Table 1 ANN-based models for SST prediction
Scholars | Model | Compared model | Case studies | Source of data | Period of data | Performance indices | Inputs |
|---|---|---|---|---|---|---|---|
(Corchado and Fyfe 1999) | FIR (ANN) | LR, ARIMA | UK Falkland Islands | Vessels | 1995 | STD, average error (AE) | Temperature of the water mass at a fixed depth |
(Corchado et al. 2001) | IBR-RBF | RBF, FIR, ARIMA, CBR, LR | UK Falkland Islands | Satellite images, Plymouth Marine Laboratory | 1997 | AE, absolute error (AbsEr) | Temperature of the water mass at a fixed depth |
(Malmgren et al. 2001) | BPNN | SIMMAX, RAM, MAT, IKTF, ANND | Atlantic Ocean, Caribbean Sea | Field data (Pflaumann et al. 1996) | 1991–1992 1992–1994 | RMSEP (RMSE) | Geography info, fossil data (plankton) |
(Peyron and Vernal 2001) | BPNN | MAT | Bering Sea, Northern North Atlantic, Arctic Sea | Field data | 1997–2001 | CC, RMSE | Sediment fossil samples |
(Ali et al. 2004) | MLP-BP |  | Arabian Sea | Buoys | October1984–October 1985 | R, RMSE | Net surface heat flux, net radiation, wind stress, SST, and dynamic height |
(Chen et al. 2005) | BPNN | IKM, SIMMAX, RAM, MAT | Western Pacific | MARGO Database Proj ects/ MARGO/data.html | 1980–2000 | RMSE, R2 | Sediment fossil samples (plankton) |
(Garcia-Gorriz and Garcia-Sanchez 2007) | FFNN |  | Western Mediterranean Sea. | ECMWF, HadiSST, MODIS, NCAR | 1960–2005 | RMSE | Mean sea level pressure, 10-m zonal and meridional wind components, 2-m air temperature, 2-m dewpoint temperature, and total cloud cover |
(Gupta and Malmgren 2009) | ANN | IKTF, WAPLS, MAT, ML | Pacific Ocean, Antarctic Ocean | National Geophysical Data Center, CLIMAP | 1900–2000 | R, RMSE | Surface sediments |
(Bhaskaran et al. 2010) | MLP | LR | Indian Ocean | WOA94 (http://iridl.ldeo.columbia.edu), OSD, PFL, NOAA | 2001–2006 | R, MSE, AE | depth of samples, patio-temporal parameters of SST |
(Mahongo and Deo 2013) | FFNN | GRNN, RBF, ARIMAX | Western Indian Ocean | HadISST | 1874–2010 | R, MAE, RMSE, MSRE, CE, Nash-Sutcliffe index, MaxAE, MAPE, R2 | Lagged SST |
(Piotrowski et al. 2015) | MLP | ANFIS, K-Nearest Neighbor, Wavelet ANN | Biala Tarnowska and Suprasl catchments, Poland | Gauging Station | 1983–2000 | MSE | Air temperature (mean, max, min), river runoff, declination of the sun |
(Patil et al. 2016) | Wavelet ANN | Â | Arabian Sea, Bay of Bengal, West of Indian Ocean, East of Indian Ocean, African Coast, South of the Indian Ocean | INCOIS, NOAA (OISSTv2 http://www.ncdc.noaa.gov/sst/description.php), CCCma (http://www.cccma.ec.gc.ca/data/cgcm4/CanESM2/index.shtml), Hadley SST | 1870-2014 | R, RMSE, MAE | Lagged SST |
(Modaresi et al. 2016) | GRNN | Â | Karkhe subbasin | NOAA (OISSTv2) http:// www.esrl.noaa.gov/psd) | 1982-2007 | R, RMSE, Nash- Sutcliffe index | Persian Gulf SST, Mediterranean Sea SST |
(Salles et al. 2016) | ANN | ARIMA, Random Walk | Tropical Atlantic Ocean | PIRATA Buoys | January 1998-March 2015 | MSE, NMSE, MAPE, SMAPE | Lagged SST |
(Liao et al. 2017) | RBF-NN | Reynolds optimum interpolation | Pacific Ocean | AVHRR SST L2p-OISST | 2014 | R, RMSE, MAE, SN | Lagged SST |
(Patil and Deo, 2017) | WNN | Wavelet Auto-Regression (WAR) | Indian Ocean | HADISST dataset, OISST, RAMA buoy, ROMS | 1870–2015 | R, RMSE, MAE | Wavelet Transform (WT) of SST |
(Patil and Deo 2017) | WNN | ROMS, persistent model | Indian Ocean | AVHRR, NOAA (http://www.esrl.noaa.gov/psd/) AMSR (buoys), RAMA, ROMS | 1981–2015 | R, RMSE, MAE | Wavelet Transform (WT) of SST |
(Zhang et al. 2017) | LSTM (Recurrent Neural Network) Deep | MLP, SVR | China | NOAA/OAR/ESRL PSD | 1981–2016 | RMSE | Lagged SST |
(Yang et al. 2017) | CFCC-LSTM(RNN) | SVM, SVR, FC-LSTM | Bohai Sea China Ocean | AVHRR | 2007–2012 | RMSE, ACC | Spatio-temporal, SST |
(Guo et al. 2017) | SOM |  | Pacific | HadISST, CRU, GPCC, NOAA/OAR/ESRL, ECMWF, | 1901–2010 |  | Lagged SST |
(Wang and Deng 2017) | ANN |  | USA | MODIS (https://ladsweb.nascom.nasa.gov/) USGS (http://waterdata.usgs.gov/la/nwis/dv/?referred_module=sw) | 2007–2014 | R, RMSE, STB, AM, Median | Normalized water-leaving reflectance for thermal infrared band 31 and 32 |
(Aparna et al. 2018) | FFNN-Quasi Newton BPNN algorithm |  | Northeastern Arabian Sea | MODIS | 2013–2015 | RMSE, MAE | Lagged SSTA |
(Foroozand et al. 2018) | ANN | Ensemble Entropy (bagging), MLR , Bayesian Neural Network | Tropical Pacific Sea | ECMWF (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era-interim | 1979–2017 | RMSE, MSE | Lagged SST |
(Liu et al. 2018) | Temporal Dependence-Based Long Short-Term Memory (LSTM) Networks (RNN) | MLPR, SVR | Global Ocean | BOA Argo data | 2004-2017 | MSE | Lagged SST |
(Nodoushan 2018) | FFNN | Bayesian Network | Honolulu, Hawaii | water quality buoy, | August 2008 to September 2014 | R, RMSE, MAE | Lagged SST |
(Ouala et al. 2018) | Bi-NN-Based Kalman Filter, | Ensemble Kalman Filter, Bi-NN-NNKF-EOF | South Africa | OSTIA product | 2008-2015 | RMSE | Spatio-temporal parameters, lagged SST |
(Patil and Deo 2018) | ANN | Â | Indian Ocean, Red Sea | HADISST (MDB, ICOADS, GTS, AVHRR) | 1990-2016 | R, NRMSE | SSTA |
(Quilodrán Casas, 2018) | DA-NN | Ensemble Kalman Filter | Atlantic | AVISO and GHRSST | 2001-2010 | R, MSE | Sea Surface Height (SSH), Sea Surface Temperature (SST), Eastward and Northward horizontal velocities |
(Davies 2018) | NN | Bootstrap, OLS | Pacific Ocean | NOAA | 1981-2014 | MSPE | Lagged SST |
(Broni-Bedaiko et al. 2019) | LSTM-NN | MIMO | World | ICOADS, NOAA ERSSTv5 //www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.html | January 1950 to October 2017 | RMSE, MAE | Lagged SST |
(Wei et al. 2019) | MLP | Â | South China Sea | OSTIA http://marine.copernicus.eu/services-portfolio/access-toproducts/) | 1985-2018 | Bias, RMSE, STD | Lagged SST |
(Wu et al. 2019) | CEEMD-BPNN EEMD-BPNN | Â | Northeastern region of the North Pacific | OISST, NOAA | 1982 -2016 | RMSE, ERR | Decomposition of lagged SST |
(Xiao et al. 2019b) | Convolutional LSTM | SVR, LSTM | East China Sea | OISSTv2 (AVHRR) | 1982–2015 | R, RMSE, MAPE, P-Value | Lagged SST, spatio-temporal parameters |
(Xiao et al. 2019a) | LSTM-AdaBoost | SVR, BPNN, LSTM, AdaBoost | East China Sea | OISSTv2 (AVHRR) | 1982–2017 | RMSE, MAE, R | Lagged SST, Spatio-temporal parameters |
(Xie et al. 2019) | LSTM | SVR, GED | Bohai Sea China, South China Sea | NOAA–OISST | 1997–2018 | MSE, MAE | Lagged SST, spatio-temporal parameters |
(Han et al. 2019) | CNN |  | Pacific Ocean | AVHRR, Argo | 2004–2015 | R2, MSE | SST, SSH, SSS |
(Saha and Chauhan 2020) | NARX-NN |  | Pacific Ocean | NOAA, TOGA (buoys) | 1985–1995 | MSE, NMSE, RMSE, NRMSE, MaxAE, Mean AE, MAPE, SDD, SRes | Lagged SST, air temperature, zonal wind Meridional Winds (lagged) |
(Xu et al. 2020) | M-LCNN | CNN, LSTM, SVR, CFCC-LSTM | Yellow Sea, Bohai Sea China | NOAA | 1981–2019 | ACC, RMSE | Lagged SST |
(Zhang et al. 2020a) | M-convLSTM | CNN, LSTM | Pacific Ocean | Argo | 2005–2017 | R2, RMSE | Horizontal and vertical temperature variations |
(Wolff et al. 2020) | LSTM | GAM ,RF, MLP, XGBoost, | North Atlanic | ECMWF, Modis | 2002–2018 | MAE, MAPE | Lagged SST |
(Yu et al. 2020b) | DGC network | SVR, SVM, ARIMA, BPNN, RBFNN, RNN, GRU, LSTM, GRU-SVM, WNN, CEEMDAN-LSTM | East China Sea Yellow Sea | NOAA-OISST | 2001-2017 | RMSE, RMSPE, MAPE, MAE, ACC | Lagged SST |
(Barth et al. 2020) | CNN |  | Southern coast of France and Corsica | AVHRR | 1985–2009 | RMSE, CRMSE, Bias | SST(Anomalies), scaled lagged SST, longitude, latitude. sine and cosine of the year |
(Zhang et al. 2020b) | GRU-NN |  | Bohai Sea | OISST (AVHRR) | 1982–2019 | R, RMSE, MAE | Lagged SST |
(Sarkar et al. 2020) | LSTM | FFBP, ARIMAX, NARX | Indian Ocean | CCCma, NOAA, HadiSST | 1870–2018 | R, RMSE, MAE | Lagged SST |
(Wei et al. 2020) | LSTM |  | China Sea | OSTIA | 1985–2014 | Bias, SD, RMSE | Lagged SST |