Modeling And Forecasting Algal Blooms Based On The Neural Network Approach

This paper focuses on the modeling and forecasting algal blooms using neural networks. While we want to further our knowledge with neural network models incorporated with sensitivity analysis, we also want to realize some forecasting models with good performance.Chapter 1 introduced first the eutrophication, algal blooms problem, and the purpose of modeling algal blooms. Then the previous researches were summarized, and the applications of neural networks in this field were introduced. In the end of the chapter, the main research methods and research efforts in this paper were outlined.The empirical research of modeling algal blooms in Lake Dianchi by BP neural network was outlined in Chapter 2. Lake Dianchi is a shallow and turbid lake, located in Southwest China. Since 1985, Lake Dianchi has experienced severe cyanabacterial blooms (dominated by Microcystis spp.). In extreme cases, the algal cell densities have exceeded three billion cells per liter. To predict and elucidate the population dynamics of Microcystis spp. in Lake Dianchi, a neural network based model was developed. The correlation coefficient (R2) between the predicted algal concentrations by the model and the observed values was 0.911. Sensitivity analysis was performed to clarify the algal dynamics to the changes of environmental factors. The results of sensitivity analysis based on the trained neural network model suggested that small increases in pH could cause significant reduction of algal abundance. And further researches revealed that the response of Microcystis spp. biomass to pH increase depended on the algal biomass itself and the pH level. With moderate or low level of algal biomass and pH in Lake Dianchi, pH increase was likely to lead to an increase of Microcystis spp. biomass. Otherwise, when pH and algal biomass were high, pH increase was likely to reduce the Microcystis spp. biomass. It is concluded that the extremely high concentration of algal population and high pH could explain the distinctive response of Microcystis spp. population to +1 SD (standard deviation) pH increase in Lake Dianchi. Further, another hypothesis presented in the paper was that the higher abundance of algal population and the higher pH, an increase of pH would be more likely to have a negative (or strong negative) influence on algal growth in an extremely high eutrophicated water such as Lake Dianchi. And the paper also elucidated the algal dynamics to changes of other environmental factors. One SD (standard deviation) increase of water temperature (WT) had strongest positive relationship with Microcystis spp. biomass. Chemical oxygen demand (COD) and total phosphorus (TP) had strong positive effect on Microcystis spp. abundance while total nitrogen (TN), biological oxygen demand in five days (BOD5), and dissolved oxygen had only weak relationship with Microcystis spp. concentration. And transparency (Tr) had moderate positive relationship with Microcystis spp. concentration.In Chapter 3, radial basis function (RBF) neural network were applied for modelling the abundance of cyanobacteria. The trained RBF neural network could predict with high accuracy the population of the two bloom forming algal taxa, Nostocales spp. and Anabaena spp., in River Darling, Australia. To elucidate the dynamics of algal population for both Nostocales spp. and Anabaena spp. Sensitivity analysis was performed. Some hypothesis could be obtained from the results of sensitivity analysis. First, total kinetic nitrogen had a very strong influence on the abundance of the two algal taxa. Second, electrical conductivity had a very strong negative relation with the population of the two algal species. In other words, more dissolved salts or ions in the water, the algal population are more likely to decrease. Third, flow was identified as one predominant factor on algal blooms after scatter plot clarified the fact that high flow could reduce significantly the algal biomass for both Nostocales spp. and Anabaena spp. in River Darling. Other variables such as turbidity, colour, and pH were less important in determining the abundance and succession of the algal blooms.Chapter 4 demonstrated first the nonstationarity of algal population behaviors by analyzing the historical populations of Nostocales spp. in the River Darling, Australia. Freshwater ecosystems were more likely to be nonstationary, instead of stationary. And nonstionarity implies that only the near past behaviors could forecast the near future for the system. However, nonstionarity was not considered seriously in previous research efforts for modeling and predicting algal population behaviors. Therefore the moving window technique was incorporated with radial basis function neural network (RBFNN) approach to deal with nonstationarity when modeling and forecasting the population behaviors of Nostocales spp. in the River Darling. The results showed that the RBFNN model could predict the timing and magnitude of algal blooms of Nostocales spp. with high accuracy. Moreover, a combined model based on individual RBFNN models was implemented, which showed superiority over the individual RBFNN models. Hence, the combined model was recommended for the modeling and forecasting the phytoplankton populations, especially for the forecasting.The last chapter concluded the paper and made some comments on further researches.