Comparative study on algae growth prediction models-A case study of Lake Honghu

根据洪湖2014—2019年水质及藻类监测数据,运用综合营养状态指数法评价了丰、平、枯3个时期的营养状态.在此基础上运用逐步回归分析法确定影响藻类生长的显著因子,并根据不同水量不同营养状态细分9种情形对藻类生长做回归预测分析,同时运用BP神经网络模型对回归预测的结果进行比较验证.结果表明:洪湖丰、平水期以蓝藻门为主,枯水期以硅藻门为主;湖泊的营养状态处于中度富营养与轻度富营养之间.分析各时期藻种生物量与影响因子的相关性,发现丰水期控制因子有水温、CODMn和透明度;平水期和枯水期控制因子有水温、总氮、总磷.以2014—2018年数据逐步回归分析得出枯水期+中营养和枯水期+轻度富营养决定系数较低,其余7种时期决定系数均在0.5以上,说明逐步回归并不适用于所有时期.使用2014—2018年的数据进行神经网络训练和验证,2019年的数据进行预测,比较BP神经网络与逐步回归的均方根误差发现全年预测时BP神经网络效果更好;枯水期+中营养和枯水期+轻度富营养逐步回归效果较好,逐步回归的均方根误差仅为1600～4000;丰水期和平水期2种方法预测效果相当.合理地选择预测模型能为湖泊水华做出预警,控制显著变量可以达到防治水华污染的效果.;According to the monitoring data of water quality and algae in Lake Honghu from 2014 2019, comprehensive trophic level index method was used evaluate nutritional status three periods wet season, normal season dry season. On this basis, stepwise regression analysis determine significant factors affecting growth algae, prediction conducted on nine situations according different amounts status. At same time, BP neural network model is compare results prediction. The showed that Cyanophyta dominant species Honghu, Bacillariophyta lake between moderate mild eutrophication. correlation number algal cells temperature, total nitrogen, phosphorus, CODMn other influencing each period analyzed. control were transparency. nitrogen phosphorus seasons. based 2018 determination coefficients + medium nutrition eutrophication low, seven all above 0.5, indicating not suitable for periods. training verification. 2019 root mean square error compared. It found better whole year 1600-4000. Stepwise effects two methods are similar. Selecting reasonable models can make early warnings bloom help variables achieve effect preventing controlling pollution.