The paper analyses seasonal changes of spectral characteristics values of satellite remote sensing when the forest fire was
happening during the season of forest fire prevention. Based on the seasonal change characteristics and spectral
characters of polar satellite bands, the forest hotspot was identified by using several infrared bands, and based on the
combination-spectral expression of mid-infrared and thermal-infrared channel, the spectral threshold would be
established for monitoring forest fire in different forest growing season in subtropical monsoon zone.
The study selected 9 factors, average maximum temperature, average temperature, average precipitation, average the
longest days of continuous drought and average wind speed during fire prevention period, vegetation type, altitude, slope
and aspect as the index of forest fire danger district division, which has taken the features of Lushan Mountain's forest
fire history into consideration, then assigned subjective weights to each factor according to their sensitivity to fire or
their fire-inducing capability. By remote sensing and GIS, vegetation information layer were gotten from Landsat TM
image and DEM with a scale of 1:50000 was abstracted from the digital scanned relief map. Topography info. (elevation,
slope, aspect) layers could be gotten after that. A climate resource databank that contained the data from the stations of
Lushan Mountain and other nearby 7 stations was built up and extrapolated through the way of grid extrapolation in
order to make the distribution map of climate resource. Finally synthetical district division maps were made by weighing
and integrating all the single factor special layers,and the study area were divided into three forest fire danger district,
include special fire danger district, I-fire danger district and II-fire danger district. It could be used as a basis for
developing a forest fire prevention system, preparing the annual investment plan, allocating reasonably the investment of
fire prevention, developing the program of forest fire prevention and handle, setting up forest fire brigade, leaders'
decisions on forest fire prevention work.
In recent years, the impacts of natural disaster are more and more severe on coastal lowland areas. Aim to the threat of climate change and sea level rise, the natural disaster reduction in coastal lowland areas is paid highly attention. Based on a number of literatures, the paper summarizes the categories and characteristic of natural disasters emerging in coastal lowland areas, such as windstorm and storm surge, hurricanes and hurricane winds, tsunamis and floods, and analyzes the most devastating natural disasters in coastal lowland in the world 2005. The paper also summarizes the effects of typhoons on the coastal lowland areas of China in 2005 and review to analyze the natural disaster mitigation measures and its researches. At last, the paper discusses the vulnerability assessment and response strategies.
Based on Landsat TM data combined with practical investigation information obtained using Global Positioning Systems (GPS), we created a training field of land use classification. Using the methods of spectral distance analysis, we analyzed spectral signature value of different training fields in TM3, TM4, TM5 and TM7 band, and compared these with the standard deviation analysis. Based on these results, we selected the best spectral bands for classification and created remote sensing interpretation marks of land use classification. Supervising classification was used with the image classification of TM and the maximum likelihood was used for parametric rule of supervised classification. We applied the method of spectral signature analysis to the individual study of land use classification of Poyang Lake region. The land use was classified into 9 classes: paddy field, non-irrigated farmland, forestland, grassland, water area, lake beach, grass beach, sandy land and residential area. Based on the data of GPS investigation, we assessed the classification accuracy. Result indicated that classification accuracy reached 91.43% and the classification effect was better than the common supervised classifying and unsupervised classifying.