Estimating rural Pashtun settlement population in Arghandab district

Document Sample
Estimating rural Pashtun settlement population in Arghandab district Powered By Docstoc
					                                        Virtual Field Exercise for the GEOINT Professional Spring 2009 Capstone Project




   Estimating rural Pashtun settlement population in Arghandab district, Zabul
        province, Afghanistan: A structured geospatial analytic approach

                                                                     Kevin Stofan*
                    Postbaccalaureate Certificate in Geospatial Intelligence Program, Pennsylvania State University, Department of Geography


Abstract

  The applicability of using areal weighted interpolation and dasymetric mapping of settlement features extracted from Quickbird panchromatic
imagery was assessed for Arghandab district, Zabul province, Afghanistan. Areal weighted values of sample settlement populations showed
significant positive correlation with observed values yielding a Pearson product correlation coefficient of 0.83. Linear regression modeling using
observed settlement population values as the dependent variable and settlement structure area as the independent variable was used to predict
settlement populations throughout the district yielding an R2 of 0.689. Additionally, the use of structured analytic techniques including multiple
scenarios generation, multiple hypothesis generation, quadrant crunching, and analysis of competing hypotheses were used to increase the number of
variables and hypotheses used in the study.

Keywords: Dasymetric, Areal Interpolation, Remote Sensing; Quickbird, Buckeye; Arghandab, Zabul; Afghanistan, Pashtun, Human Terrain.
________________________________________________________________________________________________________________________

1. Introduction                                                                   heterogeneous tribal areas as potential flashpoints, and finally
                                                                                  the targeting of insurgents likely based in rural safe havens.
1.1 Situation                                                                               Recent reports have identified the growing insurgency
                                                                                  as having a rural base (Tellis 2009, Jones 2007). Once the
          Since 2001, the United States and NATO have                             coalition forced Taliban leadership to Pakistan and established
conducted counterinsurgency operations in Afghanistan in an                       the central government in Kabul, efforts to extend the
effort to rid the country of Taliban and Al Qaeda forces and                      government’s reach to rural areas have not existed on both a
build support for the central government in Kabul. Coalition                      service or security standpoint (Jones 2007). Reconstruction
forces have seen a growing insurgency in recent years based in                    efforts in Afghanistan have focused on areas near Kabul and the
eastern and southern Afghanistan’s Pashtun tribal belt with                       Kabul-Kandahar highway. Security forces, both local national
support and safe haven in Pakistan’s northern Pashtun tribal                      and coalition, have failed to maintain a persistent presence in
areas. It has been increasingly reported that shortcomings in the                 rural areas.
coalition effort to defeat the Taliban lie in the lack of presence in                       Tribal affiliation and social dynamics do not completely
rural Pashtun areas (Tellis 2009). Additionally, the lack of                      explain the local population support for the Taliban although it
reconstruction efforts in these areas fails to legitimize the central             does play some role in explaining the insurgency (Tellis 2009).
government by providing tangible evidence of their value and                      Identifying tribal populations in rural areas can certainly benefit
authority (Johnson and Mason 2007). Lack of sustained                             coalition outreach efforts in these areas. With a long history of
presence, i.e. clear and sweep military operations, fail to                       resisting central authority, the Ghilzai tribal federation may have
legitimize local national security forces and in turn fail to secure              a predisposition to supporting the Taliban. Ghilzai tribes have
the rural populace support for the central government.                            long shrugged off central government and refused to participate
Identifying and targeting rural Pashtun populations is key to the                 politically in any Afghan central government (Barfield 2007).
success of the counterinsurgency operations in Afghanistan.                       Additionally, many Ghilzai tribes may feel disenfranchised by
Identifying and targeting rural populations provides several                      the Karzai government which is lopsidedly Populzai Durrani
benefits to current operations including the identification of                    Pashtun. On the other hand, identifying and securing support
tribal populations more susceptible to recruitment and support of                 from Durrani tribal populations, particularly isolated rural
the Taliban, the identification of tribal populations more                        Populzai and Alikozai populations is also critical.
susceptible to coalition cooperation, the need to balance                                   A large part of legitimizing the central government in
reconstruction efforts in rural areas, the identification of                      Afghanistan is providing government services to rural areas in
                                                                                  proportion to population. A large part of the failure in garnering
_________
*Corresponding Author.
E-mail address: kas673@psu.edu (K. Stofan).
                                                                                                                                                                   2
                                 Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


rural support for the central government, and therefore rural                  arid steppe climate with cold arid winters using the Köppen
areas supporting the Taliban, is the lack of services and                      climate classification system.
reconstruction efforts in rural Pashtun areas (Jones 2007).
Although Provincial Reconstruction Teams (PRTs) have
attempted to provide reconstruction and services to rural areas
they rarely reach truly rural areas. Accurately identifying rural                   Afghanistan
                                                                                                                                                             ¯
populations, providing government services, and applying
services proportionally and indiscriminately is key in gaining
support among rural Pashtuns.
         Finally, targeting rural Taliban bastions is essential in
defeating the insurgent threat. Identifying larger rural
populations where support for the Taliban has grown is a key
component of identifying areas requiring security presence.                                                 Arghandab

Establishing a permanent presence in these areas, disrupting
Taliban operations and safe havens, is essential to victory in
Afghanistan.

1.2 Purpose

          The purpose of this study is to identify a simple,                                                                                   Zabul

inexpensive, and accurate population estimate for rural Pashtun
areas of Afghanistan, effectively conducting a remote census of                 0          10         20 Kilometers


these areas. Quantifying, identifying, and mapping these rural                 Figure 1. Arghandab district overview map with provincial and country inset maps.
populations will provide coalition forces, government agencies,
and NGOs with accurate data on rural populations in                            2. Literature Review
Afghanistan.
          Through the use of structured analytic techniques, this              2.1 Pashtun Rural Culture
study will attempt to identify the variables affecting population
dynamics of rural Pashtuns and their geographic distribution                             Pashtuns in rural Afghanistan and Arghandab district
patterns. The intent of this study is to test and validate a tool to           practice an agrarian lifestyle. Settlements are typically collocated
accurately estimate population in rural Afghan settlements for                 with irrigated fields, orchards, and grazing land. Most men
use by agencies conducting operations in rural Pashtun                         perform work as farmers or herders while women and children
Afghanistan.                                                                   produce household goods such as milk and textiles. Settlements
                                                                               are often densely clustered, reflecting the gregarious nature of
1.3 Study Area                                                                 their culture.
           Arghandab district lies in north central Zabul province,
a southern province of Afghanistan (Figure 1). Arghandab
district lies nearly 25km north of the only paved highway in
Zabul province and the provincial capital, Qalat. Arghandab
district consists of 147 rural settlements (AIMS 2005)
concentrated near the Arghandab river valley which flows
northeast to southwest and bisects the district. Transportation
networks in the district consist of unimproved dirt roads. There
is no water or sewage infrastructure in the district and fresh
water is accessed via shallow wells and surface water. The
ethnic composition of Arghandab district is completely Pashtun
with the overwhelming majority of those being from the Ghilzai
tribal federation (Swintek 2009).
           Arghandab district is a mountainous area with valleys
formed from fluvial processes. Settlements in the district are
restricted to valleys where surface water is available for
agriculture and collection. The Arghandab river peaks in the late              Figure 2. Typical settlement structure in rural Pashtun Afghanistan.
spring and summer months due to snow melt from northern
mountainous areas. The Arghandab river basin is classified as an               2.1.1 Pastunwali
                                                                                                                                                  3
                                 Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


          Pashtuns in rural Afghanistan follow the pashtunwali or              provides an inexpensive and safe alternative to conducting a
“the way of the Pashtun”. The pashtunwali is a code of conduct                 census. Studies which identify homogenous land use categories
that places honor, equality, and personal autonomy above all                   are important because this study has distinct land use categories
else. The code stresses equality over the acquisition of money                 with little variation in residential land use type (i.e. single
and wealth and is not conducive to the establishment of a class                family, high density housing).
based economy or culture (Barfield 2007). It is the attributes of                        Literature reviews of population estimation methods in
pashtunwali, and the effect it has on the geographic and                       GIS and remote sensing have been conducted before including a
population patterns, that make rural Pashtuns an ideal culture to              recent survey by Wu et al. (2005). Wu et al. provides an
apply the techniques used in this study. Since most families and               excellent survey of the literature in the discipline dating back to
individuals live relatively similar lives as far as wealth and                 1936. They provide a useful framework for categorizing research
status, the physical manifestation of this lifestyle are settlements           by separating methods into areal interpolation techniques and
and communities that are very similar in demographics.                         statistical modeling methods. They further sub-categorize those
                                                                               techniques into areal interpolation methods that use ancillary
2.12 Pashtun Definition of Space                                               data or not, and statistical modeling methods that use urban
                                                                               areas, land use, dwelling units, image pixel characteristics, or
          As outlined above, rural Pashtun settlements are often               other physical or socioeconomic data as the estimation variable.
densely clustered and located near arable land. Arable land in                 For the purposes of this review the Wu et al. framework will be
Arghandab district is often found in valleys near small streams                used as a reference.
or near the Arghandab River. Most settlements reflect Pashtun                            The methodology for estimation of rural populations in
tribal patterns with a distinct settlement often representing a clan           Afghanistan used in this study does not fit perfectly into one of
or kheyl (Glatzer 2006). This is illustrated by the fact that                  the categories outlined by Wu et al., however, it can be
settlement names are often the actual name of the tribe or clan                considered a hybrid of areal interpolation and statistical
within that settlement. Pashtuns in Arghandab district refer to                modeling methods. The use of empirical census data using the
their home or space as a distinct settlement name (often the                   settlement as the basic unit of measurement and disaggregating
name of a tribe or elder from that settlement) or a group of                   total population to population density at the areal unit (i.e. m2)
settlements which share a similar tribal affiliation and are often             using features extracted from remotely sensed data is consistent
within or near the same physical feature such as a valley or                   with dasymetric mapping. The use of population density
mountain range.                                                                determined through regression analysis derived from the
                                                                               dasymetric methodology above and used to predict settlement
2.2 Geospatial Population Estimation Techniques                                population outside of the model building area is more consistent
                                                                               with statistical modeling methods correlating population with
          The following is a review of population estimation                   land use. The latter methodology is summarized as a function:
methods in GIS and remote sensing that are applicable to this
study. The intent of this review is to identify research relevant to                                          P = Σj Aj * Dj ,
the problem of estimating tribal populations in rural Afghanistan
where no consistent, fine scale census data exists. Although a                 where P is the total estimated population; Aj is the area of land
relative paucity of rural population estimation research exists in             use j (in the case of this study the only land use category); and Dj
the literature (Lo 1980), extensive literature exists on population            is the population density for land use j, determined through
estimation techniques for urban areas dating back to 1936                      regression analysis based on dasymetric mapping of the model
(Wright). Many of the following population estimation                          building subset area (Wu et al. 2005).
techniques are applicable to rural areas with many of the
pioneering research papers utilizing study areas that would today              2.2.1 Dasymetric Mapping
be categorized as rural areas.
          Although a wide variety of population estimation                               Early dasymetric mapping of populations is best
techniques were surveyed, the most important research includes                 characterized by J. K. Wright (1936) in his work in refining
work that attempted to estimate rural population at the small area             population maps of Cape Cod, Massachusetts where existing
level (i.e. distinct settlements of 50-5000 persons) using GIS or              maps showed population densities of townships which consisted
remote sensing techniques which identified homogenous land                     of large swaths of unpopulated area. Wright used an algebraic
use areas and studies which estimated populations of distinct                  function to refine these maps by disaggregating the township
settlements based on empirical census data as a model for                      population figures and applying density variation to subdivisions
estimation. Identifying studies of rural areas is important here               of the township while preserving the overall population of the
because settlements in the Arghandab district of Zabul province,               area (township). Wright’s conclusion was that dasymetric
Afghanistan are all rural and have populations ranging from 50-                mapping could be applied to ever increasing subdivisions as long
5000 persons. The use of GIS and remote sensing in studies of                  as the density for each subdivision was accurate and the overall
interest is important due to the fact that no systematic census has            density of the higher division was preserved.
been conducted in the study area. GIS and remote sensing
                                                                                                                                                4
                                Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


           Although census figures for Afghanistan do exist at the            overcome spatial considerations (i.e. is a person always standing
district and provincial level, the intent of this study is not to             at one place within a structure at all times?).
utilize dasymetric mapping to disaggregate population densities
to the settlement level (sub-district). Population figures for this           2.3 Data Overview
area of Afghanistan are outdated and the methodology for
determining those figures is under suspicion. This study does,                2.3.1 Population
however, intend to apply the dasymetric method by taking
empirically derived census data at the settlement level and                             Published population data for Afghanistan is sparse.
transferring it to the sub-settlement level by distributing the               The Central Statistics Office of Afghanistan provides population
population over the area of a settlement’s structure footprint. In            records at the district level for all of Afghanistan’s provinces
the sense of dasymetric mapping, the subdivision in this study                (Afghanistan CSO 2006), however, the methods and standards
may be thought of as the dwelling footprint, unit of measurement              used in determining those figures are not transparent and consist
(i.e. m2), or pixel.                                                          of numbers of families reported to live in rural settlements by
           Although the dasymetric method is an attractive                    elders who attended a district level meeting. The number of
technique for estimating population, it still requires an even                families was multiplied by the average family size to determine
distribution level at any subdivision (Wu et al. 2005). In the case           district populations. Similar figures were reported by the
of this study, it assumes that population is evenly distributed               UNHCR for Zabul Province at the district level (UNHCR 2003).
among structures. The only solution to this assumption is the                 The estimated population for Arghandab district is 32,875
acquisition of ever increasingly detailed population data (i.e.               individuals.
population of individual households).                                                   In order to assess the ability of structure features
           The simplest form of dasymetric mapping is the use of              extracted from remotely sensed data as an estimation tool for
a binary land use division (i.e. residential and non-residential)             population, in situ population data was collected from a sample
and is outlined by Fischer and Langford (1996) in their use of                of settlements in the study area (Tomberlin 2009). Population
classified LandSAT imagery to refine population maps for areas                figures were collected between January and April 2009 for 15
smaller than the finest census data unit. They differentiate                  settlements in the Arghandab district area. It is important to note
between areal weighted and the dasymetric method used in their                that the sample was not randomly selected from within the
study where areal weighted techniques simply allocates                        population of settlements in the district due to accessibility and
population to target zones (i.e. subdivisions) according to the               security reasons. The majority of the settlements are near
relative area of source zones (i.e. divisions). Fischer and                   coalition forces stationed in Arghandab district and this
Langford detail their process for estimating population in target             sampling scheme may present bias in the study Table 1).
zones but it is interesting to note that their technique still
requires even allocation across a pixel (i.e. 900m2).                          Settlement                                          Population
           This study utilizes Thiessen polygons calculated using              Aga Saheb Kalay                                     350
the centroid of settlements as the source zone and the actual                  Bagh  Kalay                                         100
residential features within the Thiessen polygon as the target                 De Khayr Gol Kalay                                  40
zone. Despite settlements being tightly grouped around the                     De Mohammad Osman Kalay                             200
centroid, the use of Thiessen polygons was necessary for the                   De Sorksang De Sar Kalay                            300
small amount of structure area not tightly grouped near the                    Deh Afghan                                          400
centroid but still belonging to that settlement. Census data
                                                                               Hezar Kheyl                                         200
collected in those settlements could be said to include the entire
                                                                               Kashani                                             400
Thiessen polygon, hence the need to conduct dasymetric
                                                                               Kuchi Kheyl/Dehmazang                               325
mapping and disaggregate the population data to residential
areas.                                                                         Ma Sum Kalay                                        75
           This study uses remotely sensed data with a ground                  Morghan                                             200
sample distance fine enough to resolve individual structures.                  Sayagez                                             150
Unlike data used in the studies above, pixels classified as                    Takhunak                                            125
structures in this study do represent actual residential areas. The            Tawiz Kheyl                                         300
result is that once the binary dasymetric method is used, areal                Toray                                               70
weighted interpolation is then used to evenly distribute                      Table 1. In situ population data (Tomberlin 2009).
population over residential areas, albeit at much finer scales. In
comparison to Wright (1936), only one iteration of subdivision                         The in situ population data were used to assess the
creation was needed in this study to distinguish between                      predictability of settlement structure area on population and
structures and non-structures. Of course, further iterations could            build the model for predicting population throughout the district.
be made as Wright did if ancillary data existed on a by dwelling
basis. At some point, however, temporal considerations will                   2.3.2 Imagery
                                                                                                                                                                                       5
                                                  Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


          Remotely sensed panchromatic images from Digital                                                                                                 0       25      50 Meters



Globe’s Quickbird satellite were used for feature extraction of
settlement structure surface area. The images were captured in
June 2007. The ground sample distance of the panchromatic
images are 61cm and allow the analyst the ability to accurately
extract structures within the study area (Figure 3).
                                                            0    25     50 Meters




                                                                                                Figure 4. Buckeye high resolution color image with structure inset map (AGC 2008).


                                                                                                2.3.3 LIDAR/DEM

                                                                                                         In conjunction with the high resolution Buckeye
                                                                                                imagery, LIDAR derived digital elevation models captured in
                                                                                                tandem were used to validate Buckeye feature extraction (Figure
Figure 3. Quickbird satellite image with structure inset.                                       5). LIDAR derived DEMs had a resolution of 1m and when used
                                                                                                in conjunction with color imagery allowed the analyst to
         In order to validate the accuracy of features extracted                                accurately detect structures within the validation subset.
from Quickbird satellite imagery, high resolution color
photographs of a subset of the study area were used to validate                                                                                            0       25      50 Meters



the accuracy of features extracted. The color images were
produced by the Army Geospatial Center (AGC) as part of their
Buckeye program and were captured in 2008 (Figure 4). The
images have a ground sample distance of approximately 10-
20cm and allow the analyst the ability to resolve features with
more precision than the Quickbird imagery.




                                                                                                Figure 5. LIDAR DEM sample with structure inset (AGC 2008).


                                                                                                2.3.4 Political

                                                                                                        Political boundaries and settlement location vector data
                                                                                                were obtained from the Afghanistan Information Management
                                                                                                                                                                                 6
                                                Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


Services (AIMS 2005). The Arghandab province boundary was
modified to represent the establishment of Khak-i-Afghan
province northeast of Arghandab district which annexed some of
the former district area. Thiessen polygons of the study area’s                                                                              Multiple 
settlements were created to establish the boundaries of a                                                                                    Scenarios 
settlement’s influence. The majority of settlements are densely                                                                             Generation
grouped, reflecting the gregorian nature of rural Pashtuns,
however, in a small number of cases individual dwellings were
located away from the settlement centroid. In this case,                                                                                     Multiple 
structures were attributed to the settlement of the Thiessen                                                                                Hypothesis 
polygon they were located within (Figure 6).
                                                                                                                                            Generation




                                                                                                                                             Quadrant 
                                                                                                                                             Crunching


                                                                                                                                            Analysis of 
                                                                                                                                            Competing 
                                                                                                                                            Hypotheses



                                                    0          10
                                                                    ¯     20 Kilometers

                                                                                                  Figure 7. Visualization of structured analytic techniques used in the study.
Figure 6. Arghandab district settlement Thiessen polygons with polygon boundary close up inset.
                                                                                                  3.1 Multiple Scenarios Generation
3. Analytic Methodology
                                                                                                            Multiple scenarios generation is a technique used to
          In order to identify as many variables and forces acting                                identify all of the possible scenarios and combinations of driving
on population distribution and geographic patterns of rural                                       forces at play for a given problem. The technique reduces the
Pashtun settlements as possible, several structured analytic                                      chance that an event or process will play out in a way that was
methods were used. The initial hypothesis for this study was that                                 unforeseen to the analyst (Heuer and Pherson 2009). The
rural Pashtun settlement area is significantly positively                                         technique involves defining the issue at hand, identifying all of
correlated with population.                                                                       the key factors, forces, or events influencing the issue, defining
          The reason for the use of structured analytic techniques                                the ends of the spectrum for each driver, and pairing drivers in
is to mitigate the natural biases of an analyst testing a hypothesis                              2x2 matrices. Once all permutations of drivers are identified, a
(Heuer and Pherson 2009). The use of structured analytic                                          scenario is established for each combination (i.e. each quadrant
methods in this study included several techniques outlined below                                  in the matrix) and indicators are developed to track whether one
which were used to identify extreme circumstances, variables                                      of those scenarios is in fact developing. This technique is
and factors not originally identified, alternate hypotheses, and                                  particularly useful in identifying extreme cases of interaction
interactions between variables not originally seen. In certain                                    between drivers of an issue.
cases the methodology is redundant but obtaining duplicate                                                  For this study the focal issue in question was “What
scenarios, hypotheses, and variables from different analytic                                      determines population in rural Pashtun settlements in
techniques reinforces the fact that many variables and forces                                     Afghanistan?”. The key drivers identified for this focal issue
acting on the problem had been identified. Figure 7 depicts the                                   included terrain, arable land, structures, rurality, tribal
analytic methodology used for this study.                                                         affiliation, feature extraction methods, and settlement
                                                                                                  aggregation. These seven initial drivers were further condensed
                                                                                                  into three main drivers: terrain, rurality, and feature extraction.
                                                                                                                                                                                                                                                 7
                                                                         Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


The spectrum range for each driver was identified and all drivers
were combined to form three matrices and twelve potential                                                                                                             Washed out 
                                                                                                                                                                       images of 
scenarios (Figures 8a-c). For each scenario, three settlements                                                                                                        structures 
                                                                                                                                                                     concealed by 
that fit that scenario’s description were noted as potential                                                                        Settlement with very small 
                                                                                                                                                                      vegetation
                                                                                                                                                                                                Small individual structures 

indicators or outliers in the statistical analysis. The most extreme                                                                structures concealed by 
                                                                                                                                    dense vegetation  furthest 
                                                                                                                                                                                                in dispersed settlements 
                                                                                                                                                                                                where vegetation is 




                                                                                                                                                                          Feature Extraction
scenario, identified as having the most affect on the                                                                               away from urban areas.                                      concealing structures near 
                                                                                                                                                                                                Qalat with trade to the 

predictability of remotely sensed structures on population, were                                                                                                                                urban area.


settlements near Qalat with trade to the urban area and small
structure features concealed by vegetation.                                                                              Far from urban                                                                               Near urban area, 
                                                                                                                          areas, Ghilzai                 Rurality                              Rurality                Durrani tribal 
                                                                                                                         tribal affiliation                                                                             affiliation




                                                                                                                                                                          Feature Extraction
                                             Near urban area, 
                                              Durrani tribal 
                                               affiliation
           Minority Durrani settlement                                     Minority Durrani settlement                              Small, dense settlements                                    Large, dispersed settlements 
           located in Ghilzai tribal area                                  located in Ghilzai tribal area                           with structures on the high                                 with isolated structures near 
           near Qalat (urban area) and                                     near Qalat (urban area) and                              ground where vegetation                                     Qalat and less agrarian 
           located in a narrow valley.                                     located in a large basin.                                transitions to mountain                                     economics.
                                                                                                                                    rock.                            Clear, sharply 
                                                    Rurality




                                                                                                                                                                    edged buildings 
                                                                                                                                                                      away from 
                                                                                                                                                                      vegetation


  Narrow valleys, 
     densely                                                                                        Large basins, 
   distributed                   Terrain                                   Terrain                   dispersed 
   settlements                                                                                      settlements        Figure 8c. Multiple Scenarios Generation with Rurality and Feature Extraction as drivers (note:
                                                                                                                       population predictability of the scenario decreases up and right).
                                                    Rurality




                                                                                                                       3.2 Multiple Hypothesis Generation
           Very isolated Ghilzai                                           Very isolated Ghilzai 
           population in a tightly 
           distributed settlement 
                                                                           population in a large 
                                                                           dispersed settlement 
                                                                                                                                 Multiple hypothesis generation is a tool for generating
           furthest away from Qalat. 
                                             Far from urban 
                                                                           furthest away from Qalat. 
                                                                                                                       additional alternative hypotheses for a problem or issue (Heuer
                                               areas, Ghilzai 
                                             tribal affiliation                                                        and Pherson 2009).The process involves identifying several
                                                                                                                       leading hypotheses, breaking them down into their component
                                                                                                                       parts, creating all possible permutations, and sorting the credible
                                                                                                                       products to identify additional alternative hypotheses.
Figure 8a. Multiple Scenarios Generation with Terrain and Rurality drivers (note: population
predictability of the scenario decreases up and right).                                                                          Using the lead hypothesis that “Settlement area in rural
                                                                                                                       Pashtun Afghanistan is significantly positively correlated with
                                                                                                                       population and therefore an accurate estimate of population” the
                                               Washed out 
                                                images of 
                                                                                                                       components of this hypothesis were identified and are
                                               structures 
                                              concealed by 
                                                                                                                       summarized in Table 2. Two alternatives to each component
           Settlement with very small 
                                               vegetation
                                                                           Small individual structures in 
                                                                                                                       were identified and all permutations of those components were
           structures concealed by 
           dense vegetation .
                                                                           dispersed settlements where 
                                                                           vegetation is concealing 
                                                                                                                       listed. Those permutations were analyzed for credibility and
                                                    Feature Extraction




                                                                           structures.                                 sorted to identify the leading alternative hypotheses shown in
                                                                                                                       Table 3 with the null and initial hypotheses. Although distance
                                                                                                                       to potable water source and settlement elevation were identified
  Narrow valleys,                                                                                                      as potential variables, they were excluded due to mutual
     densely 
   distributed                   Terrain                                   Terrain
                                                                                                    Large basins, 
                                                                                                     dispersed         exclusivity issues with vegetation. Elevation and distance to
   settlements                                                                                      settlements
                                                                                                                       water would vary with vegetation.
                                                    Feature Extraction




                                                                                                                        Question     Hypothesis Component                 Alternative Component                         Alternative Component
                                                                                                                        Who          Rural Pashtun                        Urban Pashtun                                 Nomads
           Small, dense settlements                                        Large, dispersed settlements 
                                                                                                                        What         Settlement Structure Area            Vegetation Area                               Elevation
           with structures on the high                                     with isolated structures                     When         Image Capture Date                   Seasonal Migration                            Population Record Date
           ground where vegetation                                         away from vegeatation.                       Where        Rural Afghanistan                    Near Arghandab River                          Near Qalat
           transitions to mountain 
           rock.                              Clear, sharply                                                            Why          Constant Density                     Variable Birth Rate                           Variable Density
                                             edged buildings                                                            How          Visible Structures                   Non‐Agrarian                                  Opium
                                               away from 
                                               vegetation
                                                                                                                       Table 2. Multiple Hypothesis Generation components of the leading hypothesis.



Figure 8b. Multiple Scenarios Generation with Terrain and Feature Extraction drivers.
                                                                                                                                                                                                     8
                                                 Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


   Hypothesis
H0 Population is randomly dispersed in rural Pashtun
   Afghanistan and not significantly correlated with
   any geospatially derived features.
H1 Settlement area in rural Pashtun Afghanistan is
   significantly positively correlated with population.

H2 Settlement vegetation area in rural Pashtun
   Afghanistan is significantly positively correlated
   with population.
H3 Distance from urban areas of rural Pashtun
   settlements is significantly negatively correlated
   with population.
Table 3. Hypotheses produced as a result of multiple hypotheses generation.


3.3 Quadrant Crunching

          The use of structured analytic techniques to identify
alternative hypotheses carried with it significant assumptions.
Quadrant crunching was used to identify the main assumptions                                   Figure 9. Screen capture of the analysis of competing hypotheses results using ach software version
                                                                                               2.0.3 (PARC 2006).
associated with each hypothesis. Quadrant crunching is a
technique used to identify the full range of assumptions
                                                                                               4. Geospatial Analysis
associated with a hypothesis. Much like multiple scenarios
generation, it identifies a full range of assumptions by creating a
                                                                                               4.1 Feature Extraction
robust set of stories from all permutations of a set of
assumptions (Heuer and Pherson 2009).
                                                                                                        The extraction of settlement structure features using on
          Due to the number of permutations created in quadrant
                                                                                               screen digitization of polygon features was critical to this study.
crunching, five key assumptions were used to identify contrary
                                                                                               Quickbird satellite imagery was used to capture 1,838 structure
assumptions. The technique yielded 40 categories of potential
                                                                                               polygon features for 106 of 147 settlements in the study area.
scenarios (Appendix A). The scenarios were evaluated for
                                                                                               Features were extracted on screen at scales ranging from 1:1000
consistency and invalid scenarios were discarded. The remaining
                                                                                               to 1:3000 and were saved to a geodatabase (Figure 10).
scenarios were further evaluated to determine scenarios that had
                                                                                               Additionally, structure polygon features were extracted using
compounding effects. These compounding scenarios were used
                                                                                               high resolution Buckeye imagery for the subset of settlements
in the determination of evidence and hypotheses in the analysis
                                                                                               used for validating Quickbird feature extraction accuracy.
of competing hypotheses below.
                                                                                               Features extracted using Buckeye imagery were captured at
                                                                                               scales ranging from 1:200 to 1:500
3.4 Analysis of Competing Hypotheses

         Analysis of competing hypotheses (ACH) is a
technique used to systematically evaluate the credibility of
hypotheses given a set of evidence. The technique attempts to
refute hypotheses rather than qualify them and determines the
best hypothesis based on the one with the least evidence against
it (Heuer and Pherson 2009). For this study, four hypotheses
developed from multiple hypothesis generation were used and
12 pieces of evidence were used to evaluate the consistency of
each hypothesis (Figure 9).
         The results of ACH yielded weighted inconsistency
scores for each hypothesis with H2 (vegetation area) having the
highest inconsistency score of -0.707 followed by the initial
hypothesis H1 (structure area) with -1.707. The null hypothesis
and H3 had much lower weighted inconsistency scores of -4.0
and -10.0 respectively.
                                                                                                                                                           9
                                            Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009




                                                                                           25 Meters
                                                                                           0


                                                                                          Figure 11. Buckeye feature extraction example.


                                                                                                   Once feature extraction accuracy was confirmed,
                                                                                          polygon area was calculated for all structures and an intersection
                                                                                          vector geoprocessing function was conducted with the settlement
                                                                                          Thiessen polygon layer as the second input dataset. Total area by
                                                                                          town was then calculated and statistical correlation was
                                                                                          conducted by town between the Buckeye and Quickbird derived
                                                                                          products.




Figure 10. Geospatial analysis flowchart.


 (Figure 11). In order to improve the accuracy of the Buckeye
derived features, a LIDAR DEM was used to confirm correct
structure identification. The LIDAR DEM was the product of a
second return LIDAR DEM subtracted by a bare earth model
using raster math. The resulting raster dataset clearly identified
structures and vegetation above ground.
                                                                                                                                                                                                                                                      10
                                                         Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


                                                                                                                    hypothesis suggesting that a log transformation may be
 25 Meters




                                                                                                                    necessary to evaluate this data using parametric statistics.

                                                                                                                                                        Quickbird                     Natural Log
                                                                                                                                                      Structure Area                Transformation
                                                                                                                                           N                              106                         106
                                                                                                                        Kolmogorov-Smirnov Z                            2.011                       0.856
                                                                                                                                           p                            0.001                       0.456
 0




                                                                                                                    Table 6. Kolmorogov-Smirnov normality test for Quickbird derived settlement area.


                                                                                                                    5.2 Pearson Product Correlation

                                                                                                                              Correlations between all variables were calculated
                                                                                                                    using Pearson’s product-moment correlation coefficient (Table
                                                                                                                    7). Critical to the use of Quickbird derived settlement structure
                                                                                                                    area were the extracted features being equal to real world
                                                                                                                    structures. Since a real world survey of the sample structures
                                                                                                                    was not available, LIDAR derived features were used as the
                                                                                                                    representation of the actual area of settlement structures. As
                                                                                                                    noted earlier, features extracted using Buckeye LIDAR and
                                                                                                                    imagery did not include the typical interior space within a walled
                                                                                                                    compound and therefore areas of Buckeye derived structures
                                                                                                                    were much less than Quickbird derived areas. In order to assess
                                                                                                                    the accuracy of Quickbird derived features, the correlation
Figure 12. Quickbird feature extraction example.                                                                    coefficient was evaluated for significance using a t-test with the
                                                                                                                    null hypothesis that ρ = 0 (no relationship between Quickbird
5. Statistical Analysis                                                                                             and Buckeye derived structure areas). The calculated t statistic
                                                                                                                    was high, 8.389, rejecting the null hypothesis at the 0.001
5.1 Descriptive Statistics                                                                                          significance level. The highly significant correlation between
                                                                                                                    Quickbird and Buckeye derived structure areas suggests that
          Descriptive statistics for all sample variables are                                                       features extracted from Quickbird imagery are an accurate
summarized in Table 4. In addition to exploratory statistical                                                       representation of real world structures.
analysis each variable was tested for normality using a                                                                       The hypotheses developed in the structured analytic
Kolmogorov-Smirnov test for normality (Table 5). In all cases,                                                      techniques above were individually tested against the null
the test failed to reject the null hypothesis that the data is                                                      hypothesis that ρ = 0 or that there is no relationship between the
normally distributed at 0.05 and 0.01 significance level (two-                                                      variable in question (structure area, orchard area, or distance
tailed distribution).                                                                                               from urban area). Although a very small positive correlation (r =
                                                                                                                    .390) existed between orchard area and population, the t-test
Population
           VARIABLE                         n
                                                  15
                                                         MEAN
                                                           215.67
                                                                        STD DEV
                                                                            122.68
                                                                                         MIN
                                                                                            40.00
                                                                                                       MAX
                                                                                                         400.00
                                                                                                                    failed to reject the null hypothesis that there was no relationship.
Quickbird Structure Area (m2)                     15     13086.08          7276.35        3867.46      24490.60     Distance from urban area showed a very small negative
Buckeye Structure Area (m2)
Orchard Area (m2)
                                                  10
                                                  10
                                                          5231.68
                                                         96089.02
                                                                           2076.14
                                                                          49072.17
                                                                                          2350.23
                                                                                         45449.77
                                                                                                        8134.44
                                                                                                      214418.56
                                                                                                                    correlation to population (r = -.348) but the t-test also failed to
Distance from Urban Area (m)                      15     34134.65          4378.07       29639.87      48129.56     reject the null hypothesis. Both Quickbird and Buckeye derived
Table 4. Descriptive statistics for study variables.                                                                settlement structure area were highly correlated with population,
                                                                                                                    r = .830 and r = .829 respectively. Both correlations were highly
                               Population         Quickbird
                                                Structure Area
                                                                     Buckeye
                                                                  Structure Area
                                                                                    Orchard Area    Distance to
                                                                                                    Urban Area
                                                                                                                    significant at the .001 and .005 significance level.
                           n          15.00               15.00             10.00           10.00              10
        Kolmogorov-Smirnov Z           0.60                0.62              0.64            0.70           0.417                                                                         Quickbird           Buckeye         Orchard      Distance to
                                                                                                                                                                       Population       Structure Area     Structure Area      Area        Urban Area
                           p           0.87                0.83              0.81            0.71           0.995
                                                                                                                    Population                 Pearson Correlation                  1             .830**             .829**       0.390           -0.348
                                                                                                                                                     Sig. (2-tailed)                               0.000              0.003       0.265             0.203
Table 5. Kolmogorov-Smirnov normality test for all variables.                                                                                                      n            15                    15                 10          10                15
                                                                                                                    Quickbird Structure Area   Pearson Correlation          .830**                     1             .948**       .650*            -.597*
                                                                                                                                                     Sig. (2-tailed)         0.000                                    0.000       0.042             0.019

         In addition to the sample data above, the population of                                                    Buckeye Strucutre Area     Pearson Correlation
                                                                                                                                                                   n            15
                                                                                                                                                                            .829**
                                                                                                                                                                                                      15
                                                                                                                                                                                                  .948**
                                                                                                                                                                                                                         10
                                                                                                                                                                                                                          1
                                                                                                                                                                                                                                     10
                                                                                                                                                                                                                                  .689*
                                                                                                                                                                                                                                                       15
                                                                                                                                                                                                                                                 -.800**

Quickbird derived settlement structure area was tested for                                                                                           Sig. (2-tailed)
                                                                                                                                                                   n
                                                                                                                                                                             0.003
                                                                                                                                                                                10
                                                                                                                                                                                                   0.000
                                                                                                                                                                                                      10                 10
                                                                                                                                                                                                                                  0.027
                                                                                                                                                                                                                                     10
                                                                                                                                                                                                                                                   0.005
                                                                                                                                                                                                                                                       10

normality due to the high positively skewed distribution value of
                                                                                                                    Orchard Area               Pearson Correlation           0.390                 .650*              .689*           1           -0.540
                                                                                                                                                     Sig. (2-tailed)         0.265                 0.042              0.027                        0.107

2.38. The Kolmogorov-Smirnov test for normality rejected the
                                                                                                                                                                   n            10                    10                 10           10               10
                                                                                                                    Distance from Urban Area   Pearson Correlation          -0.348                -.597*            -.800**       -0.540                1

null hypothesis that the distribution of settlement structure area
                                                                                                                                                     Sig. (2-tailed)         0.203                 0.019              0.005        0.107
                                                                                                                                                                   n            15                    15                 10           10              15

was normal at the 0.01 significance level. A natural log                                                            Table 7. Correlation matrix for study variables.
transformation was performed on the distribution and the test
was repeated. The test for normality failed to reject the null
                                                                                                                                                                                  11
                                                                 Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


5.3 Linear Regression                                                                                            settlements where the security situation is stable enough to
                                                                                                                 collect demographic data from settlement leadership. Applying
          Linear regression was used to predict population values                                                predictive modeling based on this sampling scheme may
throughout the study area. It must be noted that due to the                                                      invalidate the results of the predicted population. To improve the
sampling scheme the applicability of using the regression                                                        integrity of the prediction model the use of a random sampling
equation throughout the study area may be limited. The best fit                                                  scheme is needed.
linear regression model was:
                                                                                                                 5.2 Feature Extraction
                                       P = 0.014A + 32.521
                                                                                                                      Although feature extraction of settlement structures
where: P=total population in persons; A=area of total settlement                                                 followed guidelines to increase the uniformity of the extraction
structure in m2. The linear regression model was used to predict                                                 process, there is still user bias introduced to the study. The
populations throughout the study (Figure 13).                                                                    assumption in this study is that the user bias is systematic and
                                                                                                                 presents negligible error to the study. The high correlation
5. Results and Discussion                                                                                        between Quickbird and Buckeye derived features supports this
                                                                                                                 assumption. In order to improve the accuracy and remove bias
     The results of the study are very promising for the use of                                                  from the settlement structure dataset, the use of automated
remotely sensed features as an estimate for population in rural                                                  feature extraction techniques is recommended. The use of
areas. Although the study showed potential there are several                                                     unsupervised classification of land use in this study area was
aspects of the study that must be improved to verify the                                                         attempted, however, the use of earthen building materials in
significance of the structure-population correlation and the                                                     settlement structures makes the classification difficult due to
application of the linear regression model to areas outside of the                                               similar spectral characteristics of structures and the surrounding
in situ population subset. These include sampling design, feature                                                area. The use of automated building extraction using LIDAR
extraction techniques, and scale dependency.                                                                     data is a promising approach, particularly for this study area due
                                                                                                                 to the isolated and homogenous land use in the area. Until



0
            ¯10              20 Kilometers
                                                                                                                 comprehensive LIDAR datasets are available this may not be
                                                                                                                 possible.

                                                                                                                 5.3. Scale Dependency

                                                                                           -
                                                                                           -                          This study utilized the settlement as the basic unit of
                                                                                          --                     aggregation and the district political boundary as the study area
                                                                 -
                                                                          -
                                                                                         -
                                                                                         -                       boundary and extent of the prediction model’s applicability. It is
                                                                               -   ---
                                                                                    -                -
                                                             -
                                                                               --                                visibly evident that the majority of settlements within the
                                                                               -
                                                                               --
                                                                           ---
                                                                              -                                  settlement have easily discernable boundaries. Aggregating
                                                                         - --
                                                                         - -
                                                                             -
                                                                                                                 population at the settlement level is logical. To prove this
                                                                    --
                                                                   --                                            assumption, the collection of population data of individual
                                                                  --
                                                                   -
                                                         -
                                                                 -
                                                                 -
                                                                 -
                                                                               -
                                                                                                                 dwellings within a settlement would be useful.
                         -              -
                                                         -
                                   -
                                        -
                                            -
                                            - - -- --
                                             - - - --
                                                         -

                                                                  -
                                                                      --
                                                                         -
                                                                       - --
                                                                               -
                                                                                                                      Using the district boundary as a study area boundary may be
                                                -
                                     -
                                   - --
                                       --                    -                                                   inappropriate considering that population dynamics and social
                  -- - -               -                             -
                                                                                                                 behavior in rural Afghanistan may not recognize or follow
                   --
                         -      ---                                  -
                       -      -
                              -              -
                                                 -   -
                          -
                          -             -
                                                 - -                                           Population        political boundaries. Collecting data from multiple districts and
                                                                                                                 assessing the spatial variability of the population data is
                                   -
                                                                                               Total Persons
                               -
                                                                                                -   0 - 66

                         -         -                                                            -   67 - 215
                                                                                                                 important in establishing the spatial applicability of the
                               -
                                                                                                -   216 - 376    predictive model.
                                                                                                -   377 - 702

                                                                                                -   703 - 1394
                                                                                                                 6. Commentary on the Use of Structured Analytic
Figure 13. Simple map showing the results of predicted settlement populations.
                                                                                                                 Techniques

5.1 Sampling Design                                                                                                  Structured analytic techniques were critical to this study
                                                                                                                 given the heavy dependence on imagery intelligence (IMINT).
    Although there was a high correlation between the in situ                                                    Structured analytic techniques provide the ideal mechanism for
population data and settlement structure area the sampling                                                       thoroughly analyzing imagery intelligence. Multiple scenarios
design requires improvement. The sampling design for this study                                                  generation allow the imagery analyst to identify the major
could be characterized as a convenience sample. The settlements                                                  dimensions of the variables affecting the process of interest
where in situ population data were collected represent                                                           within an image. Identifying the extreme scenarios of all
                                                                                                                                             12
                                Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


dimensions allows the analyst to systematically analyze                       Heuer, R. J. & Pherson, R. H. (2009) Structured Analytic
qualitative evidence and identify the worst and best case                             Techniques for Intelligence Analysis. Draft.
scenarios.                                                                    Jones, S. G. (2008) The Rise of Afghanistan’s Insurgency: State
     Multiple hypothesis generation forces the analyst to deviate                     Failure and Jihad. International Security, Vol. 32, 7–40.
from his or her original hypothesis. The implications for this                Lo, C. P. & Chan, H. F. (1980) Rural Population Estimation
study were that, rather than testing a single leading hypothesis                      from Aerial Photographs. Photogrammetric Engineering
and potentially failing to reject the null hypothesis in favor it,                    and Remote Sensing, 46, 337-345.
multiple hypotheses were established and variables associated                 Mennis, J. (2003) Generating Surface Models of Population
with each were measured simultaneously. The simultaneous                              Using Dasymetric Mapping. The Professional
evaluation of multiple hypotheses can save the analyst time                           Geographer, 55, 31-42.
when compared to the typical iterative method of testing,                     UNHCR (2003) Arghandab District Profile. UNHCR Sub-Office.
rejecting, and identifying a new hypothesis and repeating the                         www.aims.org.af
cycle.                                                                        Swintek, P. (2008) Arghandab District Village Assessments. B
     For this study, quadrant crunching was the single most                           Co 1-4 Infantry Regiment.
important technique. Quadrant crunching forces the imagery                    Tellis. A. J. (2009) Reconciling with the Taliban? Toward aa
analyst to identify the key assumptions in their hypotheses and                       Alternative Grand Strategy in Afghanistan. Carnegie
systematically create numerous permutations of alternate                              Endowment for International Peace.
assumption dimensions for analysis and consideration. Simply                  Tomberlin, J. (2007) Arghandab District Village Assessments. B
put, the technique keeps the imagery analyst honest and diverts                       Co 1-4 Infantry Regiment.
them from qualifying their original hypothesis.                               Wright, J. K. (1936) A Method of Mapping Densities of
     Finally, analysis of competing hypotheses is an excellent                        Population. The Geographical Review, 26, 103-110.
technique when applied to imagery intelligence in that it once                Wu, S., Qiu, X., & Wang, L. (2005) Population Estimation
again forces the analyst to systematically disprove their working                     Methods in GIS and Remote Sensing: A Review.
hypotheses. IMINT is an excellent example of how an individual                        GIScience & Remote Sensing, 42, 80-96.
will qualify their original hunch or hypothesis. The number of
avenues or forces that an individual will develop from an image
is countless. Compounded by the fact that many image analysts
have not been to the area of interest in the imagery, the use of
non-qualifying techniques are essential.

References

Afghanistan Central Statistics Office (2006) District Level
       Population Estimates. www.ads-services.net/cso
Afghanistan Information Management Services (2005)
       Arghandab Geospatial Data. www.aims.org.af
Army Geospatial Center (2008) Buckeye High Resolution
       Imagery and LIDAR Data. Army Engineer Research and
       Development Center.
Barfield, Thomas J. (2007) Weapons of the not so Weak in
    Afghanistan: Pashtun Agrarian Structure and Tribal
    Organization for Times of War & Peace. Agrarian Studies
    Colloquium Series “Hinterlands, Frontiers, Cities and
    States: Transactions and Identities” Yale University,
    February 23, 2007.
Fisher, P. F., & Langford, M. (1996) Modeling Sensitivity to
       Accuracy in Classified Imagery: A Study of Areal
       Interpolation by Dasymetric Mapping. Professional
       Geographer, 48, 299-309.
Glatzer, B. (2006) War and Boundaries in Afghanistan:
       Significance and Relativity of Local and Social
       Boundaries. Weld des Islams, 41, 379-399.
Johnson, Thomas H. (2007) Understanding the Taliban and
       Insurgency in Afghanistan. Orbis: A Journal of World
       Affairs, 51, 71-89.
                                                                                                                                                                                                                                                    13
                                                         Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009




                        Settlement structures extracted from Quickbird satellite imagery are an accurate estimator of population.
                     Key Assumptions                                             Contrary Assumptions                                             Contrary Dimensions                                          Contrary Dimensions
Population density does not vary within the district.         Population density does vary within the district.              Higher birth rate due to medical facilities or other outside  Lower birth rates due to some form of population control.
                                                                                                                             factor.
Proportion of structure area to irrigated crop area is equal  Proportion of structure area to irrigated crops varies within  Herder population i.e. Kharoti (Kuchi) nomads.                Xerophytic crops i.e. almond, pomegranate.
within the district.                                            the district.
Features extracted from Quickbird are accurate                  Features extracted from Quickbird imagery are not accurate  Structures that are uninhabited.                               Vegetation and ground reflection is producing identification 
representations.                                                representations.                                                                                                           error.
Settlement centroid and associated Thiessen polygon is the  Settlement centriods and polygons are not a logical              Settlements that have multiple tribes i.e. Mukurak.           Population density varies greatly from structure to 
best aggregation of populations.                                aggregation of populations.                                                                                                structure.
Minority Durrani population has the same                        Minority Durrani population has different economy/lifestyle  Durrani settlements show charateristics of sedentary          Economy lifestyle varies with distance from closest urban 
economy/lifestyle as Ghilzai majority.                          than Ghilzai majority.                                       economy with land owners and subjects.                        area i.e. Arghandab not equally rural.
                                     Varied Population Density / Structure:Crops Varied                                                                   Structure:Crop Varied / Thiessen Polygon Incorrect Aggregation
High birth rate / Herder population                             High birth rate / Xerophytic crops                           Herder population / Multiple tribe settlement                 Herder population / Within settlement variance
Low birth rate / Herder population                              Low birth rate / Xerophytic crops                            Xerophytic crops / Multiple tribe settlement                  Xerophytic crops / Within settlement variance
                                    Varied Population Density / Feature Extraction Errors                                                                   Structure:Crop Varied / Durrani:Ghilzai Economy Difference
High birth rate / Structures uninhabited                        High birth rate / Ground interference                        Herder population / Durrani sedentary                         Herder population / Rural variance
Low birth rate / Structures uninhabited                         Low birth rate / Ground interference                         Xerophytic crops / Durrani sedentary                          Xerophytic crops / Rural variance
                            Varied Population Density / Thiessen Polygon Incorrect Aggregation                                                           Feature Extraction Errors / Thiessen Polygon Incorrect Aggregation
High birth rate / Multiple tribe settlement                     High birth rate / Within settlement  variance                Structures uninhabited / Multiple tribe settlement            Structures uninhabited / Within settlement variance
Low birth rate / Multiple tribe settlement                      Low birth rate / Within settlement variance                  Ground interference / Multiple tribe settlement               Ground interference / Within settlement variance
                              Varied Population Density / Durrani:Ghilzai Economy Difference                                                               Feature Extraction Errors / Durrani:Ghilzai Economy Difference
High birth rate / Durrani sedentary                             High birth rate / Rurality variance                          Structures uninhabited / Durrani sedentary                    Structures uninhabited / Rurality variance
Low birth rate / Durrani sedentary                              Low birth rate / Rurality variance                           Ground interference / Durrani sedentary                       Ground interference / Rurality variance
                                       Structure:Crop Varied / Feature Extraction Errors                                                            Thiessen Polygon Incorrect Aggregation / Durrani:Ghilzai Economy Difference
Herder population / Structures uninhabited                      Herder population / Ground interference                      Multiple tribe settlement / Durrani sedentary                 Structures uninhabited / Within settlement variance
Xerophytic crops / Structures uninhabited                       Xerophytic crops / Ground interference                       Multiple tribe settlement / Durrani sedentary                 Ground interference / Within settlement variance

Appendix A. Quadrant crunching results.
                                                                                                                         14
              Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009




Settlement                                             Structure Area                             Predicted Population
Abdul Qayum Kalay                                      11560                                      194
Aga Saheb Kalay                                        16772                                      267
Akhtar                                                 75157                                      1085
Alajargha                                              7512                                       138
Alaqadari Arghandab                                    27596                                      419
Alday Kalay                                            9600                                       167
Angur Kalay                                            9887                                       171
Aparan                                                 2395                                       66
Bagh                                                   8893                                       157
Balandwarkh                                            29180                                      441
Band                                                   15737                                      253
Barakat Kor                                            9656                                       168
Bargah                                                 25720                                      393
Barghantu                                              20948                                      326
Bata                                                   49                                         33
Bazidkhel                                              8114                                       146
Boruj Kalay                                            9600                                       167
Braj                                                   25880                                      395
Budaddin                                               3199                                       77
Ceray Ludin                                            5386                                       108
Chakanak                                               88875                                      1277
Cherga                                                 23421                                      360
China (2)                                              7638                                       139
Darwazagay Kalay                                       3869                                       87
Darya                                                  56772                                      827
Dawlatkhel                                             17715                                      281
De Baluc Kalay(De Bagh Khola Kalay)                    14273                                      232
De Haji Khan Mohammad Kalay                            9892                                       171
De Haji Sayed 'Omar Kalay                              10881                                      185
De Khayrgul Kalay                                      5128                                       104
De Mohammad Usman Kalay                                9688                                       168
De Nezam Kor                                           4551                                       96
De Surkhsang                                           11653                                      196
De Zyarat Kalay                                        11202                                      189
Deh Afghanan                                           19970                                      312
                               Appendix B. Arghandab district settlement predicted populations.
                                                                                                       15
            Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


Deh Bum Kalay                                        5278                                       106
Enjar                                                6051                                       117
Eraqi                                                23356                                      359
Fadzluddin Kalay                                     4026                                       89
Garnay                                               10867                                      185
Gerday                                               13933                                      228
Ghundey                                              11856                                      199
Hasakhel                                             4736                                       99
Jalu                                                 81156                                      1169
Janubi Sar Darrah                                    74597                                      1077
Janubi Sherak                                        3646                                       84
Kala (1)                                             19551                                      306
Kalagay                                              13895                                      227
Karkora                                              18904                                      297
Kash China                                           6339                                       121
Kashani                                              19482                                      305
Khanjak                                              9226                                       162
Khushkawa                                            13054                                      215
Khushkawa‐i‐Hazarkhel                                9730                                       169
Khwaja Zangi                                         36604                                      545
Kochikhel                                            18033                                      285
Korghan                                              14229                                      232
Kotana                                               16221                                      260
Lekak                                                9016                                       159
Mahmudkhel                                           19141                                      300
Malang Khune                                         10488                                      179
Malek                                                58047                                      845
Malikhel (2)                                         35879                                      535
Manakhel                                             25270                                      386
Masum Kalay                                          11532                                      194
Maydan Kalay                                         12358                                      206
Mayni Wola                                           9174                                       161
Miragha                                              8966                                       158
Moci                                                 13459                                      221
Mohmmad Rasul Khune                                  8761                                       155
Mulla Qyamuddin Chambar                              15707                                      252

                             Appendix B. Arghandab district settlement predicted populations.
                                                                                                         16
              Stofan / Virtual Field Exercise for the Geospatial Intelligence Professional Summer 2009


Murghan                                                19609                                      307
Murghawi                                               20317                                      317
Naray                                                  15247                                      246
Neknam                                                 91344                                      1311
Olgay                                                  25292                                      387
Pam Khak                                               5489                                       109
Parsang                                                31630                                      475
Petaw                                                  43308                                      639
Qarya‐i‐El                                             8146                                       147
Sagena                                                 10876                                      185
Sartezay                                               9075                                       160
Sayagaz                                                5085                                       104
Sayedkhel                                              47801                                      702
Shadu                                                  13894                                      227
Shahabudin Kalay                                       5436                                       109
Shalkak                                                5533                                       110
Shamali Sar Darrah                                     38575                                      573
Shamali Sherak                                         5644                                       112
Shaygan                                                97263                                      1394
Shimizi                                                21283                                      330
Skecha(De Zaydullah Kalay)                             5602                                       111
Slemanzi (1)                                           28067                                      425
Slemanzi (2)                                           24532                                      376
Spera                                                  29845                                      450
Ta Wola                                                7818                                       142
Takatu                                                 25464                                      389
Takhonak                                               7294                                       135
Tanor                                                  27926                                      423
Taru                                                   12396                                      206
Tawizkhel                                              22408                                      346
Taybaz                                                 12543                                      208
Tor Takuna                                             7951                                       144
Toray                                                  3867                                       87
Wali Mohammad Kalay                                    11680                                      196
Wetob(Betob)                                           1536                                       54

                               Appendix B. Arghandab district settlement predicted populations.

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:8
posted:12/18/2011
language:
pages:16