Object-based classification for analysing hydrologically relevant land use elements; a case study for the rio de la compañía and rio nexapo watersheds, mexico.

Organisator Laboratory of Geo-information Science and Remote Sensing

wo 22 oktober 2014 14:30 tot 15:00

Locatie Gaia, building number 101
Droevendaalsesteeg 3
6708 PB Wageningen
+31 317 48 16 00
Zaal/kamer 1

By Stefan Arts (Netherlands)


Land use and land cover is changing to meet the requirements of a growing world population. The change of land use and land cover has several environmental consequences of which some are better known than others. Until one decade ago not much research was devoted to the impact of land use and land cover change on the hydrology of an area. Since then, much research addresses this topic.

This research investigates the potential for object-based image analysis (OBIA) to derive hydrologically relevant land use elements (LUE) from high resolution remotely sensed imagery. This study is conducted within the Rio de la Compañia and Rio Nexapo catchment area in central Mexico, which covers an area of 1246 kilometres. An assessment of the LUE is conducted to gain insight in the hydrological, spectral and spatial characteristics of the LUE. Afterwards, OBIA and pixel-based classification (PBC) are conducted on six LUE; forest, river incision, built-up, roads, agroforestry and monoculture. A stratified random sampling scheme is applied to create validation points using the OBIA result for the stratification.

Results show that OBIA achieves higher accuracies than PBC. An overall user’s accuracy of 0.49 and a producer’s accuracy of 0.56 is obtained by OBIA. The accuracies for the full study area are lower than the accuracies obtained in the training areas. The highest accuracies are obtained by OBIA for built-up and river incisions, with user’s accuracies of respectively 0.73 and 0.52 and producer’s accuracies of 0.75 and 0.82. PBC performs better for forest and agroforestry, where OBIA results in lower accuracies.

OBIA has several other benefits for usage in hydrological modelling. Firstly, classification creates large continuous objects, in which parcel shape is preserved. Secondly, polygons are created for urban areas in which roads and built-up are further specified. This has the advantage that the curb density (CURBDEN) and fraction impervious surface (FIMP) can be computed automatically. Both are important in hydrological models for built-up areas.