Representability of training areas is primarily focused on its qualitative spectrum features. But it was revealed by this study that the representability by a quantitative application of prior frequencies of each land-cover item also plays an important role for more accurate classification in remote sensing, increasing the rate by ten to fifteen percent for the whole scene in pixel-by-pixel evaluation. Complete enumeration for accuracy assessment in test field was realized such that detailed digital land-use data of 10 m×10 m resolution, prepared by the Japanese Geographical Survey Institute, were aggregated to 50 m×50 m cell size in match for geocoded Landsat MSS pixels. Changing the sampling sizes in the merged file, coupled with eight application types of Bayesian discriminant method, produced sixteen systemat-ically-conditioned classifiers to compare, and resulted that there is a great variation of accuracies by the classifier especially in terms of the applicable suitabilities in case of separating individual land-cover items.
A new method is adopted for the evaluation of the upwelling radiation emerging grom the atmosphere bounded by non-iniform terrain, which is assumed to be composed of "n" uniform regions (each regions may be different from each other). The atmosphere is assumed to be horizontally homogeneous, but it may be vertically inhomogeneous. It is composed of aerosol, molecules and absorbent gases. Numerical simulations are undertaken for the model aerosols of the oceanic and water soluble types. In the computational procedure an iterative doubling-adding equation is expanded into a series of the radiative interaction modes between atmosphere and surface. A probability of contribution of respective regions in the radiation field is calculated based on the assumption of single scattering in the atmosphere. Then on the basis of this probability, the emergent radiation at the top of the atmosphere is calculated approximately by considering the radiative interactions with surfaces to be twice as large. Here, the effect of the multiple scattering in the atmosphere is fully undertaken in the computational procedures. Therefore, the present new version enable us to quantitavely discuss the upwelling radiation over the non-uniform surfaces even if the optical thickness (τ-2) is large and (or) surface albedo is great (A-0.7).
The development in drainage basin of near urban has been producing the quick change of landuse and the danger caused by floodpeak incresease. For taking a countermeasure to protect the danger, it is necessary to timely obtain the infermations of change of landuse, and take an examination on relationship between the change of landuse and the change of flood runoff charactaristics. This paper shows a method of classification of landuse which use the vegetation cover rate (V.C.R) as an index, and which V.C.R is calculated by Landsat data, and shows the result of a case study through the drainage basin Iruma River in Saitama Prefecture. The categorie of classification of landuse are Rice field, Forest, Urban, Mixed area of forest and urban, and Golf links.
Applicability of a scheme developed by the authors for snow covered area detection based on satellite data is studied with a series of AVHRR/ NOAA data in snowmelt season. N-LAND database and Ground Truth Data produced from AMeDAS data are combined in the region of the Echigo Mountains, 16, 000 km2. Variation of a set of detection parameters in the scheme in confirmed on two deimentional plot of ALBEDO chl, infrared ch3 or thermal infra-red ch4 of six scenes from Junuary to April in 1989. High and low clouds can also be detected by using ch4 and ch3 infra-red, respectively. The results show the threshold values increase as the solar altitude becomes high with season. Snow covered area, free-snow and cloud can be detected precisely by the scheme.