Dealing with the uncertainty and randomness in natural biosystems to reduce the risk pattern

Abstract

Statistical Models are good approach to deal with the coupling effects of governing parameters in algal bloom growth. In this thesis we have used two different statistical models which are Multiple linear regression model and Artificial neural network (ANN). ANN has significant effect on the development of algal bloom. We have simulated the growth of algal bloom under environmental factors that can lead to bloom pattern and validate our data with two different datasets which are 2 reservoirs of Moncton city and Mattatall lake Nova scotia (Canada) over the period of 3 year 2015-2018. In the classical statistical analysis, we have the Multiple Linear Regression (MLR) which can predict dependent variables (Chlorophyll-a and Phycocyanin) based on multiple input variables. multiple linear regression (MLR) model have been used and investigated to predict the chlorophyll-a (Chl-a) and Phycocyanin (PC) concentration in water of lake reservoir. The other model we used for this thesis is the type of supervised learning classifiers, named Artificial Neural Network (ANN), which has already been developed as a model utilizing the learning process algorithms and considered as an alternative tool to data processing in ecological sciences. An important step in the development of ANN is to select the input data which have the most significant impact on the model’s performance (ASCE, 2000). Models for Harmful Algal Bloom (HAB) involves the development of a functional relationship between the algal growth and environmental parameters. The ability to ANN technique to capture the behavior of non-linear complex data (Elangasinghe et al., 2014). The models type ANN are now being widely used because they can learn (training), non-linear and input relationship by using data and parameters. ANN has many characteristics like generalization, adaptation, etc. and multilayer model trained with backward propagation is the most used in modeling and optimization, prediction, function approximation (Madic & Radovanovic, 2011). The simulation results showed that ANN model is quite suitable for capturing the non-linearity of the relationship between variables and helping to determine a predictive solution for algal bloom. Results also suggested that the ANN model can be effectively used for a predictive scenario of algal bloom occurrence via the predicted values of Chl-a and Phycocyanin(PC) in terms of various input parameters like Nitrates, pH, Temperature, Dissolved Oxygen (DO) especially when we consider more environmental factors such as wind factors and light intensity in the future. MLR model did well It is essential to deal with various types of randomness and uncertainties that can improve ability of researcher to predict about different 3 needs of natural organisms in biosystem in a mannered way. Additionally, it can help to reduce market challenges and thus improve natural balance in a mannered way. Therefore, it will improve understanding of researcher about different concepts namely ANN model, algal growth, uncertainties and randomness along with natural biosystem. In addition to that, it can help to develop equal opportunities for different living organisms to share their needs and improve cooperation among them to live in a single population. The attempt to predict the HAB occurrence and proliferation under complex context of environmental conditions (nutrient, light, meteorological factors, etc.) led to many indexes for the estimation of HAB risks based on chemical components such as micronutrient factors of the waterbody that contributed significantly to algal growth. The research throughout this thesis showed an important insight: bloom patterns can only be explained and predicted by coupling effects of all involved parameters. To combine all effects of all possible parameters, only a statistical model can help us to deal with this complex issue. This approach can be a powerful tool to deal with the coupling effects of governing parameters in the bloom occurrence issue. Therefore, the more numerous factors that are combined, the better simulation and hence prediction are. Our main target is to define the declining coupling factor(s) for the pattern HAB which cannot be based on just one or two parameters, but a combination of many involving ones. The nutrients Nitrates/Phosphates, we suggested herein seems to have many practical aspects for fresh water to evaluate the algal instability state leading to the onset of bloom patterns. This research work of pattern recognition theory and framework would contribute significantly to many scientific and engineering fields, including the toxin blue-green algal bloom proliferation, thermal and water stressed areas in agriculture.