DEPARTMENT OF GEOMATIC ENGINEERING From 1m to the planet: mapping global population using Earth Observation: An Overview Jan-Peter Muller Professor of.

DEPARTMENT OF GEOMATIC ENGINEERING

From 1m to the planet: mapping global population using Earth Observation: An

Overview

Jan-Peter Muller

Professor of Image Understanding and Remote SensingMODIS & MISR Science Team Member (NASA EOS Project)

POLDER (ADEOSI/II), VEGETATION (SPOT4), GLI (ADEOSII)

Workshop on Gridding Population Data Columbia University, New York

Tuesday 2nd May 2000


Overview How can we use remote sensing to measure what is of

relevance to global population mapping? Current Space Earth Observation platforms of relevance Exemplar 1: MODIS & VEGETATION for land surface

BRDF/albedo for mapping large-scale urban structures Exemplar 2: ERS SAR interferometry for mapping urban land

use and land use change at the sub-hectare scale Exemplar 3: IRS-1C for mapping urban land use for European

cities (EU MURBANDY) Exemplar 4: IKONOS-2 for automated building detection and

extraction (including height) Exemplar 5: POLDER & MISR for mapping aerosol sources Exemplar 6: MOPITT for mapping CO CEOS-IGOS Project proposal: Global mapping of urban

population


Current Space Earth Observation platforms -

a sample Moderate resolution (≥250m IFoV)

– DMSP/OLS (1974- )– CNES/EU VEGETATION (3/99-12/03)– NASDA ADEOS-POLDER (11/96-6/97) & ADEOSII-POLDERII

(12/01-12/05)– NASA EOS/Terra+Aqua-MODIS (3/00-3/08)– NASA EOS/Terra-MISR (3/00-3/06)

High resolution (≥5m IFoV)– LANDSAT-TM (5:1984-2000; 7:1999- ): 30(/15) m– ERS-SAR (1:1991-2000; 2:1995- )– SPOT-XS/PA (1,2,3,4:1984- ): 20m/10m)– IRS-1C (1997- ): 5.8m

Very High resolution (≤5m IFoV)– IKONOS-2 (≥9/99 ->)


How can we use remote sensing to measure what is of relevance to global

population mapping? Currently civilian remote sensing can not be used to identify the location of people so have to infer anthropogenic activity from….

Land cover, land use and land use changes (including 3D)– Traditional remote sensing techniques employ classification of single or

multiple (time series) of multi-spectral images» Problem is mixed signals from gardens & parks» Requires initial training sites to identify urban signals (very costly)

– SAR (Synthetic Aperture Radar) interferometry (IfSAR) has strong reflection from “hard targets” such as buildings, roads or railways. However, visibility of objects is illumination angle dependent. IfSAR v. good for fragmentation.

– BRDF/albedo related to land surface object geometry (e.g. urban, deforestation, desertification) and so exploit multi-angle RS signatures

– Night-time lights only visible where and when cloud cover is absent and where signal sufficiently strong to be detectable & difficult to interpret

– 3D elevations with very high resolution images can be used to extract automatically individual 3D built settlement locations

Anthropogenic activity - link to global change & human health

– Aerosol PM (Particulate Matter) sources and sinks– Carbon Monoxide & Methane mapping


Exemplar 1: SPOT4 VEGETATION. Multi-angle/date BRDF retrieval Urban areas have high albedo due

to man-made materials which enhance the “urban heat island” and pose the greatest threat to human health in the short-term

Albedo determined through BRDF (Bi-Directional Reflectance Distribution Function)

Three band colour composite of Red channel for July 1999 using semi-empirical kernel BRDF model (courtesy of M. J. Barnsley, UWS)

– Nadir-equivalent (Red)– Geometric kernel (Green)– Volumetric kernel (Blue)

Urban areas well identified at 1km Extension to global unlikely due to

the high cost of the data


Exemplar 1 : MODIS BRDF-derived Nadir reflectance & Albedo. Global maps at 1km being produced every 16 days


Exemplar 2: Interferometric SAR for built environment area

mapping: The UK LANDMAP Project to create a DEM

and orthorectified LANDSAT, SPOT & ERS IfSAR Interferometric SAR excellent for

mapping of Built environment– Potential down to resolutions of 30m ~

0.3 hectares– Initial test over the British Isles as by-

product of the LANDMAP project– Potential to map settlements where

Night-light data is absent (developing world).

– Can be done daytime and/or night-time– Not dependent on cloud-free conditions

Urban areas are highly coherent especially over longer (inter-seasonal) temporal baselines

Development can be tracked on an annual (or even shorter) timescale


Exemplar 2: False Colour composite and processed classified image of 3 band IfSAR data (Doll & Muller, ISPRS Congress 2000)

Th

ree b

an

d c

olo

ur

com

posi

te o

fP

hase

Coh

ere

nce

(R

ed

)A

mp

litu

de (

Gre

en

)D

iffere

nce

in

Am

pli

tud

e (

Blu

e)


Exemplar 3: EU MURBANDY - application of traditional photo-interpretation to urban land use classification using IRS-1C for Vienna, AU.

20 EU cities selected for pilot project. http://murbandy.sai.jrc.it/


Exemplar 4: DSEM (Digital Surface Elevation Model) from simulated 1m spaceborne stereo-optical sensors with digital map data superimposed. Such DSEMs can

be used to extract single buildings

St Albans, UK≤30m AGL (Red)≈ AGL (Green)≤-3m AGL (Blue


Exemplar 4: IKONOS-2 off-nadir 79cm image over the

London Millennium Wheel (January 2000) ©Lockheed Martin


Examplar 4: IKONOS-2 derived building top height map over central London (©NPA 2000). Accuracy

expected to be ≈1m Zrms)


Exemplar 5: Aerosol Mapping and Monitoring from POLDER @6.7km -

example of DAILY air pollution (PM) maps over Mexico city


Example 5: MISR imaging geometry - multi-angular sampling for surface BRDF, aerosol retrieval and stereo

Cloud-top heights & winds

9 pushbroom cameras4 spectral bands275m off-nadir,250m nadir IFoVSwath≈380kmRepeat time ≈3-9 days


Example of MISR 3-band multi-look (all 9 angles) imagery


Example MISR nadir vs 70ºf view of aerosols


Examplar 6: Examples of anthropogenic and natural urban-atmosphere interactions from MODIS, CERES &

MOPITT

QuickTime™ and aCinepak decompressor

are needed to see this picture.


CEOS-IGOS Project proposal: Global mapping of urban population

At a recent CEOS (Committee on Earth Observing Systems) WGISS (Working Group on Information Systems) meeting held at UCL, UK it was suggested that a new IGOS (International Global Observing System) project should be initiated along the lines of either the global Carbon Inventory or the global Forest Map project. This idea is currently being debated by email.

IGOS are prototype very large-scale projects with active support from the space agencies who provide not only data but often funding for the generation of value-added products. It is not restricted to the academic community but includes government and the private sector

Given the results of this workshop (and especially this session), should we push within CEOS to establish this project?

Is there sufficient scientific justification even given the uncertainties?

DEPARTMENT OF GEOMATIC ENGINEERING From 1m to the planet: mapping global population using Earth Observation: An Overview Jan-Peter Muller Professor of.

Documents

mapping urban land use

global population mapping

land use changes

urban signals

land cover

height exemplar

land surface brdfalbedo

civilian remote sensing