of the urban ecosystem
- simulation framework for modelling impacts of urban growth and spatial planning on the
- (bio)physical characteristics of the environment
- assessment of impacts of urban growth on heat regulation for alternative spatial planning scenarios
- assessment of impacts of urban growth on water fluxes for alternative spatial planning scenarios
EXPECTED PRODUCTS AND SERVICES
- Spectral database for urban surface materials with interface for sensor-specific resampling
- Tools for spectral unmixing
- A set of tools for increasing the spectral separability and reducing the endmember variability of target land cover classes in urban mapping from remotely sensed data at different spectral and spatial resolutions (including integration of height information
- Quantitative and spatial explicit maps of the chemical and structural properties of the urban vegetation for the main study site, derived from airborne hyperspectral imagery and from real and simulated sensor data
- A LCZ mapping system based on Landsat, simulated Enmap and Sentinel-2 data
- Characteristic profiles of relevant urban climatic parameters for all LCZ at multiple spatial scale levels, assisting local policy makers in their interpretation of LCZ maps
- Distributed ecosystem service indicator maps regarding ‘urban water regulation’ for Brussels
- A simulation tool for modelling impacts of alternative urban growth and spatial planning scenarios on the (bio)physical characteristics of the urban environment
- Indicator maps characterizing environmental conditions for different urban growth scenarios in relation to urban climate and water fluxes
