Assessment of Land Use and Land Cover Change Detection Using Eleven Techniques of Satellite Remote Sensing in the Pao River Basin, Venezuela

Abstract

In this investigation the land use and land cover change detection are applied on the Pao
river basin, which is a study unit that is composed by the LULC: agricultural, rangeland,
urban, water, vegetation and degraded soil. This study includes the five stages following: 1)
data acquisition, 2) data preliminary processing, 3) change detection techniques application,
4) threshold analysis, 5) method result comparing. Eleven change detection methods have
been evaluated; ten of these are classified like a methods based on pixels and the last
corresponds to the classified object method. Eleven images are acquired from the Landsat
satellite in the period between 1986 and 2016. The comparison of the results depicted by the
area change detection methods based on post-classification using Maximum Likelihood
(ML), Artificial Neural Net (ANN) and Support Vector Machine (SVM) algorithms expressed
by the area change detection percentage according each class: a) U: Urban: 18 to 40%; 30
to 50%; 30 to 50%, b) A: Agricultural: 85 to 95% 80 to 90%, 80 to 90% c) R: Rangeland: 80
to 95%, 90 to 95%, 90 to 95%, d) W: Water: 10 to 20%; 15 to 20%, 15 to 20%, e) V:
Vegetation: 5 to 10%, 10 to 20%, 10 to 20%, f) D.S.: Degraded Soil: 55 to 60%, 15 to 40%,
15 to 40% g) C: Clouds: 85 to 100%, 95 to 100%, 95 to 100%, h) Sh: Shadows: 95 to 100%,
95 to 100%, 95 to 100% . The change detection method based on the pixel with the most
capability for estimating is the principal components using the component N° 1 compared
with the rest of the methods image difference, image ratioing, image regression and
normalized difference vegetation index. The classified object method requires to be reviewed
in the segmentation process.