Advantages of remote sensing for land evaluation

1. Relatively cheap and rapid method of acquiring up-to-date information over a large geographical area. Example: Landsat 5 covers each area of 185x160km at a ground resolution of 30m every 18 days, cost of the original digital data is $5 000 (6 200 ha $-1, each hectare contains approximately 11 observations. Even with the cost of ground truthing this is very economical.

2. It is the only practical way to obtain data from inaccessible regions, e.g. Antarctica, Amazonia.

3. At small scales, regional phenomena which are invisible from the ground are clearly visible. Examples: faults and other geological structures. A classic example of seeing the forest instead of the trees.

4. Cheap and rapid method of constructing base maps in the absence of detailed land surveys.

5. Easy to manipulate with the computer, and combine with other geographic coverages in the GIS.

Disadvantages of remote sensing for land evaluation

1. They are not direct samples of the phenomenon, so must be calibrated against reality. This calibration is never exact, a classification error of 10% is excellent.

2. They must be corrected geometrically and georeferenced in order to be useful as maps, not only as pictures. This can be easy or complicated.

3. Distinct phenomena can be confused if they look the same to the sensor, leading to classification error. Example: artificial & natural grass in green light (but infrared light can easily distinguish them).

4. Phenomena which were not meant to be measured (for the application at hand) can interfere with the image and must be accounted for. Examples for land cover classification: atmospheric water vapor, sun vs. shadow (these may be desirable in other applications).

5. Resolution of satellite imagery is too coarse for detailed mapping and for distinguishing small contrasting areas. Rule of thumb: a land use must occupy at least 16 pixels (picture elements, cells) to be reliably identified by automatic methods. However, new satellites are being proposed with 1m resolution, these will have high data volume but will be suitable for land cover mapping at a detailed scale.