Quantification of impervious surfaces in urban areas using remote sensing techniques / Quantificação das superfícies impermeáveis em áreas urbanas por meio de sensoriamento remoto

AUTOR(ES)
DATA DE PUBLICAÇÃO

2006

RESUMO

Impervious surface coverage growth has direct consequences in quantity and quality of stormwater runoff, such as floods and nonpoint source pollution. The quantification of basins imperviousness makes possible to use it as a water quality indicator for urban areas and as an important planning and managing instrument in the urban water resources management system. Furthermore, hydrologic studies need parameters, among which impervious surface percentage may be one of the most important. The diffusion of Remote Sensing techniques and their great potential motivated this research, in order to find more efficient methods for estimating impervious surface coverage in basins. Three different algorithms were used to classify satellite images and to calculate imperviousness in control areas located in the city of Brasilia-DF, Brazil. The algorithms used were the Linear Mixture Model, available in software SPRING version 4.1, the Fuzzy classifier with membership function based on Mahalanobis distance, and the traditional Maximum Likelihood classifier (MaxVer). The last two were implemented in JAVA software language as a plugin for the software Image J. The three algorithms were tested with different spatial resolution digital imageries: IKONOS, with 1 meter; SPOT, with 10 meters; CBERS, with 20 meters, and LANDSAT, with 30 meters. To compare with the reality, impervious surfaces were manually digitized over IKONOS imagery and validated with field visits, so that the imperviousness obtained was considered the ground truth. Results showed that best performances were obtained by Linear Mixture Model and by Fuzzy classifier, which had mean errors of 21,2% and 23,7%, respectively. MaxVer classifier obtained greater mean error of 31,8%. Linear Mixture Model, however, showed some deficiencies, such as high values for shadow classes, tendency to underestimate impervious surfaces, and very high computer processing times. The influence of spatial resolution was important only to the MaxVer classifier

ASSUNTO(S)

áreas impermeáveis classificação sub-pixel modelo de mistura engenharia hidraulica classificação fuzzy

ACESSO AO ARTIGO

Documentos Relacionados