Testing remote sensing data spatial resolutions for very fine water resources management requirements

Water resources management can use information gathered from very high spatial resolution imagery to feed models unexpected to be employed with this kind of data source. By example, water withdrawals for irrigation can be estimated with a variable precision according to the imagery spatial resolution. This paper regards to which resolution should be better for an application that could result in control and management of the water resources. A 1 km2 surface has been used for spatial resolution adequacy tests. Acquired imagery for tests performance was: SPOT5 multispectral 10m, panchromatic 5m and 2.5m; and QuickBird multispectral 2.8m, and panchromatic 0.7m spatial resolutions. The last one has been used as ground truth. A more general procedure was adopted to identify a selected group of features that could be employed for generic management and control purposes. The selected features and the images processed with standardized treatment procedures were submitted to remote sensing expert interpreters to visual classification and answer to evaluation questions. These answers and three other automated classifications results were compared with the ground truth for directions on image spatial resolution selection.