时空大数据驱动的蓝藻水华预测预警方法研究

Space-time data monitored in the process of cyanobacteria blooms formation is always with characteristics of massive amounts, multidimension, nonlinearity, space-time correlation and space-time heterogeneity. Existing cyanobacteria bloom prediction models driven by mechanism or data are not applicable to the analysis of the space-time big data of cyanobacteria blooms, and the risk warning of.cyanobacteria bloom is lag based on cyanobacteria biomass prediction. Forecasting and early warning methods are lack of accuracy and effectiveness for such complex ecological space-time evolution process driven by multiple influence factors coupling in existing research. Research project here focus on influence factors prediction and risk warning method of cyanobacteria blooms driven by space-time.big data based on improving and combining with methods of space-time series analysis, deep-learning and nonlinear space-time stochastic dynamics analysis. The main works will include 1) to study on key influence factors which influence the formation of cyanobacterial blooms and Biochemical and physical dynamics; 2) to study on characteristic decomposition modeling technique of key influence factor space-time series of cyanobacteria blooms, and propose prediction method of multi-factor space-time big data of cyanobacteria bloom within a deep-learning framework; 3) to study on the mechanism of nonlinear space-time stochastic dynamics of cyanobacteria bloom formation, and propose estimation method for the risk of cyanobacteria bloom in the sense of space-time stochastic bifurcation probability. This project is interdisciplinary. It will provide the important reference for the effective prevention and cure of cyanobacterial blooms.

蓝藻水华形成过程中所监测到的时空数据具有海量、多维、非线性、时空相关性与异质性等特征，现有基于机理或数据驱动模型无法满足蓝藻水华时空大数据的预测要求，且传统基于蓝藻生物量预测的水华风险预警方法存在滞后性，无法对蓝藻水华这一由多种影响因素耦合驱动的复杂生态时空演变过程进行准确、有效地预测预警。本项目通过改进并结合时空序列分析、深度学习及非线性随机动力学理论，研究时空大数据驱动的蓝藻水华关键影响因素预测与风险预警方法。内容包括1）研究蓝藻水华形成过程的关键影响因素及生化物理动力学机理；2）研究蓝藻水华关键影响因素时空序列的特征分解建模原理，提出时空大数据深度学习框架下的蓝藻水华关键影响因素预测方法；3）研究蓝藻水华暴发行为的非线性时空随机动力学机制，提出时空随机分岔概率意义下的水华风险预警方法。项目属于交叉领域课题，成果将提高蓝藻水华预测预警的准确性及有效性，为蓝藻水华的防治提供重要参考依据。

针对现有基于机理或数据驱动模型无法满足蓝藻水华时空大数据的预测要求，且传统基于蓝藻生物量预测的水华风险预警方法存在滞后性，无法对蓝藻水华这一由多种影响因素耦合驱动的复杂生态时空演变过程进行准确、有效地预测预警的问题，本项目通过改进并结合时空序列分析、深度学习及非线性随机动力学理论，研究了时空大数据驱动的蓝藻水华关键影响因素预测与风险预警方法。内容包括1）研究蓝藻水华形成过程的关键影响因素及生化物理动力学机理；2）研究蓝藻水华关键影响因素时空序列的特征分解建模原理，提出时空大数据深度学习框架下的蓝藻水华关键影响因素预测方法；3）研究蓝藻水华暴发行为的非线性时空随机动力学机制，提出时空随机分岔概率意义下的水华风险预警方法。本项目通过深入系统性研究，形成了藻类水华污染“时空大数据预测及水华暴发风险预警”的理论与技术体系，产生了丰富的研究成果，为蓝藻水华智能化和精细化防控提供了理论基础与解决方案，为城市生态文明建设提供有力技术支撑。通过实际河湖数据的实例验证表明，本项目所提的时空大数据深度学习预测方法能够处理海量、多维、高度非线性及时空分布的蓝藻水华预测问题，提高蓝藻水华关键影响因素的预测精度；所提的非线性时空随机动力学水华风险预警方法，能够处理蓝藻水华暴发的突发性风险预警问题，提高蓝藻水华风险预警的有效性。项目属于交叉领域课题，成果能够提高蓝藻水华预测预警的准确性及有效性，为蓝藻水华的防治提供重要参考依据。