Академический Документы
Профессиональный Документы
Культура Документы
views or
policies of the Asian Development Bank (ADB), or its Board of Directors, or the governments they represent. ADB does not
guarantee the source, originality, accuracy, completeness or reliability of any statement, information, data, finding,
interpretation, advice, opinion, or view presented, nor does it make any representation concerning the same.
Over the past decade, the Asia and Pacific region has shown a rapid decline in income-
based poverty as a result of pro-poor economic growth. In addition, social indicators of living
standards improved substantially.
The underlying economic development, however, has had significant impacts on natural
resources and environmental services that fuel much of this growth. Increased resource
demand and unsustainable exploitation practices results in a decline in natural resource
quality and productivity across the region, and increased the risk of natural hazards and
disasters.
Hence, environmental poverty (i.e. the spatial distribution of poverty tied to resource
dependencies and risk exposures of the poor) is likely to become the main form of poverty in
the near future. It is estimated that by 2020, more than two-thirds of the vulnerable and poor
population in the region will suffer from environmental poverty—up from less than one-half
today 2 . More investments are particularly required for rural dryland areas and urban slums. It
is estimated that worldwide USD 60-90 billion are needed annually to address environmental
poverty, and additional USD 80 billion to tackle climate change. About 60-80% of these costs
would be needed for the Asia and Pacific region.
1
Lothar Linde is a GIS expert for the ADB Greater Mekong Subregion (GMS) Core Environment Program (CEP-
BCI, RETA6289). Armin Bauer is senior economist in the Asian Development Bank..
2
See: (1) ADB (2008): Environments of the poor. New Perspectives on Development Planning.
http://www.povertyenvironment.net/?q=filestore2/download/1829/PEP13-EnvironmentofthePoor-Bauer.ppt;. (2) A.
Bauer (June 2008): The Environments of the Poor - Summary of a forthcoming book on "Environmental Poverty”.
http://www.povertyenvironment.net/pep/?q=filestore2/download/1848/Bauer%20_June%2008_%20The%20Envir
onments%20of%20Poverty%202.pdf. (3) A. Bauer et al (forthcoming): The Environments of the Poor. Manuscript
of a book for ADB.
The Environments of the Poor is a geographic concept influenced by many spatial
parameters such as terrain, climate, hydrology, and population distribution. Translating this
into a Geographic Information System (GIS) adds geographic context and improves
statistical data presentation, spatial data integration and analysis.
GIS integration of such a complex concept is not a “one-size-fits-all” approach though. Each
topic and stakeholder has different demands with regard to thematic content and detail,
spatial accuracy, and level of comprehension of and interaction with GIS data and outputs.
In the early planning stages (visions, strategies, regional plans and program) GIS outputs
primarily need to serve a better regional understanding of broader geographic dimensions
and associations. At this level, generalization (trends) and efficient distribution of the data
among large stakeholder groups is more important than detail. Further on in the planning
process (national plans and program), decision makers and planning experts need scalable
GIS tools that can accommodate increased thematic and spatial detail to aid spatial
prioritization (e.g. vulnerable areas). Finally, the evaluation of national interventions and local
investments is the domain of dedicated thematic modeling tools that can answer very
specific questions at maximum detail.
Along this line of thinking, the paper introduces three examples that translate the
Environment of the Poor concept into a Geographic Information System (Table 1).
Table 1: Stages of (Environments of the Poor) planning and corresponding GIS approaches
and tools introduced.
Each of these browsers serve as an interface to read and render file formats that actually
contain the geographic content (data). While some of them directly build on Desktop GIS
standards such as ESRI ArcGIS (Shapefiles, Grid), the geodata format most commonly used
on the net today is the Google-introduced Keyhole Markup Language (KML), which has
been specifically developed for the use with public geodatabrowsers rather than analytical
Desktop GIS software.
While most geodatabrowsers (and Desktop GIS software) can display KML data, Google has
practically built its Google Earth browser around the language. That makes it easy for
developers to prepare corresponding contents, and the intuitive user interface of Google
Earth simplifies the browsing experience. Together with Google’s omnipresence in the
internet, this has helped to quickly establish a large community of users as well as
contributors – essential for the long term success of a particular technology. As a result,
KML has quickly become the quasi-standard for sharing geodata through the internet.
Geodatabrowsers traditionally have their strength in distributing raw geographic data (e.g.
satellite image mosaics, terrain maps) and basic administrative and infrastructural contents
(e.g. borders, roads, major towns). More detailed thematic information – like different
population and environmental poverty statistics – is better displayed through a combination
of geographic location and charts. The corresponding symbology however requires more
sophisticated customization options than most geodatabrowsers support, and hence has
largely remained a domain of stand-alone maps created by desktop GIS.
Currently, both the Google Earth (Figure 2) and the Google Maps application (Figure 3, 4)
hold the following 11 thematic layers (all currently displaying the information as pie charts)
developed out of the statistical information of the Environments of the Poor research:
Population
1. Rural and Urban Population (2005, percent and million people)
2. Rural and Urban Population (Projection for 2020, percent and million people)
Poverty (rural and urban)
3. Individual national poverty lines (2005, million people)
4. USD 1.25 as per PovCalNet (2005, million people)
5. USD 1.25 as per PovCalNet (Projection for 2020, million people)
6. USD 2 as per PovCalNet (2005, million people)
7. USD 2 as per PovCalNet (Projection for 2020, million people)
Poverty (in the Environments of the Poor):
8. USD 1.25 (2005, in million people)
9. USD 1.25 (Projection for 2020, in million people)
10. USD 2 (2005, in million people)
11. USD 2 (Projection for 2020, in million people)
Figure 2: Google Earth Desktop Geodatabrowser displaying Environment of the Poor
statistical data as pie charts for each country. Clicking on a pie chart symbol triggers a
callout box with an enlarged chart including labels and legend.
Figure 3: Web-based, interactive Google Maps Geodatabrowser displaying EOP data as
pie charts using a combination of HTML, JavaScript, KML, Google Charts API, Google Maps
API and the Google Earth Plugin.
Figure 4: Clicking on a pie chart returns an enlarged version with labels and legend (both
Google Maps and Google Earth Plugin mode).
Interactive geovisualization of statistical data from literature research and expert opinions is
a critical first step towards a better understanding of spatial associations. It also provides
easier access to the data itself through web-based distribution.
Geovisualization, however, is only one step (normally the last) in a GIS workflow. Particularly
for intrinsically geographic concepts such as the Environments of the Poor, a GIS can and
should be used as an analytical tool to contribute to the generation of parts or all of the
desired statistical data – functionality that only a Desktop GIS can provide. The following
section explains such an application.
The translation of the Environments of the Poor into a GIS map starts with identifying the
geographic references contained in the concept. Two distinct geographic references can be
identified: 1) 40 countries in the Asia and Pacific Region, and 2) six distinct geographical
areas in which poor people live, of which one is related to pro-poor growth potential areas
where poverty is mainly caused through lack of pro-poor growth. Five other spatiual areas
are related to environmental poverty categories that are each defined by one or more
geographically explicit poverty driver, namely dryland poverty, upland poverty, flood affected
wetland poverty, coastal poverty, and slum poverty.
The 40 countries are the simplest form of geographic reference, with a clearly defined
geographic level of spatial disaggregation. This reference, together with statistical base
information, was used as a base for the interactive geographic browsing tool introduced in
the previous chapter. The five environmental poverty categories, in comparison, are defined
by non-administrative geographic factors, introducing flexibility of scale and room for spatial
disaggregation beyond the national level. The Environment of the Poor report defines them
as follows as follows:
• Dry-land poor are those living in arid and desert land [dryland] areas […] including
desert areas in northern Asia, especially China, Pakistan, Central Asia, and northern
India; degraded, salinized, and dry land (some caused by desertification);
• Upland poor are those living in upland areas [that are] mountainous, remote,
forested [and] inhabited by small-scale farmers and mountain people; deforestation
[and other land conversion / degradation processes contribute to poverty pressure];
• Coastal poor are those living adjacent to coasts and dependent upon coastal and/or
marine resources […] affected by degraded fish stocks (typically within five
kilometers of the sea) and areas flooded due climate change (such as Bangladesh,
and the Pacific Islands); areas frequently subject to floods and natural disasters 3 ;
• Flood-affected wetland poor are those in wetland areas who are […] experiencing
frequent flooding (such as in the Mekong and Ganges delta, or around the Yellow
river) 4 […]; and
• Slum poor are those living in substandard [urban] settlements with high exposure to
urban pollutants.
Following those compatibility guidelines – particular for such a large study area as Asia and
the Pacific – naturally limits the options on GIS layers suitable for such an assessment. For a
first approximation of the extent of the environment of the poor categories, five GIS layers
were identified that are covering key geographic drivers at sufficient thematic and spatial
accuracy (Table 2).
3
Coastal area floods referred to in the concept are assumed to be sea-born (e.g. sea level rise, marine storms
and corresponding inland surges)
4
Wetland area floods referred to in the concept are assumed to be river-born (e.g. downstream riverine floods
from upstream rainfalls)
Table 2: GIS datasets available and used for the development of the draft EOP layer.
EOP Dataset used Resolutio Coverag Data Producer / Provider
Category n e
Dryland FAO Global 10 arc min Global UN Food and Agricultural
Map of Aridity Organization (FAO). Download at:
www.fao.org/geonetwork/
Upland SRTM30 30 arc sec Global USGS Shuttle Radar Topographic
Digital Mission (SRTM). Download at:
Elevation dds.cr.usgs.gov/srtm/version2_1/SRT
Model M30
Coastal Coastline, 30 arc sec Global Extracted from the geographic extent
Distance from of: LandScan 2009™ High Resolution
coastline Global Population Data Set. Oak
Ridge National Laboratory/UT-
Battelle, LLC
Wetland Global Lakes 30 arc sec Global Center for Environmental Systems
and Wetland Research, University of Kassel,
Database Germany. Distributed by World Wide
(GLWD) Fund for Nature (WWF) at:
http://web01.worldwildlife.org/science/
data/item1877.html
Urban Global Map of 15 arc sec Global Schneider, A., M. A. Friedl and D.
Urban Extent Potere (2009) A new map of global
from MODIS urban extent from MODIS data.
Data Environmental Research Letters,
Volume 4, article 044003. Request
download at:
http://www.sage.wisc.edu/people/schn
eider/research/data.html
While each of these datasets have global coverage, there are still unavoidable small
differences in the datasets that need to be dealt with before a seamless, aggregated layer of
the Asian and Pacific region can be produced. This includes GIS processing steps to
mosaicking tiles, fill data gaps, delete data overlaps, adjust the spatial resolution, and
perform conversion and raster calculation steps detailed out for each individual layer in table
3. The processed layers ready for analysis are shown in appendix 1 (Figure 12-16).
Also, each of the GIS layers is covering the full thematic value range, i.e. both what’s
contributing to poverty as well as what’s not. For instance, the FAO Aridity Index has a scale
from 0-10, of which only a fraction is actually arid and semiarid (0-0.5). The GLWD covers
altogether 12 classes of which only 3 classes relate to flood land. And the digital elevation
data range from 1-8685m 5 , of which only the higher elevation ranges are actually
mountainous uplands.
5
The elevation dataset averages elevation values within a 1kmx1km area (pixel size), leading to lower values
than the maximum elevation (8848m)
Table 3: Raster processing steps to prepare the individual GIS layers for further analysis.
EOP Dataset Properties Processing steps
Category used
/ LandScan 40 EOP countries at • Preparation of a mask with common
2009 30 arc sec extent, resolution (pixel size) and
Country resolution (1km) projection
Raster provided with (and
aligned to)
LandScan 2009
population dataset.
Dryland FAO Aridity 10 arc minutes • Resampling (bilinear interpolation)
Index (AI) resolution, gaps / to 30 arc seconds,
overlaps • floating point to integer conversion,
• clipping along mask,
• Euclidean allocation to fill gaps
Upland SRTM30 30 arc seconds • Correction of USGS DEM NoData
Digital resolution, gaps / values,
Elevation overlaps • mosaicking of tiles,
Model • clipping along mask,
• Euclidean allocation to fill gaps
Coastal Coastline Same as LandScan • Raster to vector conversion of
Distance 2009 boundary,
• Extraction of sea boundary,
• Euclidean distance from sea
boundary
Wetland Global Lakes 30 arc seconds • Filling of NoData values within EOP
and Wetland resolution, gaps / area
Database overlaps
Urban MODIS 15 arc seconds • Raster calculation (criteria
500m Global resolution, extraction),
Urban Extent projection error • X-axis shift to correct offset (-6
pixels),
• Region group, lookup count and
elimination of small areas (<=6
pixels)
• Filling of NoData values within EOP
area
Therefore, the fraction of values which is poverty-relevant needs to be extracted from each
GIS layer. These criteria need to be defined from literature research and / or expert
consultations, and can be regionally different depending on societal conditions (cultural,
ethnical, political, economic etc.). A significant amount of country level research and GIS
data is needed to make that distinction, so for the purpose of estimating the extent of the
Environments of the Poor in Asia and the Pacific, generalized – i.e. region-wide – criteria
were used as described in table 4.
Table 4: Criteria used to extract poverty relevant values from the GIS layers.
EOP Dataset used Total value range of dataset Value range used to
Category define EOP category
Dryland FAO Aridity AI < 0.05 = Hyperarid AI <= 0.50
Index (AI) 0.05 < AI < 0.20 = Arid
0.20 < AI < 0.50 = Semi-Arid
0.50 < AI < 0.65 = Dry sub-humid
Upland SRTM30 Digital 1 to 8848 meter elevation above sea Elevation >= 600m
Elevation Model level above sea level
Coastal Coastline 0 to 29 decimal degree distance Distance <= 0.415
Distance (equals ~3500km) from coastline (equals <= 50km)
Wetland Global Lakes 1 = Lake Classes 4, 5 and 6
and Wetland 2 = Reservoir
Database 3 = River
4 = Freshwater Marsh, Floodplain
5 = Swamp Forest, Flooded Forest
6 = Coastal Wetland (incl. Mangrove,
Estuary, Delta, Lagoon)
7 = Pan, Brackish/Saline Wetland
8 = Bog, Fen, Mire (Peatland)
9 = Intermittent Wetland/Lake
10 = 50-100% Wetland
11 = 25-50% Wetland
12 = Wetland Complex (0-25%
Wetland)
Urban Global Urban 13 = Urban Class 13
Extent from
MODIS
After extracting the poverty-relevant fractions from the GIS layers, they were aggregated into
one joint GIS layer using the following order: 1) Urban, 2) Wetland, 3) Coast, 4) Dryland, and
5) Upland 6 . The resulting “Environments of the Poor GIS layer” (Figure 5) can be used –
besides for visual interpretation – to summarize the area covered by each environmental
poverty category in Asia and the Pacific. By superimposing regional, country or even
provincial boundaries, these numbers can be further spatially disaggregated, adding detail
and depth to the EOP statistical data and opening up new opportunities for analysis that
might not be available from national statistics and literature.
An additional step towards standardizing the Environments of the Poor analysis through the
use of GIS is the combination of the Environments of the Poor GIS layer with a population
distribution raster. Corresponding GIS data are produced by the Center of International Earth
Science Information Network (CIESIN) of Columbia University (www.ciesin.org) and the Oak
Ridge National Laboratory (ORNL) of UT-Battelle (www.ornl.gov). Since the ORNL
LandScan dataset has the highest resolution (1km) and is being updated and improved
every two years, it was used for this study (Figure 6). The EOP GIS layer was superimposed
over the population distribution raster and corresponding zonal statistics calculated. An
example for a sub-national application is given in table 5.
6
Pro-Poor Growth Area was defined as all area that was left after the aggregation of the other five layers.
Figure 5: Spatial approximation of poverty environments. Urban areas enlarged for better
visibility.
Mapping out national poverty statistics (chapter 3) and delineating and disaggregating
poverty environments (chapter 4) establishes the geographic context and provides additional
data on spatial aggregation and association to the Environments of the Poor analysis. This
serves – and potentially improves – the development of state and trend indicators feeding
into regional visions, strategies and plans. The GIS also simplifies data integration and
facilitates methodological standardization.
Naturally, data integration and standardization comes at the expense of some thematic and
spatial detail. Such generalization is perfectly acceptable – and even required – in early
strategic planning stages. Once planning moves on to developing local interventions and
investments however, thematic and spatial accuracy becomes more important than regional
standardization.
Detailed assessments of driving forces and their interactions are what dedicated thematic
modeling tools plug into. They contribute either into broader suitability or vulnerability
assessments (like the Environments of the Poor), they contribute local, yet very detailed
information on – for instance – floods and droughts (climate and hydrological models), soil
and ecosystem degradation (universal soil loss equation, biodiversity pressure models,
emission / pollution models), or projected land conversion (land demand models). Such
dedicated models can provide in depth information on the present and projected future
distribution of a specific environmental poverty driver.
Among potential poverty drivers, land conversion is a key driver that many other poverty
drivers (like soil erosion, biodiversity loss, ecosystem degradation) partially or fully derive
from. An example for a GIS tool that predicts future land conversion is the CLUE model
(Conversion of Land Use Change and its Effects modelling framework), developed by the
Technical University Wageningen / NL (www.cluemodel.com). Two versions of the model
exist: the “original” CLUE framework which works for global level analysis and the CLUE-s
(…at small regional extend) framework, which is an adaptation fit for land use change
analysis at a regional scale.
Task of the model is to processes and analyse a range of spatial data and non-spatial
parameters and criteria to provide a spatially explicit estimate of near-future land use
changes. To do so, the CLUE-s framework works with four knowledge components (Figure
7):
a) Land use requirements: projected land demand in the future, researched from past
sector development trends and national targets,
b) Land use type specific conversion settings: a matrix that identifies typical land
conversion sequences,
c) Spatial policies and restrictions: identifying areas that are restricted for future land
conversion, such as protected areas, but also army land or land already allocated to
private sector investments, and
d) Location characteristics: a set of underlying “explanatory” layers which strong
associations to individual land use types.
While the first both are non-spatial “qualitative” components broadly defining the
development trends in the target region, the latter are location specific components that
provide the “quantitative” dimension of the model and the necessary spatial correlations.
Based on these inputs, the land use change allocation procedure calculates a future land
use map.
A good example to demonstrate the usefulness of a land demand model for province level
impact / vulnerability assessments is the Strategic Environmental Assessment of the Quang
Nam 7 Land Use Plan 2011-2020 8 . Quang Nam Province was chosen over other provinces
because of its challenges to harmonize between rapid economic growth and protecting
underlying natural resources and dependent livelihoods.
In the East of Quang Nam province, the coastal plains have been transformed into intensive
farming systems and aquaculture. As coastal population increases, so does competition for
additional land for subsistence farming and agro-businesses. Parallel, the neighboring city of
Da Nang has emerged as Central Vietnam’s largest business hub, catalyzing particularly the
development of manufacturing industry (e.g. furniture) and related transport networks.
Much of these developments profoundly change the forest ecosystems that the upland poor
depend on. Conversion of forests into agriculture in steep terrain might compromise
agricultural productivity (soil degradation, erosion) and increase disaster risk (landslides from
exposed slopes, flash floods). Mining activities, often illegal, contribute to water
contamination affecting freshwater fish stocks. Immigration from the coastal plains seeking
new land for sustenance farming and agro-businesses even exposes the upland poor to
competition for their ancestral land. Without providing appropriate livelihood alternatives to
mitigate these pressures, land conversion might aggravate upland poverty rather than
reducing and eventually eradicating it.
Therefore, being able to anticipate where total projected land demand is likely to trigger land
conversion can help identifying which upland poor communities are potentially facing
particularly high land conversion pressure in the future, based on which field surveys can
scope out appropriate mitigation measures.
CLUE-s in the SEA of the Quang Nam Land Use Plan 2011-2020
After introducing and raising awareness on the model with provincial authorities, GIS base
data were collected and processed to populate the spatial components of the CLUE-s
model. Present land cover / use was derived from AWIFS satellite imagery of 2007 (eight
classes, Figure 8). A layer of spatial restrictions was generated from information on
protected areas (special use forest) and biodiversity conservation corridors. The location
characteristics were configured with eight layers (Figure 9): elevation, slope, aspect, cost-
distance to coast, cost-distance to road, cost-distance to rivers, population density, and cost-
distance to settlement.
7
8
Quang Nam Province, Viet Nam.
This activity is supported by the Asian Development Bank Greater Mekong Subregion (GMS) Core
Environment Program (CEP‐BCI, ADB RETA6289) to provide the Vietnamese government with technical support
and training to achieve socially inclusive economic growth through environmentally sound development
planning.
Figure 8: Input land cover / use of 2007, derived from AWIFS satellite imagery.
Figure 9: Input land cover / use of 2007 and underlying layers used to establish
correlations explaining the distribution of the individual land use types.
Individual consultations with the LUP writing team identified past development trends and
present sector demands, based on which two broad land demand scenarios were
formulated: A ‘business as usual’ baseline scenario, focusing on maximised agricultural
development and assuming that spatial restrictions – despite existing – are not enforced
efficiently. As an alternative, an “environmentally and socially balanced” scenario puts
emphasis on lowering agricultural land demand through productivity enhancement
measures. It also internalizes demand in environmental and social services for economic
tools such as ecotourism and PES/REDD (e.g. protected areas, biodiversity corridors).
This information was used to configure and run the CLUE-s model. Projected land
conversion (business as usual scenario) until 2020 (Figure 10) shows land conversion rates
steeply increasing at the fringe between the coastal lowlands and the upland areas, and then
again decreasing towards more remote mountain areas, particularly low along the Laotian
border (Figure 11). This general pattern is locally overridden by the alignment of the
mountain valleys, acting as entry points for land conversion. According to the first model
outputs, 74.8% (or 92.500ha) of the total land converted until 2020 (business as usual)
would fall into upland poverty communes with a poverty incidence of 60% or higher. More
disaggregated figures (districts, communes, watersheds) can be calculated using GIS zonal
statistics, opening up interesting new data and possibilities to social and environmental
assessments.
Figure 10: CLUE-s output: Projected land cover / use in 2020 (business as usual.
Figure 11: CLUE-s output: Areas projected to be converted until 2020, overlaid on
upland poverty communes (>60% poverty incidence).
References
Figure 12: FAO Global Map of Aridity, adjusted to the EOP target area.
Figure 13: SRTM 30 Digital Elevation Model, adjusted to the EOP target area.