180904_A novel single-parameter approach for forecasting algal blooms

1. Title, Journal and Authors

- Title: A novel single-parameter approach for forecasting algal blooms

- Journal : Water research

- Authors : Xi Xiaoa,b,c, Junyu Hea,b,c, Haomin Huangb,c,d, Todd R.Millere, George Christakosa,

Elke S. Reichwaldtf, Anas Ghadouanif, Shengpan Ling, Xinhua Xub, Jiyan Shib,c,*

- Affiliation :

a: Ocean College, Zhejiang University, Zhoushan, PR China

b: College of Environmental & Resource Sciences, Zhejiang University, Hangzhou, PR China

c: Key Laboratory for Water Pollution Control and Environmental Safety, Zhejiang Province, PR China

d: School of Environment and Energy, South China University of Technology, Guangzhou, PR China

e: Joseph J. Zilber School of Public Health, University of Wisconsin - Milwaukee, USA

f: Aquatic Ecology and Ecosystem Studies, School of Civil, Environmental and Mining Engineering, The University of Western Australia, Crawley, WA, Australia

g: Department of Integrative Biology, Michigan State University, East Lansing, MI, USA


2. Summary

- This paper presents a new approach to forecast algal blooms by combining wavelet analysis with artificial neural networks (ANN).

- The dataset used in this paper was time series data based on daily online monitoring datasets of algal density.

- The methodology suggested in this paper is a single parameter approach. The details of the approach are following as:

  • The original data was decomposed in to four approximation series (1 low frequency and 3 high frequency) with discrete wavelet transformation   
  • The decomposed data were applied to each different ANN algorithms.
  • The final prediction result is calculated as the sum of the outputs of each ANN model.
  • The optimal ANN structure is selected based on the result of repeated simulation with ANN models of different structures.

- Compared to other methods, including ARIMA and ANN models, the suggested model here gave a better performance.


3. Application to research

- The methodology to decompose the time series data can be applied to predict water quality with time series data.

- However, since the proposed method requires time series data measured in a short period, it can’t be applied in Korea where the monitoring period is different.

- Therefore, future research should focus on advancing the model to be used with datasets with different and low monitoring frequencies.


4. Contact

Heewon Jeong (Ph.D. program)

Environmental Systems Engineering Lab.

School of Environmental Science & Engineering

Gwangju Institute of Science and Technology

1 Oryong-dong Buk-gu Gwangju, 500-712, Korea

Email :  gua01114@gist.ac.kr

첨부 (0)