摘要
以象山港养殖区监测数据为例,在充分考虑养殖区各个环境因子的基础上,应用贝叶斯网络模型来判断网箱转移时期,以期为开发养殖和环境协调发展的健康养殖模式提供科学参考和决策依据。贝叶斯网是图形表示和概率知识的有机结合,是复杂联合概率分布的图形表示方式,它提供了一种自然的表示因果信息的方法,用来发现数据间的潜在关系。在这个网络中,用节点表示变量,有向边表示变量间的依赖关系,以条件概率表示变量之间影响的程度。凭借因果知识就能在了解一些行为和结果的关系下,有原因地解释可能出现的结果,进行预测和决策。使用贝叶斯网建模已成为表达不确定性知识和进行不确定性推理的强有力工具。海水养殖水环境系统是一个多因素多水平耦合作用的复杂不确定系统。首先通过主成分分析选取水温、pH、DO、COD等10个模型指标,然后结合专家知识和给定数据,利用K2和MCMC算法的优势,采用模型平均思路构造了海水网箱养殖水环境各水质因子与网箱是否移动之间的关系模型,进而确定在一定水质条件下网箱移动的概率,为网箱移动情况进行预测及诊断,结果表明:与网箱移动直接有关的是COD、NO2--N、pH、微生物降解能力、叶绿素a和DO这6个指标,它们是影响网箱移动的主导因素,而水温、无机磷、NO3--N和NH4+-N这4个指标对网箱移动不产生直接影响。以200条数据对模型进行验证表明本文建立的网箱移动预测模型其预测精度达94.23%,Kappa指数为0.868,这说明应用贝叶斯网络模型来判断网箱转移时期是可行的,具有非常好的应用前景。
Now net-cage culture has become the main form of marine fish culture,which promotes the economic development and improves the people＇s living standards,but also results in a series of environmental problems.A large number of pollutants generated by the farming activities worsen the quality of the farming water and the surrounding water,which in turns leads to the reduced quality of cultured fish and the spread of different diseases and mortality of the cultured objects.These problems have seriously affected the health and sustainability of the aquaculture.In order to alleviate the worsening of the breeding environment,the environmental capacity of the net-cages culture has become the focus of many scholars.The reduction of the number of cages will definitely affect the production.If the net-cages could be moved to other areas in the course of aquaculture at a time,the contaminated water at the original farming area might be restored by self-purification and water exchange,which guarantees the quality of aquaculture and protects the environment.But the time when the net-cages should be moved has to be judged on a scientific basis.Based on the observed data from the culture zone,Xiangshangang Bay and taking into full consideration of different environmental factors,Bayesian network model is used to determine the transform time of the net cage,with an aim to provide scientific basis for the development of a healthy culture model and decision-making for the coordinated development of farming and environment.Bayesian network,a combination of a graphical representation and probability,is a graphical representation of the complex joint probability distribution.It provides a natural way to express the causal information and is used to discover the potential relationship among the data.In this network,the node represents a variable and the directed edge represents the dependencies between variables,while the conditional probability represents the extent of influence between variables.With causal knowledge the relationsh
出处
《海洋学研究》
2009年第4期102-107,共6页
Journal of Marine Sciences
基金
国家科技部资助项目(2007DFA21300)
宁波市海洋渔业局资助项目(甬海办2005/331-6)
关键词
网箱养殖
模型
贝叶斯网
MCMC
K2算法
象山港
cage aquaculture model Bayesian network MCMC K2 algorithm Xiangshangang Bay
