Further spatial analysis by using landscape metrics underlined the evidence of changes in landscape characteristics with an increase in values o f num ber o f patch[r]
Trang 1V N U Journal of Science, E arth Sciences 28 (2012) 251-263
Spatio-temporal dynamics and evolution o f landscape pattern
in coastal areas o f central region, Vietnam
M Kappas'’*, Nguyen Hoang Khanh Linh'’^
^Dept o f Cartography, GĨS & Rem ote Sensing, Georg-August-University Goettingen,
Goldschmidtstr 5, 37077 Goettingen, Germany
^Faculty o f Land Resources ẵ Agricultural Environment, H ue University o f Agriculture & Forestry,
Ỉ02 Phung Hung, Hue City, Vietnam
Received 05 October 2012;
Revised 26 Octobcr 2012; accepted 02 December 2012
A b stract Studying temporal changes o f land use and land cover from satellite images has been conducted in Vietnam several years However, few studies have been done to consider seriously the changes and landscape fragmentation, especially in coastal region, one o f the ecologically vulnerable regions due to the intensive human activities and urbanization processes Hence, analyzing the changes o f landscape pattern helps revealing the interactions between anthropogenic factors and ứie environment, through which planning actions could be effectively supported The present study aimed to examine these changes in the suưoundings o f Da Nang City, Vietnam from
1979 to 2009 based multi-temporal imagery viz LANDSAT MSS, TM, ETM +, and ASTER satellite images The IR-MAD (iteratively re-weighted M ultivariate Alteration Detection) transformation approach was employed for processing Land cover change maps with six classes
including agricultural land, urban, baưen land, forest, shrub and water body were created by the
supervised classification method based on maximum likelihood algorithm Post-classification comparison was chosen as change detection method for four periods as 1979-1996, 1996-2003, 2003-2009, and 1979-2009 From which key landscape indices were applied by using FRAGSTATS software The results showed that during the whole study period, there was a notable decrease o f forest, shrub, agricultural land and baư en land while urban areas expanded dramatically Further spatial analysis by using landscape metrics underlined the evidence of changes in landscape characteristics with an increase in values o f num ber o f patches and patch density while the value o f m ean patch size decreased during the span o f 30 years which indicated landscapes o f Da Nang city have been becoming more fragmented and more heterogeneous
Keywords: landscape pattern, change detection, coastal region, Vietnam.
expected to continue for the n ext decades
As stated in C om petitive C ities in the A ccording to the U nited N ations, roughly h a lf
G lobal Econom y [1] and State o f the W o rld ’s o f the w o rld ’s population lives in urban areas, Cities 2008/2009: H arm onious C ities [2], and in 2030 it w ill be reached at 60%
E-mail: mkappas@uni-goettingen.de
251
Trang 2252 M Kappas, N.H.K Link / V N U Journal of Science, Earth Sciences 28 (2012) 25Ĩ-263
2030 [3] U rban areas concentrate not only
people but also econom ic density and
productivity [4] This is often the reasons o f
changing in lifestyles, high consum ption o f
energy, fransportation, infrasừiicture, and
production o f w aste, etc [5-12] U rbanization is
believed one o f the m o st prevalent
anthropogenic causes o f the losing arable land,
devastating habitats, and the decline in natural
vegetation cover [13] As a consequence, rural
areas have been converted into urban areas w ith
an unprecedented rate and m aking a noted
effect on the natural functioning o f ecosystem s
[14] Consequently, a profound understanding
o f land use change is very im portant to have a
prop er land use planning and sustainable
developm ent policies [15]
A ccording to M yint and W ang [16], such a
sustainable urban developm ent m ust be
sum m arized from num erous decisions, which
ex ừ acteđ based on huge data sources, viz
physical, biological and social param eters o f
urban areas in the continued specừ um o f spatial
and tem poral dom ains T herefore, to understand
urban land-use and land cover change (LU LC)
and to predict the change o f L U L C in future, it
is im portant to have an effective spatial
dynam ic tool N ow adays, rem ote sensing
technologies have proven its capacity in
providing accurate and tim ely inform ation on
the geographic disừ ibu tion o f land use,
especially for region areas [17] W ith the
support o f G eographical Inform ation System s
(G IS), satellite im ages can be used effectively
for estim ating and analyzing changes and
L U L C trends [18]
D ue to the fact that the rapid LU LC change
o f one certain area is considered as the driving
force o f environm ental and /o r ecological
changes, w hich is continuously transform ing
landscape pattern, thereby a need for
com prehensive assessing and analyzing the change in landscape at broad scales is required
Im portantly, understanding the changes in spatial contribution o f landscape p attern helps revealing the critical im plication o f com plex relationship betw een anthropogenic factors and environm ent [19] T o describe fragm entation and spatial disừ ib utio n, a range o f landscape
m eừ ics w as calculated for each land use/cover class from satellite classification results by
FR A G STA TS [20]
The Earth's coastal zone is know n as home
o f diverse ecosystem s, such as estuaries, sea- grass, coral reefs, lagoons, bays, tidal flats,
e tc It plays a crucial part for socio econom ic developm ent and national security This zone is quite sensitive and vulnerable because o f hum an developm ent activities, especially, the tropical coast As consequcnces, these activities causes loses o f living environm ent o f sea species, degradation o f drinking w ater, changes o f hydrological cycles, depletion o f coastal resources and m any other
im pacts to the global clim ate change Therefore, the m anagem ent o f m arine and coastal zone has particularly received great attention from
m anagers as w ell as scientists all around the world The urgent dem ands should be set as lop national sừ ategic m issions and should be caư ied out w ith scientific fundam entals
A fter the adoption o f the D oimoi (R enovation) p olicy in econom y o f the national assem bly since 1986, D a N ang city has developed in m any aspects In addition, it was separated from Q uang N am Province m 1997 and has officially becom e an adm inistration unit that directly belongs to the governm ent Since then, D a N ang city has asserted as the
im portant position at nation level and the crucial factor o f the key area econom y o f Central region This has caused the incessant
Trang 3M Kappas, N.H.K Link / V N U Journal of Science, Earth Sciences 28 (2012) 251-263 253
land use/cover change in D a N ang for over past
20 years Through exploring the land use map
extracted from satellite data o f different
periods, the aims o f the present study w ere to
detect, quantify and characterize the changes o f
land use/cover and landscape fragm entation in
Da N an g city
2 S tudy area
D a N ang city is located in Central region o f
V iet N atn, betw een the 15°55’ 19” to
16°13’20 ”N and 107°49’ 11” to 108°20’20”E
(F igure 1) It is a long-stretching narrow region
and w ell know n as a dynam ic city o f the Key
E conom ic Zone in central V iet N am The area
consists o f hiils and m ountains in the northw est
and the Eastern Sea in the east The altitude
varies from 400 m eters to 1,524 m eters above
sea level; next to is the upland with low
m ountains and the delta takes 'Á areas in the
southeast; it covers an area o f 1,283.42 square kilom eters, including H oang Sa archipelago district o f 305 square kilom eters
D a N ang city has typical tropical m onsoon clim ate The average annual tem perature is about 26°c, average rainfall is about 2,505 mm per year and average hum idity is 83.4% T here are two main seasons annually: the w et (A ugust-D ecem ber) and the dry (January-July)
In 2009, the total population is about 887,070 and the population density is 906.7 persons per square kilom eters D a N ang city is know n as one o f the m ost densely populated and urbanized area in V ietnam W ith the econom y developm ent and population increasing, the local LULC in Da N ang city has changed seriously
Figure 1 Location o f Da Nang city in Vietnam
Trang 42 5 4 M Kappas, N.H.K Link / V N U journal o f Science, Earth Sciences 28 (2012) 251-263
3 D ata and m ethods
3.1 D ata so u rces and Im age p rep ro cessin g
L A N D S A T and A S T E R satelliteim ages
w ere chosen for this study The follow ing
criteria w ere considered for choosing proper
data: (1) the im ages should be long tim e enough
for detecting th e land use change; (2) study area
should not h av e cloud cover U nfortunately, the
study area is located near coastal D ue to the
influence o f clim ate, there are not m any data
satisfied b o th conditions T he im ages alw ays
have som e th ick cloud cover or haze In
addition, the study area is not entirely contained
w ithin one scene o f L A N D S A T either A STER
T herefore, h av in g acquisition im ages near
anniversary d ates for changing detection as
Jensen m entioned [21] w as unavailable In this
study, three periods o f satellite im ages were
selected to classify study area: LA N D SA T-3
M SS July 24, 1979; LA N D SA T-7 ETM +
M arch 04 and A pril 14, 2003 (dow nload free at
h ttp ://earth ex p lo rer usgs.gov/ and
h ttp ://gloviS.usgs.gov/); and A S T E R A pril 02,
2009 T he details o f data w ere described in
T able 1 F or th is study, the reference data were
also used, included: (1) topographic m ap at
scale o f 1/50.000 conducted in 2001; and (2)
land use m aps at scale o f 1/25.000 conducted in
1997, 2003 an d 2010
B ecause L A N D S A T and A ST E R im agery
w ere collected at level IT and IB respectively,
im ages w ere acquired at different spatial resolution and pro jectio ns T h erefo re, all
im ages w ere first rectifie d to U niversal
T ransverse M ercator (U T M ) coo rdin ate system ,
D atum W G S 84, Zone 48 N o rth for m atching the geographic pro jection o f th e referen ce data
Im ages w ere also co -reg istered to g eth er w ithin
25 well distributed G C P s (ground control points) and polynom ial Is d by m eans o f
O rthoEngine provided b y P C I G eom atica 10.3 software RM S < 0.5 w as receiv ed In addition,
N earest N eighbour resam p lin g w as set for not changing heavily the rad io m etric characteristic
o f image
In this study, the iterativ ely re-w eighted
multivariate alteration detection (IR-
M AD ) fransfonnation w as u sed for autom atic radiom etric n o m a liz a tio n for all im ages by
m eans o f E N V I 4.7 so ftw are; see [22-24J
A ST ER 02/04/2009 w as ch o sen as reference image H ow ever, this im age d o es not cov er all the region o f study area, th erefo re a subset o f
1800 X 1100 pixels w ith 30m spatial resolution including 968.17 square k ilo m eters w as created for all im ages for fu rth er studying This territory w as chosen to ensu re the specific study area w as in the an alysis im age B esides the requirem ent o f the sam e d im en sio n, im ages
m ust have the sam e spectral reso lu tion H ence, the com posite o f stand ard false colours was used for this study: L A N D S A T M SS (754);
L A N D SA T T M /E T M + (4 32); A S T E R (321) geom etric co rrection do n ot require H ow ever,
Table 1 Characteristics o f satellite data used in study area
T ype o f sen so r Spatial resolution (m) B and D ate P a th Row A v erag e cloud coverage (% )
Although the average cloud coverage of LANDSAT-7 ETM+ is very high, there is almost no cloud in study area at that time
Trang 5M Kappas, N.H.K Linh / V N Ư Journnl o f Science, Earth Scienccs 28 (2012) 251-263 255
3.2 L U L C c la ssific a tio n a n d C h a n g e d etectio n
Six land u se/co v e r classes w ere defined for
image classification based on the m odified
A nderson land use/co v er schem e level I [25],
included; (1) w ater, (2) forest, (3) shrub, (4)
agriculture, (5) barren and (6) urban land
A nderson classificatio n schem e w as chosen
because o f the m a jo r land use/cover classes
using im ages w ith differences in spatial
resolution, w h ich are L A N D S A T M SS,
LA N D SA T T M , L A N D S A T ETM + and
ASTER S u p erv ised classification using
m axim um lik elih o o d approach in EN V I 4.7 was
individually ap p lied for each im age o f study
area to classify land use/cover M axim um
likelihood alg o rith m w as p refeư ed because this
rule is con sidered to have accurate results
because it h as m o re accurate results than other
algorithm s [26-28]
B ecause o f v ario u s im age acquisition dates,
training areas for the im ages o f the years 1979,
1996, 2003 and 2009 w ere different during the
classification In addition, the iTaining areas
w ere verified by references data A s the next
step, post-classification com parison change
detection alg o rith m w as selected to detect
changes in L U L C from 1979 to 2009 in study
area in order to m inim ize the problem in
radiom etric calib ratio n o f im agery o f two
different dates F o r com parison o f the
classification results o f two dates, a change detection m atrix w as created based on pixel-by- pixel [21] T hereby, each type o f from -to LU LC change is identified
3.3 L a n d sc a p e fra g m e n ta tio n
For quantifying landscape pattern and landscape fragm entation, FR A G S T A T S w as applied because this spatial statistic prog ram offers a com prehensive choice o f landscape
m etrics This program w as created by decision maker, forest m anager and ecologists therefore
it is appropriate for analyzing landscape fragm entation or describing characteristics o f landscape, com ponents o f those landscapes [29] H ow ever, landscape pattern s w ere com plicated; hencc, to clarify the relationship
o f spatial pattern and process it cannot use single m etric alone [19, 30]
B ased on the scale o f study area (i.e the district level) and its characteristic as w ell, six related landscape m etrics w ere selected: (1) Percentage o f landscape (PLA N D ), (2) N um ber
o f patches (N P), (3) Largest patch index (LPI), (4) M ean patch area (A R E A _M N ), (5) Patch density (PD ), and (6) P roxim ity index (PR O X _M N ) A b rie f description o f those landscape m etrics used in study w as given in Table 2 T hose descriptions could be also found
at u ser’s guide o f FR A G ST A T S™ [31]
Table 2 Landscape pattern m eừics description [29, 31]
PLAND
NP
P ercentage o f landscape-equals the sum o f the areas (m^) o f all patches o f the corresponding patch type, divided by total landscape area (m^), multiplied by 100 to convert to a percentage
N um ber o f patches-equals the number o f patches of the coưesponding patch type (class)
Largest patch index-equals the area (m^) o f the largest patch o f
percent
none
0<PLAND<100
NP>1, no limit
LPI the corresponding patch type divided by total landscape area
(m ), m ultiplied by 100 to convert to a percentage
percent 0<LPI<100
Trang 6256 M Kappas, N.H.K Link / VNU journal of Science, Earth Sciences 28 (2012) 251-263
no limit PD
Patch density equals the num ber o f patches o f the coưesponding patch type divided by total landscape area (m),
num ber per
no limit
PROX_M N
multiplied by 10,000 and 100 (to convert to 100 hectares)
Mean proxim ity equals the sum o f patch area (m^) divided by the nearest edge-to-edge distance squared (m^) between the patch and the focal patch o f all patches of the corresponding patch type whose edges are within a specified distance(m) o f the
focal patch; Average proxim ity index for all patches in a class
hectares
no limit
4 R esults and discussion
4.1 L a n d U se/ C over C hanges
B efore doing any other interpretations,
th em atic LU LC m aps (1979, 1996, 2003 and
2009) w ere assessed their accuracy through four
m easurable m eans o f error m atrix: overall
accuracy, p rod ucer’s accuracy, u se r’s accuracy
and K appa coefficient A total o f 300 sfratified
ran do m pixels w as taken for each LU LC m ap
and then checked w ith reference data
A ccording to the accuracy assessm ent results o f
classified maps, the overall accuracy for
L A N D S A T M SS 1979, L A N D S A T ETM +
2003 and A ST ER 2009 w as 92.15% , 80.33% ,
84.44% and 89.00% respectively; the K appa
C oefficient o f those m aps reached at 0.9021,
0.6921, 0.7534 and 0.8005, respectively The
results showed that LU LC m ap derived from
A S T E R has higher accuracy than the others
T his could be explained by the better spatial,
specừal and radiometoic resolution o f ASTER data
T h e LU LC m aps o f study area w ere
generated for all four years (Figure 2) and
classification area statistics w ere sum m arized in Table 3 T he classified areas w ere m easured by
m ultiplying the n um b er o f pixel w ith spatial resolution o f rem ote data (i.e 30 m eters), in
w hich the pixel nu m ber w as determ ined after applying post-classification analysis And then changes w ere defined based on the difference o f pixel num ber betw een tw o dates Based on Table 3, forest and urban areas w ere the dom inant LU LC classes m spatial distribution pattern A ccordingly, forest area w as counted for about 64.0% , 60.0% , 61.4% and 59.8% o f the total area in 1979, 1996, 2003 and 2009 respectively; m eanw hile urban area w as occupied 6.5% , 8.0% , 12% and 17.9% o f the total area in 1979, 1996, 2003 and 2009 respectively T h e surface w ater body covers
about 2.5Vo, 2.6% , 2.9% and 3.1% o f the total
region study in 1979, 1996, 2003 and 2009, respectively The results also show ed that from
1979 to 2009 LU LC units under shrub, agricultiưe and barren decreased from 10.1% to 9.9% , 12.4% to 7.5% and 4.5% to 1.8%, respectively
Trang 7M Kappas, N.H.K Linh / V N U Journal of Science, Earth Sciences 28 (2072) 257-263 257
Legend
i m m water
H urban
m forest
^ 9 shrub
, _! barren
agncuiture
Figure 2 Land use/cover maps o f Da Nang city area
Table 3 Results o f and use/cover classification for 1979, 1996, 2003 and 2009 images
A rea (ha) (% ) A rea (ha) (% ) A rea (ha) (% ) A rea (ha) (% )
To provide a further com prehensive
calculation in losing and gaining am ong the six
LU LC classes, the from -to change m afrix o f
land use/cover in D a N ang city w ere created in
three intervals, 1979-1996, 1996-2003,
2003-2009 and 1979-2003-2009 (Table 4) In cross tabulation, unchanged pixels w ere located along the m ajor diagonal o f the m afrix w hile conversion values o f classes w ere aư an g ed in descending order As can be seen from the
Trang 8258 M Kappas, N.H.K Link / V N U Journal of Science, Earth Sciences 28 (2012) 251-263
T ables 4, there w ere sm all differences o f area
coverage o f a particu lar class because o f used
different spatial resolutions for calculating
L U LC change from 2003 to 2009 (e.g., forest
coverage in 2009 is 57936.2 hectares in T able 3
and 57935.79 h ectares in T able 4c) It resulted
because o f using different spatial resolutions for
calculating L U L C change from 1979 to 2009
In fact, the 2009 A S T E R im age w as re-sam pled
to a spatial reso lu tion o f 30 m eters
D uring the first period (1979-1996), results
show ed that forest, agriculture, and barren
decreased strongly w hile urban area, shrub and
w ater body increased, notably the raising o f
shrub area T ab le 4(a) indicated that the
expansion o f shrub area w as the m ost dram atic
changes in the region w hereas forest area
decreased, w hich w as the result o f deforestation
m ainly caused by the increasing dem and o f tim ber products U rban area grew up ju st 1476.2 hectares, representing 13.4% o f net increase o f urban area
In 1990, the policy no tim ber exploitation o f
governm ent, w hich could help to continue supplying m aterials for tim bers and paper industry C onsequently, forestry productions
w ere exploited from forest plantation [32], Therefore, in the second period (1996-2003) forest cover extent had been slightly increased
by reforestation program s w ith 1340.01 hectares As can be seen from Table 4b, urban area prom ptly grew up 3838.5 hectares after separating from Q uang N am province and becam e a cenfrally governed city
Table 4 Land use/ land cover ừansform ation mafrices o f study area from 1979 to 2009
(Unit: hectares) 1979
(a) 1979-1996
(b) 1996-2003
Trang 9M Knppns, N.H.K Link / V N U joiirm l o f Science, Earth Sciences 28 (20Ĩ2) 251-263 259
Agriculture
Baưen
Urban
Forest
Shrub
Water
2003 Total
1858.68 86.76 3188.7 1036.17 1833.21 108.27 8118.09 Change 2003-2009 -823.41
177.66 121.86 1188.27 231.93 656.01 105.48 2487.15 -778.23
711 148.14 9025.29 414.99 808.56 460.89 11629.98 5668.56
2880.63 860.58 739.35 52503.66 2364.21 46.71 59466.78 -1530.99
1645.38 464.04 2673.81 3556.26 3851.46 138.33 12335.94 -2760.12
15.03 24.93 458.55 95.85 51.3 2104.29 2778.84 224.73
7294.68 1708.92 17298.54 57935,79 9575,82 3003.57
(c) 2003-2009 2009
Agriculture
B aưen
Urban
Forest
Shrub
Water
1979 Total
1979
1779.21 353.07 2975.04 3787,38 2895.48 257.85 12048,03 Change 1979-2009 -4753.35
991.26 78.3 1933.56 227.52 747.45
334 08 4312.17 -2603.25
110.79 91.8 5096.7 221.58 430.47 182.97 6314.85 10983.69
2394.99 933.93 3898.26 51584.22 2834.19 326.43 61972.02 -4036.23
1950.3 240.48 2789.37 1928.79 2589.48 286.74 9785.16 -209.34
61.83 8,73 581.04 89.37 67.68 1575.9 2384.55 619.02
7294.68 1708.92 17298.54 57935.79 9575.82 3003.57
(d) 1979-2009
W hich w as 35% o f n et increase o f urban
area W hereas from 1996 to 2003, w ithin ju st
seven years, agriculture area reduced 2298.6
hectares, thus representing o f 19.1%
In the third period, from 2003 to 2009,
forest area decreased once again (1.6% o f total
area in D a N ang City) due to the rapid
urbanization A griculture area reduced 823.41
hectares w ithin six years, w hich represented o f
6.8% C onversely, urban area incessantly
increased and gained 5668.5 hectares, w hich
contributed 51.6% to net increase o f urban area,
experienced a rem arkable change o f urban area
w ith a rapid scale
A ccording to Table 4d, for 30 years,
although forest extent fluctuated variously in
d ifferent periods, this area decreased in general
R esults show ed that the forest area lost 10387.8
hectares o f Its 1979 area to other classes, in
w hich 37.5% (3898.26 hectares) converted to urban, 27.3% (2834.19 hectares) to shrub and 23.1% (2394.99 hectares) to agriculture From
1979 to 2009, agriculture area strongly decreased 4753.35 hectares (Table 5d), representing a net decrease o f 39.5% , the change o f agriculture area altered considerably
in different periods o f tim e T h e loss o f agriculture from 1979 to 2009 w as m ainly caused by the encroachm ent o f urban and forestation A ccording to Table 5d, agriculture area lost 2975.04 hectares to urban area and 1392.39 hectares to forest, rep resen tin g 60.3% and 29.3% o f total decrease in agriculture land use, respectively B ased on statistic, 10983.69 hectares o f urbanized area in this p eriod w as calculated, w hich w as nearly tw ofold the coverage o f urban area in 1979, thus representing an increase o f 140% (10983.69 hectares) A nalyzing the com ponent o f the
Trang 10260 M Kappas, N,H.K Link / V N U Journal of Science, Earth Sciences 28 (2012) 25Ĩ-263
conversion o f grow th in urban area, 33.5% was
converted from forestry, 26.1% from
agriculture an d 21.5% from shrub This also
resulted because o f the grow th o f econom ic
after applying D oim oi policy As can be seen in
Figure 5 gross dom estic product (G D P) o f Da
N ang city increased steadily from 1990 to 2009,
w ith an annual grow th o f G D P o f 10.3%
(h igher than n a tio n ’s annual grow th o f GDP
7.2% ) In addition, the increase o f population in
D a N ang city could be seen as another reason
for urban expansion, in w hich population
increase from 679.7 thousand in 1997 to 890.5 thousand in 2009, representing an increase o f 31% Based on Figure 3, the difference o f spatial distribution o f urban area could be clearly observed by the years In 1979, the urban area dispersedly located along the costal line By 2003, this area w as expanded more concentrated along coastal zone and moved tow ard Sontra peninsula From 2003 to 2009, the urban expansion changed the direction from costal tow ard in land
^ rỹ> rỹ rỹ rỹ fỹì rỹ> rỹì rỹ
Y ears
Figure 5 Gross domestic product and its growth in Da Nang city from 1990-2009
4.2 F ragm entation A nalyses
From L U L C m aps in 1979 and 2009, three
m ost changing classes (agriculture, urban and
forest) were chosen to com pute spatial
landscape m atrices at class level b y m eans o f
FR A G ST A T S softw are (Table 5) In D a N ang
city, forcsừy area presented as the dom inance
class o f landscape T his could be identified by the largest patch index (LPI), a specific m easure used for observing the dom inance o f a land cover type C om pared to agriculture and urban area, the largest patch index (LPI) o f forest area
is highest at rate o f 29.4% and 29.5% in 1979 and 2009, respectively T he statistic o f forestry show ed that the percentage o f landscape