Red tide is an abnormal ecological phenomenon, and it severely affects marine ecosystem, economic growth and human health. The mechanism of red tide occurrence has been studied for many years, and a lot of indicators and techniques have been developed. Although these studies have effectively expanded the knowledge of mechanisms of red tide occurrence, the combination effect of water quality factors of different sea areas on red tide occurrence has not received enough attention. In this paper, a water quality factor dissimilarity network model (WDN), which can systematically reflect water quality of different sea areas, is presented. According to this model, a series of WDN networks are constructed with the data collected from monitoring stations of Bohai Sea and North Yellow Sea, and a novel red tide detection approach based on the outlier analysis of node strengths is proposed. Based on the analysis of the results obtained using this approach, the locality, the seasonality and the alternation of red tide occurrences are revealed. The validation from actual red tide events shows that our model and approach perform effectively and contribute to expanding the knowledge of mechanisms of red tide occurrence.