The Tsunami classification model with real-time sensors placed at different locations and at different depths is proposed. To exclude the artifact effects in the sensor values, a wavelet-based denoising scheme is integrated in the model. In addition, a downsampling approach has been proposed to achieve maximum flat delay response, and the present results are compared with the Pc McClellan method. Various parameters such as conductivity, salinity, pressure, temperature, and dissolved oxygen are measured and sensed using multisensor grid architecture. Our results show that by sensing the above parameters and subject them online, it is possible to clearly distinguish the pre- and post-tsunami behaviors.