Examining the Relationship between Phytoplankton Community Structure and Water Quality Measurements in Agricultural Waters: A Machine Learning Application

Phytoplankton community composition has been utilized for water quality assessments of various freshwater sources, but studies are lacking on agricultural irrigation ponds. This work evaluated the performance of the random forest algorithm in estimating phytoplankton community structure from in situ water quality measurements at two agricultural ponds. Sampling was performed between 2017 and 2019 and measurements of three phytoplankton groups (green algae, diatoms, and cyanobacteria) and three sets of water quality parameters (physicochemical, organic constituents, and nutrients) were obtained to train and test mathematical models. Models predicting green algae populations had superior performance to the diatom and cyanobacteria models. Spatial models revealed that water in the ponds’ interior sections had lower root mean square errors (RMSEs) compared to nearshore waters. Furthermore, model performance did not change when input datasets were compounded. Models based on physicochemical parameters, which can be obtained in real time, outperformed models based on organic constituent and nutrient parameters. However, the use of nutrient parameters improved model performance when examining cyanobacteria data at the ordinal level. Overall, the random forest algorithm was useful for predicting major phytoplankton taxonomic groups in agricultural irrigation ponds, and this may help resource managers mitigate the use of cyanobacteria bloom-laden waters in agricultural applications.