Monitoring agricultural lands in Egypt with multitemporal Landsat TM imagery: How many images are needed?

Multitemporal imagery is needed to monitor agricultural landscapes characterized by high temporal variability. With the use of a database of 10 Landsat Thematic Mapper images dating from 1984 to 1993, the effects of the number and timing of images on area estimates of productive and nonproductive lands are evaluated for agricultural lands in Egypt. Average underestimation of non-productive lands in the Nile Delta ranges from 5% with data sets of nine images to more than 300% with data sets of two images. Average underestimation of reclaimed lands in the Western Desert ranges from 4% with six image data sets to 42% with one-image data sets. Generally, data sets including more high-growth-season imagery result in higher accuracies, although in some cases a mixture of high- and low-growth-season imagery is more accurate. A single peak-growth image dominates results in both regions; data sets including this image are consistently more accurate than those without it. The results indicate the critical significance of using extensive multitemporal data sets to monitor agricultural landscapes. (C)Elsevier Science Inc., 1997.