Datasets for the surrogates were generated from the OGS model based on different inflow–extraction scenarios for a management period of 60 years. Water table heights were extracted from all nine MOLs whereas only the first seven MOLs were considered for the salinity because of the fact that salinity was negligible at the last two MOLs. One inflow rate, one extraction rate, nine water table heights, and seven salinity values from any given time step were considered to be the inputs for that time step whereas nine water table heights and seven salinity values from the next time step were used as outputs. Table 5 shows the first three rows of any dataset with superscripts indicating the time steps. As is seen, for each time step there are 18 inputs from that time step and 16 outputs from the next time step. Training based on the pattern of the data shown in Table 5 makes the surrogates suitable for variable-time simulations.