Elaborate processes are required for land cover classification using multiscene high spatial resolution satellite images, like selecting training area on each scene with sufficient a priori knowledge. The classification method proposed in this paper is assumed to use both high spatial resolution images and temporal low spatial resolution images. It can automatically produce training data set on each scene, optimized considering land cover characteristics to the scene. Moreover, it prevents from deteriorating into low classification accuracy, by referring to and checking consistency to the class candidate information derived from temporal low spatial resolution images. Experiments were conducted that used twenty-eight scenes of Landsat TM and NOAA AVHRR images as high and low spatial resolution images, respectively. Validation results by using three visually interpreted images demonstrate the optimization of training data set improved the classification accuracy from 59.0% to 66.2%, and the class candidate information improved from 61.9% to 66.2%.
In the present paper, the authors tried to estimate the influx loads into a double-closed bay (Omura Bay) using forty-nine scenes of Landsat5/TM from 1984 to 2000, and statistical data for many years. From the TM data, NDVI (Normalized Difference Vegetation Index) values were calculated for four seasons in each year. The outcomes without sea areas were classified into five categories, which were water region (pond, lake, etc.), forest area, agricultural land (plow land and paddy field), uncovered ground, and urban area. The influx loads were calculated from the total of surface areas of each category and the statistical data in each year. The relationship between the influx loads and the sea water quality changes of the bay was evaluated. Adequate correlation was estimated between phosphorus and nitrogen influx load, which were calculated from TM data, and the actual values of total phosphorus and nitrogen in the bay, respectively. However, in the case of COD and chlorophyll-a values, mutual relationships were not estimated from the results.
Recent developments of digital photogrammetric technology are bringing elevation data with high spatial resolutions. This study proposes a new approach to three-dimensional visualization of pastoral woodlands, known as “sato-yama” in Japanese, using a digital elevation model (DEM) and a digital surface model (DSM) . Subtraction of the DEM from the DSM provides information on tree heights over the targeted area, which enables highly realistic landscape simulations. A comparison between two visualization techniques, the conventional“draping” technique and the new“planting”approach featuring the above subtraction operation, shows that the latter can provide outputs with higher qualities. Realism of landscape images produced by this new technique is suitable for analytical and monitoring purposes. The new approach proposed in this paper is relevant especially to environmental assessments for woodlands, forests and other vegetated areas.
Ice concentration derived from satellite passive microwave sensor data is a key parameter for monitoring the global sea ice cover areas. Since the sea surface heat flux is quite sensitive to the existence of thin ice, accurate calculation of ice concentration within thin ice area is very important for heat exchange between ocean and atmosphere. In this study, two sea ice concentration algorithms, namely NASA Team algorithm and Bootstrap algorithm, were compared for the thin sea ice area of the Okhotsk Sea to evaluate the effect of sea ice thickness to the calculation of ice concentration from SSM/I data. Firstly, 100% sea ice concentration areas were extracted by using simultaneously collected NOAA/AVHRR/2 data. Then the ice concentrations of those areas were calculated with using SSM/I. The result shows that the NASA Team algorithm was likely to underestimate the ice concentration within the 100% thin ice area. The ice thickness of those areas was estimated with using AVHRR/ 2 data, and they were compared with the measured brightness temperatures of each SSM/I channels. It has become clear that the wavebands of 19GHz and 37GHz at horizontal polarization were sensitive to the thickness of thin ice. This explains why ice concentrations of thin ice area were underestimated with the NASA Team algorithm.
The Digital Map 50m Grid (Elevation) of Japan (J-DEM-50m) is a digital elevation model generated from 1: 25, 000 scale topographic maps by the Geographical Survey Institute of Japan. Since its publication in 1993, it has contributed considerably to various fields such as topography, geography, geology and hydrology. J-DEM50m is, however, accompanied with an unrealistic relief and spurious pits caused by shortcomings of its generation algorithm, In this paper, a new algorithm is proposed. It is characterized by (1) an increase in the number of angles from 8 to 16 for searching the steepest slope direction, (2) a composition of interpolation interval with piecewise steepest slope segments and (3) an improved identification method of slope grids. Our evaluation of contour lines reconstruction and identified slope grids from generated DEM shows that the new algorithm provided excellent results for the test areas.
A new spatial feature description method based on Wavelet Descripters is proposed. It is found that the proposewd method shows twice much better quality of the restored spatial feature depending on the support length of the Wavelet transformation in comparison with the Fourier Descripter.