The data used to train our model is drawn from the National Oceanic and Atmospheric Administration's National Center for Environmental Information Database (access the database below). This publicly available database provides hundreds of thousands of sound velocity profiles taken worldwide. Profiles of other attributes besides soundspeed such as salinity, temperature, conductivity, and density are also included with certain profiles in the database. For our model we used only data points taken near the US coast. To train our model we used 25,000 profiles. The map below generated by our MATLAB data parsing script shows the location of each profile used to train the model. Predictions using locations near areas of high data density will have better accuracy when using our model.
To access the online database yourself, just click the NOAA logo below!
