Evapotranspiration is an important factor in hydrology cycle and regional climate. For example, rice paddy field absorbs and stores huge amount of energy, which is released to the air through evapotranspiration. As a result, ambient temperature is adjusted and the strength of the heat island effect is mitigated in the urbanization areas. The aim of the current study is to develop an evapotranspiration retrieval algorithm for rice paddy field by remote sensing technique in order to provide long-term and large-area observation and to investigate its role in heat island effect. The proposed retrieval algorithm is based on energy balance at the land-air interface. Surface heat fluxes are estimated by remote sensing data with in situ surface meteorological measurements. The initial values of latent heat and sensible heat fluxes are obtained using an unstable atmospheric correction method. By iterating the energy balance equation, radiation and evaporation controlled lines are determined, and hence, the net available energy is redistributed into latent and sensible heat fluxes, respectively. In this study, both airborne and satellite imageries are utilized. The former imagery covers a small area with high spatial resolution. The satellite imagery acquired by MODIS is utilized to estimate the evapotranspiration over the entire Taiwan. The proposed retrieval algorithm has provided a fairly good estimate of evaporation fraction by airborne imagery with bias of 6.8% when compared with the in situ measurements, while performed poorly with the satellite imagery. To improve the evapotranspiration estimate by satellite imagery, the Paulson and Dyer’s corrections to unstable atmosphere were modified. The correlation coefficients for latent heat flux and sensible heat flux and the corresponding in situ observations were found to be 0.66 and 0.76, respectively. The root mean square errors for latent heat flux and sensible heat flux were 97.81 Wm-2 and 124.33 Wm-2, respectively.

The spectral characteristics of the vegetation canopy, including four dominant vegetation types, Leucaena leucocephala, Vitex negundo, Acacia confuse and Pandanus odoratissimus, were measured from November 2006 to October 2007 using a ground handheld spectroradiometer in the Kenting National Park of Taiwan. The purpose of this paper was conducted to evaluate the potential of multispectral and hyperspectral reflectance in the visible to near infrared spectral range, for discriminating four dominant vegetation types every sampling date. Differences observed in the multispectral analysis showed that four dominant vegetation types were possible to distinguish due to phenology in March 2007 (dry season) and July 2007 (wet season), and in hyperspectral were statistically significant in March 2007, which facilitates their discrimination. Our results suggest that mapping four dominant vegetation types is feasible using multi-temporal imagery, and hyperspectral sensors, taking into account the plant phenological stages.

GPS provides all-weather observations, and highly accurate positions. With a GPS model using carrier-phase observations containing integer-valued and real-valued unknown parameters, this paper presents a near-real-time data processing technique based on Bayesian statistics. The Bayesian approach takes the maximum–likelihood posterior solution of positioning parameters as the estimator with highest posterior density function of ambiguity. Moreover, for the sake of the fact that restricting the prior distribution of ambiguities is not consistent with the real data in Bayesian view, a theory of mixture model based on Bayesian inference was advanced and the index for decision-making was derived. The index was used here to determine whether the ambiguity search space is expanded or not. The main aims of this research are to perform the GPS positioning by using the Bayesian approach with the index for decision-making and determine the confidence regions for positioning parameter by using a Monte Carlo method based on the Bayes estimation and covariance matrix. The experimental results show that the correct solution can be obtained and the accuracy of positioning results can be improved in real time and automatically by using the Bayesian approach. Furthermore, the confidence regions for GPS positioning parameter can be determined by using the Monte Carlo method.

Data Mining can be applied to the satellite images and can be combined with machine learning theory. This technology is used to discover knowledge from large mounts of data. CART is a kind of methods to acquire the knowledge of rice paddy classification. Although this method needs to select the samples, it can reduce the effects which are caused by the selection of samples on the results of classification, for example, the narrow area and the uneven characteristics. The machine learning method, Boosting, can solve the problem of characteristic changes that are caused by the re-selection of samples. This method can increase the accuracy of the classification. The samples which have the low accuracy of classification are organized by Boosting Method. Boosting Method analyzes the samples and acquires the samples that have the higher classification accuracy in order to avoid the re-selection of samples. And, Boosting method co-operates with CART classification method to improve the accuracy of classification. According to the result of this study, Boosting Method can improve the maximum likelihood method and the CART method on the classification accuracy which rise 5% and 3% individually.

The indices of evaluating the performance of GNSS are composed of satellite availability, positioning accuracy, integrity and reliability. The future GNSS including stand alone modernized GPS, Galileo or GPS/Galileo integrated system will support more satellites, more observable frequencies and better signal quality comparing to current GPS. Therefore, the positioning accuracy, reliability and efficiency can further be enhanced. For surveying communities, those improvements are very critical. Since the compatibility between modernized GPS and Galileo is enhanced in the design phase and both of them will share two frequencies (L1/E1, L5/E5a), therefore, the receiver supporting dual-systems and dual-frequencies will dominate the geodetic grade GNSS receiver market in the future based on the accuracy enhancement and cost anticipated. This study aims at using the GNSS software simulator developed in the department of Geomatics, NCKU to analyze the three dimensional positioning accuracy, internal reliability (MDB) and external reliability (MDE) in Taiwan region based on three different scenarios including stand along modernized GPS, Galileo and GPS/Galileo integrated system.

The point cloud obtained from airborne LiDAR is composed of ground points and non-ground points. The density of ground points is an important index for characterizing the quality of DEM. When ground point density is high and distributed evenly, topography can be presented with high fidelity. There are a number of factors which may influence the ground point density. In addition to the density of the original point cloud, penetration rate is another essential factor. Penetration rate is influenced by land-cover types, incidence angle, flying height, etc. This study investigates the influence of incidence angle on penetration rate. Partitioning a flight line into five segments along the flight direction, the penetration rates are computed. It is shown that the penetration rate is higher with smaller incidence angle. The central segment of the flight line is 1.6 times of the right most and left most sections, and 1.3 times penetration rate of next right and next left sections.