Application of hurdle Poisson model to predict the abundance of toxic cyanobacteria Microcystis in reservoirs

Abstract

The blooming of toxic cyanobacteria Microcystis in eutrophicated reservoirs causes serious difficulties for water supply worldwide. For the appropriate management of such reservoirs, a prediction model of toxic cyanobacteria Microcystis can be a useful tool. Therefore, this study aims to develop a Bayesian hurdle Poisson model for statistical prediction of toxic Microcystis from only two predictors, air temperature and trophic state index (TSI) calculated from chlorophyll-a. The gene copy number of the mcyB gene was used as a surrogate of toxic Microcystis cell density. The data on mcyB gene and chlorophyll-a were collected from 20 reservoirs in Nagasaki Prefecture (Japan). The daily average air temperature was downloaded from the local meteorological stations and a mean for 30 days before sampling date was calculated. The results showed that higher temperature and larger TSI accelerate the growth of toxic Microcystis. Furthermore, this model successfully predicted mcyB gene copy number as a surrogate of toxic Microcystis cell density for different conditions of air temperature and TSI with sufficient accuracy. Therefore, the proposed model has the potential to be a useful prediction tool for toxic cyanobacteria Microcystis in the integrated management of reservoirs.