Fig. 7: Noise waveforms and convolutional neural network (CNN) tsunami forecasting results with and without noise. | Nature Communications

Fig. 7: Noise waveforms and convolutional neural network (CNN) tsunami forecasting results with and without noise.

From: Early forecasting of tsunami inundation from tsunami and geodetic observation data with convolutional neural networks

Fig. 7

a Sea-level observations from GPS buoys before the 2011 tsunami event. The grey and red lines represent the raw observations and the 15 min running average of the raw data, respectively. The elapsed time is from a day before the earthquake occurrence. b Noise waveforms extracted from the observations. The deviation from the averaged data (the red lines in a) is extracted as noise. The figure shows the noise waveforms during the same observation period for the actual tsunami observations, but a day before the tsunami. c Noise waveforms for CNN input. The extracted noise waveforms before the event are added to the actual tsunami observations to prepare the noise inputs. d Inundation forecasting results with and without noise. The grey and red lines represent the forecasting results without and with noise, respectively. Because the difference between the two waveforms is small, the result without noise is illustrated with a thicker line. e Offshore tsunami forecasting results with and without noise. The grey and red lines represent the forecasting results without and with noise, respectively. Because the difference between the two waveforms is small, the result without noise is illustrated with a thicker line.

Back to article page