Analysis of Methods to Detect Changes in Coverage from Water Reservoirs of the Pao River Basin, Venezuela

Abstract

In this paper, four methods of detecting changes in land use and coverage are applied to
three reservoirs on the Pao river basin, Carabobo and Cojedes States, Venezuela: 1)
difference of reflectance images, 2) reflectance images ratio, 3) difference of index images of
vegetation and 4) difference of images of principal components. Eight Landsat satellite
images corresponding to L5TM (1986, 1991, 2001), 2) L7ETM (2000, 2002, 2003) and 3)
L8OLI (2015 and 2016) are used. The bitemporal change detection methods are: the
difference in reflectance in the near-infrared region, the normalized difference in vegetation
index, the principal components based on reflectance in the optical and infrared spectral
regions and the reflectance ratio in the near infrared region. The global accuracy index
varies between 86.67 and 100%, the Kappa coefficient between 0.6 and 1. The water quality
changes are estimated in terms of the presence of only sediments, a mixture of sediments and
algae; coverage from water to vegetation; the reflectance in the near infrared region and the
concentration of sediments vary between 1 and 3%; 50 to 100 mg / l; 0.5 and 1%; 50 to 250
mg / l; an increase from 1% to 40%, respectively.