Geostatistical Analyses of Landsat Thematic Mapper–Derived Surface Temperature during Winter Nights

Abstract

Geostatistical analyses were used to clarify the effects of the spatial structure of terrain on the spatial variability of satellite-derived surface temperatures on clear winter nights in central Japan. The clarification describes the mechanism for the formation of thermal belts found on mountain slopes. In addition to a digital elevation model on a fine scale, satellite images that were obtained by the Landsat Thematic Mapper on four different occasions were used. In the thermal belts, spatial interdependence between the surface temperatures and the digital elevations was different from that in other regions such as mountains and plains. Cross-variograms between these two attributes were represented by a double-exponential model comprising both short- and long-range structures. The cross-variograms with regard to the short-range structure depended on the kinds of heat sinks (i.e., atmosphere or surface) that control the cooling of surface air. When the sink was the upper cold part of the atmospheric boundary layer, the cross-semivariances were negative, indicating that surface temperatures were negatively correlated with elevation. When the sink was the ground surface that was cooled by radiative cooling, on the other hand, the cross-semivariances were positive. The ranges estimated from autovariograms of surface temperature were correlated with those of elevation.