Prediction of Chlorophyll-a content using hybrid model of least squares support vector regression and radial basis function neural networks

Eutrophication has become a serious environment problem in many parts of the world and Chlorophyll-a concentration is one of the important parameters for the characterization of water quality, which reflects the degree of eutrophication and algae content in the water body. So establishing a forecasting model to predict the chlorophyll-a concentration in evaluation of eutrophication become more urgent. In this paper, a hybrid model of least squares support vector regression optimized by improved particle swarm optimization and radial basis function neural networks (IPSO-LSSVR-RBFNN) was proposed, which effectively modifying the forecasting accuracy by extracting the useful information in the error term of the traditional methods. A real monthly dataset that collected from a typical reservoir in China during 2010–2012 and two public datasets were used to evaluate the performance of the proposed hybrid model. From the experiment results, we can see that the proposed model of IPSO-LSSVR-RBFNN achieve a higher accuracy rate compared with other models.

I. Introduction

Eutrophication has been one of the most significant causes of water quality impairment and a major threat to the health of estuarine, land waters and marine ecosystems for more than four decades [1]–[2]. We can assess the degree of eutrophication by chlorophyll-a concentration, because the trophic state of the water body is related strongly to the chlorophyll-a concentration and the chlorophyll-a concentration effects the aquatic phytoplankton biomass and abundance [3]–[4]. So estimating chlorophyll-a concentration is one of most important content for evaluating monitoring eutrophication and aquatic ecosystems.