Models for identifying significant environmental factors associated with cyanobacterial bloom occurrence and for predicting cyanobacterial blooms

Compared with microscopic indices such as biomass, inverted satellite images can reflect cyanobacterial blooms from a macroscopic perspective, can provide planar information for blooms, and can more definitely reflect the occurrence of visible cyanobacterial blooms. We therefore adopted inverted images (from MODIS imagery) to judge whether cyanobacterial blooms had occurred in a water area at a given time. We constructed two probit models for identifying significant environmental factors related to cyanobacterial bloom occurrence and for short-term forecasts of bloom occurrence. The models used the index of cyanobacterial bloom occurrence as the dependent variable and the predicted variable, respectively, and used three categories (water quality, hydrology, and weather) of monitoring variables as the independent variables (or predictive variables). We used the Hill Dagong water area of Lake Tai in China as a case study of the new methods. The results produced by the identification model are consistent with the general conclusions in this research field indicating the validity of the model. The mean relative error of the forecast model is 13.5%, which is close to or lower than that of two previous models. Compared with the previous models, our forecast model also has advantages in terms of spatial and temporal precision. The new models have both practical applicability and the ability to be generalized and can, therefore, be easily adapted for the prevention, control, and prediction of cyanobacterial blooms in other bodies of water.