Prediction of algal blooms via data-driven machine learning models : an evaluation using data from a well-monitored mesotrophic lake

• With increasing lake monitoring data, data-drivenmachine learning (ML) models might be able to capture thecomplex algal bloom dynamics that cannot be completely described in process-based (PB) models. We applied two MLmodels, the gradient boost regressor (GBR) and long shortterm memory (LSTM) network, to predict algal blooms andseasonal changes in algal chlorophyll concentrations (Chl) ina mesotrophic lake. Three predictive workflows were tested,one based solely on available measurements and the othersapplying a two-step approach, first estimating lake nutrientsthat have limited observations and then predicting Chl usingobserved and pre-generated environmental factors. The thirdworkflow was developed using hydrodynamic data derivedfrom a PB model as additional training features in the twostep ML approach. The performance of the ML models wassuperior to a PB model in predicting nutrients and Chl. Thehybrid model further improved the prediction of the timingand magnitude of algal blooms. A data sparsity test based onshuffling the order of training and testing years showed theaccuracy of ML models decreased with increasing sampleinterval, and model performance varied with training–testingyear combinations.

Ämnesord

NATURVETENSKAP  -- Geovetenskap och miljövetenskap -- Oceanografi, hydrologi och vattenresurser (hsv//swe)

NATURAL SCIENCES  -- Earth and Related Environmental Sciences -- Oceanography, Hydrology and Water Resources (hsv//eng)

Nyckelord

Biologi med inriktning mot limnologi

Biology with specialization in Limnology

Publikations- och innehållstyp

ref (ämneskategori)

art (ämneskategori)