Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 121
6
Land-Use/Land-Cover and Population
Dynamics, Nang Rong, Thailand
Barbara Entwisle, Stephen ]. Walsh,
Ronald R. Rindfuss, and Aphichat Chamratrithirong
This chapter describes ongoing research in Nang Rong, Thailand that joins
social, biophysical, and spatial perspectives, data, and tools to study population
and environment there. To date, this research has addressed two sets of ques-
tions:
.
Did land use/land cover in the 1970s and early 1980s affect the subse-
quent out-migration of young adults (Rindfuss et al., 1996a)? More specifically,
did the availability of relatively undeveloped forested land close to villages act as
a brake on out-migration that otherwise would have occurred? Did fragmentation
of land use and potential competition from residents of other nearby villages have
an effect?
· Did population change between 1984 and 1994 including rates of growth
(or decline), household formation, and net in- and out-migration affect land
use/land cover in 1994 (Walsh et al., in press; Entwisle et al., 1997b)? If so, what
was the nature of these effects?
The purpose of this chapter is to use our research experience to consider the
challenges of the multidisciplinary and integrative approach we have taken. Ac-
cordingly, we discuss the history of our projects, the intensive focus on a rela-
tively small geographic area, the development of a prospective multilevel survey-
based social data set, the creation of a complementary satellite time series, the
choice of villages as the primary link between data sources, and other issues of
scale compatibility.
12
OCR for page 122
22
LAND-USE/LAND-COVERANDPOPULATIONDYNAMICS, THAILAND
NANG RONG, THAILAND: LABORATORY FOR STUDY
Nang Rong district, Thailand occupies approximately 1300 km2 in the south-
ern part of the Korat Plateau in Buriram Province in the Northeast of the country
(see Figure 6-1~. The idea that this relatively small geographical area (approxi-
mately the size of a U.S. county) should be the focus for intensive study over a
broad range of interrelated topics is key to our work. The "laboratory" approach
has been used quite successfully by others interested in social change (for ex-
ample, the Middletown studies [e.g., Caplow et al., 19821) or demographic change
(for example, the Matlab project [Phillips et al., 19881), although it is the excep-
tion rather than the rule in social demographic research. Our goal is a compre-
hensive account of social, economic, demographic, and environmental change
within Nang Rong district. Our approach is cumulative. One element at a time,
we have built a database integrating diverse sources (social surveys, administra-
tive data, map products, remote imagery) that is substantively broad, includes
multiple levels of observation, and is unusual in its temporal depth. This data-
base is continually expanding as new sources are identified and new data ac-
quired. In parallel fashion, the knowledge base is also in a state of continual
expansion, as each study builds on preceding ones. We believe the laboratory
approach has much to offer for the study of population and environment.
Why was Nang Rong selected? The origins of our projects lie in a develop-
ment project and in data collection to evaluate its impact. In 1984, the Population
and Community Development Association (PDA) began a Community-Based
Integrated Rural Development (CBIRD) project in selected villages in the dis-
trict. The CBIRD project was designed to (1) improve skills and productive
capacity in agriculture, animal husbandry, and cottage industry, and (2) upgrade
waste disposal facilities, increase year-round availability of clear water, and pro
mote improved individual health care practices. Nang Rong was selected be-
cause it was (and is) a relatively poor district in an historically poor region of
Thailand. "Nan" Rong" means "sit and cry."
The CBIRD project and associated data collection provided the base for our
projects, but what sustained our interest was the multifaceted nature of change
under way in Nang Rong district. The past few decades have witnessed a major
fertility decline, increased use of contraception, electrification, the extension and
improvement of the road network, increasing frequency of bus service, more
widespread use of tractors (mostly "walking tractors") for land preparation, an
increase in the number of mechanized rice mills, and improvements in sanitation
and water storage. Initially, we were interested in contraceptive choice within a
changing environment; we later expanded our scope to consider migration and
fertility in the context of change. With the broadening of our substantive inter-
ests, along with additional data collection (described below), the work that began
as an evaluation of the CBIRD project grew until Nang Rong became a laboratory
for the study of social and demographic change.
OCR for page 123
BARBARA ENTVEISLE ET AL.
.
MYANMAR ,~
~LAOS
~0 '\
.~.2',.:;. -,i.:.,2;';~,.:
:.:.:..-.-, ~;~ It
· I; :. ~ , ~
~ i i.
: -34ndaman
.. ~
., .
. . ~ ~ ,. .
~ :~ - Sea ~
,
123
VIETNAM ~ --
~'fentian^e\
THAILAND
,Bangkok :7
~ ~. ~:'~.-~ . ~
So {.- ~.~,-,.,~.',.-.:,,
f'~::':':~:'~'~"-' ~'-~.,--.:.-
\~ ~ ~ ~ .~6 ~ -; ~-~ ;
~\,;-. -,~11. ~. ,, ;
~- ~ .~''~. i'''
~ \,-'
Rong ~ ~
.~4
)
CAMBODIA ~
,~1
.Phnom Penh
_^,,f
>~ ':'';
At, :~
~ MALAYSIA-\ ~
FIGURE 6-1 Location of Nang Rong study area.
a,. ~-
OCR for page 124
24
LAND-USE/LAND-COVERANDPOPULATIONDYNAMICS, THAILAND
Given the concept of Nang Rong as a laboratory, extending our work to
include environmental concerns was a natural step. The area that makes up the
district is noted for its undulating landscape; in the lower parts there is paddy
land, and in the higher parts open-canopy forest and upland for dry field crops
(Fukui, 1993~. As with the Northeast as a whole, the region is composed of
Cretaceous sandstone, shale, and siltstone overlaid with Tertiary and Quaternary
alluvial deposits that have been eroded to form a succession of low-lying terraces
(Pendleton, 1963; Donner,1978~. As is typical of the Korat Plateau, the soils are
dominated by fine sandy loams that form one of the most infertile soil groups in
Thailand. Cultivation of wet rice occurs in shallow depressions, low terraces, and
alluvial plains, while dry dipterocarp savanna forests or drought-resistant crops
are grown on the upland sites (Parnwell,1988~. Annual rainfall is low, leading to
low rice yields and interannual instability (Limpinuntana et al., 1982; Fukui,
1993~. Precipitation is also variable. While nearly 80 percent of the annual total
falls within a 5-month period from May to September, the rains may arrive in
early March or well into June or July. September may be a dry month, or it may
be so wet that extensive flooding occurs. Droughts are not uncommon in the
middle of the rice-growing season, June to September. Precipitation extremes are
the rule rather than the exception in Nang Rong.
Major changes in land use have occurred over the past several decades,
paralleling social, economic, and demographic change. Agricultural productivity
grew rapidly between 1960 and 1980, mostly because of the rapid extensification
of agricultural land into formerly forested and upland areas (Siamwalla et al.,
1990~. In the Northeast as a whole, 60 percent of the land was forested in 1960,
but only 15 percent in 1982 (Panayotou and Sungsuwan, 1989~. Cropping pat-
terns also changed. Marginal lands were cleared initially to grow cassava, ex-
ported to Europe as a calorie rich ingredient for livestock feed (Phantumvanit and
Sathirathai, 1988~. In recent years, extensification has slowed as a result of
preventive deforestation policies, demands for surplus labor in the rapidly grow-
ing manufacturing and urban sectors (Tambunlertchai, 1990), and the availability
of only marginal lands for further agricultural expansion. The interaction of
population and environment through forces within and exogenous to the region
has created a dynamic landscape mosaic.
Nang Rong thus offers an important context in which to study interrelation-
ships between population and environment, and the already existing data and
knowledge base helped justify the considerable additional investment required to
establish a geographic information system (GIS), acquire and process remote
imagery, and integrate social survey and remotely sensed data in analyses of
population change and land use/land cover.
PROJECT HISTORY
In the fall of 1992, we collaborated on our first grant proposal, submitted to
OCR for page 125
BARBARA ENTWISLE ET AL.
125
the National Science Foundation (NSF), to integrate spatial data, methods, and
perspectives into ongoing social and demographic research in Nang Rong. Sara
R. Curran, now of Princeton University, contributed importantly to this effort.
With NSF funds, we accomplished several important goals. First, we developed
the basic elements of a GIS to link spatial and social survey data. Second, we
purchased, processed, and evaluated five frames of Landsat satellite data from the
1970s. Details are given below, but for purposes of describing the project his-
tory, suffice it to say that the satellite data provided an initial look at land use/land
cover in the district, showing, for example, the dominance of rice cultivation,
spatial variability in crop mix and landscape pattern, and change over time. The
promise of Nang Rong as a site for research on population and environment was
thus borne out in the initial remotely sensed data. Third, and more substantively,
we initiated a study of out-migration in relation to land use/land cover (Rindfuss
et al., 1996a).
While the NSF proposal was under review, we developed a related initiative
to integrate social and spatial data in the evaluation of family planning programs.
Funded as a subproject under the EVALUATION Project, sponsored by the U.S.
Agency for International Development, with additional funds from the Mellon
Foundation, this effort, led by Amy Tsui, explored the GIS as a means of organiz-
ing and integrating family planning program data from multiple sources and time
points. Alternative measures of family planning accessibility were derived from
spatial data using spatial network analysis, and then incorporated into statistical
analyses of contraceptive choice that drew on measures based on social survey
and administrative data as well (Entwisle et al., 1997a). Although not specifi-
cally focused on population and environment, these analyses helped further the
GIS and our collaborative strength in the marriage of social and spatial data,
techniques, and perspectives.
Consequently, when the National Institute of Child Health and Human De-
velopment (NICHD) issued its Request for Applications on "Population and
Environment" in June 1994, we were well positioned to compete for funds. We
had already created a data set for Nang Rong that was unique in its social, spatial,
and temporal coverage, and a program of social demographic research based on
these data was well under way (Entwisle et al., 1996; Rindfuss et al., 1996b).
Instrumental to this latter research was an initial grant received from the National
Institutes of Health (NIH) to study contraceptive choice in a changing environ-
ment based on 1984 and 1988 data, which was followed by a competing continu-
ation grant to collect additional social data in 1994-1995, investigate social net-
works in relation to other characteristics of social context, and expand the scope
to include migration as well as fertility. A GIS was in place. We had processed
satellite data and could demonstrate their relevance to a study of land-use/land-
cover and population change. An interdisciplinary team was in place that had
collaborated successfully and developed some common language.
Our project, "Population Dynamics and Changes in the Landscape," was
OCR for page 126
26
LAND-USE/LAND-COVERANDPOPULATIONDYNAMICS, THAILAND
funded and is under way at the time of this writing. The first goal was to expand
the time series of optical satellite images for Nang Rong from 1972 through 1996,
and also to incorporate radar data and aerial photography extending back to the
1950s. The other goals of the project revolve around use of the satellite time
series, combined with data from social surveys, administrative records, and digi-
tal maps, to explore dynamic interrelationships among land use, population, and
social and economic change. As indicated earlier, these explorations have fo-
cused thus far on land use/land cover in the 1970s and early 1980s as an influence
on subsequent out-migration of young adults, and on population change between
1984 and 1994 as an influence on land use/land cover in 1994. Next steps include
the development of a household model linking land use and migration, a study of
the consequences of past high fertility for the current size of landholdings and for
the fragmentation of agricultural plots, and the construction of models for assess-
ing landscape organization and use of these models to simulate trajectories of
future change.
A related project, funded by the MacArthur Foundation, has focused specifi-
cally on villages as the link between measures derived from the satellite time
series and those derived from social survey and administrative data. The maps
available to that project are based on different concepts of "village," and we are
just now completing the field work necessary to resolve this problem. The 1984
topographic map identifies clusters of dwellings as villages, whereas the 1993
planning map identifies administrative units. This difference in concept, com-
bined with real change over the decade separating the maps, led to substantial
discrepancy between the two maps. Each map shows locations for approximately
300 villages, but only 200 of those villages appear in both maps. Accounting for
and resolving these differences is vital for analyses of land use/land cover and
population over time.
We have described the history of our Nang Rong projects in terms of each
project building on previous ones, but this is not the whole story. An expanding
research infrastructure for spatial analysis paralleled the progression of indi-
vidual research projects. There is a division of labor between the Carolina Popu-
lation Center (CPC) and the Spatial Analysis Laboratory in the Department of
Geography at the University of North Carolina with respect to data processing,
but the nature of this division of labor has changed over time. Initially, social
data processing took place at CPC, and spatial data processing took place at the
Spatial Analysis Laboratory. Gradually, CPC's Spatial Analysis Unit was devel-
oped, first with start-up funding from the university's Vice Chancellor for Health
Affairs and then with funding from an NICHD "Center" grant. We now use both
spatial analysis units, which has increased efficiency and allowed us to draw on a
wider range of expertise.
OCR for page 127
BARBARA ENTVEISLE ET AL.
127
DATA
Social Survey Data
Key to our research on land-use/land-cover and population change is pro-
spective longitudinal and multilevel survey data, with observations for individu-
als, households, and villages covering the 1984-1994 period. As mentioned, the
longitudinal data collection began as an evaluation of the CBIRD Project (Insti-
tute for Population and Social Research, 1984:3~. To evaluate the success of the
initial 4-year project, the Institute for Population and Social Research (IPSR),
Mahidol University, Thailand, with financial assistance from the Canadian Inter-
national Development Research Center, conducted a complete household census
in each of 51 villages in 1984, a 1988 follow-up survey of households containing
a woman of reproductive age, and community surveys in the 51 villages at both
time points as part of a comprehensive research effort. These are known as the
CEP (Community Evaluation Project) surveys. No U.S. researchers were in-
volved in the 1984 surveys, but Ronald R. Rindfuss and David K. Guilkey of
CPC collaborated with Aphichat Chamratrithirong and other IPSR researchers to
design the 1988 data collection. The idea was to piggyback a study of contracep-
tive choice onto the CBIRD evaluation already planned (Entwisle et al., 1996;
Rindfuss et al., 1996b). The linked data sets have also been used to study
migration (Curran, 1994; Sawangdee, 1995~.
Likewise, the 1994 surveys (funded by an earlier NIH grant) built on the
foundation of the 1988 survey panel. These surveys, known as the CEP-CPC
surveys, followed up individuals, households, and villages included in the 1984
data collection, but with a greatly expanded research agenda. The 1994 data were
intended to support prospective migration studies, detailed descriptions of life-
course transitions in the young adult years, and studies of social networks at
multiple levels of observation (individual, household, village) in relation to de-
mographic behavior. Four different theoretical perspectives guided this effort:
Davis' (1963) multiphasic theory, which stresses multiple demographic responses
to social change (especially the interconnectedness of fertility and migration);
multilevel approaches to examining social contexts and demographic behavior
(Entwisle and Mason, 1985; Bilsborrow et al., 1987; Findley, 1987~; the life-
course perspective, with its interest in individual role trajectories over time
(Clausen, 1972; Elder, 1985, 1991~; and social network theory (e.g., Freeman,
1979) as applied to demographic phenomena (Rogers, 1983; Massey and Garcia
Espana, 1987~. There are three components to the 1994 data collection: a
community profile, a household survey, and a migrant follow-up.
The 1994 community profiles were done in all of the Nang Rong villages,
including but not limited to ones that were initially part of the CBIRD evaluation.
The community profiles done in 1984 and 1988 were fairly extensive. The 1994
data collection obtained much of the same detail, including population size and
OCR for page 128
28
LAND-USE/L~AND-COVERANDPOPULATIONDYNAMICS, THAILAND
composition; cropping (e.g., specific crops grown, prices received, marketing);
water sources (location, quality, and sufficiency); agricultural technology (use of
various kinds of equipment, fertilizers, and pesticides); electrification; transpor-
tation and communication; health and family planning services; and village groups
and committees. In addition, the 1994 community profile covered some new
topics, including links to other villages by virtue of shared temples, schools, and bus
routes; labor exchanges between villages; and perceptions about deforestation.
Although not specifically designed to be used with remotely sensed data, the
village profiles provide data useful for many purposes. First, survey information
about cropping, use of fertilizer, water sources, and deforestation provides a
cross-check on the interpretation of remotely sensed data (and vice versa). Sec-
ond, information about links and exchanges between villages can shed light on
questions of scale (see below). Although the village seems to be an appropriate
level at which to conduct both social and spatial analysis, it is possible that some
clustering of villages may be better. Social network analysis can be used to
develop a basis for such aggregation. Third, village profile data can be linked
with variables derived from remotely sensed data, and statistical analyses can
then be undertaken. We have conducted some preliminary analyses of popula-
tion, land use/land cover, and landscape diversity along these lines (Walsh et al.,
in press; Entwisle et al., 1997b).
A household survey was the second component of the 1994 data collection.
In fact, the survey consisted of a complete household census in each of the 51
villages that participated in the 1984 and 1988 data collections. The 1994 survey
addressed the topics covered previously, and considerably more: substantial
detail on the whereabouts and current characteristics of 1984 household mem-
bers; visits and exchanges of goods and money with former household members;
social and demographic facts about current members; yearly life history data for
those aged 18-35, including information about migration patterns and occupa-
tion; sibling networks for those aged 18-35; and household characteristics, in-
cluding plots of land owned and rented, use of agricultural equipment, crop mix,
planting and harvesting of specific crops (rice, cassava, and sugarcane) and house-
hold debts.
These data have quite a bit to offer when used in conjunction with remotely
sensed data. Aggregated to the village level, household data offer an additional
perspective on the interpretation of satellite images. Data on crop mix and on the
timing of planting and harvesting can be used to verify land-use/land-cover clas-
sifications based on satellite data for the same time period. It is also possible to
estimate the total amount of land cultivated by village households, and this infor-
mation can be used to adjust spatial boundaries around villages. As explained
below, one of the challenges faced in joining social survey and remotely sensed
data is the identification of a common unit of observation. Thus far, the village
has served as our common unit. This approach has worked reasonably well, but
a difficulty is that the spatial boundaries of a village are not so easy to determine.
OCR for page 129
BARBARA ENTWISLE ET AL.
129
Tom Evans of the University of North Carolina Department of Geography is
working on a dissertation that will help clarify questions about boundaries in the
Nang Rong setting. In the meantime, however, clues provided by the aggregated
household data as to which villages are quite large spatially and which are quite
small add realism to some of the arbitrary procedures we must employ.
Another key strength of the household data is the prospective design. Using
the 1984 household rosters as a basis, survey interviewers in 1994 attempted to
find out about all households and household members present in 1984. Identify-
ing information was obtained so that the 1994 and the 1984 records could be
linked. This means that for the 51 villages in which the household censuses were
done, detailed data are available not only about population change over the 10-
year period, but also about changes in population composition (for example, age,
education, occupation, assets). Further, it is possible to differentiate between
changes to households and household members present in 1984 and changes due
to new individuals moving into the village and to the formation of new house-
holds. Most important, it is possible to examine population processes prospec-
tively. We have conducted a preliminary analysis in which the out-migration of
young adults between 1984 and 1994 is examined in relation to the availability of
undeveloped land (based on remote imagery from the mid-1970s), the fragmenta-
tion of land use (also based on remote imagery), and the potential for competition
from other villages (Rindfuss et al., 1996a). During this period, marginal lands
were cleared to grow cassava in Nang Rong (and elsewhere in the Northeast) as
part of the rapid extensification of agricultural land into formerly forested and
upland areas. Consistent with this picture, we found that the availability of
forested land in the 1970s was associated with diminished out-migration in the 51
villages where household data were collected. The prospective design has also
made it possible for us to consider household formation, as well as changes in
household size and composition, as a potential influence on land use/land cover
(Entwisle et al., 1997b).
The final component of the 1994 survey is a follow-up of out-migrants from
22 of the 51 villages. These 22 villages were selected randomly within strata
created by cross-classifying general location (quadrant) and distance from major
paved roads in 1984. Persons resident in 1984 but no longer resident in 1994
were candidates for follow-up if they had gone to Bangkok and surrounding
areas, the Eastern Seaboard (a focus of rapid growth and development), Korat (a
regional city), or Buriram (the provincial city). We succeeded in finding and
interviewing about 70 percent of the migrants who still had family in Nang Rong,
a remarkably high proportion (cf. Bilsborrow et al., 1984~. In Nang Rong, migra-
tion of young adults is extremely common. Some return, while some do not. An
understanding of migration patterns thus requires an understanding of the factors
influencing some temporary out-migrants to become permanent out-migrants.
Such an understanding in turn requires prospective and retrospective data on
OCR for page 130
130
LAND-USE/LAND-COVERANDPOPULATIONDYNAMICS, THAILAND
those who leave, as well as those who stay or have returned. The Nang Rong data
are unique in providing such a complete picture of migration.
As we have stressed, the 1994 surveys were not designed specifically for use
with satellite datai (and we would have done some things differently had we had
the wisdom of hindsight), but collectively, and combined with the 1984 and 1988
data collections, the survey data offer excellent opportunities for analysis in
conjunction with remotely sensed data. There are many features of the survey
data that make their integration with satellite imagery especially exciting. One is
coverage. As noted earlier, the 1994 community profiles were done in all of the
villages, while data were collected from all households in a subset of 51 villages.
A second important feature is the multilevel structure of the data: there are
observations for individuals, households, and villages. With this structure, even
though we cannot link individual people to individual pixels, it is still possible to
study individuals and households in contexts characterized in part by satellite
data. This is important because many land-use decisions are made by individuals
and households. A third important feature is temporal depth. Ten years is not
such a long time for interrelationships between population and environment to
manifest themselves, but it is a longer period than that spanned by many social
demographic data sets, especially prospective ones. A final and related feature is
that the data follow migrants as well.
Remotely Sensed Data
Remote images obtained from the Landsat Multispectral Scanner (MSS) and
Thematic Mapper (TM) and from the Systeme pour ['Observation de la Terre
(SPOT) Multispectral (MX) and Panchromatic (PAN) sensors are also key to our
research on land-use/land-cover and population change in Nang Rong. The
following discussion briefly describes the approaches we are following to assess
land use/land cover and plant biomass through satellite spectral responses. GIS
techniques used to derive biophysical and geographical variables are also de-
scribed, as are pattern metrics for assessing the composition and spatial organiza-
tion of land use/land cover derived from the digital classifications.
Using a derived local crop calendar, we constructed a satellite time series
that incorporates Landsat MSS, TM, and SPOT MX and PAN data. The MSS
data include scenes for 18 December 1972, 28 February 1973, 24 November
1975, 17 January 1976, and 7 October 1979; the TM data include scenes for 17
November 1988, 6 February 1989, 15 December 1992, 2 February 1993, 3 No-
vember 1994,6 January 1995,7 February 1995,6 November 1995,24 December
1995, and 25 January 1996; and the SPOT data include MX and PAN scenes for
7 April 1994. The time series was configured to provide (1) historical coverage
prior to the 1984 population survey and subsequent to the 1994 survey; (2)
multispectral, multispatial, and multitemporal representation; and (3) interannual
and interseasonal characteristics for defining the nature of land use/land cover
OCR for page 131
BARBARA ENTWISLE ET AL.
131
and the magnitude, pattern, and directionality of change. Additional Landsat TM
and SPOT MX scenes will be acquired as our research proceeds.
Based on our experiences with processing of images for the region, the
importance of a satellite time series for land-use/land-cover mapping has become
clear. Variation in local and regional phenologies, planting schedules across the
district, and land-use/land-cover types suggests the need for a minimum of two
and preferable three images per analysis period for satisfactory land-use/land-
cover separation. Following data preprocessing to correct for geometric, radio-
metric, and topographic distortions in the data, we used empirical normaliza-
tion procedures to balance and correct responses from data acquired on
different dates and therefore under different solar angles and atmospheric
attenuation characteristics. The image time series was then classified for
land-use/land-cover extraction. An unsupervised classification approach was
used to map land use/land cover at multiple dates, with emphasis on 1979
and 1993. Level 1 and Level 2 mapping classes were defined through the
classification process, with emphasis on the pattern of deforestation, agricul-
tural intensification in rice-growing areas, and agricultural extensification of
cassava in upland sites. Plate 6-1 (after page 150) provides an example.
Supervised classification is also being used, as well as hybrid approaches, as
our ground control information expands areally and across land-use/land-
cover types, and as we seek to identify specific cover types for selected
villages, landscape strata, or time periods.
The accuracy of a classification is summarized through an error matrix,
Kappa statistic, and omission and commission errors by comparing satellite-
based land-use/land-cover categories against aerial photography and/or field
samples. We are using two approaches. The household and village surveys
contain information on related variables, such as the total area planted in various
crops, which can be compared with the Level 1 and Level 2 classes. Also, as a
consequence of prior trips to the district, a ground-control GIS directory has been
created and expanded over time. This directory contains assorted ground-control
coverages that relate to various land-use/land-cover verification efforts conducted
for various villages within the district. Locations are defined through Global
Positioning System (GPS) values and through delineated plot locations on scaled
Landsat TM and SPOT PAN satellite output.
Our initial work with these land-cover classifications has involved ex-
amining the relationship between migration patterns and patterns of land-
cover change (Rindfuss et al., 1996a; Entwisle et al., 1997b), as well as more
general relationships between population and land-cover patterns (Walsh et
al., in press). As part of this work, we are interested not only in types of land
cover, but also in the way the land was compositionally and spatially orga-
nized, so we can begin to assess the nature of human-environment interac-
tions through the scale, pattern, and process paradigm that is emergent in
ecology and other natural sciences.
OCR for page 134
LAND-USE/LAND-COVERANDPOPULATIONDYNAMICS, THAILAND
134
FIGURE 6-3 Units of observation and
links among them.
FIGURE 6-4 Fishbone pattern: Stylized
relationship between place of residence and
land ownership and use.
FIGURE 6-5 Place of residence and land
ownership and use in Nang Rong.
People
T
| Household ~ At
:~3
Pixel
shown as Figure 6-4. It is conceptually straightforward to link people, plots, and
pixels given this pattern, although there are complications in practice, as indi-
cated in the chapter by Moran and Brondizio in this volume. The settlement and
farming pattern in Nang Rong departs from the stylized fishbone pattern in sev-
eral important ways, as illustrated in Figure 6-5. Household residences are clus-
tered in villages; the location of a household provides few clues about the loca-
tion of the land farmed by that household. Further, households farm multiple
plots (as many as 11), and these plots may be and often are scattered. If we had
the spatial coordinates of all the plots farmed by all households in a village, as
OCR for page 135
BARBARA ENTWISLE ET AL.
135
well as a clearly defined village territory, it would be possible to move among all
the units of observation indicated in Figure 6-3. Since we do not have these
spatial coordinates, we link at the village level instead.
The population surveys conducted in 1984, 1988, and 1994 at the household
and village levels are represented in the GIS as discrete point locations mapped at
the village centroid. Such a spatial representation is appropriate given the nuclear
settlement pattern of Thai villages. Biophysical and geographical information
derived through remote sensing techniques, however, represents space as a con-
tinuous surface summarized at the pixel level. Integration of social and environ-
mental data therefore requires a transformation of space whereby a polygon
representation is used to denote the pattern and variability of landscape condi-
tions associated with discrete village locations. This transformation involves
defining village boundaries, a complex issue in a region where political delinea-
tions change as village populations grow, and social, cultural, biophysical, and
geographical parameters combine to influence the geometry and areal extent of
village boundaries.
Our initial approach is fairly simple and involves generating radial buffers
around the nuclear village centroids at distances of 2 and 3 km. This model is
simple to implement and has particular relevance to the study area in that it (1)
assumes travel distances from households to field plots appropriate to travel
conditions during the 1984-1994 study period, which involved primarily walk-
ing; (2) provides for overlapping village boundaries associated with the density
and location of villages; and (3) represents the nuclear village settlement concept,
whereby residents disperse from village centers to engage in various land activi-
ties. Plate 6-2 (after page 150) shows an example of 3-km buffers for our 51
study villages overlaid on a 1993 TM land-cover classification. This figure
makes it clear that villages may be competing with one another for available land,
and we include measures of competition in our models (Entwisle et al., 1997b;
Rindfuss et al., 1996b).
We are also beginning to explore other approaches for setting village bound-
aries, including Thiessen polygons, population-weighted Thiessen polygons, and
the Triangulated Irregular Network (TIN). Such approaches produce nonover-
lapping village boundaries and irregular boundary dimensions and orientations.
Thiessen polygons maintain a one-to-one relationship between points and poly-
gons. Any location within the defined polygon is closer, in Euclidean distance, to
the associated point than to any of its neighboring points, and polygons vary in
size inversely with point density. Plate 6-3 (after page 150) shows an example
using a Thiessen polygon approach incorporating all villages in Nang Rong over-
laid on a 1993 Landsat image.
In addition, doctoral research by Tom P. Evans noted earlier is examining the
concept of distinct and fuzzy boundaries for representing territories of Thai vil-
lages. A fuzzy boundary can be created, for example, by land-ownership pat-
terns, a product of the degree of land fragmentation around a village. Land
OCR for page 136
136
LAND-USE/LAND-COVERANDPOPULATIONDYNAMICS, THAILAND
fragmentation may develop as a consequence of intradistrict migration, land-
inheritance practices, land-tenure patterns, and environmental variation. Fuzzy
boundaries between villages suggest, in the context of Thai culture and society,
the concept of functional boundaries that overlap in space and vary by social,
cultural, biophysical, and geographical domains.
Availability of Exogenous Variables
We are interested in the interrelationships between population growth and
change on the one hand and land-use and land-cover change on the other. While
it is possible to examine associations, we also want to be able to examine causal
relationships. The difficulty here is that the settlement of Nang Rong predates
our data. Even at the earliest point in our time series, the land-cover pattern has
been influenced by human behavior, and human behavior has been influenced by
land-cover patterns. To draw causal inferences, one must be able to identify
variables exogenous to the system of interrelationships. One such variable is
elevation, which thus far in Nang Rong has not been influenced by human behav-
ior or changes in land cover (Entwisle et al., 1997b).
The contour lines and point elevations on our 1 :50,000 scale base maps have
been scan-digitized and attributes attached, edge-matched, and processed to yield
digital elevation models (DEMs) of the study area. Figure 6-6 shows the topo-
graphic inputs for the DEMs. The DEMs are being used to characterize the
percentages of alluvial plain/lower terrace, middle/high terrace, and upland in the
2- and 3-km boundaries associated with each of the district villages.4 The con-
struction of DEMs is important for the characterization of topographic site and
situation relationships associated with land-use/land-cover potentials (Entwisle
et al., 1997b). For example, most recent agricultural production, occurring on
low terraces and alluvial plains, has seen a substantial amount of the agricultural
intensification related to paddy rice production.
In addition to obtaining elevation from the DEMs, we plan to work with
several other derived measures to see if we can obtain exogenous measures of
land suitability for various types of land cover and land use. One of the most
common indices derived from DEMs, used to indicate potential wetness for a
site, is the Topographic Convergence or Wetness Index (B even and Kirby, 1979;
Moore et al., 1991; Wolock and McCabe, 1995~. This index uses calculations of
slope and upslope contributing area (the area that drains through a given location)
to generate a dimensionless index of potential wetness, which provides a simple
but effective model used to characterize channels, gullies, and moisture sinks
(Allen and Walsh, 1996; Townsend and Walsh, 1996~. A digitized hydrography
layer was used to support the calculation of the index. Topographic curvature
was generated to integrate planform and profile curvature in order to portray
localized topographic situations (Moore et al., 1991) within the soil moisture
potential surface. The variable distinguishes slopes having a convex versus a
OCR for page 137
BARBARA ENTVEISLE ET AL.
Contour Lines (10m interval)
t
A.
Topographic Sinks
137
~ . . . . .
~e~5~e-~
IN · ee ~ en ar .. A- :
Spot Elevations (1m vertical resolution)
Linear Surface Drainage Patterns
FIGURE 6-6 Topographic data inputs for generation of digital elevation models, Nang
Rong, Thailand.
concave slope. Profile curvature is the vertical or downslope concavity/convex-
ity and portrays slope steepness. Slope curvature affects the acceleration and
deceleration of gravity flows. Solar radiation affects the soil moisture, tempera-
ture regimes, and photosynthetic capacity of the landscape. The potential direct
solar insolation at any location will be represented through grid locations through-
out the study area, given DEM-derived elevation, slope angle, and slope aspect
information. The procedures are modeled after Bonan (1989) and Montieth and
Unsworth (1990~. Shadowing by adjacent terrain is an important factor influenc-
ing potential insolation. Solar radiation potential is derived by integrating the
OCR for page 138
138
LAND-USE/LAND-COVERANDPOPULATIONDYNAMICS, THAILAND
effects of shadows, direct solar insolation, and insolation variability throughout
the normal growing season within the study area. Values can be calculated for
various time steps on an hourly, daily, and seasonal basis, and in future work we
plan to explore the most appropriate time periodicity.
Scale of Analysis
The effects of spatial scale, pattern, and process are inherently involved in
landscape studies. Processes shaping the landscape function only at certain ranges
of spatial and temporal scales, and their effects vary with scale (Woodcock and
Strahler, 1987; Nellis and Briggs, 1989~. An important consideration in linking
survey and image time series data is thus the scale of analysis. The spatial
resolution of remote sensing systems is the pixel, whose dimensions vary with the
sensor system and reconnaissance platform. As a consequence of the continuous
nature of remotely sensed data organized within a raster data model, data aggre-
gation approaches are available for altering the pixel dimension to a coarser grain
size where appropriate. Scale-dependent analyses use such an approach to define
the range of spatial scales of remotely sensed data and continuous GIS coverage
that may be highly autocorrelated in an effort to interrelate scale, pattern, and
process.5 Integration of data collected from different satellite systems (e.g.,
SPOT MX, Landsat TM, the National Oceanic and Atmospheric Administration' s
[NOAA] Advanced Very High Resolution Radiometer [AVHRR]) offers an ad-
ditional approach to spatial resolution through information scaling. While popu-
lation data can be transformed from discrete point locations to a continuous
surface through spatial interpolation and spread approaches, natural population
units (e.g., household, village, district) may be more suitable. The summarizing
of satellite data at the village level is subject to difficulties related to the setting of
village boundaries, whereas the use of agricultural plots as the unit of measure is
subject to the spatial resolution limitations of the satellite systems and the con-
straints of the population survey that make it difficult to link households to
geographic lot locations.
The temporal dimension of remotely sensed data is achieved by acquiring
images for single time slices that are combined into a time series. Population-
based life histories and retrospective analyses based on population surveys offer
a distinctly different approach for representing the temporal dimension as seen
through the creation of an image time series.
While questions of information scaling within social and biophysical sys-
tems need to be addressed more effectively, the scaling of social and biophysical
relationships across temporal and spatial domains and within similar and differ-
ent environments offers unique research challenges. In our research, we have
defined a scale continuum for the population, environment, and geographical
domains that includes, respectively, (1) household, village, and village clusters;
(2) field plot, terrain strata, village, and watershed; and (3) pixel, aggregated
OCR for page 139
BARBARA ENTVEISLE ET AL.
139
pixel, village, terrain strata, and watershed. The temporal range extends from the
1970s to the present, with particular emphasis on 1984, 1988, and 1994. The
temporal scale of analysis is the season, year, and decade. Research is ongoing at
various scales along the continuum, and other research is planned. A challenge is
to link population-environment data at various scales, define important variables
along the scale continuum, scale the variables from one resolution to another, and
assess the important processes and relationships across domains and across spa-
tial and temporal scales.
CONCLUSIONS
In conducting the research described above, our research team has benefited
from expertise across a range of substantive and methodological specialties. In
our case, long-standing research programs in social and biophysical topics have
been integrated to address population-environment questions through a synergis-
tic relationship among sociology, demography, geography, and spatial techniques.
While the exchange of language and concepts among disciplines is still occur-
ring, team members assumed leadership roles in various substantive and method-
ological areas early in our research relationship to move quickly up the learning
curve. Literature exchanges, processing discussions, conceptual dialogue, loca-
tion at the same university, standing in one's scientific community, similarity of
research styles and modes of work, and early success with funding agencies
contributed to our collaborative efforts. Other advantages of our research col-
laboration included team proficiencies in a broad-based set of research methods,
as well as an appropriate computer infrastructure to support remote sensing im-
age processing, GIS, spatial analysis, GPS technology, and statistical modeling.
The availability of population survey data collected for the study area in 1984,
1988, and 1994 was another essential ingredient of both our start-up efforts in
linking spatial and social data and our sustained program of inquiry. The survey
data and its intensive preprocessing at IPSR, Mahidol University, and the CPC
provided the temporal and longitudinal depth for our ongoing analyses. Finally,
especially in the early years of the project, we benefited from the willingness of
funding agencies and program administrators to take a risk on an unproven re-
search plan. Consider these comments by a reviewer of one of our earliest
proposals:
This research proposal is complex and ambitious....It is highly innovative and
interdisciplinary. This presents pitfalls as well as exciting opportunities....It is
difficult to assess the potential value of the research; it is an experiment, and we
won't know until it's been tried....My reaction is to forgive the pitfalls and to
embrace the experiment as plausible and worthwhile (but risky). I suspect that
serious efforts to incorporate environmental factors into social science will re-
quire spatial and interdisciplinary research.
Fortunately, the funding agency followed the recommendation of this reviewer
OCR for page 140
140
LAND-USE/LAND-COVERANDPOPULATIONDYNAMICS, THAILAND
and provided crucial start-up money. This sort of vision on the part of funding
agencies and program administrators is still needed.
Remotely sensed data provide a valuable landscape perspective that is sub-
stantially enhanced through the availability of corresponding social data. Satel-
lite time series offer the best opportunity for landscape analyses over time and
space. The use of remotely sensed data for international sites, however, is prob-
lematic in that archived data are often limited, multitemporal data for examining
interannual or interseasonal variations are difficult to assemble and acquire, and
only a few commonly available satellite systems offer alternate views of the
landscape. Finally, while remote sensing systems and data accessibility could be
enhanced, population surveys could also benefit from the routine geocoding of
landscape features (e.g., households and village centroids) as part of the survey,
the formulation of survey questions that directly examine spatial pattern and
geographic location (e.g., spatial networks and field plot positions), and the ex-
amination of population-environment issues across a range of spatial and tempo-
ral scales.
NOTES
1 The planning for the 1994 survey began in 1990, well before Aphichat Chamratrithirong,
Barbara Entwisle and Ronald R. Rindfuss were aware of the potential for linking the social data to
satellite data.
2 Turner's (1990) SPAN (Spatial Analysis) algorithm was one of the first widely distributed
landscape pattern programs. Users of the Geographic Resource Analysis System (GRASS) use a
module developed by Baker and Cal (1992) for examining multiscale landscape patterns in the "r.le"
supplemental programs. DeCola and Montagne (1993) have published results from the PYRAMID
system, a package designed to analyze multiscale spatial structure in raster data sets, and McGarigal
and Marks (1993) developed FRAGSTATS, a widely used package of composition and pattern
metrics. Other recently developed packages for assessing composition and pattern include APACK,
a free-standing package for spatial analysis of raster data; DISPLAY, which calculates landscape
structure through a set of metrics that can be processed in batch mode and implemented sequentially;
RULE, which assesses landscape organization according to defined neighborhood dimensions; and
HISA (Habitat Island Spatial Analysis), used to calculate a host of landscape metrics from digital
land-cover maps.
3 The 1984 maps showed nucleated settlement patterns as "villages," irrespective of the num-
ber of households in the village. The 1992-1993 planning maps showed administrative village
locations, based on an administrative system that aims for villages of approximately 100 households.
Needless to say, with two different concepts behind the mapping procedures, the two maps do not
agree, even taking into account the time differences between them. At the time of this writing, field
work is in progress that will result in our being able to move back and forth between these two
conceptually distinct types of villages.
4 The TOPOGRID module in ARC/INFO was used to interpolate the DEM from the scanned
contour data. TOPOGRID generates a DEM from contour lines, point elevation data, and hydrogra-
phy information using an iterative finite difference interpolation technique developed by Hutchinson
(1989). The algorithm interpolates elevation values iteratively using a thin plate spline, thereby
developing an optimal flow model that maintains the integrity of the input data while simultaneously
ensuring surface continuity (Hutchinson, 1993). The algorithm is optimized by using digital hydrog-
raphy data to constrain TOPOGRID to identifying the appropriate low elevations along drainages.
OCR for page 141
BARBARA ENTVEISLE ET AL.
141
5 Semivariograms and fractals will be used to assess the nature and degree of the autocorrelation
among landscape features at different spatial scales. Walsh has explored the scale dependence
among elevation, slope angle, and slope aspect, derived from GIS-based DEMs, and plant biomass,
defined through satellite digital data in an alpine environment in Montana (Bian and Walsh, 1993).
Bian and Walsh addressed (1) the effective range of spatial scales within which plant biomass and
terrain variables were spatially dependent; (2) optimum spatial scales for representing terrain and
plant relationships; and (3) the degree of spatial dependence of these relationships. Walsh et al.
(1997) report that the spatial organization of the Normalized Difference Vegetation Index (NDVI) at
finer scales is an important indicator of the NDVI at coarser scales.
REFERENCES
Allen, T.R., and S.J. Walsh
1996 Spatial and compositional pattern of alpine beeline. Glacier National Park, Montana.
Photogrammetric Engineering and Remote Sensing (in press).
Baker, W.L., and Y. Cal
1992 The r.le programs for multiscale analysis of landscape structure using the GRASS geo-
graphical information system. Landscape Ecology 7(4):291-302.
Beven, K.J., and M.J. Kirby
1979 A physically-based variable contributing area model of basin hydrology. Hydrological
Sciences Bulletin 24(1):43-69.
Bian, L., and S.J. Walsh
1993 Scale dependence of vegetation and topography in a mountainous environment of Mon-
tana. Professional Geographer 45(1):1-11.
Bilsborrow, R.E., T.M. McDevitt, S.A. Kossoudji, and R. Fuller
1987 The impact of origin community characteristics on rural-urban out-migration in a devel-
oping country. Demography 24:191-210.
Bilsborrow, R.E., A.S. Oberai, and G. Standing
1984 Migration Surveys in Low Income Countries: Guidelines for Survey and Questionnaire
Design. London, England: Croom Helm.
Bonan, G.B.
1989 A computer model of the solar radiation, soil moisture, and soil thermal regimes in boreal
forests. Ecological Modelling 45:275-306.
Caplow, T., H.M. Bahr, et al.
1982 Middletown Families: Fifty Years of Change and Continuity. Minneapolis: University of
Minnesota Press.
Clausen, J.A.
1972 The life course of individuals. Pp. 457-514 in Aging and Society. Vol. 3: A Sociology of
Age Stratification, M.W. Riley, M. Johnson, and A. Foner, eds. New York: Sage.
Curran, S.R.
1994 Household resources and opportunities: The distribution of education and migration in
rural Thailand. Ph.D. dissertation, Department of Sociology, University of North Caro-
lina.
Davis, K.
1963 The theory of change and response in modern demographic history. Population Index
29:345-366.
DeCola, L., and N. Montagne
1993 The PYRAMID system for multiscale raster analysis. Computers and Geosciences
19(10): 1393-1404.
OCR for page 142
42
LAND-USE/LAND-COVERANDPOPULATIONDYNAMICS, THAILAND
Donner, W.
1978 The Five Faces of Thailand: An Economic Geography. London, England: Hurst Publish-
ing.
Elder, G.H., Jr.
1985 Perspectives on the life course. Pp. 23-49 in Life Course Dynamics: Trajectories and
Transitions, 1968-1980, G.H. Elder, Jr., ed. Ithaca, N.Y.: Cornell University Press.
Elder, G.H., Jr.
1991 Making the best of life: Perspectives on lives, times, and aging. Generations 15:12-17.
Entwisle, B., R.R. Rindfuss, S.J. Walsh, T.P. Evans, and S.R. Curran
1997a Geographic information systems, spatial network analysis, and contraceptive choice. De-
mography 34: 171-187.
Entwisle, B., S.J. Walsh, and R.R. Rindfuss
1997b Population Growth and the Extensification of Agriculture in Nang Rong, Thailand. Paper
presented at the annual meetings of the Population Association of America, Washington,
D.C.
Entwisle, B., R.R. Rindfuss, D.K. Guilkey, A. Chamratrithirong, S.R. Curran, and Y. Sawangdee
1996 Community and contraceptive choice in rural Thailand: A case study of Nang Rong.
Demography 33(1):1-11.
Entwisle, B., and W.M. Mason
1985 The multilevel effects of socioeconomic development and family planning programs on
children ever born. American Journal of Sociology 91:616-649.
Findley, S.
1987 Rural Development and Migration: A Study of Family Choices in the Philippines. Boul-
der, Colo.: Westview Press.
Freeman, L.C.
1979 Centrality in social networks: I. Conceptual clarification. Social Networks 1:215-239.
Fukui, H.
1993
Food and Population in a Northeast Thai Village. Monographs of the Center for South-
east Asian Studies, Kyoto University. English-Language Series, No. 19. Honolulu: Uni-
versity of Hawaii Press.
Hutchinson, M.F.
1989 A new procedure for "riding elevation and stream line data with automatic removal of
spurious pits. Journal of Hydrology 106:211-232.
Development of a continent-wide DEM with applications to terrain and climate analysis.
Pp. 393-399 in Environmental Modeling with GIS, M.F. Goodchild, B.O. Parks and L.T.
Steyaert, eds. New York: Oxford University Press.
Institute for Population and Social Research (IPSR)
1984 A Demographic, Socioeconomic, and Health Profile of a Rural Community in Nang Rong.
Bangkok, Thailand: Institute for Population and Social Research, Mahidol University.
Limpinuntana, V., A. Patanothai, T. Charaoenwatana, G. Conway, M. Seetisarn, and K. Rerkasem
1982 An Agroecosystem Analysis of Northeastern Thailand. Khon Khaen, Thailand: Faculty of
Agriculture, Khon Khaen University.
Massey, D.S., and F. Garcia Espana
1987 The social process of international migration. Science 237:733-738.
McGarigal, K., and B.J. Marks
1993 FRAGSTATS: Spatial Pattern Analysis for Quantifying Landscape. Version 1.0. Forest
Science Department, Oregon State University.
Montieth, J.L., and M.H. Unsworth
1990 Principles of Environmental Physics. London, England: Edward Arnold.
Moore, I.D., R.B. Grayson, A.R. Landson
1991 Digital terrain modelling: A review of hydrological, geomorphological, and biological
applications. Hydrological Processes 5:3-30.
OCR for page 143
BARBARA ENTWISLE ET AL.
143
Moran, E. F., E. Brondizio, and P. Mausel
1994 Integrating Amazonian vegetation, land-use, and satellite data. BioScience 44:329-338.
Nellis, M.D., and J.M. Briggs
1989 The effect of spatial scale on Konza landscape classification using textural analysis. Land
scape Ecology 2(2):93-100.
Panayotou, T., and S. Sungsuwan
1989 An Econometric Study of the Causes of Tropical Deforestation: The Case of Northeast
Thailand. Harvard Institute for International Development Discussion Paper No. 284,
Harvard University.
Parnwell, M.J.G.
1988 Rural poverty, development and the environment: The case of North-East Thailand.
Journal of Biogeography 15: 199-208.
Pendleton, R.L.
1963 Thailand: Aspects of Landscape and Life. New York: Dwell, Sloan, and Pearce.
Phantumvanit, D., and K.S. Sathirathai
1988 Thailand: Degradation and development in a resource rich land. Environment 30:11-15,
30-32.
Phillips, J. F., R. Simmons, M.A. Koenig, and J. Chakraborty
1988 Determinants of reproductive change in a traditional society: Evidence from Matlab,
Bangladesh. Studies in Family Planning 19:313-334.
Rindfuss, R.R., S.J. Walsh, and B. Entwisle
1996a Land Use, Competition, and Migration. Paper presented at the Population Association of
America Meeting, New Orleans, La.
Rindfuss, R.R., D.K. Guilkey, B. Entwisle, A. Chamratrithirong, and Y. Sawangdee
1996b The family building life course and contraceptive use: Nang Rong, Thailand. Population
Research and Policy Review 15:341-368.
Rogers, E.M.
1983 Diffusion of Innovation. New York: Free Press.
Sawangdee, Yothin
1995 Multi-Level Analysis of Factors Influencing An Adult Temporary Migrant's Destination: A
Case Study From Nang Rong, Thailand. Unpublished M.A. thesis, University of North
Carolina.
Siamwalla, A., S. Setboonsarng, and D. Patamasiriwat
1990 Agriculture. Pp. 1-117 in The Thai Economy in Transition, P.G. War, ed. New York:
Cambridge University Press.
Tambunlertchai, S.
1990 A Profile of Provincial Industries. Bangkok: Thailand Development Research Institute
Foundation.
Townsend, P.A., and S.J. Walsh
1996 Spatial variability of a wetness model to soil parameter estimation approaches. Earth
Surface Processes and Landforms 21(4):307-326.
Walsh, S.J., T.P. Evans, W.F. Welsh, R. R. Rindfuss, and B. Entwisle
in Population and environmental characteristics associated with village boundaries and
press landuse/landcover patterns in Nang Rong district, Thailand. Proceedings of Pecora 13,
Symposium on Human Interactions with the Environment: Perspectives from Space.
Bethesda, Md.: American Society for Phtogrammmetry and Remote Sensing.
Walsh, S.J., A. Moody, T.R. Allen, and D.G. Brown
1997 Scale dependence of NDVI and its relationship to mountainous terrain. Pp. 27-55 in
Scaling in Remote Sensing and GIS D.A. Quattrochi and M.F. Goodchild, eds., Boca
Raton, Fla: CRC/Lewis Publishers.
OCR for page 144
44
LAND-USE/L~AND-COVERANDPOPULATIONDYNAMICS, THAILAND
Wolock, D.M., and G.J. McCabe, Jr.
1995 Comparison of single and multiple flow direction algorithms for computing topographic
parameters in TOPMODEL. Water Resources Research 31: 1315-1324.
Woodcock, C.E., and A.H. Strahler
1987 The factor of scale in remote sensing. Remote Sensing of Environment 21:311-332.
Representative terms from entire chapter:
remotely sensed
