An Exploration of using Nighttime Satellite Imagery from the DMSP OLS for Mapping Population and Wealth in Guatemala Paul C.Sutton [email protected] Department.

An Exploration of using NighttimeAn Exploration of using NighttimeSatellite Imagery from the DMSP OLSSatellite Imagery from the DMSP OLSfor Mapping Population and Wealth for Mapping Population and Wealth

in Guatemalain Guatemala

Paul C.SuttonPaul C.Sutton

[email protected]@du.edu

Department of GeographyDepartment of Geography

University of DenverUniversity of Denver

Outline• Motivation: Why do this? Is is worthwhile?

• Brief Summary of DMSP OLS image processing

• How Nighttime images can be used to map urban areas and estimate urban populations.

• How Nighttime images can be used to estimate and map economic activity

• How Nighttime images may be used to estimate Human Impact on the environment

Why Use Nighttime Imagery to Map/Model Why Use Nighttime Imagery to Map/Model Demographic and Socio-Economic Phenomena?Demographic and Socio-Economic Phenomena?

• Social, Economic, and Behavioral Demographic Data are the major gaps to be filled in globally integrated geo-information

• Existing Information is degrading due to increasing human mobility, and the fact that a growing proportion of the earth’s population live in developing countries which can’t afford to conduct accurate censuses

• Spatially referenced demographic information is a vital component of studies of: Hazard Planning and Response, Sustainability and Development Issues, and countless other cross-disciplinary investigations

Two sun-synchronous polar orbiting satellites (865 km orbit)

Observations at 1) ~ Dawn & Dusk, 2) ~ Noon & Midnight

Pixel Size: smoothed ~2.4 km2, fine ~ 0.5 km2, Swath Width ~3000 km

Two Bands: 1) Panchromatic VNIR, 2) Thermal Infrared

Dynamic Range: VNIR more than 4 orders of magnitude larger than traditional sensors optimized for daytime observation

(e.g. sees light from reflected moonlight to reflected sunlight)

Data available from early 1970’s to Present, Digital Archive est. in 1992

Data Products derived from imagery (hyper-temporal mosaicing):

% cloud cover, % light observed, Fires, Lantern Fishing, Gas Flares, City Lights, Radiance Calibrated City Lights, Atmospherically corrected radiance calibrated city lights

System OverviewDefense Meteorological Satellite Program Operational LineScan System (DMSP OLS)

Example of Cloud Screening over ItalyExample of Cloud Screening over Italy

VNIR over Italy Thermal over Italy

A comment on aggregation & scale: This is a 1 km2 pixel in Denver, Colorado

Nighttime Satellite ImageryNighttime Satellite Imagery

Fires, Fishing, Flares, & City Lights

Lantern FishingIn Japan City Lights along the Nile

Forest fires in AustraliaGas Flares in the

Persian Gulf

Mapping Population• Ln(area) vs. Ln(pop) regression method for

Estimating the Population of Urban Clusters• Intra-Urban measures of population density:

Light Intensity as a proxy for Population Density

• Works better in countries with high % of population in urban areas.

• Rural Electrification in Guatemala probably reduces utility of these methods.

Light Intensity from DMSP OLS imagerymatched with Photographs using GPS

1 ) C e n tr a l G u a te m a la C ity (D N = 5 0 0 ) 2 ) S a n J u a n Ix c o y, H u e h u e . (D N = 4 6 )

3 ) S a n ta E u la lia , H u e h u e . (D N = 3 1 ) 4 ) C o b a n , A lta Ve r a p a z (D N = 1 8 1 )

5 ) F lo re s , P e te n (D N = 3 0 1 )

6 ) N a h u a la , S o lo la (D N = 1 4 8 )

7 ) S o lo m a , H u e h u e . (D N = 8 7 )

8 ) N o r th e r n G u a te m a la C ity (D N = 2 1 5 )

Night Lights and Pop Den in Guatemala

L ig h t I n te n s ity fr o m D M S P O L S

Pop

u lat

ion

De n

sity

f r

om L

and S

can

R = .592

L n(population) = 3.047 + 1 .1463L n (A rea)

P ix e l R e g ress io n(P t s a re u P ix e ls )

D M S P O L SN ig h tt im e Im a g eo f G u a tem a la

P o p u la t io n D en sityo f G u a tem a la fro mL a n d S ca n

Mapping Economic Activityand GDP per Capita

• Just as Night Lights are a proxy measure of population they are also a proxy measure of Economic Activity.

• Again, relationship is far from perfect (see next slide); however, Light intensity can be used as a proxy measure of GDP.

• GDP of Guatemala ~50 Billion; 25% apportioned to dark area to account for agriculture, remaining 75% apportioned based on light intensity.

• Using LandScan Population Density dataset and dividing it into this map of GDP produces a map of GDP per Capita.

Scatterplot of Night Light Energy & PPP of GDP for 208 nations

Global map of Marketed Economic Activity as measured by Nighttime Satellite Image Proxy

Dividing‘map’ of

GDP from

DMSPby PopDensity

from Landscan

0 -9 9 91 ,0 0 02 ,0 0 03 ,0 0 04 ,0 0 05 ,0 0 06 -1 0 ,0 0 011 -5 0 ,0 0 05 1 -1 0 0 ,0 0 0O v e r 1 0 0 ,0 0 0

G u a te m a la’s G D P p e r C a p ita

(U .S . D o lla rs )

G D P per C ap ita in G u a tem ala

a t ~ 1 k m sp atia l re so lu tion

2

Using Nighttime Imagery to Create an “Environmental Sustainability Index”

• Measure Environmental Endowment of Nations using Ecosystem Service Value of Nation’s Lands

• Measure Human Impact of Nation from DMSP OLS nighttime Image

• Divide The above measures to create and ESI (Environmental Sustainability Index)

Measuring Human ‘Impact’• What data can be used in the I = P*A*T formulation?• If you use Population for P, GDP/Capita for Affluence,

and CO2 Emissions/GDP for Technology, then ‘Impact’ simplifies to total CO2 emissions

• Daily & Ehrlich used Energy Consumption per Capita to capture the A*T

• “Impact” is a function of both population size and individual consumption levels

• Nighttime Imagery from the DMSP OLS correlates with Population, Energy Consumption, CO2 emissions, and GDP and may be the best spatially explicit, single variable, measure of ‘Impact’

Ecosystem Service Valuation:IGBP to Nature Conversion Table

IGBP Class Nature Paper Interpretation Value IGBPcode MoneyEvergreen Needleleaf Forest Temperate Forest 302 1 302Evergreen Broadleaf Forest Tropical Forest 2007 2 2007Deciduous Needleleaf Forest Temperate Forest 302 3 302Deciduous Broadleaf Forest Temperate Forest 302 4 302Mixed Forest 25% Tropical, 75% Temperate 728.25 5 728.25Closed Shrublands Grass/Rangelands 232 6 232Open Shrublands Grass/Rangelands 232 7 232Woody Savannas 50% Temperate Forest, 50% Grass/Rangelands 267 8 267Savannas Grass/Rangelands 232 9 232Grasslands Grass/Rangelands 232 10 232Permanent Wetlands 50% Tidal Marsh/Mangrove, 50% Swamp/Floodplain 14785 11 14785Croplands Cropland 92 12 92Urban Urban N/A 13 0Cropland Natural Vegetation Mosaic 50% Cropland, 50% Grassland/Rangeland 162 14 162Snow and Ice Ice/Rock N/A 15 0Barren or Sparsely Vegetated Desert N/A 16 0Water Bodies Lakes/Rivers 8498 17 8498

Global map of ‘Non-Market’ economic activity from ecosystem services

Deriving The Eco-Value / Night Light Energy Environmental Sustainability Index

This index is similar to the inverse of population density e.g. ‘square kilometers of land per person’

However; ‘square kilometers of land’ is adjusted by the land’s ecosystem service value; and, ‘per person’ is measured by the nighttime satellite imagery provided by the DMSP OLS

N a tio n a l In d e xVa lu e

Va lu e o f g iv en N ation ’s E co sy stem S erv ice s as es tim atedb y C o stan za a n d m ea su red b y

U S G S 1 k m 2 G lob al L an d C ov er G r id

A m ou n t o f L igh t E n ergy seen in N ig h ttim e S a te llite Im ag ery fro m

D efen se M eteo ro lo g ica l S a te llite P ro gram ’sO p eratio n a l L in escan S ystem (D M S P O L S )

A representation of the datasets used to calculate Eco-Value and Impact from around Central America

G lo b a l 1 k m I G B P L a n d -C o v e r D a ta se t 2 D M S P -O L S ‘E a r th a t N ig h t’ d a ta s e t

G u atem ala

E l S a lva d or

H on d u ra s

N icara gu a

C os ta R ica

E v e rg r ee n N ee d le lea f F o re stE v e rg r ee n B ro a d lea f F o res tD ec id u o u s N e e d le le a f F o r es tD ec id u o u s B ro a d le a f F o re stO p en S h r u b la n d sC los e d S h r u b la n d sW o o d y S a v a n n a sG ra s s la n d sP e rm a n e n t W etl an d sC ro p la n d sU rb a nC ro p la n d / N a tu r a l Ve g et a t ion W a te r

C o u n t r y P o p u l a t i o n ( 1 9 9 6 ) E c o - V a l u e / N i g h t L i g h t L o c a l R a n kB e l i z e 2 2 4 , 0 0 0 2 6 1 , 3 0 6 6

N i c a r a g u a 4 , 3 5 1 , 0 0 0 1 8 4 , 3 0 8 5H o n d u r a s 5 , 7 5 1 , 0 0 0 9 7 , 0 9 3 4

G u a t e m a l a 1 1 , 2 4 1 , 0 0 0 6 2 , 0 8 5 3C o s t a R i c a 3 , 4 6 6 , 0 0 0 2 4 , 9 5 9 2

E l S a l v a d o r 5 , 9 3 5 , 0 0 0 9 , 8 9 6 1

Conclusions• DMSP OLS nighttime imagery shows a great

deal of promise for myriad applications such as population estimation, mapping of economic activity, and measuring human impact on the environment.

• More Validation and fine tuning of models is needed.

• Issues of spatial scale of measurement still problematic.