Study On Spectral Information Mining And Application For Oil Spill Remote Sensing Monitoring

Oil spill at sea normally results in negative economic and social consequences over time. It is therefore essential to acquire information about oil film accurately and timely in order to reduce risks. Remote-sensing surveillance constitutes an important component of oil spill monitoring and contingency management system. This dissertation aims to explore the spectral information on sea surface oil spill detecting mainly by optical remote sensing, and to improve its application. The research focuses on oil spill information extraction combined with temporal and spatial information, oil film thickness estimation with multi-platform spectral data, coupling relationship between oil spill and red tide, and the cooperation of surveillance and trajectory under oil spill emergency response. In general, the main results and some of the advantages are summarized as follows:1. To develop a credible and effective response to oil spill in open waters and ice-infested waters, the reflectance spectra of oil film including crude oil, light diesel, and heavy diesel, were collected through field experiments. A portable analytical spectral devices (ASD) was used to collect in-situ reflectance spectra (0.35-2.5μm), with initial consideration of major processes affecting oil spill during the initial period after the spill. The initial spectra library collected comprises three groups of spectrum samples-oil film at clean seawater, ice-infested water, and oil-water emulsions at clean seawater-of which the latter were taken from the U.S. Geological Survey Open-File Report. The spectra features of the collected samples help in detecting oil spill through multi-platform remote-sensing data.2. Relying on the support of the National Key Technologies R&D Program (project number:2006BAC11B01) as well as experiments and practice of oil spill emergencies, this dissertation examines the characteristics and applications of sensors for oil spill monitoring in China. Quick information extraction technology and work flow were set up, and key information distinguishing criteria in the data process and information extraction was proposed. The final remote-sensing monitoring results were introduced in the Dalian Xingang Port oil spill accident, providing decision makers critical guidance in deciding on how best to protect resources and direct cleanup.To improve the accuracy of extracted information from optical imagery, a method based on the fusion of spectral and directional textural information was proposed in this dissertation, taking the PL19-3 oil spill incident as an example. A principal components-based algorithm first extracts all spectrum information on oil on water. A directional gradient algorithm then acquires the edge distribution of oil-contaminated areas. Finally, the proposal method were tested with eight scenes of HJ-CCD data and compared with the conventional method based on singe spectrum using the Jeffries-Matusita separability index. The results show that the introduction of the directional texture analysis is effective for the edge detection of contaminated zone and the identification of thick-thin oil distribution, which is feasible in the oil spill monitoring based on HJ-CCD data.3. Passive optical instruments can provide valuable complementary information by exploiting the effects of oil thickness. A simple radiative transfer model of surface oil film was proposed for multispectral data, and the corresponding optical property of radiance increases with oil thickness in the red to near-infrared band was analyzed. This effect was estimated based on spectrum samples from field experiments. Finally, the oil film identification model for multispectral data was proposed in terms of such effect. Based on the above model and in-situ experiment by an airborne multi-channel scanner, the film thickness distribution was analyzed and verified to be practical.Hyperspectral data can provide intrinsic spectral information on oil film and background seawater, and have the potential for detailed identification of film thickness. In this dissertation, based on a spectrum reconstruction program, a spectrum-matching model for the identification of film thickness was proposed. Under the spectrum reconstruction program, on-site synchronized data of airborne hyperspectral imaging instrument and field ASD device should be obtained first and then reconstruct the spectrum of oil films using the spectral library as reference,. The model can identify the diagnostic organic C-H absorptions at 1.2,1.7, and 2.3μm, and peaks in reflectance to represent the correlation coefficient between the reconstructed spectrum and the spectrum of imagery to be checked, which can be used to determine oil thickness.4. It is challenging to assess the potential impacts of oil spill on the marine ecosystem, but modem satellites are expected to provide consistent and synoptic estimates of several biooptical properties of the surface ocean over multiple years. In this dissertation, using MODIS FLH, SST measurements, and other statistical data, some initial results were presented. The FLH values in the spilled area showed an "anomaly" area:the values are higher than in any July data since 2006 and the highest ever since June 2011 in an area of Bohai. Analyses of ocean environmental data suggest that this anomaly may be attributed to oil spill. The anomaly showed positive correlation with the quantity of oil film, which was relatively stable against CDOM changes and could be used as a better proxy for oil spill detection.5. Using the features of sensor techniques for oil spill surveillance and mathematical models for oil spill trajectory, the cooperation of surveillance and trajectory under oil spill emergency response was presented in this dissertation. Based on a simple integration of equation simulating an oil spill trajectory, the method for remote-sensing application time was described, and an optimization of processing time described in the application time equation was proposed. This cooperation research, which is integrated with remote-sensing and mathematical models, has been found to be a powerful tool for real-time contingency planning in the Dalian Xingang Port oil spill accident. According to these findings, the method allows spills emergencies alerts to make the best decision to choose remote-sensing data, considering effective temporary resolution Te.