MULTITEMPORAL ANALYSIS OF LANDSAT DATA FOR AGRICULTURAL LAND USE

The application of multitemporal Landsat MSS data in a temporal profile approach for agricultural land use/cover pattern mapping in the vicinity of Green Bay, Wisconsin, has been investigated.; The feasibility of identifying land use/cover types on the basis of the shape of temporal profile was evaluated. Six dates of 1976 Landsat MSS data were radiometrically and atmospherically corrected and registered. The registered data were transformed by means of ratioing and the Green Vegetation Index transformation to emphasize the development of greenness in vegetation. Sample points values, extracted from the transformed data, were used for construction of temporal profiles. The ground covers were then identified on the basis of the shapes of the profiles and knowledge of the local crop calendar. There is good agreement between the identification results and the aerial photo interpretation results.; In the investigation of new techniques for temporal data classification, four dates of 1979 Landsat data were chosen. The data were subjected to the same preprocessing and transformations described above. Then the transformed temporal data, and one set of single date data, were subjected to two classification processes: supervised and unsupervised. The unsupervised technique is a clustering algorithm sensitive to image texture. A spectral variance threshold for each three by three pixel window in the image data was used to isolate homogenous areas for cluster center definition. A temporal profile for each cluster center was constructed, labelled and categorized. The classification results thus produced were superior in accuracy and more consistent in comparison with those generated by other classifiers using either multitemporal or single date Landsat data.; To demonstrate the application of remote sensing to a practical problem. The Land use/cover information in digital map form generated by the classifications was superimposed on top of the soil and slope digital maps to form a data base. Parameters were derived from the data base to run a hydrologic model--LANDRUN, which simulates runoff and sediment under various conditions.