Resumen de Cartografía del uso del suelo en la subcuenca Huaquechula, Puebla, México, con un índice combinado de imágenes de satélite

Julio Cesar Flores Cesareo, Ángel Bustamante González, Samuel Vargas López, Lenom Cajuste Bontemps, Francisco J. Escobedo, Mario Valadez Ramirez

• español

    La generación de mapas actualizados de uso del suelo en regiones de agricultura de regadío es importante tanto para el monitoreo productivo como para la aplicación de modelos hidrológicos o de otro tipo. Por requerimientos de costo y tiempo, las herramientas de percepción remota y Sistemas de Información Geográfica son de gran ayuda para tal fin. En este estudio se delimitaron los usos del suelo, con énfasis especial en las tierras de agricultura de riego, utilizando imágenes Landsat 8 de tres periodos con condiciones de lluvias diferentes durante 2017; para cada imagen se obtuvo el Índice de Vegetación de Diferencia Normalizada (NDVI, por sus siglas en inglés). Los NDVI se combinaron para formar una imagen Red-Green-Blue- NDVI (RGB-NDVI). Con el programa ERDAS Imagine 2014 se obtuvieron las firmas espectrales para los usos del suelo y vegetación identificadas con el programa Google Earth. Se ubicaron puntos al azar para verificar si el uso del suelo del mapa obtenido corresponde con el uso verdadero. Además, se comparó el mapa obtenido a partir de imágenes de satélite con el mapa de Uso del Suelo y Vegetación, Serie VI, considerado el mapa oficial más actualizado en México. Los mapas se evaluaron utilizando matrices de confusión e índices de confiabilidad. El mapa generado con imágenes Landsat 8 tuvo mayor confiabilidad que el de Uso del Suelo y Vegetación, Serie VI. El mapa obtenido es útil para dar seguimiento a los cambios de uso del suelo y es insumo para modelos hidrológicos, como el Soil and Water Assessment Tool (SWAT), con el fin de estimar los efectos de prácticas agrícolas sobre las corrientes superficiales. 

• English

  The elaboration of updated land-use maps in irrigated agriculture regions is essential for the monitoring of production as well as for the application of hydrological or other models. Due to cost and time constraints, remote sensing and Geographic Information Systems are valuable tools for this purpose. This study evaluated the applicability of an index that combines satellite imagery for mapping land uses, focusing on irrigated agriculture lands in the Huaquechula sub-basin, state of Puebla, Mexico. The study area is characterized by a complex array of irrigated crops with broad spatial variations and succession of crops throughout the year, making it difficult to delimit irrigated agricultural land. Addressing this complexity requires the application of methodologies to identify the presence/ absence of irrigated crops throughout the year. As an alternative to the delimitation of land uses, we used Landsat 8 images for three dates: 23 January 2017, 19 August 2017, and 23 November 2017. For each image, we obtained the Normalized Difference Vegetation Index (NDVI) with the program ERDAS Imagine 2014; the indices for the three dates were combined to make a Red-Green-Blue NDVI image (RGB-NDVI). The spectral signatures for land uses and vegetation identified with the Google Earth program were obtained for this image. A total of 250 random points were located with ArcGis 10.4 and the Hawth's Analysis Tools extension to check the correspondence between of land uses in the map and real uses; 155 selected points were located in natural vegetation; 32, in irrigated agriculture; 52, in agriculture; 11, in urban use; and 1, in a water body. The map obtained from satellite images was compared with the Inegi Map on Land Use and Vegetation, Series VI, considered to be the most up-to-date official map in Mexico. Maps were evaluated using confusion, omission and commission matrices, and overall user and producer reliability indices. The values obtained for confusion matrices indicate that the map produced from satellite images was better at delimiting land uses compared to the Inegi Map of Vegetation and Land Use, Series VI. For the irrigated agricultural land use, the primary delimitation objective, the map produced was 81.82% reliable for the user (probability of selecting a random point in one category that really belongs to that category) for irrigated land, versus 78.57% for the Series VI map. The produced map also yielded a higher reliability for the producer, relative to the reference map (84.38% vs. 61.11%, respectively), as well as less omission errors (18.18% versus 21.43%) and commission errors (15.63% versus 38.89%). We concluded that the map generated from Landsat 8 images using the RGB-NDVI index, provided a higher reliability than the map of Land Use and Vegetation, Series VI. For regions with high spatial and temporal complexity of crops, such as the study area, the procedure involving the RGB-NDVI index is an expeditious low-cost alternative for the continuous updating of land-use maps. The map obtained here is useful for monitoring changes in land use, and as an input for hydrological models, such as the Soil and Water Assessment Tool (SWAT), to estimate the effects of agricultural practices on surface streams.