Research On Key Technologies Of Spatio-Temporal Analysis And Prediction Of Marine Ecological Environment In Zheiiang Coastal Waters

Zhejiang coastal waters is the intense area between land and sea. With the rapid development of coastal exploration, there are challenges for sustainable development of the marine economy with the problems of water pollution, ecosystem degradation and red tide. Therefore, research on key technologies of spatio-temporal analysis and prediction of marine ecological environment in Zhejiang coastal waters, which aims at developing coastal resources more scientifically and reasonably, has become a major problem we face, which is of practical significance.Due to highly complex and nonlinear characteristics of marine ecological environment in Zhejiang coastal waters, its internal mechanism is still not completely unknown. It is often limited to make analysis and forecast of the ecological environment based on the model of mechanism or traditional mathematical statistical analysis. In recent years, as the formation of three-dimensional monitoring network of buoy, ocean station, investigation ship, satellite remote sensing and so on, which has gradually formed a large scale, complex and high dimensional monitoring data ocean, it becomes a new way that data-driven technique is used to study of marine ecological environment. Under this background, taking marine ecological environment in Zhejiang coastal waters as the research object, this paper makes a systematic study on key technologies of spatio-temporal analysis and prediction of marine ecological environment, using the method of geographic information science and data mining. Our aim is to explore a feasible and complete framework for the comprehensive analysis and decision support of marine ecological environment. The work of research mainly includes the following aspects:(1) The data organization and management of marine ecological environment for analysis and prediction are studied. A triple unified data model fusing ocean phenomenon, spatio-temporal process and evolution mechanism is proposed, achieving a unified organization and storage for all kinds of marine ecological phenomena at multi-scale of space-time and of different evolutionary factors. A multi-stage tandem exchange model and dynamic integration framework for multi-source monitoring data are also designed to implement the exchange and integration within different departments in near real-time, supporting the analysis and prediction dynamicly.(2) Spatial-temporal pattern analysis of marine ecological environment based on EA-SAA-VA ternary hybrid framework is studied. Methods of exploratory analysis, spatial sutocorrelation analysis and variability analysis are discussed to match the complex characteristics of marine ecological environment. A ternary hybrid framework for Spatial-temporal pattern analysis of marine ecological environment is designed. Base on this framework, spatial-temporal pattern analysis is conducted for the fundamental elements of sea eutrophication in Zhejiang coastal waters, from the aspects of distribution characteristic, distribution pattern and variability.(3) Marine ecological environment spatial-temporal association rules are studied based on the effect of spatial-temporal domain and context constraint. Marine ecological environment spatial-temporal mining framework based on association rules is established to explore the correlation between the marine phenomenon and environmental elements. The mining strategy of considering the effects of spatial-temporal domain and context constraint is applied. A case study of red tides in Zhejiang coastal waters has carried on for analysis, to mine the leading factors of red tides, providing the basis for forecasting and early warning of the red tide.(4) Prediction model for red tide based on deep learning is studied. As the highly complex and nonlinear characteristics of the red tide, a dynamic prediction model based on deep learning is put forwarded, which overcomes the prediction limitation of traditional shallow learning such as artificial neural network. In addition, particle swarm optimization is adopted to optimize parameters of the model structure dynamicly, improving the generalization ability of the model in different waters.Based on the above research work, this paper designs and realizes a prototype system of comprehensive analysis and decision support for marine ecological environment in Zhejiang coastal waters, forming an interlocking service chain from data integration management, analysis and prediction. Through the practical application of this system in the bay of Dachen of Taizhou, it shows the feasibility and practicability of this research. This work provides comprehensive analysis and decision support for marine management, which can be a reference to the protection of marine ecological environment in other coastal waters.