THE AMAZONIA INFORMATION SYSTEM
Diogenes S. Alves (I), Luiz Gylvan Meira Filho, Julio C. L. d'Alge,
Eliana K. Mello, Jos~ C. Moreira, Jose S. de Medeiros
National Institute for Space Research
C.P. 515 - CEP 12201 -Sao Jose dos Campos - SP - Brazil
(1) e-mail: dalves@dpi.inpe.br
ISPRS Commission VI
ABSTRACT
A research initiative for the assessment of the impact of human-induced
changes in Brazilian Amazon rain forests is presented. Data from satellite
imagery, topographic, vegetation and other maps are combined inside
AMAZONIA, a geographic information system covering the entire region. As a
result, the extent and the rate of deforestation are being more accurately
estimated and the refinement of models to assess the impact of deforestation
in different physical processes was made possible by the availabitily of
georeferenced data.
Key-Words: Geographic Information Systems, Deforestation, Amazon
BACKGROUND
The Brazilian Amazon is often refered to
as Legal Amazonia
the part of the
territory formed by the states, of Acre,
Amapa,
Amazonas,
Mato
Grosso,
Para,
Rondonia, Roraima, Tocantins and the part
of Maranhao west to 44 degrees (figure
1). A region of more than 5 million
square
kilometers
including
forests,
savannas and some areas of intensive
human activity, the Legal Amazonia has
been object of different studies.
In 1980, results of the first wall-towall assessment of deforestation based on
MSS data were
published by INPE and
former' Brazilian Institute for Forest
Development (IBDF) (Tardin ,et al, 1980).
The extent and the rate of deforestation
in Brazilian Amazonia has interested many
researchers in the last decades (see for
exemple, Watson et aI, 1992). Several
estimates of deforestation were produced,
most of them based on satellite imagery,
a unique source of data due to the
difficulties in access and the dimensions
of the region.
ASSESSMENT OF DEFORESTATION
Starting
in
1988,
INPE
has
been
developing a series of studies using
Landsat TM imagery, which allowed to
estimate the rate of deforestation over
the last decade, as shown in table 1.
The use of MSS and TM imagery for the
assessment of deforestation has allowed
to measure both the extent and the rate
of deforestation for comparatively short
time intervals, in some cases, on an
yearly basis. particularly I TM 30-meter
resolution and the high geometric quality
of
its
imagery
has
proved
to
be
appropriate
for
that
purpose.
Also I
Landsat imagery recorded at INPE since
1973 gives complete yearly coverages of
the entire area.
The
size
of
the
region
makes
the
assessment of deforestation a costly,
laborious
task.
For
example,
a
comprehensive survey based on Landsat
Thematic
Mapper
(TM)
imagery
would
require at least 229 scenes to cover the
region for one single year, as shown in
figure 2.
It could be noted that ,data from other
orbital sensors has been used in some
studies.
However,
aVp,ilable data
are
either scarce, as in the case of the
Satelli te Pour l ' Observation de
(SPOT),
or do not have an
Notwithstanding,
comprehensive
surv~ys
are unvaluable for the assessment of
deforestation on an yearly basis as well
as for the study of the impact' of human
action
on
the
different
ecosystems.
Besides,
the
scarceness
of
socioeconomical data difficults selection of
representative areas for the adoption of
non-comprehensive, sampling approaches.
la
Terre
adequate
resolution for the identification of many
small areas deforested each year, like
1.1-km NOAA Advanced Very High Resolution,
Radiometer (AVHRR)"
Combinations of Thematic Mapper bands 3
(red), 4 (near infrared) and 5 (short
wave infrared). are used to identify
deforested areas. Also, this combination
allows to distinguish and map areas of
forest, non-forest (typically savannas)
and major water bodies.
The availability of Landsat Multispectral
Scanner (MSS) and TM imagery at INPE made
it possible to perform comprehensive
surveys of deforestation in the Amazon.
259
TABLE 1
Deforestation rates in Brazilian Amazon
Rate of deforestation (km 2 )
Period (years)
1978 - 1989
21,500
1988(*)-1989
18,800
1989 - 1990
13,800
1990 - 1991
11,100
(*) - included some 1987 imagery
INPE/IBDF team (Tardin et all, 1980).
VEGE and ZOPOT datasets are the result of
The scale adopted for analysis of the
imagery and mapping is 1:250,000. The 229
TM scenes (figure 2) that cover the area
are coverted into the 334 1:250,000 maps
of the region (figure 1). These maps are
the basic units of the geographically
referenced data base.
digitization of existing maps. Finally f
was created by importing available
digital data from IBGE,
the
Federal
Government mapping agency.
MUNI
It is expected to extend the piimary
dataset to incorporate more information
necessary for a better understanding of
the
impacts
of
deforestation.
Also,
derived datasets can be produced from the
primary ones in data analysis procedures.
BUILDING
A GEOGRAPHICALLY REFERENCED DATABASE
Deforestation
may
have
different
consequences
and
impact depending on
local
conditions
such as
ecosystems I
soils,
vegetation
types
or
climate.
Because of this I the first question to
ask after how much forest have been
cleared each year is where are the
cleared areas.
The
technology
of
DATA ANALYSIS
Considering the volume of data already
gathered inside AMAZONIA, there are at
least three major problems that can be
investigated by analysis of the datasets:
Geographical
* the
provides means'
to investigate where deforestation is.
GIS
are tools dedicated to storing "
analyzing and visualizing geographically
referenced data.
One of their basic
functions is to combine different types
of data and find relationships among
them, such as how much deforestation is
occuring in different types of vegetation
or soils, how much of it is close to
rivers and water bodies, etc.
Information
Systems
estimation of the rate of
deforestation
how
much
* the
assessment
of
deforestation has been occuring
inside the different types of
vegetation
(GIS)
* the
assessment
by municipios
of
deforestation
To have data on deforestation in a
georeferenced
form,
INPE
started
to
develop the AMAZONIA Information System,
a GIS that integrates deforestation areas
and several other data for the study of
the changes in Amazon rain forests and
its impacts.
Each of these problems has to be solved
considering its specificity.
Different
methods of data analysis and, also, GIS
techniques have to be used in each case.
Several aspects of analyzing data within
the spatial context are not sufficiently
explored and may be challenging for
research on statistics applied to the
problems of environment and spatial data.
AMAZONIA is based on INPE developed SGI,
Estimation of the rate of deforestation
a geographical information system that
handles
vector
(e.g.
administrative
boundaries) and raster (such as satellite
imagery) data and provides functions for
data acquisition, storage, analysis and
presentation (de Souza et all, 1990).
The
rate
of
deforestation
can
be
estimated
by
determining
how
deforestation
changes
over
time.
Theoretically, having a complete coverage
of images for a series of years, the rate
of deforestation between any pair of
years
could
be
calculated
as
the
difference between the total deforested
area divided by the time interval.
At
present,
five
primary
datasets
comprehend AMAZONIA as shown in table 2.
The TM dataset is formed mainly by data
extracted from TM imagery. After analysis
of the images, all identified features
(deforested areas,
forest,
non-forest,
major water bodies) are digitized to
create maps at the 1: 250,000 scale. MSS
assembles maps elaborated in 1980 by the
In the practice, however, two problems
with data (image) sampling have to be
dealt with:
260
TABLE 2
AMAZONIA data sets
TM
1:250,000 scale
UTM projection
Contents:
Deforested areas from TM 1984-1991 imagery
Forest domains from TM imagery and vegetation maps
Water bodies from TM imagery and existing maps
State boundaries from existing maps
Clouds from TM imagery
MSS
1:500,000 scale
Lambert projection
Contents:
Deforested areas from MSS 1975-1978 imagery
VEGE
1:1,000,000 scale
Lambert projection
Contents:
Vegetation types from RADAM maps
ZOPOT
1:2,500,000 scale
polyconic projection
Contents:
Vegetation types from IBGE/SUDAM map
MUNI
1;2,500,000 scale
polyconic projection
Contents:
Municipal boundaries from IBGE digital map
Having different dates for the images
requires
an
understanding
of
how
deforestation behaviors over time and the
estimation of the amount of deforestation
accumulated at a defined moment.
* frequently,
images are partially
covered by clouds that can hide
deforested areas
* to avoid too much clouds the best
image for each year is selected
for analysis, causing the moment
of sampling to be distributed all
over the year for the different
images
Another issue to be investigated is how
to calculate deforestation rates over
periods of time covered by both TM and
MSS imagery (TM and MSS datasets). The
poor geometric quality of MSS imagery and
the differences of scale (TM at 1:250,000
and MSS at 1:500,000) causes a series of
errors that difficult integration of the
two datasets.
Cloud presence has
to be considered
because
area
under
clouds
is
not
surveyed. As a result, the amount of
deforestation under clouds has to be
estimated
under
four
different
circumstances:
Finally,
there are other sources
of
errors that have a direct effect on
estimates of deforestation: errors of
image
analysis,
digitization
and
georeferencing due to the poor quality of
some maps available in the Amazon.
* the area under clouds was never
observed;
* the
area under clouds was not
observed in the initial of a
series of observations;
Deforestation versus Vegetation types
The Amazon forests are formed by various
types of vegetation cover I that shelter
different species and exchange energy and
ele~ents
differently with the
nearby
envlronment. Also, deforestation is often
associated
to
biomass
burning
which
varies . according
to
the
type
of
vegetatlon.
Therefore, deforestation may
have different impacts depending on the
type of vegetation in which it occurs.
* the
area
under
clouds
was
observed
in
past
and
recent
surveys but there are clouds in
the middle of the series;
* the area under clouds was
not
observed in the last of a series
of observations.
261
'The relationship between vegetation types
and the areas of deforestation could be
assessed by superposing maps with each
kind
of
information.
This
can
be
accomplished by using the classical GIS
functions available in AMAZONIA.
Some of the aspects that have to be
considered in this kind of analysis are
the
difference
in
scales
(TM
at
1: 250,000, MSS at 1: 500 ,000 and VEGE at
1:1,000,000), and imprecisions in the
boundaries of RADAM vegetation maps.
Also, in some cases it may be valuable to
model the distribution of some pointcollected data over a region of study.
Data on biomass may be one interesting
example of this. Due to the scarcity of
data this is one challenging problem
related to this kind of analysis.
Deforestation by municipios
The amount of deforestation inside each
municipio may be important for different
socio-economical studies. Municipios are
administrative units for which several
socio-economical data are collected, such
as population and crop and harvested
areas.
Considering the datasets available in
some aspects that have to be
considered in this type of data analysis
are
the very
reduced
scale of the
available
municipios
boundary
map
(1:2,500,000), the 5-year interval for
some socio-economical data and the fact
that socio-economical data are collected
for the entire municipio and not only for
the areas of forest as are collected
deforestation data.
AMAZONIA,
REFERENCES
de Souza, R.C.M., Alves, D.S., Camara
Neto, G., 1990 "Development of Geographic
Information and Image processing Systems
at
INPE"
(original
in
Portuguese),
Proceedings,
Brazilian
Symposium
on
Geoprocessing, Sao Paulo, May 1990
IBGE,
1990
"Projeto
Zoneamento
das
Potencialidades dos Recursos Naturais da
Amazonia Legal, Anexo 3: Vegeta9ao", Rio
de Janeiro, 1990
IBGE, various years,
project reports
RADAM/RADAMBRASIL
et
all,
1980
"Projeto
Tardin,
A.T.
Desmatamento" , Report, INPE-1649-RPE/103,
January 1980
watson,
R.T.;
Meira
Filho,
L.G.;
Sanzuela, E.;
Janetos, A.
"Greenhouse
Gases: Sources and Sinks", Al - Climate
Change, Pre-pubnlished Copy, 1992
262
GS-
;.......
2 C;
T-A
III
€..
I~
'4
P.TJ" V III
Il"
\
,
I)
10
1(11
III
17
~
II\' ~ ~
22
au
~-
Ie
III
..II
47
41
~,
110
'1
ez
II!
34',
III-A
III
,~
42
1'1
.
4 ..
H
l1li
III
87
8~
,H...'
TO)
TI
n
'1'3
74
715
711
f7
-I
811
110
III
til
III
"
118
lit!
110
..
,
.t!~
110
,,
,ij'\
'7e-_
if'
1/1211
54°
41
'--
\It.
if'
4°
II-A
~2S
I~
7.
10
iii I
17
III
1111
100
II.
1111
120
121
IH
142
1411
144
14S
141
147
I,.
i4d1V~
P
102
lOll
11
UtI
12,'
112
115
114
118
IU\
117
~
IIUI
1114
1118
1M
IlI7
1M
1111
140
141/
F
IS7
'SI
IS.
1110
1111
112
114
leG
)"
1117
ill.
'S"
ITO
171
Vilill
114
'88
I ••
117
III
...
In
'110
Itl
1111
IItS
'114
IIlI!
'II'
'117
11l1li1
211
112
2UI
114
218
III
ZI7
218
2111
>.z2O
H'
HZ
223
224
225
131
2i....
240
24'
142
~;a
Ml6
He--
~~
"lI1O....
~7
......... -
i~ ~I
l!S'l
72°
I
~
HO-A
!~
....,
Vlll2
-
~-.
ITO
:171
n!!
'I~
2114
~
114
lII5
SIC
~
lIali
••
12°
03'\
6ft>
~
146
1147..
1t41
172
2711
274-
287
IN
IN
: aI'!'
1117
;
K
-11M
SI.
Ii'"
I
.. ~
I.l~~' 'ITli
H6
401
11'/'0
""""402""
60°
l!49
250
2S1
2112
-275-
-276_
J!'.
_!tn
soo
801
1102
I
I
eOIl
I
1111.
III'
1120
821
IIH
~IIU
,~5
IH
11111
140
1141
142
1166
586
1157
IIH
11M
lI'I'l
sn
an
1114
515
M'
11110
j?.
SIT
40lS
...
404
405
..,;
7
I
104
114/
1110
-!lOll
1124
1146
!lG~ ~ ~
576
407
~ ~ ~ \;20
4U
Figure 1 - Map of the Legal Amazonia showing the 9 states and the 334
map sheets that cover the region
~
128
1411
,SO
174
-
12111('"
In
"'..,
In
~Oo
I.........-
263
;~
~!
2155
1144
cP
eo
101
\
_I~!-
ra
IIII-A
I
45°
10'1.
III
111....-
12
~ F/" II;; ~
~ ~U lr/....{ 86
110
I
48°
~-
fll
72°
4°
ZlIG
II '"
~IO
1\'
40
lUI
.~
-r
j
.,..!-
M
~~
u
lIZ
r-..'
1.41
I,.-( c
2~
2\1
~
~8H
1)-
178
.<z.
o
ro
~
o
z
;~----
o
~-}----------------------~=--~~~~~~~~-------------------\
0
~
~
N
o
~
z
Figure 2 - Landsat scenes that cover Legal Amazonia according to Landsat
Thematic Mapper World Reference System (WRSTM)
264
BOLIVIA
I
I
•
F
NF
Figure 3 - MIR-293 map sheet showing areas of forest (F) and no-farest (NF)
265
BOLIVIA
-
DE F
Figure 4 - Deforested areas through 1990 (DEF) in MIR-293 (Rondonia state, close
to border with Bolivia
266