Forest Ecology and Management, 57 ( 1993 ) 1-16

Elsevier Science Publishers B.V., Amsterdam

Rates and patterns of deforestation in the

Philippines: application of geographic
information system analysis

Dawning, S. Liu a'b, Louis R. I v e r s o n a'b and Sandra Brown a

aDepartment of Forestry, University of Illinois, W-503 Turner Hall, 1102 South Goodwin, Urbana,
IL 61801, USA
bIllinois Natural History Survey, 607 East Peabody Drive, Champaign, IL 61820, USA
(Accepted 26 August 1992 )

ABSTRACT

Liu, D.S., Iverson, L.R. and Brown, S., 1993. Rates and patterns of deforestation in the Philippines:
application of geographic information system analysis. For. Ecol. Manage., 57: 1-16.

Land-use maps for 1934 and 1988, and a 1941 road map of the Philippines were digitized into a
geographic information system. These maps were then analyzed to determine the rates of deforesta-
tion and their relationship with factors such as the distance of forests to roads and forest fragmenta-
tion (measured by perimeter-to-area ratio ( P / A ratio) of forest patches) for this country. The Phil-
ippines lost a total of 9.8 million ha of forests from 1934 to [Link] presence of major roads was a
very important factor affecting deforestation. The closer a forest was to roads, the higher the rate of
deforestation. Nearly 78% of the 2.1 million ha of forests within 1.5 km of roads in 1934 was removed
by 1988. By contrast, only 39.5% of forest was lost between 15.0 and 16.5 km from roads. The density
of roads per unit area did not predict deforestation as well as the distance of forest patches to roads.
The P/A ratio of forest patches was also useful for assessing their rate of clearing. The larger the P/A
ratio of a forest patch, the more likely it was to he cleared. Forests with P/A ratios greater than 65 m
ha-~ in 1934 had all disappeared by 1988. Forests with large P / A ratios were also characterized by
small area and the presence of adjacent agricultural lands in 1934. They were readily cleared and
never reforested.

INTRODUCTION

Because of the ecological importance of tropical forests, their rapid destruc-

tion has raised increasing concern in the global community. Many studies
have investigated rates of forest clearing and their respective causes. These
studies were mainly accomplished by analyzing a time-series of satellite im-
agery (e.g. Malingreau and Tucker, 1987; Nelson et al., 1987; Stone et al.,

Correspondence to." D.S. Liu, Department of Forestry, University of Illinois, W-503 Turner Hall,
1102 South Goodwin, Urbana, IL 61801, USA.

© 1993 Elsevier Science Publishers B.V. All rights reserved 0378-1127/93/$06.00

2 D.S. LIU ET AL.

1991 ) or a time-series of land-use maps (e.g. Sader and Joyce, 1988; Harri-
son, 1991; Iverson, 1991; Brown et al., 1992). Several of these studies have
shown that roads have been a very significant factor influencing deforesta-
tion. Road building increased the likelihood of deforestation in Costa Rica
(Sader and Joyce, 1988) and Brazil (Stone et al., 1991 ). It is commonly be-
lieved that the closer a forest is to roads, the more likely it is that it will be
cleared. However, no study has reported on the relationship between defores-
tation rate and where it occurs in relation to road networks.
Other studies have also revealed that the perimeter-to-area ( P / A ) ratio of
forest patches was often correlated with the degradation of forest. For exam-
ple, Brown et al. (1992) found that an increase in the P / A ratio of forest
patches between 1972 and 1982 in Peninsular Malaysia was associated with
a decrease in biomass density (biomass per unit area) of the forests. Yet, no
evidence has been reported on how deforestation correlates with the P/A ratio.
The Philippines is one of the countries undergoing severe forest loss (Food
and Agriculture Organization (FAO) 1981 ), but the mechanisms of loss are
generally unknown. The objectives of this research were to use geographic
information system (GIS) techniques to understand the factors leading to
deforestation in the Philippines. Specific questions that were addressed are:
( 1 ) how fast have the Philippine forests been cleared? (2) how is the distance
of forests from roads correlated with the rate of forest clearing? (3) how are
the P / A ratios of forest patches correlated with forest clearing? Understand-
ing how these factors influence tropical deforestation is crucial for a more
accurate assessment of the problem and for guiding decision-makers on land-
use planning.

THE PHILLIPINES AND ITS FORESTS

The Philippines, located between the equator and the Tropic of Cancer in
Southeast Asia, is one of the largest archipelagoes in the world. With approx-
imately 7100 islands, the land area of this country is 299 400 km 2 (FAO,
1981; Baconguis et al., 1990 ). It comprises three major island groups: Luzon,
Visayas and Mindanao. The main types of forest vegetation of the Philippine
islands are dipterocarp, mangrove, pine and mossy (or cloud forests). Most
of the commercially valuable species are members of the Dipterocarpaceae
(Ooi, 1987; Garrity et al., 1992).
Disappearance and degradation of forests are among the most significant
issues of natural resources management in the Philippines. The country was
90% forested when the Spaniards conquered the islands in the middle of the
sixteenth century (Westoby, 1989). By 1900, forest cover was reduced to 70%
(Garrity et al., 1992 ). The establishment of plantations of export crops led to
a tremendous loss of forests in the nineteenth century. During World War II,
forest logging decreased because of the decline of the timber industry. On the
GEOGRAPHIC INFORMATION SYSTEM ANALYSIS AND DEFORESTATION IN THE PHILIPPINES 3

other hand, forest clearing increased because a large number of Filipinos took
refuge in the forests and practiced farming therein. The overall effect of these
factors on forest cover is unknown (Garrity et al., 1992 ).
During the post-war years from 1945 to 1953, the Philippines experienced
economic reconstruction with concomitant rapid and widespread forest ex-
ploitation and deforestation (FAO, 1989). Forest harvesting was unre-
stricted at that time; log exportation was unrestricted after the late 1940s. The
only forest management control imposed was that the minimum diameter of
the trees to be cut was 50 cm. When logging was mechanized, even this minor
regulation was ignored. This vast forest destruction continued until the 1950s,
when the selective logging system was initiated. Even at that time, many peo-
ple still thought that forests were inexhaustible no matter how much volume
was removed from them.
In the early 1970s, over 170 000 ha of forest lands was destroyed annually,
but less than 5% of that area was reforested (Bunge, 1984). In the late 1970s,
the emphasis began to shift from timber harvesting and utilization to the pro-
tection, rehabilitation, and development of forest lands. The Philippines' Bu-
reau of Forestry Development of the Ministry of Natural Resources pre-
scribed an annual allowable cut for each logging concession. As a result, from
1975 to 198 l, log production steadily decreased according to official statis-
tics, but illegal logging was widespread (Kummer, 1992 ). In 1982, the Bureau
further reduced the annual allowable cut for most timber concessions and log
export was also restricted. However, illicit logging still occurs (Abate, 1992 ).
The main causes of deforestation and degradation of forests in the Philip-
pines have been attributed to shifting cultivation, unorganized encroachment
on forest lands, squatting, migration of landless lowlanders into upland for-
ested areas, and government-sponsored settlement schemes (Ooi, 1987).
However, more recent analysis suggests that a combination of commercial
logging followed by 'frontier' agriculture is the main cause (Kummer, 1992 ).
Much of deforestation is associated with the conversion of forest lands to
agriculture and mangroves to aquaculture lands. For example, 75% of the
mangrove vegetation which existed in 1918 had been converted to fishponds
and other landuses by 1987 (Baconguis et al., 1990). Rapid population
growth, landlessness, poverty and lack of unoccupied arable lowland created
pressure for the immigration from lowland to upland areas (Acosta, 1989;
World Bank, 1989). Now, about one-third of the Filipinos are upland dwell-
ers, half of whom are farming on former forest lands.

METHODS

D a t a sources

Two land-use maps were acquired for determining the rate of forest clear-
ing. The first map was compiled at a scale of 1 : 1 600 000 by the Philippine
4 D.S. LIU ET AL.

Bureau of Forestry, Department of Agriculture and Commerce, based on 1934

survey data (Fischer, 1934). No further written records on this map were
available. The second map was compiled by the Swedish Space Corporation
(SSC) and was based on the interpretation of 187 images from Satellite Pour
l'Observation de la Terre (SPOT) recorded during 1987-1988, and on ground
reference data (SSC, 1988 ). Most of the images used for this map were from
1987, and the ground reference data were acquired in the 1987 dry season by
aerial reconnaissance and ground surveys. The map scale is 1:2 000 000. An-
other current forest map is available (German-Philippine Project), but we
decided to use the Swedish one because it was based on satellite imagery col-
lected over about a 1-year period versus 6 years for the German-Philippine
product.
The 1934 land-use map identified seven types of land, three of which were
forests: commercial forest; non-commercial forest; pine forest. The 1988 land-
use map had four groups of land cover: forest; extensive land use; intensive
land use; non-vegetated lands and other areas. The forest group consisted of
five classes: closed dipterocarp (mature trees covering > 50%); open dipter-
ocarp (mature trees covering < 50%);mangrove; pine; mossy (SSC, 1988).
Also shown on the 1988 land-use map were the boundaries of 13 administra-
tive regions.
Finally, a 1941 official road map of the Philippines at a scale of 1 : 2 000 000
was obtained. This map is the closest one available in time to the 1934 land-
use map. In addition to railroads, this map shows four types of roads: first
class; second class; third class; trail (Bureau of Public Works, 1941 ). It was
produced using road diagrams and maps submitted by the district engineers.
Data on trails were derived from various sources.

Building the GIS data base

The two land-use maps, the 1941 road map, and the boundaries of the 13
administrative regions shown on the 1988 land-use map were manually digi-
tized using a vector-based GIS software package. They were then transformed
to the Albers conic equal-area projection coordinate system and indepen-
dently stored in digital format in the computer.
The next step was to unify the classification systems used on the two maps.
The three forest types on the 1934 map and the five forest types on the 1988
map were treated as one class, and all other classes were grouped as one. This
aggregation left two broad classes, forest and non-forest, on each of the land-
use maps. This grouping enabled us to use these maps for determining the
temporal trends of deforestation by avoiding the disagreement of different
classification schemes.
Four types of transportation corridors (first class road, second class road,
third class road and railroad) on the 1941 road map were also digitized. The
GEOGRAPHIC INFORMATION SYSTEM ANALYSIS AND DEFORESTATION 1N THE PHILIPPINES 5

trials were not digitized because their impact on deforestation was not com-
parable with that of the other four types and the source of information for
mapping trials was not very reliable (Bureau of Public Works, 1941 ).

Deforestation and roads

A buffer is an area around a road within a certain distance. In this study,

17 buffers were created at 1.5-km intervals around the 1941 road network.
The largest buffer, which included the land area within 25.5 km of roads,
covered almost all of the Philippine land area except remote islands. These
buffers were then overlaid on the 1934 and 1988 forest/non-forest maps, re-
spectively. Overlaying is a technique used to combine two layers of coverage
into a single coverage containing the information from both layers.
The concept of a buffer ring was developed. A buffer ring is the area re-
maining after subtracting a smaller buffer from a larger buffer. The buffer
rings of roads can be used to study the pure effects of roads on deforestation.
For example, the buffer ring created by subtracting the 1.5-km buffer from
the 3.0-km buffer can show the impact of roads on forests located 1.5-3.0 km
from roads. On the other hand, the 1.5- or 3.0-km buffers consider the entire
area from the road to the buffer distance and only show the cumulative im-
pact of roads on deforestation.
Seventeen buffer rings were created this way. The forest areas for 1934 and
1988, as well as the changes and percentage changes in forest area between
1934 and 1988, were calculated for each buffer ring. Least-squares regression
analysis of percentage of forest clearing versus the distance of buffer rings to
roads (the distance from road to the central line of a buffer ring) was used to
test the hypothesis that the closer a forest is to a road, the greater the rate of
deforestation.
Next, the boundaries of the 13 administrative regions, in digital format,
were overlaid on the road map and land-use maps of 1934 and 1988. With
the overlaid maps, road densities (length of road per unit area of land), the
forest area in 1934, and the forest area in 1988 were calculated for each ad-
ministrative region. These data were then used to analyze the relationship
between deforestation and road density by administrative region.

Deforestation and perimeter-to-area ratio

A polygon stored in a vector GIS format has two primary geometric mea-
surements, perimeter and area. We hypothesized that for a forest patch of a
given size, the longer the perimeter (that is the more edge exposed to hu-
mans) the more likely it is to be cleared. The perimeter-to-area ratio ( P / A
ratio) is therefore used as a spatial measure expressing how much edge per
unit area is exposed to humans. However, P / A ratios are not fixed, and they
6 D.S. LIU ET AL.

TABLE 1

Area loss for different geometric shapes with different areas but the same perimeter-to-area ratio ( P /
A = 1.585)

Encroaching Circle (area = 5.00) Equilateral triangle (area = 8.27 )

distance
Absolute Relative area Absolute Relative area
area loss loss (%) area loss loss (%)

0.1 0.76 15.22 1.25 15.22

0.2 1.45 29.19 2.41 29.19
0.3 2.09 41.90 3.46 41.90
0.4 2.66 53.36 4.41 53.36
0.5 3.17 63.55 5.25 63.55
0.6 3.62 72.50 5.99 72.50

vary with both shape and area. Two different conditions can result in high P /
A ratios: the patch may have an extremely irregular boundary, regardless of
size, or the patch may simply have a very small area, regardless of shape.
Nonetheless, we can show through hypothetical examples that the P / A ratio
is still a valid measure to describe patterns of area loss. With a simple com-
puter program, we can simulate and calculate the area loss in a polygon as its
size is gradually, but uniformly, reduced (encroaching distance). Examples
of this simulation for two different patch shapes (a circle and an equilateral
triangle) are shown in Table 1. These polygons were constructed to have the
same P / A ratio, but they have different areas in order to have the same P / A
ratio. For any given encroaching distance, the absolute area losses are differ-
ent, but the relative area losses are exactly the same for both shapes. This rule
was tested for several size classes and P / A ratios, and it held true in each case.
These analyses show that P / A ratio can be used as a valid measure of relative
area loss of forests regardless of the area's shape and size.
Perimeter-to-area ratios were calculated for the forest patches shown on the
1934 land-use map. The 1934 map was then overlaid with the 1988 land-use
map. On the resultant map, we determined which forests in 1934 had been
cleared by 1988 and which remained as forest. With the P / A ratios of forests
in 1934 and the losses of forests between 1934 and 1988, the relationship
between the P / A ratio and the relative forest clearing was assessed.

RESULTS A N D D I S C U S S I O N

General rate of deforestation

The total land area calculated from the digitized map was 29.78 million ha.
This value is 0.5% less than the officially reported land area of the Philippines
because some of the remote, tiny islands in the 1934 and 1988 land-use maps
G E O G R A P H I C I N F O R M A T I O N SYSTEM ANALYSIS AND DEFORESTATION IN THE PHILIPPINES 7

were not digitized. In 1934, the total area of all types of forests was calculated
to be 17.07 million ha, or 57% of the total land area. In 1988, the calculated
area for all types of forests was 7.30 million ha, or 24% of the total land area.
Therefore, 9.77 million ha of forest was cleared from 1934 to 1988, with an
average annual deforestation rate of 180 959 ha. More than half (57.2%) of
the forests in 1934 had disappeared by 1988 (Fig. 1 ). Annual deforestation
rates have not been constant over this time-period. For example, prior to 1969,
deforestation was estimated to be < 2% per year compared to 2.5% per year
during the 1970s to 1980s; annual rates up to 3.5% per year were reported for
1976-80 (Kummer, 1992).

Scale

Luz

;7
oo i0

Fig. 1. Patterns of deforestation in the Philippines during 1934-1988. The forest area in 1934
is the sum of forests present in 1988 and the forests cleared between 1934 and 1988.
8 D.S. LIU ET AL.

Deforestation and buffer distancefrom roads

The percentage of forest loss decreased linearly with increasing distance of
the buffer ring to roads for the areas within 16.5 km of roads (Fig. 2 ). Nearly
78% of the 2.1 million ha of forests within 1.5 km of roads in 1934 was cleared
by 1988, while only 39.5% of the forests between 15 and 16.5 km from roads
was cleared. Over 17% of the total forest clearing occurred within 1.5 km of
roads, and more than 50% was within 6 km (Table 2 ).
For areas beyond 16.5 km, no significant relationship was found and the
relative forest loss fluctuated slightly at about 40% (Fig. 2). Clearly, some-
thing besides the impact of roads was more effective in explaining deforesta-
tion in these areas. Inspection of the overlaid map of land use and road buff-
ers revealed that at the buffer width of 16.5 km, the buffer of the 1941 road
network covered most of the land area of the Philippines ( > 84%, Table 2).
Buffer rings beyond this distance had a significant portion of their area falling
onto the seas. Further inspection of the land-use maps of 1934 and 1988
showed that deforestation also advanced inward from the seas; the closer an
area was to the sea, the more deforestation (Fig. 1 ). These deforestation ac-
tivities were probably associated with the clearing of mangrove and beach
vegetation and their conversion to fishponds.

Deforestation and road density

Although nearness to roads affected forest loss, the relationship between
road density (length of road per unit area) and relative forest loss was not

8O
- \ ¥=[Link]-2.64x if x ~ 1 5 . 7 5 R 2 = 0 . 9 9 2 (P<O.01)

== 70

-~ 6o
.E

_o
50

o
._>m 40 "NO • . ,I .
\

C~

30 . , , , ,
5 10 15 20 25 0

Distance of buffer ring to road ( k i n )

Fig. 2. T h e relationship between the relative forest losses in buffer rings during 1934-1988 a n d
the distancesof bufferrings to roads for the Philippines. The relativeforest loss in a bufferring
is definedas the percentageof forest loss relativeto the forest area in the bufferring in 1934.
GEOGRAPHIC INFORMATIONSYSTEMANALYSISAND DEFORESTATION IN THE PHILIPPINES 9

TABLE 2

Forest area and forest loss in buffer rings during 1934-1988 for the Philippines. (Total land area
covered by all the buffer rings in this table was 93.6% of the country's total land area ).

Distance of Land area 1934 1988 Forest loss Forest loss (%)
buffer ring (kin 2) forest area forest area (km 2)
to road I ( km 2 ) (km 2 ) Relative Relative
(km) to total to buffer
loss 2 ring 3

0.75 68911 21496 4735 16761 17.15 77.97

2.25 42553 17513 4614 12900 13.20 73.65
3.75 30258 15628 4821 10807 11.05 69.14
5.25 22300 13482 4538 8944 9.15 66.34
6.75 21369 14023 5482 8541 8.74 60.90
8.25 16310 11399 5062 6337 6.48 55.58
9.75 13017 9658 4651 5007 5.12 51.84
11.25 11888 9062 4723 4339 4.44 47.88
12.75 10.267 8190 4418 3771 3.85 46./)4
14.25 8755 7058 4122 2936 3.00 41.59
15.75 7209 5944 3595 2350 2.40 39.52
17.25 6153 5155 3116 2038 2.08 39.54
18.75 5377 4577 2781 1795 1.83 39.22
20.25 4426 3823 2292 1530 1.56 40.03
21.75 3882 3393 1989 1404 1.43 41.38
23.25 3393 2010 1810 1200 1.22 39.87
24.75 2608 2355 1371 985 1.00 41.80

~As measured to the central line of the buffer ring.

-'The forest loss within a buffer ring relative to the total forest loss during 1934-1988.
3The forest loss within a buffer ring by 1988 relative to the forest area within the buffer ring in 1934.

significant (R 2 = 0.20, P < 0.12 ). Regions 9, 11 and 12 were among those hav-
ing the lowest road densities, but they had very high percentages of forest loss
(Table 3, Figs. 1 and 3 ). Conversely, Regions 1 and 3 had high road densities
but relatively low percentages of forest loss.
Several factors can help explain this apparent lack of relationship. ( 1 ) The
effect of roads on deforestation depends not only on the density of roads, but
also on their distribution. If roads are too densely distributed in one area but
spread out in another area, the impact of these roads will not be as severe as
the density suggests because some areas are repeatedly covered by road im-
pact while other areas are remote from roads. The density and the distribu-
tion (or pattern) of roads together may explain deforestation better than den-
sity alone. For example, in Regions 9, 11 and 12, roads were evenly spread
across the regions (Fig. 3), making their impact more severe and leading to
high relative forest loss (Table 3, Fig. 1 ). In contrast, Region 3 had a much
higher road density, but most of the roads were concentrated near Manila,
and the relative forest loss in this region was lower than that in Regions 9, 11
and 12. (2) The amount of forest to clear is also important when considering
10 D.S. L1U ET AL.

Scale

/L~

/
/

///~//

l,/~ / /"

!""S //

x -- .-- \ ae

Fig. 3. The Philippineroad networkin 1941 and administrativeregions.

deforestation. For example, in 1934 the forest cover in Region 3 was only
39.7%, compared to 79.3, 86.3 and 78.7% in Regions 9, 11 and 12, respec-
tively (Table 3 ). (3) The four classes of road quality were assumed to have
equal impacts on deforestation in this study. In reality, one would expect dif-
ferences among road types because good-quality roads might facilitate greater
human activity, such as logging (Van der Meer and Ajaloos, 1962 ).

Deforestation and P/A ratio

The relationship between P / A ratios of the forest patches in 1934 and the
relative forest losses during 1934-1988 was significant with a second-order
polynomial regression (Fig. 4). Forest patches with a P / A ratio less than 5 m
GEOGRAPHIC INFORMATION SYSTEM ANALYSIS AND DEFORESTATION IN THE PHILIPPINES 11

TABLE 3

Road density and forest loss in each region during 1934-1988 for the Philippines

Region Region Total Road 1934 1988 Foresl

area road density Joss
( km 2 ) length ( km k m - 2 ) Forest Percent Foresl Percent !%1
(kin) area of land area of land
(km 2 ) area (km 2) area
(%) (%,)

1 22961 2404 0.1047 7096 30.9 3768 16.4 46.9

33914 1417 0.0418 22007 64.9 14428 42.5 34.4
3 17502 1797 0.1027 6942 39.7 3422 19.5 50.7
4 33600 3148 0.0937 14429 42.9 5779 17.2 59.9
5 17769 1645 0.0926 5902 33.2 1156 6.5 80.4
6 20514 2393 0.1167 5676 27.7 1538 7.5 72.9
7 14520 2096 0.1443 4192 28.9 362 2.5 91.3
8 21343 1767 0.0828 12854 60.2 5434 25.5 57.7
9 17469 646 0.0370 13852 79.3 3178 18.2 77.1
10 26028 1420 0.0546 18012 69.2 9747 37.4 45.9
II 32900 1222 0.0372 28383 86.3 11.343 34.5 60.0
12 23543 1259 0.0535 18531 78.7 5.412 23.0 70.8
13 14678 663 0.0451 12868 87.6 7458 50.8 42.0

Total 170744 73025 57.2

110

100

90
A
v
80

70
o
60
o
o 50 y=92.44-45.28/x R2=0.53 (p<O.01)
>= 40

4
0 30
OE
20

10

0 J , i • i • i t
,®. i i

10 20 30 40 50 60 70 80 90

P/A ratio (m/ha)

Fig. 4. The relationship between the relative forest losses during 1934-1988 and the perimeter-
to-area ( P / A ) ratios o f the forest patches in 1934 for the Philippines. The circled point was an
outlier and not used in the regression (see texl).
 12 D.S. LIU ET AL.

h a - ~ in 1934 were associated with relatively low percentages of forest clear-

ing. One outlier to this trend (marked by a circle in Fig. 4) was found and
was not used to develop the regression. This outlier consisted of two small
forest patches whose total area was 100 ha in 1934. A comparison of the 1934
land-use map with the 1988 land-use map showed that these patches fell into
a reforested area in 1988. This pattern was very atypical.
Less than 9% of the total forest patches in 1934 had P / A ratios less than
5.0 m h a - ' (Fig. 5 ), but they contributed over 93% to the total forest area in
1934 (Fig. 6). This indicates that these patches were of considerable size,
with very little edge (Table 4). Generally, large forest patches had small P/A
ratios, and small forest patches had high P / A ratios (Table 4).
There were major changes in the pattern of forest patches in the Philippine
forest landscape from 1934 to 1988. In 1934, the forest patches were charac-
terized by various sizes and shapes, as was reflected by the number and area
of forest patches grouped into P / A ratio classes (Table 4, Figs. 5 and 6). By
1988, the small forest patches (those with large P/A ratios) which existed in
1934 had all been cleared, and only the large blocks of forest remained. All of
the remaining forest patches in 1988 had P/A ratios less than 15.0 m ha-~.
The remaining patches of forest had less edge per unit area and were located
further from roads; they were therefore less likely to experience continued
high rates of deforestation.
During 1934-1988, the number of forest patches with P / A ratio less than
5.0 m h a - ' increased by 240% (Fig. 5 ), while the area of the forest patches
for this P/A ratio range decreased by 61% (Fig. 6, Table 4). The average size

150

¢.)
100 [] 1934

[] 1988
o

"6
50

E
z

. . . . . . . . ._

, , , , , , , , , , , , , , , ,

Range of P/A ratio (m/ha)

Fig. 5. The number of forest patches by perimeter-to-area (P/A) ratio classes in 1934 and 1988
for the Philippines.
GEOGRAPHIC INFORMATION SYSTEM ANALYSISAND DEFORESTATION IN THE PHILIPPINES 13

1000000 "

,ooooo. i
10000"ii [] 1934

'
, . . . . . . . . . . . . . . H M
o , , , , , , i , i i i , , i , ,

Range of P/A ralio (m/hal

Fig. 6. T h e a r e a o f f o r e s t p a t c h e s b y p e r i m e t e r - t o - a r e a ( P / A ) r a t i o c l a s s e s in 1934 a n d 1988 for

t h e P h i l l i p i n e s . N o t e t h a t t h e v e r t i c a l scale is l o g a r i t h m i c .

TABLE 4

Average size of the forest patches in 1934 a n d in 1988 for the P h i l i p p i n e s

Range of P / A ratio ( m ha -~ ) Average patch size ( k m 2 )

1934 1988

0. 0- 5. 0 3711.70 606.09
5. 0- 10.0 66.43 89.10
10.0-15.0 17.92 35.14
15.0-20.0 7.29 -
20.0-25.0 4.13 -
25.0-30.0 2.66 -
30. 0- 35. 0 1.83 -
35. 0- 40. 0 1.31 -
40.0-45.0 1.00 -
45.0-50.0 0.78 -
50. 0- 55. 0 0.67 -
55. 0- 60. 0 0.63 -
60.0-65.0 0.50 -
65.0-70.0 0.43 -
70. 0- 75. 0 0.40 -
75. 0- 80. 0 - -
80. 0- 85. 0 0.36 -

of a forest patch with a P / A ratio of 0.0-5.0 m ha-1 decreased from 3711.7

to 606.1 km 2. This meant that the large forest patches in 1934 were en-
croached upon and/or fragmented and became smaller forest patches by 1988.
14 D.S. LIU ETAL.

Although the P / A ratio is theoretically a good indicator for analyzing pat-

terns of forest change, there are some practical problems. For example, the
theoretical conclusion that a large P / A ratio leads to a high percentage of
forest clearing, as was found in this study, is based on the assumption that
forests are encroached upon from around all the edges of a forest patch. This,
however, is not always true. The distance of a forest to roads is also impor-
tant. Furthermore, the presence of preserved forest reserves leads to excep-
tions to these trends (Bunge, 1984). These factors were not included in this
study.

CONCLUSION

In summary, the accessibility and proximity of forests to human activity

are two mechanisms that explain the depletion of tropical forests in the Phil-
ippines between 1934 and 1988. The distance of a forest to roads was the
most important factor because roads provide the means to readily transport
people into a forested area.
The presence of transportation networks in tropical forest areas has always
led to deforestation, partly because road networks in colonial countries were
designed to extract resources not to trade within the country. Before the con-
struction of large road networks in the tropics, rivers were generally used to
gain access to the interior of forests, as witnessed by the development of pop-
ulation and agricultural production centers along major rivers, e.g. the Ama-
zon (Bates, 1989). As colonization of tropical lands spread, road networks
were constructed to connect towns, production centers, etc., providing greater
access to forests with concomitant increases in deforestation. Today, environ-
mental groups often oppose road construction into tropical areas because of
the increased potential for deforestation to occur in these areas
The relationship between roads and deforestation is further illustrated in
logging operations in tropical forests. Logging is often cited as a cause for
deforestation, but in reality it is often the construction of logging roads and
poor forest protection that is the cause. Logging roads enable landless people
to enter the previously inaccessible areas and clear these logged forests, often
through slash and burn activities.
With respect to the Philippines, one of the largest forest tracts still present
is in Palanan, northeast Luzon (Fig. 1 ), due in part, most likely, to the ab-
sence of roads (Fig. 3, Abate, 1992). Logging interests are pressing the gov-
ernment to build a road into the area, but many believe that this would be a
disaster and lead to increased deforestation (Abate, 1992 ).
Fragmentation of forests is another important factor because small patches
which have large P / A ratios and are surrounded by agricultural or other non-
forest lands are more readily cleared than large blocks of forest. Higher rates
G E O G R A P H I C I N F O R M A T I O N SYSTEM ANALYSIS A N D DEFORESTATION IN T H E PHILIPPINES 15

of forest degradation have been shown to be linearly related to increasing

fragmentation in Peninsular Malaysia (Brown et al., 1992 ).
Results from this study have practical implications for managing and mon-
itoring tropical forest landscapes. Large blocks of forest with small P/A ratios
and limited access by roads stand a good chance of being little disturbed by
humans (e.g. Palanan region in the Philippines). However, restriction of road
construction is often counter to development schemes in tropical countries.
Therefore, the next best thing is to ensure that fragmentation is minimized
and that forests designated for conservation are not fragmented (have small
P/A ratios).
This study also demonstrates how remote sensing can be used to identify
areas susceptible to deforestation. If major roads can be identified and P/A
ratios of forest patches can be calculated by interpreting imagery from satel-
lites or aircraft, then the forest patches close to roads and having high P/A
ratios are most likely to be cleared. Remotely sensed images have advantages
over land-use maps because the data contained in the former are more timely
and they can be more readily entered into a computer for processing and
analysis.

ACKNOWLEDGEMENTS

We would like to thank Thomas Frank, Elizabeth Cook, Steven Sader, Larry
Page, Eric Lambert and David K u m m e r for their reviews of this manuscript,
Lian Cheng for digitizing some of the maps and typing the manuscript, and
Helena Mitasova, Anantha Prased and Sharon Baum for their assistance and
help during this work.
Support for this project was provided by the University of Illinois Depart-
ment of Forestry, the Illinois Natural History Survey and the USDOE through
grant DE-FG02-90ER61801 to the University of Illinois, Sandra Brown, Louis
Iverson and A.E. Lugo, co-principal investigators.

REFERENCES

Abate, T., 1992. Into the northern Philippines rainforest. BioScience, 42:246-251.
Acosta, R.T., 1989. The Philippines forestation program. Canopy Int., 15 ( 3 ) 1-7.
Baconguis, S.R., Cabahug, Jr., D.M. and Alonzo-Pasicolan, S.N., 1990. Identification and in-
ventory of Philippine forested-wetland resource. For. Ecol. Manage., 33/34: 21-44.
Bales, H.E., 1989. The Naturalist on the River Amazon. Penguin Nalure Library, Penguin, New
York, 383 pp.
Brown, S.L., Iverson, L.R. and Lugo, A., 1992. Land use and biomass changes of forests in
Peninsular Malaysia during 1972-1982: use of GIS analysis. In: V.H. Dale (Editor), Effects
of Land Use Change on Atmospheric CO2 Concentrations: Southeast Asia as a Case Study.
Springer, New York, in press.
16 D.S. LIU ET AL.

Bunge, F.M., 1984. Philippines, a Country Study. The American University, Washington, DC,
368 pp.
Bureau of Public Works, 1941. Official Road Map of the Philippines. Department of Public
Works and Communications, Manila.
Fischer, A.F., 1934. Forest Map of the Philippine Islands. Department of Agriculture and Com-
merce, Manila.
Food and Agriculture Organization, 1981. Tropical forest resources assessment project. Forest
resources of tropical Asia. UN 32/6.1301-78-04, Tech. Rep. No. 3, FAO in cooperation with
UN Environment Program, Rome, Italy.
Food and Agriculture Organization, 1989. Review of forest management systems of tropical
Asia: case studies of natural forest management for timber production in India, Malaysia
and the Philippines. FAO, Rome, Italy.
Garrity, D.P., Kummer, D.M. and Guiang, E.S., 1992. The upland ecosystem in the Philip-
pines: alternatives for sustainable farming and forestry. A study commissioned by the Na-
tional Research Council Project on Agricultural Sustainability and the Environment in the
Humid Tropics. National Academy Press, Washington, DC.
Harrison, S., 1991. Population growth, land use and deforestation in Costa Rica, 1950-1984.
Interciencia, 16: 83-93.
Iverson, L.R., 1991. Forest resources of Illinois: What do we have and what are they doing for
us? I11., Nat. His. Surv., Bull., 34: 361-374.
Kummer, D.M., 1992. Deforestation in the post-war Philippines. University of Chicago Press,
Chicago, IL.
Malingreau, J.P. and Tucker, C.J., 1987. The contribution of AVHRR data for measuring and
understanding global processes: large-scale deforestation in the Amazon basin. Proc. 1987
Int. Geoscience and Remote Sensing Symp., Ann Arbor, MI. Environmental Research Insti-
tute of Michigan, Ann Arbor, MI, pp. 443-448.
Nelson, R., Homing, N. and Stone, T.A., 1987. Determining the rate of forest conversion in
Mato Grosso, Brazil, using Landsat MSS and AVHRR data. Int. J. Remote Sensing, 8:1767-
1784.
Ooi, J.B., 1987. Depletion of the forest resources in the Philippines. Field Rep. Ser. No. 18,
Institute of Southeast Asian Studies, Singapore.
Sader, S.A. and Joyce, A.T., 1988. Deforestation rates and trends in Costa Rica, 1940 to 1983.
Biotropica, 20:11-19.
Stone, T.A., Brown, I.F. and Woodwell, G.M., 1991. Estimation, by remote sensing, of defores-
tation in central Rondrnia, Brazil. For. Ecol. Manage., 38: 291-304.
Swedish Space Corporation, 1988. Mapping of the Natural Conditions of the Philippines. Swed-
ish Space Corporation, Solna.
Van der Meer, C. and Ajaloos, B., 1962. Twentieth century settlement of Mindanao. Pap. Mich.
Acad. Sci., Arts Lett., 47: 537-548.
Westoby, J., 1989. Introduction to World Forestry. Basil Blackwell, Oxford, 228 pp.
World Bank, 1989. Philippines: Environment and Natural Resource Management Study. World
Bank, Washington, DC.