PE&RS October 2002

Highlight Article

Assessing Forest Resources In Mexico: Wall-To-Wall Land Use/Cover Mapping
Jean-François Mas, Alejandro Velázquez , José Luis Palacio-Prieto, Gerardo Bocco, Armando Peralta, and Jorge Prado

Mexico, encompassing a continental territory of nearly two millions square kilometers, is one of the five biologically richest countries, therefore considered as megadiverse. However, the country is undergoing rapid processes of land use /cover change with rates of deforestation of 0.25 and 0.76 % per year during the last decades for temperate and tropical forests respectively (Velázquez et al. 2002, Mas et al. 2002). In this context, forest assessments are crucial in providing us with accurate and updated data for supporting decision-making procedures at the governmental and social levels. Nowadays, these assessments are meant to support increasing demands for additional information on topics such as land use/cover change and deforestation, biodiversity conservation and environmental services, so that detailed spatially explicit products are key to overcome new demands on environmental management.

In the year 2000, the Mexican Secretariat of the Environment put the Institute of Geography of the National University of Mexico (UNAM) in charge of carrying out the wall-to-wall land use/cover mapping of Mexico as the first step of the National Forest Inventory in order to provide a reliable, detailed and rapid forest assessment. Limiting conditions such as time (wall-to-wall mapping had to be produced in 8 months), budget, personal, hardware and software guided the planning of this work. Because of the high diversity and complexity of the vegetation, previous cartographic efforts were considered fundamental to obtain a more detailed and accurate forest resource assessment with its correspondent cartographic representation. It appeared that the best solution was to base visual interpretation of remotely-sensed images upon existing cartography from the National Institute of Statistics, Geography and Informatics (INEGI), the official Mexican mapping agency. The advantages of this approach were: 1) the cartographic input was derived from aerial photograph interpretation and intensive field work; 2) the classification scheme was compatible with existing vegetation classification systems and previous cartography, allowing a reliable land use/cover change estimation and; 3) INEGI data were available in digital format.

Land Cover Classification Scheme Design

In order to design the classification scheme, we conducted five workshops, where the most important experts in vegetation mapping in the country convened. We looked for a classification scheme which could satisfy the needs of different end-users, and was a trade-off between what is discernible in the remotely-sensed imagery and pre-existing data.

The classification scheme was obtained by regrouping the INEGI land use/cover classificatory legend used for vegetation mapping which considers over six hundred classes based upon physiognomic, floristic, phenologic, and degree of disturbance attributes. In order to fit the classification scheme to the scope of the Landsat data, we reduced the number of categories from 642 to 75. The classification scheme includes four levels. Table 1 shows the first two levels (formations and types). In the following level (communities), sub classifications of the above types were performed. For example, the type "conifers" is divided into 4 categories : 1) Pine forest, 2) fir forest, 3) conifer shrub-land dominated by Pinus culminicola and Juniperus and, 4) Cedar forest dominated by Juniperus. The 4th level involves information about the level of disturbance of the vegetation, and only applies to 28 categories.

Image Processing and Analysis
We used 126 cloud-free Enhanced Thematic Mapper plus (ETM+) sensor data of the dry season, acquired between November 1999 and April 2000. The images were registered and mosaiked. Different color composites were tested in order to select the best option for land cover discrimination. Simultaneously, the original data from INEGI were organized on the basis of the classification scheme obtained during the workshops. Final printing of the mosaics and the vectors was performed at 1:125,000 scale to ease actual visual mapping procedures.

Visual interpretation was performed by displaying the digital maps from INEGI (printed on transparent paper) on top of image mosaics. This process was conducted by a few interpreters in order to enhance consistency, and was supervised by three external experts: 1) interpreter's coordinators, 2) expert botanists in specific vegetation regions and, 3) INEGI staff that elaborated the previous cartography. The new boundaries (vectors) were manually digitized and integrated into a GIS database. Then, quality control procedures, such as searching unlike adjacent categories, were carried out.

Figure 1. Aerial survey design. Photographs were acquired along lines crossing the whole country and covered about 10% of the entire country.

Accuracy Assessment
To assess accuracy, we acquired high-resolution digital aerial photography. The aerial survey consisted of a grid with two perpendicular sets of flight lines that covered the entire country from side to side. The array was oriented so that one of the axes followed the general direction of the main range of mountains, while the other crossed them perpendicularly. The spacing between lines was 50 km and 100 km for the regions where the vegetation units are larger and more homogeneous (Figure 1).

Since photographs were collected along flights lines, the design could not be classified as a probabilistic sample of the full territory. However, flight lines were distributed in a systematic manner and sampling units were selected randomly from a large sample size, so that no bias was introduced. In other words, we assumed that the reduced population (lines) was representative of the full population (all the territory) (Stehman and Czaplewski 1998, Stehman, 2001).

We carried out the accuracy assessment for a region of about 480,000 km2 in the North of Mexico because the entire set of photographs was not available at the end of the project. A subset of about 1200 photographs was selected by a stratified random sampling and was analyzed. Results indicated accuracy over 70 percent for most of the categories at the community level (third level of aggregation).

Products
A total of 121 imagemaps, including a synoptic legend and relevant information on major topographic features and human settlements, and 121 land use/cover maps were produced for the entire country at 1:250,000 scale (Figures 3 and 4). Statistics of area per category, computed at national level, at state level and by hydrologic region can be found in Palacio-Prieto et al. (2000). Other products were a dictionary which described each category of the classification scheme precisely and the metadata.

Click any of the following images to view a larger version in a new window

Figure 2. Surface covered by formations and other land use coverage considered in the current Mexican National Forest Inventory 2000.

 

Figure 3. Land Use/Cover Map at scale 1:250,000. The continental Mexican territory is covered by 121 maps.

Figure 4. Image maps at scale 1:250,000. The image-maps have been elaborated using the same color composites as those used during the image interpretation.

Concluding Remarks
Updating a existing cartography allowed us to take into account information obtained with higher resolution images (aerial photographs) and intensive field work. If compared to a classification based only on remotely-sensed image interpretation, this approach allowed us to improve the accuracy and the precision of the classification. This is especially true in countries like Mexico, with a large diversity of vegetation types, with similar spectral responses, and highly complex and fragmented landscapes. Unfortunately, such cartography, with an acceptable level of accuracy and consistency for the entire country, does not exist in many tropical countries. This approach is also rapid and cost-effective: eight months and US$ 1,200,000 were devoted to the project which represents a mapping cost of US $ 0.6 per square kilometer.
T

e land use/cover digital map has been used in several nation-wide applications. Every state government must elaborate the land use planning of its territory. The digital land use/cover map is used as an important source for this process. The updated land use/cover map was integrated into a GIS database with two previous maps from INEGI in order to elaborate a spatially explicit multi-date database on land use/cover change. As the data are compatible with respect to scale and classification scheme, and because the maps are successively updated, this database allowed an accurate land use/cover change assessment during the last decades (Mas et al., 2002). The digital land use/cover map was also used as an input into a spatial explicit model for identifying a large number of environmental processes, such as collected fuel wood, likely hot spots for biodiversity conservation, deforestation fronts, fragmentation and desertification processes, and eventually a number of forecoming assessments on environmental services (Masera et al., 2002, Velázquez et al., 2002). Spatially explicit forests resource assessment is, therefore, fundamental to turn decisions into operating management practices, which is an issue to be further explored (Velázquez et al. in press).

Acknowledgements
We gladly thank F. Takaki, A. Victoria and G. López from INEGI, and M. García and F. Tudela from former SEMARNAP and the technical staff of the Institute of Geography that was in charge of image processing, image-interpretation, GIS database elaboration and cartographic design.

References
Groombridge, B., and Jenkins, M. D., 2000. Global biodiversity. Earth's living resources in the 21st century. 246 p

Mas, J.F., A. Velázquez , J.R. Díaz, R. Mayorga, C. Alcántara, R. Castro, and T. Fernández, 2002, Assessing Land Use / Cover Changes in Mexico, Proceedings of the 29th International Symposium on Remote Sensing of Environment (CD), Buenos Aires, Argentina, 8-12/04/2002.

Masera, O.R., G. Guerrero, A.Velázquez, J.F. Mas, M.J. Ordóñez, and R. Drigo, A Spatial Explicit Approach for Identifying Fuelwood "Hot Spots" Using Wisdom: A Case Study for Mexico, in press, FAO.

Palacio-Prieto, J. L.; G. Bocco; A. Velázquez, J.F. Mas; F. Takaki-Takaki; A. Victoria; L. Luna-González; G. Gómez-Rodríguez; J. López-García; M. Palma; I. Trejo-Vázquez; A. Peralta; J. Prado-Molina; A. Rodríguez; R. Mayorga-Saucedo, and F. González, 2000. La condición actual de los recursos forestales en México: resultados del Inventario Nacional Forestal 2000, Investigaciones Geográficas, 43:183-203.

Stehman S.V., and R.L. Czaplewski, 1998. Design and Analysis for Thematic Map Accuracy Assessment: Fundamental Principles, Remote Sensing of Environment, 64: 331-344.

Stehman, S. V., 2001. Statistical rigor and practical utility in thematic map accuracy assessment, Photogrammetric Engineering & Remote Sensing, 67(6): 727-734.

Velázquez, A. , J.F. Mas, J.R. Díaz, R. Mayorga-Saucedo, P.C. Alcántara, R. Castro, T. Fernández, G. Bocco, E. Escurra, and J.L. Palacio, 2002. Patrones y tasas de cambio de uso del suelo en México, Gaceta ecológica, INE-SEMARNAT, 62:21-37.

Velázquez, A., J.F. Mas, J.L. Palacio-Prieto, and G. Bocco. In press. Land cover mapping to obtain a current profile of deforestation in Mexico. Unisylva, FAO, Rome, Italy.

Authors:
Jean-François Mas*, Alejandro Velázquez , José Luis Palacio-Prieto,
Armando Peralta, and Jorge Prado
Instituto de Geografía, UNAM
Circuito exterior - Cd Universitaria
A.P. 20-850 CP 01000 México DF MEXICO
E-mail :jfmas@igiris.igeograf.unam.mx
Phone / Fax : (52) 443 324 71 49
* corresponding author

Gerardo Bocco
Instituto de Ecología, UNAM
Universidad Académica Morelia
AP 27, sucursal 3, Xangari, 58089 Morelia Mich. MÉXICO
Present address : Instituto Nacional de Ecología
Periférico 5000, Col. Insurgentes Cuicuilco,
C.P. 04530, Delegación Coyoacan, México, D.F.

Top

Home