Studies On Hyperspectral Characteristic Of Soil, And Land Use Investigation By Remote Sensing In Coastal Region

The coastal soils are the main land resource in support for Zhejiang province. The exploitation, utilization and its changes of the coastal soils not only relate with the sustained development of local economy, but also impact the subsequent exploitation and available land use plan. Thus, timely and reliable information with regard to the nature, extent and physic-chemical characteristics of the coastal soils is essential.Combined with a project supported by National Natural Science Foundation of China, a large, reclaimed with long time and representative area in Zhejiang province as the study area (located in Shangyu city and south of Hangzhou gulf), this thesis focused on the studies of hyperspectral characteristic of the coastal soils, estimating reclamation level of soil using laboratory hyperspectra and land use investigation by remote sensing with multi-type and multi-hiberarchy remote sensing data.The main study contents and corresponding results were as follow:1. After the natural drying and sieving of soil samples collected from field, physic-chemical property analysis and hyperspectral measurement were carried out in laboratory. The results showed that coastal soils had low organic matter, high electrical conductivity and sand content, and some soil properties changed with reclamation years. The Person correlation analysis showed that there exist good relationships between nine absorption bands and selected soil properties, such as soil sand and silt content.2. The study area was divided into four sub-zones with different historical years of reclamation. Stepwise Discrimination Analysis (SDA) was applied to estimate the reclamation levels of coastal soil using the nine absorption bands. The results showed that coastal soils with different reclamation years could be classified with satisfied accuracy using selected absorption bands from hyperspectra.3. Land use investigation by remote sensing were implemented with Multi-temporal Landsat TM and new generation Enhanced Thematic Mapper Plus (ETM+). First, dykes were extracted using a line feature enhanced and resolution sharpen ETM+ image. Basing on the extracted dykes, six sub-zones were defined according to different reclamation year, in order to facilitate subsequent classification of various land use types. Then, different classifiers were applied to the land useclassification for different sub-zones. Finally, a detailed land use map based on a modified land use classification system was produced. The results showed that the dominant agricultural land use type of every sub-zone was changed with the reclamation years, from barren land to aqua-farm pond, to cotton field, and to paddy field and orchard.4. In the condition of no visible-light images, only multi-temporal ERS-2 PRI SAR images were applied for land use investigation. After the SAR pre-processing, pseudo-color synthesizing of three-temporal SAR images, extracting dykes and partitioning the study area into six sub-zones according to the reclamation years, an ISODATA unsupervised classifier and a BP neural network classifier were used for the land use classification for different sub-zones. The total classification accuracy was up to 77.34%. The study results showed that SAR remote sensing, instead of multi-optical remote sensing, has presented great potential on the agricultural land use investigation in such a coastal region.