GIS, Spatial Analysis
and the Nang Rong Projects
Within the Nang Rong Project, a Geographic Information Sysytem (GIS) was
developed that includes a number of base coverages as well as derived
coverages for the study area. An invaluable tool for spatial
analysis, GIS combines the spatial elements of a map with the powerful
information management, querying, and updating capabilities of a
database. These data are integrated within a GIS for Nang Rong
district, Thailand, that is unique in its temporal coverage of
social, demographic, spatial, and biophysical variables. The GIS
provides for data input, storage, management, retrieval, analysis,
visualization, and statistical and cartographic output. In
addition, the GIS allows for the representation and analysis of
data at multiple spatial and temporal resolutions. When
incorporated into a GIS, data is stored in layers that can be analyzed
separately or together, allowing the user to gain a deeper
understanding of multifaceted issues. What is more,
GIS allows for the synthesis of various data types, facilitating
research and analyses that investigate the complex interplay and
feedbacks between spatial and socio-economic data.
|
|
A GIS is more than just the
layering of maps, however. The other principal component of
GIS is the ability to assign attributes to the features on the map.
This capability means that dots on the map are more than just
households, areas more than simple soil polygons, and lines more than
mere transportation routes. A wealth of information about each
household can be associated with every object in the layer. For
instance, the household layer could contain information on the
number of people in the household, their ages, genders, in addition
to a host of other dempgraphic variables. This wealth of
information, which would be available in the other layers, would allow
for quite sophisticated analysis. This combination of mapping and
attribute analysis means that researchers have a powerful tool that
facilitates in the understanding of the biophysical and demographic
aspects of the project.
Within a GIS, various geospatial and demographic data
layers, including vector and raster layers, as well as remotely sensed
imagery and photos, may be created, modified, and analyzed. A
sampling of the GIS techniques, methods, and analyses employed by
the Nang Rong Project is listed in the table below.
|
Aerial Photograph Mosaic |
For 4 of the 5 year
sets for which the Nang Rong Project possesses air photos, aerial
photograph mosaics have been created. Using tie-points and areas
of overlap, photos have been combined into seamless images. For
more information, visit the Aerial Photography page in the Remote Sensing section of this website.
|
Agent Based Modeling |
Agent
based modeling (ABM) has been used to simulate the development of
villages based on elevation, the distance to nearest water, and
distance to the nearest pre-existing village. Representing
individual agents within the models, village agents simulate the
establishment of villages. A longitudinal survey was used to
define village establishment dates for observed villages, an image
animation was used to visualize the pattern of observed village
establishment on an annual time step, and a GIS to represent
topography, and road and water networks as village attractors.
|
Cellular Automata (CA) |
Cellular
Automata (CA) is a valuable method currently under development in our
research to simulate LULC dynamics. CA models are composed of a
regular grid of cells, each in a finite state, that are iteratively
updated in discrete time steps. Growth or transition rules allow
systems to grow from initial conditions, vary their rates of change, or
reverse directions in a recursive sequence of iterations. In the
Nang Rong setting, transition probabilities may be developed that
depend on cell-level resource endowments, such as soils, slope,
hydrology, or terrain, demographic characteristics, such as number of
households, degree of land competition, or population density, and its
geographic proximity to water, villages, roads, or markets.
The models allow for the spatial simulation of LULC patterns,
examination of likely future LULC scenarios, and the examination of the
role of houshold and community-level social and demographic factors in
the alteration of trajectories of LULC change, which may result in
shifts in the composition and spatial structure of the landscape.
|
Change Detection |
Following
classification into LULC classes, the character and magnitude of change
on the landscape was analyzed using post-classification change detection
methods. Various types of change were investigated, including
intra-annual (seasonal), inter-annual, and decadal change.
However, a number of additional image change-detection approaches can
be used to characterize landscape dynamics, including change vector
analysis, binary masks, Principal Components Analysis, and image
algebra. See the Aerial Photograph and Image Classification section of this site for additional information.
|
Cloud Masking |
Prior
to image classification, pre-processing steps were undertaken to remove
clouds and cloud shadows from the satellite imagery. Clouds are
masked to reduce classification error and confusion, and to increase
classification accuracy.
|
Cluster Analysis |
LULC
classification accuracy was improved, while classification error was
reduced, through Cluster Analysis, or "Cluster Busting". Following
image classification, LULC classes that exhibited high degrees of
confusion and misclassification were futher processed in order to
extract a finer level of detail about the landscape. Following
cluster analyses, the land cover types take on a more speckled
appearance, which more accurately mirrors the Nang Rong landscape.
|
Euclidean Distances |
Euclidean,
or straight-line, distances are useful for determining the features
nearest to a theme of interest. The following nearest distance
measures have been calculated:
- Village-to-Village:
For each village, the nearest village. This has been performed
for various subsets of villages (original villages, villages
settled prior to 1950, survey study villages, non-survey villages, etc.).
- Village-to-Transportation Network:
For each village, the distance to the nearest transportation
route. The distance to the major east-west (Chok Chai-Det Udom)highway was also calculated.
- Village-to-Hydrographic Features: For each village, the distance to the nearest perennial water feature.
- Village-to-Health Facilities: For each village, the distance to the nearest health center and hospital.
|
Historical and Future Road Extraction
|
Through
analysis of road network, digitized from the 1984 topographic map,
coupled with the aerial photograph mosaics from the 1950s, 1960s and
1990s, efforts are underway to update the network. For the 1990s,
road arcs are added that appear on the 1994 mosaic. For the
historical updates, the major task is to trace the evolution of
roads from the 1950s and 1960s to the 1980s thorough the subtraction of
road segments.
|
Land Use/Land Cover Classifications |
LULC
classifications for 37 images were performed using a combination of
supervised and unsupervised approaches. Following classification,
the quality of the imagery was quantified using accuracy assessment
tenchiques, which involved comparing ground collected GPS points and
interpretation of aerial photography. Aerial Photograph classification
was performed on a limited scale for intensive study sites in Nang
Rong. This was a less automated endeavor, and instead relied on
manual interpretation of digital air photos. Click here for more information about the classification process.
|
Network Distances |
Network distances, such as
travel time and total distance travelled along a transportation
network, are helpful for determing the perceived and actual distances
of features to a theme of interest. In order to enable such
calculations, the road network was attributed with average travel
speeds and road segment length. The following network distances
have been calculated:
- Village-to-Health Facilities: For each village, the network distances to the nearest health center and hospital for the survey years of 1984, 1994, and 2000.
- Village-to-Village: For each village, the network distances to the nearest village. These have been performed
for various subsets of villages (original villages, villages
settled prior to 1950, survey study villages, non-survey villages, etc.).
- Village-to-Major Highways: For
each village, the network distances to the major east-west (Chok
Chai-Det Udom) highway that bisects the study site was calculated
- Village-to-Hydrographic Features:
For each village, the network distances to the nearest perennial water
feature. Additionally, the flow distance from each village to a
water body was calculated, derived using the digital elevation model
(DEM) and slope.
|
Pattern Metrics |
In order to characterize and
assess the structure and composition of the landscape for various
dates, pattern metrics were calculated for the LULC
classifications. Metrics include measures that assess the make-up
of the landscape, such as number of patches, in addition to measures of
the structure, such as connectivity, patchiness, and contagion.
Various intensive study sites were examined in order to examine the
impact of landscape context on strcuture and composition.
|
Pixel Trajectories |
Combining
GIS, remote sensing and statistical programming methods, the history of
each pixel in our study region will be tracked. LULC
trajectories, or “pixel histories”, are under development,
enabling the examination of space-time patterns of LULC change.
The unit of observation is the pixel, and its “life
history” is constituted by the values derived from the component
images of a satellite time-series. For each pixel, values derived
from satellite images captured at multiple points in time are
concatenated into a sequence of values in a panel data, or pixel
history, approach. Two primary benefits of the pixel history
approach are that the current landscape can be understood through the
dominant trajectories of landscape change, and that periodic patterns
of LULC dynamics can be distinguished.
|
Radiance Diagrams |
Radiance
diagrams, or "Star Diagrams", are another method employed by the Nang
Rong Project to understand the village territory, settlement patterns,
and land use decisions. Since land use and ownership takes place
at the parcel level, the configuration and distribution of parcels is
paramount in understanding village territory. For the parcels
that were used by housholds within a village, a variety of links were
constructed at the village level, including parcel-to-village center,
and parcel-to-household. Valuable information about
household-level land use decisions was facilitated through this type of
analysis.
|
Standard Deviational Ellipses |
Standard
Deviational Ellipses (SDEs) assist in the analysis and understanding of
village territory. The orientation, areal extent, and
distribution of land in a village can be graphically depicted through
the calcualtion of SDEs, which are based on the geographic mean center
of the parcels in a village that are used or owned. At the
village level, SDEs offer insight and information about the impact of setllement on LULC.
|
Village Level Animations |
Animations
have been created for a host of village level attributes. In a
dynamic region such as Nang Rong that has undergone striking changes
over the past 50 years, animated datasets have been helpful in
visualizing and understanding the complex and multifaceted nature of
settlement, expansion, and evolution within our study site. Some
of the animations that have been created include the following:
- Date of Settlement: Villages appear at the discrete point in time that they were established.
- Administrative Village Splits: As
the region grows and evolves, villages that were formerly one village
split into two or more administrative villages, reflected through
animations.
- Village Electrification: Following
years of isolation in the frontier, Nang Rong became increasingly connected with the
rest of Thailand. The development of the electrical grid that
accompanied the integration of Nang Rong into the broader political,
social, and economic framework of Thailand was animated, showing the sequence of electrified villages.
|
Watershed Calculations
|
As
a means of understanding village level decision making, various
watershed areas were calculated for each village in order to ascertain
the influence of biophysical resource endowments on LULC. Access
to water is critical in parts of Nang Rong, owing to the reliance on
rice growing.
|
|
|