| Cyanobacteria blooms mostly occur in eutrophic waters.When the climate and hydrological conditions are favorable for the growth of algae,cyanobacteria outbreak propagation and aggregation,and reach a certain concentration,which is a natural phenomenon caused by a combination of multiple reasons.There are many factors affecting the occurrence of cyanobacteria blooms:water environment,meteorology,topography,exogenous pollution,human activities and the growth of algae itself.The outbreak mechanism is not clear.For nonlinear,multi-factor,and randomly occurring cyanobacteria blooms,a variety of data-driven non-mechanical early warning research methods have been developed.This paper chooses BP artificial neural network,which has become more and more mature and based on its non-linear approximation function,which can directly describe the dynamics of the objects with uncertain influencing factors,to carry out the research and analysis of cyanobacteria blooms prediction and early warning in Sanshiliujiao Lake of Pingtan,and it has important practical significance to protect people’s drinking water safety.In this paper,through the combination of field observation sampling,laboratory monitoring analysis and model analysis,the influencing factors of cyanobacterial blooms in the 36-foot lake were studied,and the corresponding models were constructed to warn the blooms of water blooms.The time period of the study is divided into two periods,from January 1,2016 to May 30,2017 and June 2018 to July 2018 and November to December,from January 1,2016 to 2017.The online monitoring data obtained during the period of May 30 was used for the determination of the impact factors of the blooms and the construction of the early warning model.The data from June 2018 to July 2018 and November to December were used for the water cyanobacteria.Chinese analysis and verification of the model.According to the characteristics of reservoir water body and related literature reports,the BP artificial neural network was identified as an early warning model for cyanobacterial blooms.The spatial geometric mean was used to determine the threshold of chlorophyll a in the cyanobacterial blooms in the lake area.Through the analysis of the water body analysis and early warning model of the lake area,the following conclusions are drawn:(1)There are spatial differences in nutrients in the waters of the lake area,but the overall nitrogen and phosphorus nutrients in the water are rich.The nutrient salt is not the limiting factor of the blooms,and the cyanobacterial bloom in the lake area is the distribution of phytoplankton affected by the wind direction..(2)Through correlation analysis and principal component analysis(PCA),the environmental factors affecting the species and biomass of phytoplankton in the 36-foot lake were analyzed.The results showed:temperature,wind speed,sunshine hours,water temperature and conductivity.The dissolved oxygen has an environmental factor with a high coefficient of the main component.(3)BP artificial neural network is used to input meteorological and water quality factors as input,and chlorophyll a concentration is used as output to optimize the calculation.Finally,the optimal model input factor combination is determined as temperature,wind direction,water temperature and conductivity.The model output of the combination is obtained.The fitness(R~2)reached 0.97,and the standard deviation ratio(RSR)and root mean square error(RMSE)were 0.16 and 0.04?g/L,respectively.(4)Considering the spatial difference of the lake area,calculate the spatial geometric mean of the output factor chlorophyll a,and set the chlorophyll a threshold of 35.5?g/L for the Sanshiliujiao Lake water bloom warning model.(5)The model of Fuqing Dongzhang Reservoir was selected to calculate the applicability of the model,and the verification accuracy reached more than 75%.The results of this study provide a reference for the cyanobacteria bloom warning in water source. |
PDF Full Text Request