Work Package 1 – Análisis of Deforestation

							                    Work Package 1 – Analysis of Deforestation
                                15 January 2003



PLANNED ACTIVITIES 2002-2003 (Text in italics represent notes taken from
presentations in Oaxaca in December, 2002)

1) Universidad Austral de Chile (Contact: Cristian Echeverría)

Cristian Echeverría is working on his Ph.D. thesis (University of Cambridge) on the
fragmentation and deforestation and its effect on the diversity of woody species in the
Maule and Lake Regions in Chile.

This study has the following objectives:

1. To analyse the spatial and temporal patterns of deforestation in selected counties
   in the Maule and the Lake Regions (Regions VII and X)
2. To identify and analyse the determinant factors that explain the temporal changes
   in forest cover in both study areas.
3. To develop predictive models spatially explicit.

Methods:

The analysis will be done from Landsat images of the following dates:

VII Región 1975, 1990 y 2000.
X Región: 1976, 1985 y 1999.

The images have been already analysed with the PCI geomatic software, for
geometric, atmospheric and topographic correction (September, done in collaboration
with Partner 8 in Alcalá de Henares). After corrections, the images will be classified
according to different land use/ land cover. The fragmentation will be analysed through
the study of the spatial configuration of forest stands, and the factors that explain
deforestation. With this information for the changes occurred between 1975 and 2000
for Region VII and 1976-1999 for Region X, a spatially explicit predictive model will be
developed.

Studies done by other partners

Within WP 1, Luis Cayuela, a Ph.D. student from U. de Alcalá de Henares (Partner 8)
will carry out a fragmentation and deforestation study in Chiapas, using similar
methods to those being used in Chile. This study will imply a close collaboration among
Partners 2, 8 and 6 (Colegio de la Frontera Sur, ECOSUR).

Kerry Wilson a Ph.D. student from University of Melbourne is conducting her thesis
“Logistic model for the identification of vulnerable areas to deforestation in Chile”,
working with Adrian Newton (Partner 1), and with the collaboration of Partner 2.

Milestones

-   Correction of Images for the study areas in Chile
-   Analysis and classification of images
-   Development of GIS data base with land use-land cover categories and
    environmental variables
Deliverables

-    GIS Data base on a CD-ROM
    Map of forest cover cross-time
Acquired LANDSAT ETM mages from UNEP. For two regions of Chile; currently being
analysed by UAL & UACH for three periods:
1975-76
1985-86
1999-2000

They plan to model fragmentation and express the need to discuss how to do this with
the best approach possible. To do this we need to liaise between partners regarding
development of method.

2) ECOSUR (Contact: Dr. Susana Ochoa)

       En cuanto al uso del SIG en ECOSUR se prevee la participación de la Dra.
Susana Ochoa Gaona desde la Unidad Villahermosa de ECOSUR, con el apoyo de
algún becario de licenciatura o posgrado que pueda realzar el trabajo de campo
necesario en acuerdo con el personal de la Unidad San Cristóbal en Chiapas. Ella ha
establecido contacto con el grupo de UAL para definir los detalles de su colaboración
con el uso de SIG y teledetección. De ello ha dado cuenta el líder de UAL en su
resumen de actividades propuestas.
       El objetivo principal es calcular las tasas de deforestación, los procesos
implicados y los patrones obtenidos entre los años setenta y el 2000 en la totalidad del
área de estudio en Chiapas, a partir del uso de varias imágenes de satélite y
posiblemente fotografías aéreas. Susana Ochoa ya ha publicado dos antecedentes
importantes para regiones más pequeñas que las implicadas en el trabajo propuesto
(Ochoa-Gaona y González-Espinosa 2000 y Ochoa-Gaona 2001).
Productos: mapas de usos del territorio, mapas de deforestación, un manuscrito
publicable y una tesis de licenciatura o maestría en ciencias.

LANDAT MSS images procured and analysed for 1970, 1980, 1990, 1996. Work
already done but will continue under this project. The analyses to be included are an
analysis of fragmentation it’s causes including socio-economic factors. Patterns will be
documented during 1974, 1995, and 2003. The study will focus on mountains of north
of Chiapas although there is concern that Los Altos should be included as well since
this is were many study sites are located. They need to get hold of new maps from
INEGI including DEMs (20 m resolution). ECOSUR is currently working on creating
land use mapas from LANDSAT 2000 images for Los Altos – Duncan Golicher.

3) Instituto de Ecología, A.C. (Contact: Dr. Robert H. Manson)

       We have acquired previously classified land use / vegetation maps based aerial
photographs and satellite images (INEGI, CONABIO, Instituto de Geografía, UNAM),
from different years 1980 and 1996 (scale 1:1,000,000) as well as 1984 and 2000
(scale 1:250,000). We will be using the higher resolution maps for estimates of
deforestation on a regional scale but finer scale analysis of fragmentation is probably
not possible with these maps. Environmental and demographic data will be collected
from the National Censuses of 1990 and 2000 (INEGI). One possible constraint is the
availability of up to date and fine-scale images with which to work (the most recent
photos we have are from 1993; scale 1:250,000). We are currently negotiating with
various investigators at the Instituto de Ecología, A.C. and state and federal agencies
about the possibility of collaborating in the purchase of 1 m2 resolution color images
from the IKONOS satellite. These images cost $US 20.00 per km2 with a minimum
purchase of 100 km2 and come with the spectral bands necessary to perform a
supervised classification of the vegetation in the region. These or similar images will be
orthorectified and georeferenced using ARCINFO and ARCVIEW software. During the
first six months, students and collaborators will use handheld GPS receivers to
measure the size and shape of each new forest fragment. In addition, note will be
taken on land use around each fragment to aid in the estimation of the degree of
connectivity of each fragment.

Products: Ph.D. thesis, two publications, CD with spatial data of the study area for
decision makers and NGOs interested in forest conservation.

They are working on the regional level in three areas of cloud forest with 15 study sites
all located in one of these. Analyses are planned at two spatial scales (i) The
mountainous region (regional scale analysis of deforestation) and (ii) the landscape
scale around study sites so it can relate with other aspects of project eg edge effects,
and diversity impacts. Therefore, land use information is needed at two spatial scales.
They need some maps at high resolution to link with other work packages. The maps
already acquired of vegetation and land use came from governmental organisations
such as CONABIO, INEGI, etc. were classified previously by other groups of
investigators and include:

1980, 1996 1: 1,000,000
1984, 2000 1: 250,000

One problem is reconciling different classification systems used on different maps. IE
does not have the capacity to analyse / process satellite images that may have to be
acquired for comparisons between regions of the project. This includes LANDSAT or
MMS impages from 1975, 1985 & 2000. In a previous meeting in Xalapa it was agreed
that work on defining changes in patterns of land use should span 30 years, that
classifications should be reduced and integrated between Chiapas, Chile and Mexico,
that questions and data developed should be appropriate for each question, and that
each of regions will have a representative for the purpose of communication between
partners regarding this work package.

*** Please refer to the report of the meeting held January 2002 in Mexico for more
details; there will be a subgroup of specialists who will liaise on the remote sensing
analysis – Jose Maria Benayas, Robert Manson, Susana Ochoa, and Cristian
Echeverria

4) University of Acalá (Contact: Dr. José Maria Rey Benayas)


Estos objetivos están coordinados por la Universidad Austral de Chile. Se ha
establecido una interacción entre la UACH, el WCMC y la UAL por medio del
doctorando Cristian Echeverría para el estudio en las áreas de Chile, y entre Robert
Manson de Xalapa, Susana Ochoa de SOG-ECOSUR y la UAL para el estudio de las
áreas en México. Los profesores de la UAL encargados de estas investigaciones son
José María Rey Benayas y Javier Salas.
El objetivo principal es calcular las tasas de deforestación, los procesos implicados y
los patrones obtenidos. Estamos analizando en las 4 regiones de estudio tres
horizontes temporales: ca. 1975 con imágenes MSS, ca. 1985 con imágenes TM y ca.
2000 con imágenes ETM+. Los análisis de imágenes serán completados con mapas
de cobertura vegetal y distintas características físicas y de geografía humana del
territorio. Ya se ha comenzado a analizar las 2 regiones de Chile gracias a la visita de
Cristian Echeverría a la UAL durante el mes de septiembre. En Chile se persigue una
clasificación última que distinga bosque nativo (primario y secundario), plantaciones
(adulta, joven y recién cosechada), y áreas no forestadas (tierras agrícolas, etc.). En
Xalapa y Altos de Chiapas se distinguirán, a priori, la siguiente clasificación
simplificadas: bosques no perturbados, bosques perturbados (plantaciones, fincas de
café o bosque secundario), pastos inducidos o naturales, cultivos (incluyendo milpas,
caña de azácar, etc.), zonas antropogénicas (zonas urbanas, caminos, etc.), otros
usos de suelo.

Productos: mapas de usos del territorio, mapas de deforestación, 2 publicaciones y 1
tesis doctoral.

Have assisted with classification of images in Chile. Offered help for Mexican groups.
Using ThematicMapper for map classifications.

5) Instituto Politécnico Nacional (Contact: Dr. Rafael F. del Castillo)

This partner is not involved officially in this work program but they have an interest in
assessing the dynamics of land cover change for assessing impacts on population
viability. They have air photos of their study area as well as a satellite image for 2000,
but image is very coarse. They are also looking at what INEGI can supply. In
conclusion, they are searching for information on pattern of land cover change but not
planning a detailed assessment of deforestation. They are interested in linking with
other partners regarding how apply remote sensing tools to address this question.
CIIDIR does not have much capacity for remote sensing analysis but has some
capacity in GIS.

6) UFZ Center for Environmental Research (Contact: Dr. A. Huth)

This partner is not officially involved in this work program but is providing some support
to UACH (Cristian Echeverría) with predictive land use modelling.

7) UNEP World Conservation Monitoring Centre (Contact: Dr. Lera Miles)

This partner is providing support to C. Echeverría and access to images for the
analysis of deforestation and fragmentation in Chile.

8) Universidad de Chile (Contact: Dr. Juan Armesto)

This partner’s needs are being addressed by partners at UACH.


KEY QUESTIONS / ISSUES TO ADDRESS
(Elaborated during the December 2002 meeting in Oaxaca)

1. The standardisation of methods (e.g. modelling of deforestation & fragmentation)
There are two main phases of this research 1) the need to document patterns of
deforestation that have already occurred in the region (this will relate to the
classification scheme), and 2) the need to develop models for predictive analysis of
future landscape change. Various groups are working on such models, using a variety
of statistical approaches, including ECOSUR, IE, UACH, UAL, WCMC. All agreed that
the methods for mapping deforestation and modelling future changes needs to be
discussed further. During the next meeting we should hold a series of working groups
or mini-workshops focused on these and other technical issues related to Work
Program 1. In the first year of the project we should focus our efforts on calculating
deforestation rates and wait on the modelling of future changes in forest cover until the
second and third years.

2. How do we describe patterns of forest fragmentation? This is a very important issue
for other work packages, which will use this information. In order to make comparisons
between regions, there is a need to standardise the methods of measuring / describing
fragmentation. ECOSUR partners have used the FRAGSTATS package to this end and
suggest that all partners might consider a similar approach to standardise results.
However, there are many new packages that could also be used. These programs
provide indices of matrix and edge quality and quantity, as well as the degree of
isolation of fragments. There are three main things to measure – the size, shape,
connectivity of fragments. The FRAGSTATS program can provide this information but
is not very flexible. We need to research what other options / programs are available.
The WCMC has developed an approach for a combined index that could also be
applied. A suggestion was made that fragmentation should be analysed in the context
of other variables (eg roads). All partners agreed to discuss subsequently which
software should be used via email interchanges. The main task for all groups is to
research the potential methods available for measuring fragmentation, and then
discuss the advantages / disadvantages of each to come to an agreement on the best
methods that everyone should use.

3. How classify images? How verify / ground-truth? Duncan suggested that all partners
use NDVI as a way to standardise the classification scheme for creating land use maps
from satellite imagery. This technique is easy to use and will insure that all maps are
created in the same way; however, one potential problem is that an unsupervised
classification may generate many errors. Another issue is that IE will be using maps
that have already been classified and therefore, they will have to explore the
classification methods used by the groups that created these maps to insure that they
are compatible with the other regions of the study. The UACH will use the supervised
classification to generate the land-use maps for each period. The most important issue
is to compare images that are already classified. Through email partners will compare
the classification schemes that all partners have and explore whether they can develop
a common classification scheme. Initially there should be six classes of land use: The
UACH suggests the following basic classes: grassland, shrubland, native forest, urban
areas, bare grounds, other land-use. Depending on each study areas some classes
may be sub-classified. “Other land-use” category may include particular land-uses (e.g.
snow, commercial plantations, old –fields, etc.) that are not present in the other study
areas. . When possible it’s advisable to compare with aerial photos as a double-check
of the classification scheme developed. These six classes will be used to make initial
comparisons between regions. A finer level of classification (different types of forest)
will allow calculations of the loss of different types of native forests but this depends on
the needs of each partner and region.

4. Which sort of images to use – standardise, or use a variety of image types?
Regions should be comparable if IE can obtain images from 1975 (MSS) and 2000
(LANDSAT). Three periods over thirty years (1970s, 1980s and 1990s). LANDSAT
ETM+ and TM have a pixel size of 28.5 square meters, which should provide the level
of resolution needed for other work packages.

5. Obtaining images? The UNEP-WCMC has offered to help IE acquire MSS and
LANDSAT images but needs details of which scenes needed & dates. Only IE needs
images; all other partners now have them.
6. Need to undertake analyses at two spatial scales – There is some concern about the
need for high resolution images / maps to provide information relevant to the other
work packages of the project. What is the scale at which the different participants are
working? The conclusion of partners was that the images from LANDSAT 1985, 2000
(pixel size 28.5 m resolution per pixel. So this should be sufficient? May need a digital
elevation model to include in analyses.

7. Do we need to undertake analyses at same locations as locations of work on other
work programs? In Chile and Xalapa this is the objective. However, in Chiapas the
focus of fieldwork will be Los Altos where landscape-scale information is lacking. Land
use maps will be needed undertake deforestation assessments in this area also. There
was general agreement that 15 should be the minimum number of fragments to study
for WP2.

8. What do we need from this work packages for the other work packages? There will
be a need to scale up to link the different analyses. It remains to be seen if there are
limitations in resolution of remote sensing data.
a.