Environment Canada, Meteorological Service of Canada, 1 Meteorological Research Branch 2...

13
Environment Canada, Environment Canada, Meteorological Service of Meteorological Service of Canada, Canada, 1 Meteorological Research Branch Meteorological Research Branch 2 Environmental & Emergency Response Environmental & Emergency Response Div. Div. A. A. Lemonsu Lemonsu 1 , A. Leroux , A. Leroux 2 , , S. Bélair S. Bélair 1 , S. , S. Trudel Trudel 2 and J. Mailhot and J. Mailhot 1 CRTI Project # 02-0093RD

Transcript of Environment Canada, Meteorological Service of Canada, 1 Meteorological Research Branch 2...

Environment Canada, Environment Canada, Meteorological Service of Canada, Meteorological Service of Canada,

11 Meteorological Research Branch Meteorological Research Branch

22 Environmental & Emergency Response Div. Environmental & Emergency Response Div.

A.A.LemonsuLemonsu11, A. Leroux, A. Leroux22, , S. BélairS. Bélair11, S. Trudel, S. Trudel22 and J. Mailhot and J. Mailhot11

CRTI Project # 02-0093RD

An urban canopy parameterization was recently implemented in the GEM and MC2 Canadian mesoscale models

GEM and MC2 currently use a 1 km global classification including 1 urban class imported from DCW (Danko, 1992)

Specific urban cover classifications are required in order to describe the spatial distribution and spatial variability of urban areas

Scientific Context

ObjectiveDevelop a general methodology to provide new urban land-use land-cover classifications for major Canadian cities (extended to North American cities)

The methodology is based on a joint analysis of remote sensing data and digital elevation models in order to take both surface characteristics and building height into account

Mesoscale urban modeling applications (and possibly future applications to weather forecasting) require:

Free of charge or low cost data

Data covering large urban areas

Data available for major North American cities

Possibility of future updates

General Approach

Data Sources

City Data Spatial Res. Date

OKCASTER L1B 15 m 2001-07-21

NED (for US) 1/3 arc-sec

SRTM-DEM (for the world) 1 arc-sec

MTL

Landsat-7 30 m + 15 m (Pan) 2001-06-08

CDED1 (for Canada) 1:50,000

SRTM-DEM (for the world) 3 arc-sec

VAN

ASTER L1B 15 m 2001-08-09

CDED1 (for Canada) 1:50,000

SRTM-DEM (for the world) 3 arc-sec

Classifications were produced for Oklahoma City (OK, US), Montreal (QC, Canada) and Vancouver (BC, Canada)

NED National Elevation DatasetSRTM Shuttle Radar Topography MissionCDED Canadian Digital Elevation Data

15-m building height database

SRTM-DEM minus NED/CDED1

Elevation for built-up pixels

General Methodology

15-m unsupervised classification

ASTER/Landsat-7

40 built and vegetated simple elements regrouped in 11 simple elements/criteria

1-Excluded covers2-Water3-Trees4-Low vegetation5-Grass6-Soil and rocks

7-Roof 8-Concrete 9-Asphalt10-Veg/built 111-Veg/built 2

60-m LULC classification

1-Excluded2-Water3-Soils4-Crops5-Short grass6-Mixed forest7-Mixed shrubs

1-High buildings 2-Mid-heigh buildings 3-Low buildings 4-Very low buildings 5-Industrial areas 6-Sparse buildings 7-Roads and parkings 8-Road mix 9-Dense residential10-Mid-density residential11-Low-density residential12-Mix nature and built

60-m aggregation of classification

criteria

1- Excluded covers 2- Water 3- Trees 4- Low vegetation 5- Grass 6- Soil and rocks 7- Roof 8- Concrete 9- Asphalt10- Veg/built 111- Veg/built 2

12-Built

13-Built214-Height

Decision tree

Unsupervised classification

ExcludedWaterTreesLow vegetationGrassSoil and rocks

RoofConcreteAsphaltVeg/built 1Veg/built 2

OKC ASTER image 2001-07-21

N

Building heights Bald Earth’s topography provided by NED (US) and CDED1 (Canada)

Total elevation provided by SRTM-DEM

Height of roughness elements estimated from: SRTM-DEM minus NED SRTM-DEM minus CDED1

Building heights estimated by considering only built-up pixels

Offset correction

Threshold of 3 m applied to minimum building heights

OKC SRTM-DEM minus NED

N

15-m building height database

SRTM-DEM minus NED/CDED1

Elevation for built-up pixels

General Methodology

15-m unsupervised classification

ASTER/Landsat-7

40 built and vegetated simple elements regrouped in 11 simple elements/criteria

1-Excluded covers2-Water3-Trees4-Low vegetation5-Grass6-Soil and rocks

7-Roof 8-Concrete 9-Asphalt10-Veg/built 111-Veg/built 2

60-m LULC classification

1-Excluded2-Water3-Soils4-Crops5-Short grass6-Mixed forest7-Mixed shrubs

1-High buildings 2-Mid-heigh buildings 3-Low buildings 4-Very low buildings 5-Industrial areas 6-Sparse buildings 7-Roads and parkings 8-Road mix 9-Dense residential10-Mid-density residential11-Low-density residential12-Mix nature and built

60-m aggregation of classification

criteria

1- Excluded covers 2- Water 3- Trees 4- Low vegetation 5- Grass 6- Soil and rocks 7- Roof 8- Concrete 9- Asphalt10- Veg/built 111- Veg/built 2

12-Built

13-Built214-Height

Decision tree

Decision tree

DECIDUOUS BROADLEAF TREES

SHORT GRASSAND FORBS

LONG GRASS

CROPS

MIXED WOODFOREST

VEGETATION CLASSES

LAND/SEA MASK

WATER<80%

BUILT>10%

WATER

ROADS>80%

VEGETATION

BUILT2>20%

ELEVATION>20 m

ROADS ANDPARKINGS

ELEVATION>30 m

ELEVATION>10 m

HIGHBUILDINGS

MID-HIGHBUILDINGS

LOWBUILDINGS

VERY LOW BUILDINGS

RESID + MIX>40%

ELEVATION>30 m

HIGHBUILDINGS

ELEVATION>20 m

MID-HIGHBUILDINGS

ELEVATION>10 m

LOWBUILDINGS

BUILT2>20%BUILT>45%

VERY LOWBUILDINGS

ROADBORDERS

BUILT>70%

BUILT>70%

BUILT2>20%

BUILT>40%

ELEV>10 mBUILT2>20%

ELEV>10 mBUILT2>20%

SPARSEBUILDINGS

LOW VEG+GRASS>60%

DENSERESIDENTIAL

MID-DENSITYRESIDENTIAL ELEVATION

>20 m

ROADBORDERS

ELEVATION>30 m

ELEVATION>10 m

HIGHBUILDINGS

MID-HIGHBUILDINGS

LOWBUILDINGS

VERY LOWBUILDINGS

SPARSEBUILDINGS

ROADBORDERS

LOW-DENSITYRESIDENTIAL

BUILT &NATURE MIX

ROOF>20%

Urban classificationOKC60-m resolution classification

N

High buildingsMid-high buildingsLow buildingsVery low buildingsSparse buildingsIndustrial areasRoads and parkingsRoad mixDense residentialMid-density residentialLow-density residentialMix of nature and built

SoilsCropsShort grassMixed forestMixed shurbsWater

Excluded

N

N

Montreal60-m resolution classification

Vancouver 60-m resolution classification

High buildingsMid-high buildingsLow buildingsVery low buildingsSparse buildingsIndustrial areasRoads and parkingsRoad mixDense residentialMid-density residentialLow-density residentialMix of nature and built

Zoom

Interest of the methodology

Limited number of data sources

Large availability of the databases

Time processing of about 1 week

General and robust approach applicable to any Canadian city (can be extended to North American cities)

Urban classifications

Horizontal resolution adapted to meso-gamma-scale modeling

Good identification of the major urban landscapes

Number of urban classes allowing a satisfying representation of urban cover variability

Conclusions

New approach based on analysis of the vector National Topographic DataBase (NTDB):

Cover of the entire Canada

High resolution: scale of 1:50000 and 1:250000

Large number of urban features

No manual processing/correction

No interpretation

Future works

Montreal, NTDBExample of urban features

Roads

Bridges

Highways

Rails

Sparse buildings

Buildings

Residential areas

Vegetation

Golf

Water