Due to the low accessibility of high mountains in Taiwan, there is little knowledge of the geologic conditions in these areas. This paper is a part of the study sponsored by Remote Sensing Development Committee of Agriculture Council to collect geologic information in NengKaoShan Area, with a size around 1700 square kilometers. The area covers also a part of the ridge line of Central Mountain Range. The results show that there are more than 4000 landslides in this area. The area is in the Central Mountain Geologic Province with two old geologic formations, i.e. The Tananao Schists and Bilusan/Lusan Formation. The Tananao Shists Formation (incl. Taluko Belt and Yuli Belt) spans from Paeleozoic to Mesozoic time. The Pilushan/Lushan Formation was in Paeleo-Tertiary Period. Eight lithologic units can be readily identified from both the satellite images and aerial photographs. They are denotated as Y, S, K, T, TM, Eh, M, and Qa. The K unit exists to the noth of SanChanSi, whereas the Y unit takes place to the south of SaPoDan­Si. In this year, GPS is used to position the outcrop when field works were conducted. It is concluded that GPS linked to a notebook computer is a good tool for improving the efficiency and potentially an important tool for automating field work.

Both the simple random sampling and three forms of forest-type oriented stratified random sampling are the four designs in this comparison study. Samewise, three forms of circular plot sampling, i.e. 0.0125ha, 0.0250ha and 0.0500ha, and two forms of horizontal point sampling i.e. BAF4 and BAF2, are enumerated with the common center or point in every observations. The 140 observations are allocated to the strata according to the degree of precision desired for the particular strata, i.e. 32 for conifer-hardwood plantation, 48 for coniferplantation, 20 for natural hardwood forest, 40 for natural conifer forest. For time and cost concerned, the two forms of horizontal point sampling are superior to three kinds of circular plot sampling whatever the simple random sampling design or strati­fied random sampling design might be; Among the sampling designs, the horizontal point sampling of BAF 2 is the best, hori zontal sampling of BAF 4 comes next and the circular plot sampling of 0.050ha be the worst; Compared with the simple random sampling, three forms of stra tified random sampling gain in time/cost in corresponding form of horizontal point sampling respectively. The "R1" (BAF2) on 3-stratum stratification is with the least survey time/cost among four different sampling designs and five field enumeration methods.

Synthetic Aperture Radar (SAR) is an active microwave detection system which can be used day or night and rain or shine. For both civilian and military concerns, high - resolution imagery resulting from SAR represents an important source of remotely sensed data. At first, basics on the forming of SAR imagery and its geometric and radiometric characteristics are explained. Next, we describe the radargrammetric conditions while adopting, respectively, second-order and linear prediction modeling for the time-varying exterior orientation parameters. A linear scale factor is utilized in cross-track ranging direction for the horizontal positioning of points on the ground with the SAR imagery. An experiment area, about 12 km by 12 km was selected near Chungli City; the maxi­mum height difference was 40 meters. Flight direction was north and south; an image scale estimated at 1/39,047. Conjugate points on the image were measured with X-VIEW of an HP735 Workstation. The corresponding object-space points were digitized on 1/5,000 or­thophoto maps. At the end,the root mean square errors of independenf check points serve as the indices of our accuracy evaluation.

'Facets stereo vision' is a method for the automatic and simultaneous reconstruction of the geometrical and radiometrical object surface using digital image data. It can perform strict error computation and it has some advantages, e.g. it can integrate different types of data and/or conditions from the image and/or object spaces to the computation of the object reconstruction. Nevertheless, it needs many computation operations so that it is relatively time-consuming. In order to enable a fast object surface reconstruction using this method, new model and algorithm will be briefly presented in this paper. An orthogonal object grey value model was developed from the basic concepts of the theory of wavelets. This new model gives us a very simple submatrix corresponding to the object grey value model pa­rameters in the normal equations. Furthermore, the high resolution of 2 x 2 pixels per height facet is applicable using this model. In addition a new algorithm was proposed to reduce the total computing time in a large area and to enable the object surface reconstruction in larger windows. Generally speaking, larger windows contain more image informations (e.g. image textures) than smaller windows. Test results using aerial images with the image scale 1:4000 and pixel size 15μ m show that the method of facets stereo vision aided by the pro­posed model and algorithm is capable of a high precision (0.2 ~ 0.6 pixel) object surface reconstruction.

Global Positioning System (GPS) has successfully developed and broad-gauge in­to many disciplines' applications in our life. Because GPS have static and kinematic mode positioning function, contrast with conventional work it can saving time and labors consume for forester. The purposes of this study are attempt to apply the GPS navigation ability to guide forester to the forest permanent plot location, and discussion the factors that should be consideration in Taiwan. The results indicated that GPS could guide forester to forest permanent plot location fast and accuracy. In mountain area because the sheltering precluded to acquisition the satellite signal hasn't resolved, it seems need other instruments assistant to work together for insurance the work could perform smoothly.