Time-series forecasting of chlorophyll-a using Attention-based Encoder-Decoder Recurrent Neural Network

Authors: Zhongyi Wang, Zhejiang University, Zhenhong Du*, Zhejiang University, China, Shuyu Zhang, Zhejiang University
Topics: Coastal and Marine, Environment
Keywords: time-series forecasting, attention-based recurrent neural network, algal blooms, chlorophyll-a
Session Type: Paper
Day: 4/11/2018
Start / End Time: 1:20 PM / 3:00 PM
Room: Astor Ballroom I, Astor, 2nd Floor
Presentation File: No File Uploaded


Eutrophication and algal blooms have become worldwide water quality problems in estuaries and coastal waters. Chlorophyll-a (Chl-a) concentration is commonly used as the main measurement of phytoplankton biomass. Forecasting and understanding the level of Chl-a are helpful to coastal ecosystem management and can be used as an important emergency management measure for potential algae blooms. In our study, a Chl-a prediction model based on an attention-based encoder-decoder recurrent neural network (AEDRNN) is proposed. The AEDRNN model is compared with ARIMA, feed-forward neural network, Elman RNN, LSTM and GRU models. The performance of the proposed model was examined with experimental data collected from 2015 to 2017 in coastal areas of Zhejiang, China. The results showed that the AEDRNN model outperformed ARIMA and the other ANN models and significantly enhanced the accuracy of Chl-a prediction.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login