Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Mexique → Evaluación de métodos para la cartografía digital de clases de tierra campesinas

Colegio de Postgraduados (COLPOS) Consejo Nacional de Ciencia y Tecnología (CONACYT) 2009

Evaluación de métodos para la cartografía digital de clases de tierra campesinas

Cruz Cárdenas, Gustavo

Titre : Evaluación de métodos para la cartografía digital de clases de tierra campesinas

Auteur : Cruz Cárdenas, Gustavo

Etablissement de soutenance : Colegio de Postgraduados (COLPOS) Consejo Nacional de Ciencia y Tecnología (CONACYT)

Grade : Doctorado 2009

Résumé
The digital mapping of soils is to use computer algorithms and predictive variables representing the generation of soil maps. There is evidence that these maps are reliable. However for the mapping of farmland classes using digital techniques, with slight information and maps produced are of low quality because they have used only the reflectance values of the farmland classes and algorithms to predict limited in its configuration. For these reasons, this research evaluated the quality of the maps of farmland land classes generated in Mexico in contrasting environmental conditions (arid, temperate and tropical), from techniques used in digital mapping of soils with the aim of generate a methodology applicable in different regions. Six classifications were used : decision trees, artificial neural networks, minimum distance, parallelepiped, maximum likelihood and inverse distance. Thus, as remote sensing data and topographic attributes as predictors. The results showed that the most influential variable to enhance the precision and accuracy of the maps was the elevation. As for the algorithms, the inverse of the distance was the best compared to maximum likelihood, artificial neural networks and decision trees. It is also a reduced amount of complex to configure, does not require the predictors and is more efficient by requiring a reduced amount of sampling points for suitable configuration. This last factor is also very important in the digital mapping of farmland classes because it requires that the information is as reliable as possible for the training of the algorithm is accurate, which is achieved by making this information in the field and through interviews with farmers, this step is very relevant and can be established that the field sampling should not be ignored. However, if we can use a scheme that maximizes the sample, recommending systematic spatial design

Mots clés : Precisión y exactitud de mapas. Imágenes de satélite. Atributos topográficos. Inteligencia artificial. Tamaño y diseño de muestro espacial. Edafología. Precision and accuracy of maps. Remote sensing data. Topographic attributes. Artificial intelligence. Size and spatial sampling design.

Présentation

Version intégrale (10,57 Mb)

Page publiée le 12 février 2015, mise à jour le 11 août 2017