The use of satellite images as spacial recolution tools for

Transcripción

REMOTE SENSING IN THE CONSERVATION
OF WETLANDS IN MERCOSUR
Graciela Canziani, Diego Ruiz Moreno,
Manuel David Vargas Russo, Federico Dukatz, Rosana Ferrati
Instituto Multidisciplinario sobre Ecosistemas y Desarrollo Sustentable
Facultad de Ciencias Exactas
Universidad Nacional del Centro de la Provincia de Buenos Aires
Paraje Arroyo Seco – 7000 Tandil – Argentina.
Introduction
The use of satelital images as tools of spatial resolution for mathematical models
applied to landscape dynamics for the conservation is essential in the vast wetlands that
characterise the Del Plata basin, because they allow the possibility of studying the area in a
integral manner.
Here, a methodology applied to the analysis of the Ypoa Wetlands located in
Paraguay, is described as part of the REGWET project.
Considering the framework of the methodology developed in the INCO project “The
Sustainable Management of Wetland Resources in MERCOSUR” for the study of the Ibera
wetlands, the REGWET project was oriented to adapt and extend this methodology to the
neighbouring Paraguayan wetlands. This means the introduction of refinements based on the
different conditions of the particular region. The newly developed models use different
techniques to optimise landscape classification and the calculus of vegetation indexes. Also a
model to estimate tendencies is developed, which improves the capacity of monitoring the
temporal evolution of the vegetation. Finally, coupling the exploration of changes in time and
the results of indexes models, makes possible to detect potential anomalies and landscape and
cover variations.
Materials
The development of the project requires the use of satelital images as main material. In
order to process the images, ad-hoc software for Geographic Information System (GIS) was
required. In particular ERDAS IMAGINE 8.5 and ArcView GIS 3.2a were used here.
Images were taken from SAC-C1, Landsat-7 and Landsat-5 satellites, all of them part
of the AM constellation. They were provided by CONAE2 as part of INCO-DC project “The
Sustainable Management of Wetland Resources in Mercosur”, contract ERB IC18-CT980262 and of the present REGWET project.
SAC-C Mission
Multipurpose satellite SAC-C, is oriented to the study of the terrestrial surface. This
satellite is equipped with a special sensor (Multiespectral Medium Resolution Scanner MMRS) with the capability of analysing crops, forest as well as coast and continental water
1
2
Satélite Argentino de Aplicaciones Científicas.
Comisión Nacional de Actividades Espaciales.
Instituto Multidisciplinario de Ecosistemas y Desarrollo Sustentable
1
monitoring. It was developed by CONAE and has a 175-meter resolution and a scanner with a
band width of 360 km.
MMRS System
The MMRS (Multiespectral Medium Resolution Scanner) is an electronic scanner
specially designed for the study of ecosystems. It detects the radiation from the Earth surface
in five electromagnetic spectrum bands. The period of SAC-C is 16 days and the data
transmission to the terrestrial station is made in real time and whenever it is required.
Table 1. Technical information
Band
Wavelength (nm)
Spectral range
Resolution
(meters)
1
2
3
4
5
480 – 500
540 – 560
630 – 690
795 – 835
1550 – 1700
Green-Blue
Green
Red
Near Infrared – NIR
Short-wave Infrared – SWIR
175
175
175
175
175
Table 2. Other SAC-C radiometric data.
Total Dynamic Range
Capture Dynamic Range
Noise
Radiometric Calibration Precession
Inter-band Calibration
>
>
<
<
<
2000 : 1
256 : 1
2 DN (para todas las ganancias)
7%
5%
Landsat Mission
The Landsat Project is a joint initiative of the U.S. Geological Survey (USGS) and the
National Aeronautics and Space Administration (NASA) to gather Earth resource data. The
purpose of the Landsat program is to provide the world's scientists and application engineers
with a continuing stream of remote sensing data for monitoring and managing the Earth's
resources. Landsat 7 is the latest NASA satellite in a series that has produced an uninterrupted
multispectral record of the Earth's land surface since 1972.
The next table shows the images list from the Landsat 7 and 5 satellites, provided by
CONAE. Each row has a number, the satellite of origin, the acquisition date, some features,
like row and path, sensor and processing level (Full Fast Format NN TM-WGS84). The last
column details the geo-correction performed on the scenes.
Table 3. LANDSAT images used in the project
Nº
Satellite
Acquisition
Date
Origin Specifications
Geocorretion
Specifications
1
2
Landsat 7 ETM
10-Jul
2002
226/78 ETM 4x FFF NN TM-WGS84
Landsat 7 ETM
11-Agu
2002
UTM-21-WGS84
UTM-21-WGS84
3
Landsat 7 ETM
27-Agu
4
Landsat 5 TM
04-Sep
2002
UTM-21-WGS84
2002
5
Landsat 7 ETM
28-Sep
2002
UTM-21-WGS84
6
Landsat 7 ETM
17-Dec
2002
UTM-21-WGS84
7
Landsat 7 ETM
23-Mar
2003
UTM-21-WGS84
8
Landsat 7 ETM
24-Apr
2003
UTM-21-WGS84
2
9
Landsat 7 ETM
10-May
2003
UTM-21-WGS84
10
Landsat 7 ETM
26-May
2003
UTM-21-WGS84
Methodology Description
The next paragraph shows the steps followed
in the project and some considerations.
All the analyses were performed over 10
scenes. Acquisition dates for those scenes range from
2001 to 2003. The first step was the selection of the
satellite images. The next step was the image
rectification and a geo-correction. This procedure
involves the selection of discernible ground control
points (GCP’s) in the images. These points are then
assigned to an appropriate reference information from
a determined geographic referencing system, zone 21
UTM co-ordinates in our case. This reference data is
obtained from topographic sheets. Prior to analysis, a
sub-image of the region was extracted based on
topographic and hydrologic knowledge.
Figure 1. Diagram of the methodology
After obtaining the area of interest, a set of
transformations for enhancing spectral information of
satelital data, such as the Tasseled Cap model and the NDVI3 model. were applied
Further on, an unsupervised classification was performed over the selected area. In a
parallel way, a transformation to detect temporal dynamics on the vegetation index was used.
Satelital Images Acquisition
As mentioned earlier, satelital images were obtained from CONAE as part of the
INCO-DC Project: “The Sustainable Management of Wetland Resources in Mercosur”.
Figure 2. General view of the studied zone.
3
CONAE satelital data constitutes the base data layer for the landcover map. The images
belong to the sector determined by the Path 226 Row 78. This sector includes the region of
study which is bounded by the Paraguay River (West), the Tebicuary River (South) and the
100m elevation contour line (Northeast).
Satelital Images Geo-correction
Once the best scenes were selected relative to the amount of clouds, a process of geocorrection was performed over each scene. With this purpose, a mosaic of seven 1:100,000
DSGM-NIMA4 topographic sheets was used. Beforehand, the sheets were digitized and
geocorrected to UTM5.
Table 4. Topographic maps
Nº Name
1
Villeta
2
Paraguarí
3
San Juan Bautista Alberdi
4
Villa Oliva
5
Villa Florida
6
San Juan Bautista de
Ñeembucú
7
San Juan Bautista
Edition
1-NIMA
1-NIMA
1-NIMA
1 Edición DSGM
1-NIMA
1-NIMA
Serie
H642
H642
H642
H641
H642
H642
Sheet
5369
5469
5268
5368
5468
5367
Country
Paraguay
Paraguay
Paraguay
Paraguay
Paraguay
Paraguay
1-NIMA
H642
5467
Paraguay
Every Landsat 7 satellite image is processed with the following operations:
·
The image is imported considering it as a TM Landsat-7 Fast-L7A EROS. The result is an
image with a Gauss Krüger (WGS’84, datum: WGS-84) geographic projection (Lat/Lon).
·
The result is then rectified to UTM – Spheroid : WGS 84 – Zone : 21 – Datum : WGS 84,
with a first order polynomial model. This task uses a set of cartographic sheets, digitised
and forming a mosaic. These are taken as the reference image, while the satelital image as
an object to be rectified.
·
Finally controls points that are visible on both images are defined. The process was done
using 82 reference points.
Definition of Area of Interest
Once the images are referenced, an Area of Interest is defined according to some
criteria. From here on, only the area contained in this zone of interest will be processed,
allowing a major certainty in results.
3
Normalized Diference Vegetation Index.
Servicio Geográfico Militar (D.S.G.M.) Asunción Paraguay, and the National Imagery and Mapping Agency
(NIMA).
5
UTM – Spheroid : WGS 84 – Zona : 21 – Datum : WGS 84
4
4
Figure 3. AOI definition
In order to define the cutting of the area, the following limits were taken into account:
to the South the limit was set to be the Tebicuary River’s flooding plane;
to the Northwest the contour was defined by the Paraguay River’s flooding plane.
to the Northeast the limit was chosen to be the 100m contour line.
Both valleys were defined through satelital image exploration and topographic sheets.
The criteria are based on hydrological and topographic information. Besides it was taken into
account the satelital images boundaries.
From here on, all the indexes and models where obtained by applying the
methodology over the resulting area only. It is important to note that for the detection of
anomalies, smaller sub-areas were separated and later analysed individually.
·
·
·
Application of Vegetation Indexes
As we have seen, the images have several layers, each one represented by a
electromagnetic frequency intensity matrix.
We apply vegetation indexes to the images to simplify and make specific their
visualisation by reducing the quantity of image layers according different criteria of
composition. For example, the NDVI transform the pixels of an image into a value matrix
(between –1 and 1), then these values can be visualised in a grey scale. These values represent
the vegetation level. This is -1 when there is no vegetation and it increases as the vegetation
increases its density or intensity.
NDVI
The NDVI (Normalised Difference Vegetation Index) is a ratio-based index.
Generally, we referr to this index when we say Vegetation Index. The algorithm quantifies the
green leaf vegetation concentration, based in the wavelength and light intensity reflected by
Earth in visible and infrared bands.
The function used to calculate the value that corresponds to the image pixel is:
NDVI =
NIR - red
NIR + red
where NIR is the Near Infrared [0,795 - 0,835]mm, and Red is band 3, red light [0,630 –
0,690] mm.
Characteristics:
· This index has the advantage of varying between –1 and 1, while the ranges of other
indexes, such as RVI, vary from 0 to infinity.
5
·
·
·
NDVI is a standard index, is simple, and has one of the bests dynamic range and the best
sensitivity to changes in vegetation cover.
It is moderately sensitive to the soil background and to the atmosphere except at low plant
cover.
NDVI is very useful for taking a quick qualitative look at the vegetation cover in an
image, unless the subject is an area with poor plant cover.
Tasselled Cap
The Tasselled Cap Transformation attempts to reduce the amount of data layers
needed for the interpretation or analysis.This enhancement uses mathematical equations to
transform the original n-multispectral bands into a new n-dimensional space.
The Tasselled Cap index relates six TM bands (1-5 and 7) to measures of vegetation
(greenness, GVI index), soil (brightness, SBI), and the interrelationship of soil and canopy
moisture (wetness). Information present in the 6 original bands is compressed into 3 Tasselled
Cap Transform bands: greenness, brightness, and wetness.
Originally the Tasselled Cap transformation is proposed for Landsat images. But since
the SAC-C imaginery doesn’t provide the band 7, the model was modified to solve the index
with only 5 bands by colleagues of the University of Siena (Loiselle and Brachini).
Table 5. Transformation matrix for SAC-C image:
Bands
1
Wetness
2
3
4
5
0.1761
0.3322
0.3396
-0.621*2
GVI – Greenness
-0.2728
-0.2174
-0.5508
0.7221
0.0733*2
SBI-Soil brightness
0.2909
0.2493
0.4806
0.5568
0.4438*2
Table 6. Transformation matrix for Landsat-7 image:
1
2
3
4
Wetness
Bands
0.1446
0.1761
0.3322
0.3396
GVI – Greeness
-0.2728
-0.2174
-0.5508
SBI-Soil brightness
0.2909
0.2493
0.4806
5
6
7
-0.6210
-0.4186
0.7221
0.0733
-0.1648
0.5568
0.4438
0.1706
Figure 4. Samples of TC and NDVI indexes
Supervised and Unsupervised Classifications
Once we have applied the transforms (NDVI and TC) to the images, we obtain new
images that are simplified relative to the original images. As well as the unprocessed images,
the transformed ones can be used to perform a classification on them. Basically, a
6
classification may be performed in two ways: “Supervised” or “Unsupervised”. The result of
a classification is the identity of the points with a specific category (for example: forest,
crops, urban zones, etc.).
In order to perform a Supervised Classification it is necessary to do some terrain
exploration fieldwork. During these campaigns, with help from Global Positioning Systems
(GPSs), the exact geographic positions of some terrain features are defined (for example, a
big wetland surface). After that, these surfaces are located on the satelital images using the
spatial coordinates obtained with the GPS (there must be a correspondence between the
campaign’s date and the image’s date), and by observing the digital values of this region
reflectance in the satelital image different layers. With this values the Spectral Signature is
constructed. The Spectral Signature describes the object or the observed region features in all
bands of the employed satellite. Once that the necessary spectral signatures are available (for
example to define forest, wetlands, crops, roads, plantations, etc), a classification process is
performed on all points or pixels of the satelital image. This process defines to which category
each point belongs (forest, wetland, crop, etc.); by using an approximation algorithm (a
multidimensional distance minimiser between the point values and the spectral signatures).
Unsupervised Classification is performed without having any terrain information to
produce spectral signatures. This classification defines the category by clustering the points in
an image. So, two points with a “spectral distance” close to zero belong to the same category
(because they are practically the same). In order to develop this process, it is necessary to
define the amount of expected classes. The result of this process responds to the aggregation
of the digital data.
Classification Process
Since data gathered in the field
was insufficient, it was not possible to
define enough spectral signatures to
develop a supervised classification in this
region. For that reason, this possibility
had to be discarded.
The
chosen
unsupervised
classification
algorithm
was
the
ISODATA algorithm. Regarding the
parameters, a configuration in 10 classes,
24 iterations, and a convergence
threshold of 0,95 were chosen.
Since the algorithm was applied
to the Tasselled Cap Transformation
output, the Unsupervised Classification
algorithm input is a three band image
(SBI, GVI, Wetness). The resulting
output image is a single band with ten
classes. The first class corresponds to
free water and the last one to bare soil.
Lower-middle classes correspond to
water and vegetation, while upper-middle
ones to soil and vegetation.
Figure 5. Classification over a TC Transformation
7
Temporal Difference Index
One way to detect changes in ecosystems, specifically the vegetation intensity
variation, is using ad-hoc indexes. In our case, we propose a Temporal Difference Index
(TDI). This index is defined by means of a linear regression. This method is performed
simultaneously over the vegetation index data from several images of the same geographic
sector taken in different dates.
The method is applied using data
obtained from any chosen vegetation index
on several images from one sector, in a
temporal succession. These methods may be
applied simultaneously with the unsupervised
classification, because it needs the same
input data, in our case the GVI band of the
Tasselled Cap Transformation, but it can also
be applied on NDVI images. This kind of
transformation produces one new image from
a set of images, where each pixel records the
Figure 6. TDI definition
magnitude of every change through a
coefficient that can be represented visually in a scale of colors.
Each input pixel represents a vegetation index value, for example, for the GVI layer,
this value is between 0 and 255. Each pixel value can be represented with respect to time on a
Cartesian axis system (on the y-axis), time being represented on the x-axis, in this case the day
number, zero being the first day.
For each pixel, this Cartesian representation will have as many points as images that
are being analysed. The slope of the linear regression that fits the data furnished is taken as
the value for the variation in each pixel. The least-squares fitting criteria is that which
minimises the expression:
n
S = å ( y i - ( ax i + b )) 2
1
where a is the slope of the line and b is the y-intercept.
In this case, only the slope value is relevant, because it determines how each pixel
varies in time, increasing (positive slope) or decreasing (negative slope) its brightness. The
slope is obtained when the derivative of S with respect to a and to b are zero. So, an equation
system in two unknowns is solved to find the value of a and b. In our case, we obtain
a=
N å xi yi - å xi å yi
N å x 2i - ( å x i ) 2
The final step is that of generating a new image with the value of each calculated slope
a for each one of the pixels. The resulting image shows darker points where the slope values
are negative, this is where the vegetation index is decreasing , and it has brighter pixels where
the vegetation intensity is increasing.
8
Anomalies and Variations
Once the classification and the TDI index is performed, the necessary information for
detecting areas that exhibit general variations and anomalies, and particularly vegetation
intensity variations, is at hand. In order to detect anomalies, zones where TDI has extreme
values (very positive or very negative) must be identified. Then, the analysis is done by
returning to the original images and focusing on these zones to study these changes. Another
way to detect changes is to look for differences in the specific zones of the classification
images.
Case Studies in Ypoa Wetlands
The identification of areas where anomalous processes take place requires a welltrained eye. Even then, when vast regions are examined, changes in small areas might be
unnoticed. We will show zones of Ypoa Wetlands that exhibit variations of some kind.
First Anomaly
As a result of a TDI transformation, we obtained a new image that shows, in different
shades, the magnitude of the change observed in each pixel. In this example, the green regions
show zones without change, while the shades of red indicate that the vegetation has suffered a
major variation, in this case, an increase in the vegetation.
Figure 7. TDI Results
In Figure 7, the resulting scene shows a large anthropic variation (specific geometry)
in the vegetation, an increase in this case. The unprocessed images are shown at the right,
each one with its date. Although it is easy to see the vegetation variation in these close-ups, it
was detected only by the TDI transformation in the complete satellite image.
9
Second Anomaly
This anomaly was so extensive that could be detected even before processing the
images. On bands 4, 5, and 6 it is possible to observe a shadow at the North side of Ypoa lake.
This anomaly has a reflectance pattern associated with a lack of brightness on the NDVI
index, as well as the TC GVI index. Its area covers approximately 4.980 ha from North to
South.
Zona 2
Zona 1
Figure 8. Second Anomaly view
The first image where this anomaly can be seen was taken on 26/5/2003. When this
image is contrasted with the scene from 10/5/2003, we can see that this shadow is absent.
Observing the images, it is possible discover the dynamics of many zones with the
same behaviour. As first hypothesis, the origin of that anomaly may be fire. Besides, when a
TDI Transformation is applied, this area shows a negative variation on greenness. Even more,
the same pattern repeats itself in other places of Ypoa.
Analysis
Performing the NDVI and TC transforms allows retrieving more information from this
situation. NDVI shows that the zones exhibit a remarkable absence of brightness. It means
less leaf vegetation, similar to that of water bodies. TC Transformation confirms the previous
result, not only through a low response on GVI index, but from low values in Wetness and
SBI.
Figure 9. Image acquired on 26/5/2003.
at left: NDVI, at right: TC transformation.
Since both results are consistent with the consequences of fire, a search on fire foci
detected in the wetland region between these dates was performed in order to define if the
10
observed area was affected by fire. The information was acquired from the Queimadas Group
(INPE/CPTEC) in Brazil. They monitor the zone with NOAA 12, NOAA 16, TERRA 01,
AQUA 01 and GOES 12 satellites. It was easy to confirm that there was a fire focus in the
spot where the anomaly was detected.
Lat
Long
Lat GMS
Long GMS
Date
Time
Satellite
Town
State
Country
Vegetation
Susceptibility
Days without
Precipitation
Risk
-25.7850
-57.4550
S 25 47 6.00
W 57 27 18.00
2003-05-21
19:17:52
NOAA-12
Carapegua
PARAGUARI
Paraguay
Undetermined
High
15
0.678
More Information about the
selected fire focus.
Figure 10. Fire foci in the zone
From 2003-05-10 to 2003-05-26
At this point, it is possible to study in depth this anomaly caused by fire. In order to
execute this operation, the Normalised Burn Ratio, NBR, is used. To perform the algorithm
that requires band 7, it was necessary to use Landsat 7 ETM+ sensors. The ratio evaluation
was performed by calculating the length of the index and the difference between dates,
specifically before and after the fire event.
The NBR is a radiometric measure of the burn severity, and it is formulated from the
reflectance in TM band 7 and TM band 4. Those bands exhibited the greatest reflectance
range in response to fire. It is hypothesised that a ratio of those bands would be most
discriminating for burn effects, and hence the Normalised Burn Ratio is constructed from
their transformed reflectance values:
NBR =
NIR - TM 7
NIR + TM 7
TM7 increased with fire, while NIR decreased, and those trends are accentuated in the
normalised ratio.
The process has the same inconvenient than NDVI, and the GVI of TC, namely the
same responses to water bodies and fire effects. Both water bodies and fires are very
important in this kind of wetlands.
On a second stage of the process, the NBR difference is performed with two images of
zone 2, the invariant zones look dark grey, and the damaged areas show a light grey or white
colour.
11
Then, we separate the burnt area and perform an unsupervised classification. This
classification has a parameter of 10 clusters. We make this classification in order to quantify
the damage on classes and relate them to particular locations.
NBR-Difference
NBR-Classification
Figure 11. Burning Damage
Other Analysis Methods
We can use TDI to identify zones where the vegetation rapidly changes. If we perform
a TDI on several images, we can find areas where the vegetation has a great increase/decrease
in a short period. The next example shows a TDI using three images:
§
24-Apr-2003
§
10-May-2003
§
26-May-2003
Then, an Unsupervised Classification can be performed on this TDI Transformation
(six classes on three TDI images); so that a classification of variation levels can be obtained.
A wide red zone can be clearly seen, indicating a sharp reduction in vegetation, as well as a
region with vegetation increase due to anthropic activities.
12
Figure 12. Unsupervised Classification on a TDI Transformation
At this point, the analysis can be continued in two additional steps. The first step is
exploring the area of interest searching for same patterns. Then, a new iteration of the
methodology over the new area can be performed.
The following figures show other possible locations of the same phenomena:
Figure 13. Other zones with possible burned areas
Right image: Villa Florida, Sheet: 5468, latitude 26º 25´ y 26º 32´, longitude 57º 37´ y 57º 13´.
Left image: Villa Florida, Sheet: 5448, latitude 26º 28´ y 26º 35´, longitude 57º 33´ y 57º 21´.
The zones show less reflectance on the GVI index.
These zones show a decrease on the vegetation index (black colour) and a negative
slope in the regression model.
13
Conclusion
The proposed methodology allows an appropriate processing of satelital images in
order to explore a wetland zone. This methodology simplifies the analysis of important
features by focusing on
· an appropriate definition of the area of interest in order to eliminate external noise;
· the vegetation dynamics to discover behaviour patterns and evolution tendencies;
· the physical properties of water bodies, such as surficial temperature.
In particular, soil brightness, burn severity, and vegetation intensity are identified as
important aspects for the definition of vegetation indexes. Based on vegetation indexes, the
methodology can be applied to the classified map in order to detect automatically the intensity
of temporal changes over the wetlands vegetation. Then, specific areas can be isolated and
further classified for the definition of particular changes that enhance the analysis of
disturbances. For example, the anomalies detected in the Ypoa wetlands include vegetation
increments due to agricultural activities, as well as vegetation reduction due to land clearing
and marks of possibly burnt areas. It is clear that the analysis needs further in situ exploration
to corroborate hypotheses.
Wetland zones are very difficult to study because of very particular features, such as
heterogeneous and variable vegetation and water distributions. The satelital image processing
is a very appropriate tool for detecting temporal changes in vegetation, temperature as well as
other anomalies that occur in wetlands, which are isolated, extended and, usually, difficult to
access. The application of different transformations to a temporal sequence of images
generates an advantageous method for problem detection through remote sensing.
Presentations in Scientific Meetings
“Sensores Remotos para la Conservación de humedales del Mercosur”
Graciela Canziani, Diego Ruiz Moreno, Manuel Vargas Russo, Federico Dukatz, Rosana
Ferrati. Seminario de la Constelación Matutina. CONAE - NASA.
Buenos Aires, Argentina. 3-5 de Diciembre de 2003.
“Temporal evolution in wetland features”
Diego Ruiz Moreno, Federico Dukatz, Manuel Vargas Russo, Rosana Ferrati, Graciela
Canziani. Seminario de la Constelación Matutina. CONAE - NASA.
Buenos Aires, Argentina. 3-5 de Diciembre de 2003.
“Satelital Data Analisis”
Aplicación de los Sensores Remotos en los Humedales del Mercosur, Yberá y Ñembucú
Federico Dukatz, Manuel Vargas Russo, Graciela Canziani.
Taller sobre Manejo de Humedales en Mercosur. Universidad Nacional de Pilar
Pilar, Paraguay, Diciembre de 2003.
“The Sustainable Management of Wetland Resources in Mercosur”
Graciela Canziani, Rosana Ferrati, Diego Ruiz Moreno, Paula Federico, Florencia Castets.
Taller sobre Manejo de Humedales en Mercosur. Universidad Nacional de Pilar, Paraguay.
Diciembre de 2003.
“Need of an Original Approach to the Management of Wetlands and
Lagoons of the Pampas Region in Argentina”,
14
Graciela Canziani, Rosana Ferrati, Fabián Grosman, Pablo Sanzano, Fernando Milano, Diego
Ruiz Moreno
Taller sobre Manejo de Humedales en Mercosur. Universidad Nacional de Pilar, Paraguay.
Diciembre de 2003
Acknowledgements
We thank CONAE for providing the images used in these projects, and Raul Colomb
and Ida Nöllmann for their unlimited collaboration. Florencia Castets (Personal de Apoyo
CICPBA) helped design graphics. Jose E. Pérez Pérez, Departamento de Medio Ambiente,
Cuerpo de Bomberos Voluntarios del Paraguay; provided the information about fire in
wetland zones.
The European Commission financed the projects "The Sustainable Management of
Wetland Resources in Mercosur" (Contract IC18 CT98 0262) and "Regional Aspects of the
SustainableManagement of Wetland Resources " (Contract A4-DEV-ICFP-599A4 AM01).
References
Canziani, G.; C. Rossi; S. Loiselle; R. Ferrati (Eds) 2003. Los Esteros del Iberá. Informe del
Proyecto “El Manejo Sustentable de Humedales en el Mercosur”. Fundación Vida Silvestre
Argentina, Buenos Aires, Argentina. 258 pp.
Loiselle, S.; L. Bracchini. Cozar A., C. Rossi, A. Cognetta. 2001. Determination of water
optical parameters in a large wetland using TM bands. Sustainable Management of Wetland
Resources in MERCOSUR.
Ruiz Moreno, D., 2002. Herramientas para el Manejo de Recursos Naturales: Modelos
Metapoblacionales Espacialmente explícitos en la óptica de los Autómatas Celulares. Thesis.
Universidad Nacional del Centro de la Provincia de Buenos Aires.
http://www.exa.unicen.edu.ar/investigacion/ecolmate.htm
INCO-DC Project :"The Sustainable Management of Wetland Resources in Mercosur".
http://www.exa.unicen.edu.ar/~wetland/
CONAE Argentina
www.conae.gov.ar
CPTEC/INPE Brazil
http://www.cptec.inpe.br/products/queimadas/
15

The use of satellite images as spacial recolution tools for

Transcripción

Documentos relacionados

Grass monoculture

RACING SERVICE