MAPPING OF SECCHI DISK TRANSPARENCY DISTRIBUTIONS IN LAKE KASUMIGAURA AND OTHER LAKES WITH LANDSAT-3 MSS DATA

Abstract

The author tried to estimate the water quality (Secchi disk transparency) of Lake Kasumigaura and other lakes in the northern Kanto plain by means of multivariate analyses and computer mapping techniques. The results of this estimation were as follows. 1. The water quality estimation models were derived from the multiple regression analysis of LANDSAT and ground truth data in Lake Kasumigaura (Jan. 19, 1980). 2. Among these models, the Secchi disk transparency estimation model retained the highest ratio of contribution and the signal in band 6 was adopted independent variable of the model. 3. It is quite difficult to fit this model to the data of the oligotrophic lakes for such reasons as weak energy penetration of the near-infrared wave-lengths and restriction in the model itself. 4. On the other hand, it is possible to apply this model to the data in Lake Kasumigaura and other eutrophic lakes, of which the transparency are below 1.59m and located in the lowland. 5. In Lake Kasumigaura the estimated transparency (Secchi disk transparency) reduced from the center or the areas adjacent of the outlet to Tsuchiurairi and Takahamairi estuaries. 6. In Lake Kitaura the estimated transparency reduced from the southern part of the lake and its value was relatively higher than that of the center of the lake Kasumigaura. 7. In Lake Hinuma the estimated transparency had uniform pattern with the same value as that of the central part of Lake Kasumigaura. 8. There might be two approaches in the future application of regression analysis into satellite remote sensing of lake water pollution. The one is to get optimal regression using numerous ground truth data in various lakes and to amend the model for each lake depending upon each circumstance of the lake. The other is to cluster various lakes into a few groups which have similar physical and chemical conditions before the application of the analysis, and store the water quality estimation models for each group. For the practical purpose, the latter seems to be more preferable than the former. 9. The ground truth observations might be carried out efficiently after the derivation of the model, but the collection of a few kinds of data would be required for the corrections of the video data.