Slides presented as part of my PhD Confirmation of Candidature.
The project is about evaluating the cooling effectiveness of green infrastructure in urban environments. Skills demonstrated include GIS, data grids, image processing, machine learning, data processing and visualization, environmental modelling,
1. P3 “Green infrastructure and Microclimate”
Confirmation of PhD Candidature
Darien Pardiñas Díaz
Supervisors:
Jason Beringer, Nigel Tapper and Matthias Demuzere
Evaluating the cooling effectiveness of
green infrastructure as a heat
mitigation strategy
2. • Motivation
• Knowledge gaps and research questions
• Research objectives and approach explained
• Summary
• Progress today and timetable
Structure
3. Motivation: The Problem
EHE
Anthropogenic
Heat
CO, SO2, NOx, PM generation
from fossil fuel combustion
Indoor
Cooling
NOX and VOC → O3
Positive
feedback
ENERGY
DEMAND
REDUCED
HUMAN HEALTH
AND COMFORT
Urban planning policies
and regulations
RAPID &
UNPLANNED
URBANISATION
AIR
POLLUTION
Urbanisation will continue in the next decades
Look for long-term solutions to
minimise the negative impacts of
urbanisation in climate
4. Motivation: Solutions and Challenges
?
Urban forestry have proven to be a
cost-effective way to reduce urban
temperatures
Challenges associated:
Implementation of UHI MS demands
initiative and important investments
Climate benefits of a particular MS are
difficult to quantify because they depends on
many factors difficult to consider in depth
UCM, RS and GIS techniques
can help us to ensure that
implementation practices
report MAX benefits AT the
MIN costs.
There are a range of technologies that can
be applied to reduce the UHI intensity
5. Climatic benefits (cooling) of UHI MS based on vegetation depends on:
Extent and scale of implementation
Spatial arrangement of existing urban features
Geographic zone and regional climate (rainfall, humidity,
temperatures, etc.)
Vegetation is irrigated or not
-Results should not be extrapolated across scales or different cities
-Climate knowledge has to be in correspondence with the spatial scale
and scope of urban planning actions
Limitation of previous studies
Time scales studied do not always satisfies the long-term climate
information that urban planners and policy-makers often demand
Rough estimates of land surface changes are usually employed in
urban climate runs → Unrealistic MS
Knowledge gaps
6. How effective is the urban forestry as a heat mitigation
strategy at local scale and how this effectiveness varies
spatially and temporally in Australian cities?
1. How well can urban climate models simulate the observed climate? Can daily and
seasonal climate be reproduced well at different densities of urbanisation? How
sensitive is the model to prescribed vegetation cover parameters?
2. What is the current LULC of the urban landscape and what are the opportunities for
implementation of urban forestry, considering urban physical constraints?
3. How much cooling can be achieved by extensive implementation of urban forestry as
a heat MS? Is the urban forestry a viable alternative for cooling under different
climatic conditions?
Assess how the cooling effectiveness varies among different seasons of the year and in EHE
Assess the cooling effectiveness across periods of different rainfall regimens
Compare the cooling effectiveness in two cities of different climate characteristics.
Develop case studies in support of forestation programs (“Greening the West”)
Research Questions
7. Summary of the Research Approach
Melbourne
&
Brisbane
UCM/LSM
validation,
sensitivity
and
selection
Surface
parameterisation
K↓, L↓,
Ta, Qa, Psurf,
Ws, Rainfall
Current LC
maps
T [°C]
Modified LC
maps
Planning
zones
T [°C]
Cooling
[°C]
Mitigation
Strategies
Remote
Sensing
OBJECTIVE 1
UC/LSM
UC/LSM
OBJECTIVE 2
OBJECTIVE 3
City-wide simulations
Atmospheric Forcing
Model
outputs
8. OBJECTIVE 1
To evaluate the ability of existing models as a tool to
assess cooling from heat mitigation strategies.
Urban climate models have strengths and weakness that need to
be considered when employing them in particular urban climate
problems
9. Objective 1: Validation sites
Preston Armadale Surrey Hills
Geometrical parameters:
Building height
Wall-to-plan area ratio ~ h/w
Roof fraction
Roughness length
Radiation Parameters:
Albedo for roof, wall and roads
Emissivity for roof, wall and roads
Thermal parameters:
Volumetric heat capacity of roof, walls and roads.
Thermal conductivity of roof, walls and roads.
Vegetation Parameters:
Natural surface fractions of trees and grass
Monthly green vegetation fraction, LAI, roughness
length and emissivity
Shortwave and NIR albedos
Minimum stomatal resistance
Root depths and distribution
Etc.
Soil parameters:
Soil texture (% of clay and sand)
Slope index
1-furb
furb
10. Objective 1: Simulation results (Preston)
TEB_GARDEN vs. SLUCM_NOAH
Summer Autumn Winter Spring
12. The performance is similar in general but it varies across the seasons of the year and
time of the day.
Systematic underestimation of QE in most seasons:
Surface parameters for vegetated surfaces could be improved (e.g. z0 in urban
conditions etc.);
Although Melbourne was under Stage 1 water restriction (Coutts et at. 2007) no
irrigation whatsoever was considered.
Patchy vegetation may transpires at a relative higher rate than a completely
vegetated surface (Offerle et al. 2006). Vegetation is really patchy in Preston.
Objective 1. Preliminary remarks
Validate models in Armadale and Surrey Hills
Sensitivity to vegetation parameters
(evapotranspiration)
Select the most appropriate model
configuration to estimate cooling
13. OBJECTIVE 2
To obtain the current LC data suitable for climate
modelling and to derive realistic UHI MS based on
urban forestry
The spatial heterogeneity of the urban landscape requires very high
resolution LC information to estimate the implementation
opportunities of MS.
14. Objective 2: Derivation of LC fractions
High resolution land cover
data (small area)
Multi-spectral Remote Sensing
Imagery (Landsat TM)
900
m
15. Objective 2: Accuracy of LC estimation
Site Cover Type
Expert
Classification
Manual
Classification
Average
Landsat TM
classification
Armadale
(37°51’S
145°1’E)
Trees 0.21 0.19 0.20 0.20
Grass 0.18* 0.11 0.15 0.15
Impervious 0.61 0.70 0.67 0.65
Preston
(37°43’S
145°0’E)
Trees 0.29 0.16 0.23 0.18
Grass 0.11 0.2 0.15 0.18
Impervious 0.60 0.64 0.62 0.64
Surrey Hills
(37°49’S
145°5’E)
Trees 0.27 0.31 0.29 0.34
Grass 0.19 0.16 0.18 0.18
Impervious 0.52 0.54 0.53 0.48
Tree cover Impervious cover Grass cover
30m 60m 900m 30m 60m 900m 30m 60m 900m
Pearson’s r 0.788 0.794 0.968 0.827 0.830 0.989 0.742 0.744 0.926
MAE 0.068 0.017 0.014 0.121 0.030 0.023 0.100 0.025 0.028
MBE 0.000 0.000 -0.004 0.001 0.000 0.01 0.001 0.000 -0.006
In City of Melbourne (LGA)
In flux tower sites (radius 500m)
16. Objective 2. Derivation of MS based on vegetation
Current Land cover Maps
Modified Land cover Maps(Mitigation)
Analysis by planning zones to derive ‘realistic’ mitigation scenarios
based on feasible increases of the amount of vegetation in urban areas
No. Planning zone classes
Total cover
fraction
Tree cover
(%)
Grass cover
(%)
Impervious
cover (%)
1 Business zone 4.6 % 9.0 15.9 75.1
2 Industrial zone 10.4 % 13.1 19.6 67.3
3 Low density residential zone 4.0 % 36.0 32.4 31.6
4 Public parks / Recreational zones 7.3 % 24.1 40.3 35.6
5 Public use zones 1.7 % 15.6 27.6 56.8
6 Road zone 4.8 % 21.9 22.2 55.9
7 Rural use zone 13.4 % 35.4 37.5 27.1
8 Residential zone 48.5 % 24.1 22.4 53.5
9 Special use zones 4.3 % 15.6 36.0 48.4
Water bodies 1.0 % - - -
17. Objective 2. Derivation of mitigation scenarios
… …
Original LC Modified LC
More
vegetated
Least
vegetated
Planning
zone pi
18. OBJECTIVE 3
To understand how the cooling from vegetation
varies at different spatial and temporal scales.
Run city-wide simulations in Melbourne and
Brisbane. Process simulation outputs to answer
the main research question
City-wide, long-term simulations of the current and improved urban
climate can provide the data to understand how the cooling
effectiveness varies across different spatial and temporal scales of
the urban climate
19. Surface parameterisation (city wide simulations)
Current cover fractions
Water bodies
Impervious
Deciduous
Evergreen
Grass
Modified cover fractions
Water bodies
Impervious
Deciduous
Evergreen
Grass
Radiation and thermal parameter Units Symbol Source
Layers thickness for roof, wall and road [m] Δzi, i = 1:4 Defaults from Loridan et al. (2011)
Albedo for roof, wall, and road [-] αroof, αwall, αroad Defaults from Loridan et al. (2011)
Emissivity for roof, wall, and road [-] εroof, εwall, εroad Defaults from Loridan et al. (2011)
Volumetric heat capacity for roof, wall and road [MJ/K/m3] croof, cwall, croad Defaults from Loridan et al. (2011)
Thermal conductivity for roof, wall and road [W/K/m] λroof, λwall, λroad Defaults from Loridan et al. (2011)
Geometrical parameters
Mean building height [m] h From urban planning zones aggregation
Roof width [m] r From urban planning zones aggregation
Road width [m] w From urban planning zones aggregation
Roughness length for momentum [m] z0town From urban planning zones aggregation
Vegetation parameter Units Symbol Source
Green fraction [fraction] σf (monthly) Time varying remote sensing NDVI and cover fractions
Leaf area Index [m2/m2] LAI (monthly) Profiles from literature adjusted to southern hemisphere
Roughness length for
momentum
[m] z0 (monthly) Tree height by planning zones and seasonal considerations
Shortwave albedo [-] αveg (monthly) Literature review
Emissivity [-] εveg (monthly) Literature review
Minimum stomatal
resistance
[s/m] RSmin Literature review
Soil class [-] Soil Harmonized Wold Soil Database
Slope class [-] Slope WRF default global dataset
Deep soil temperature [K] Tbot WRF default global dataset
Cells of 900m
20. City-wide Atmospheric Forcing data
Melbourne Brisbane
K↓, L↓,
Psfc, Tzref, Qzref,
Wspd,
Rainfall
Select surface station that have 30m meteorological data in the domain of interest and fill
single 30-min gaps
Gap filling using the nearest station in the period of interest
Derive K↓, L↓ using cloud cover, T2m and Q2m (NARP parameterisation)
Complement Rainfall with daily outputs from AWAP dataset
Adjustment of forcing Tair and Qair at the forcing height zref = 40 m form T2m and Q2m by an
iterative process based on bias correction (Lemonsu 2009)
30 years @ 30 min
21. Urban forestry is an effective way to mitigate heat in urban
areas but its effectiveness needs to be quantified in
Australian cities
The cooling effectiveness of UHI MS depends of several
spatial and temporal factors
UCM/LSM can help to quantify the cooling effectiveness
of heat mitigation scenarios but their fit-to-purpose should
be assured.
Modifications in the landscape as a result of UHI MS must
be represented as accurately as possible considering
urban physical constrains
Summary
22. Planning (Jun 2011-Apr 2012)
Literature review (70%)
Assimilation of models and set-up (100%)
Data request (90%)
Writing of the CoC report (100%)
Objective 1 (Dec 2011-Mar 2013)
Validate of an UCM/LSM pair (80%)
Assess the fit-to-purpose as a heat mitigation
strategy assessment tool (20%)
Publish relevant results (0%)
Objective 2 (Dec 2011-Mar 2013)
Derivation of current land cover (Melbourne
only) (95%)
Derivation of heat mitigation scenarios (25%)
Surface parameterisation for baseline and
scenarios (0%)
Publish relevant results (20% -> ICUC8 Dublin
2012)
Objective 3 (Jun 2012-Dec 2013)
Prepare forcing data to run city-scale climate
simulations in Melbourne and Brisbane (10%)
Perform grid-based simulations with current
and modified landscapes for the domains of
Melbourne and Brisbane (0%)
Analyse model outputs to respond research
questions related (0%)
Run proposed neighbourhood cases (0%)
Publication of results and thesis writing (0%)
Thesis revision and
submission
(Jan 2014 – May 2014)
Progress today and time frames
24. Objective 1. Urban Canopy / Land Surface Models
Q* + QF = QH + QE + ΔQS [W/m2]
Urban canopy model ≈ urban energy balance
fgarden
froadfroof
za
zT
zR
Ta
TS garden
TS wallTS wall
TS road
TS roof
Tcanyon
Ti bld
Tcell = furbTurb + (1 – furb)Tnature
Soil hydrology and thermodynamics
Direct evaporation from soil and canopy
Evapotranspiration
Radiation trapping in the canyon
Heat storage by the urban fabric
Anthropogenic heat release, etc.
25. Objective 1: Parameters prescription
Monthly LAI [m2/ m2]
αnir
[-]
αvis
[-]
RSmin
[s/m
]
Jan Feb
Ma
r
Apr May Jun Jul Aug Sep Oct Nov Dec
Deciduous trees 4.2 4.8 5.6 4.4 2.4 1.8 1.5 1.2 1.1 2.2 3.1 3.5 0.25c 0.05c 100c
Evergreen tress 3.2a 3.2 3.2 3.2 3.2 3.2 3.2 3.2 3.2 3.2 3.2 3.2 0.45d 0.12d 250ac
Grass 1.0b 1.0b 1.0b 1.0b 1.0b 1.0b 1.0b 1.0b 1.0b 1.0b 1.0b 1.0b 0.3c 0.1c 40bc
a Value taken from Peel et al. (2005)
b Values taken from Lynn et al. (2009)
c Default values in ECOCLIMAP-II natural parameters for such cover type (Champeaux, Masson et al. 2005)
d Average taken from Figure 3 in Lewis (2002) for Eucalyptus spp. and Acacia spp. spectral profile.
Australian native
and exotic trees
Thermal parameters
of urban materials
Most evergreen tree are native (Eucalyptus and Acacias spp.) (Frank 2006)
26. Models’ performance has been found to be similar in general
The integrated approach (TEB_GARDEN) did not report any evident improvement.
Geometries of residential areas in Melbourne do not form well defined urban
canyons
Systematic underestimation of QE in most seasons:
Surface parameters for vegetated surfaces could be improved (e.g. z0 in urban
conditions etc.);
Although Melbourne was under Stage 1 water restriction (Coutts et at. 2007) no
irrigation whatsoever was considered.
Patchy or sparse vegetation transpires at a relative higher rate than a completely
vegetated surface (Offerle et al. 2006). Vegetation is really patchy in Preston.
Objective 1. Preliminary remarks
Validate models in Armadale and Surrey Hills
Select the most appropriate model configuration for
cooling calculation
28. For every urban planning zone class 𝑝𝑖:
Calculate the cover fractions intersected with a grid of resolution X.
Given a function y(ftree, fgrass) that weighs the cooling obtainable from grass and trees
fractions, sort the land cover composition by y.
Given a threshold of implementation (𝜆 𝑝 𝑖
) [0..1] obtain the land cover composition
𝑓𝑡𝑟𝑒𝑒, 𝑓𝑔𝑟𝑎𝑠𝑠
𝜆 𝑝 𝑖
for every given class whose position in the sorted array divided by
the number of samples is equal to 𝜆 𝑝 𝑖
.
Replace the existing land cover composition on every cell of which 𝑦 𝑓𝑡𝑟𝑒𝑒, 𝑓𝑔𝑟𝑎𝑠𝑠 <
𝑦 𝑓𝑡𝑟𝑒𝑒, 𝑓𝑔𝑟𝑎𝑠𝑠
𝜆 𝑝 𝑖
Aggregate the modified cover fractions back to the urban climate model resolution.
Objective 2. Derivation of mitigation scenarios
y(ftree, fgrass) = βftree + fgrass
… …
Original Modified
More
vegetated
Least
vegetated
Business
zone
29. Seasonal parameters of vegetation are important in
simulations of long periods.
Deciduous species of occupy an significant percentage of
the total urban forestry (Frank 2006)
Assume that evergreen trees and grass present similar
properties during all seasons, then estimate
fexotic = α(NDVIleaf-on – NDVIleaf-off)
Better than assuming the same fraction
of deciduous trees city wide
Objective 2. Seasonal variability of vegetation
parameters
30. Parameters with ambiguous definitions have to be
prescribed (e.g. h/w)
Objective 1. Models limitations and data uncertainties
31. Validate models in Armadale and Surrey Hills
Make further analysis of performance to determine the
causes of limitations (e.g. underrated QE)
Test other vegetation approaches (NOAH-MP Ball
Berry)
Sensitivity analysis to vegetation parameters
Selection of the most appropriate model
configuration for cooling calculation
Objective 1. Next steps
32. Surface Parameterisation
32
Geometrical parameters:
Mean Building height [m]
Wall-to-plan area ratio [-] ~ h/w
Roof fraction [-]
Roughness length [m]
Radiation Parameters:
Albedo for roof, wall and roads [-] ~0.1 – 0.2
Emissivity for roof, wall and roads [-] ~0.85 – 0.98
Thermal parameters:
Volumetric heat capacity of roof, walls and roads.
Thermal conductivity of roof, walls and roads.
Vegetation Parameters:
Vegetation fractions of trees and grass[-]
Monthly green vegetation fraction [-]
Monthly LAI [m2/m2]
Monthly roughness length [m]
Monthly emissivity [-]
Shortwave and NIR albedos
Minimum stomatal resistance [s/m]
Other curve-fitting parameters (RGL, HS, …)
Soil parameters:
Parameters derived from the soil texture
fimperv
fpervious
33. 33
Preston Site (2004):
Impervious fraction:
62% → 50%
Tree fraction:
23% → 40%
Grass fraction:
15% → 10%
Significantly dry
summer (33mm in
the period assessed)
Discussion of scales
Cooling effectiveness calculation