Acoustic backscatter data from multibeam sonars are increasingly being used for seafloor classification. In order to improve the classification one emerging trend is to use multifrequency backscatter data (for example in this case 40 kHz to 300 kHz). While this ideally should reduce ambiguities in seafloor classification, it requires data collection by either simultaneous operation of multiple sonars or operation of a single sonar simultaneously at multiple frequencies. For a given configuration multiple sectors and swaths are used to maintain motion stabilization and to obtain an even sounding density. Furthermore as the sensor altitude varies, all of the pulselength, sector spacing, and center frequency changes. All of these add complications in removing the radiation overprint on the backscatter data.

In this study, strategies for radiation pattern removal from multispectral backscatter data are analyzed. The dataset utilized was collected using new EM710 and EM2040 multibeam sonars on the NOAA ship Thomas Jefferson. Different methods to remove radiation patterns are tested on the dataset and their relative impact on the fidelity of seafloor classification is assessed. Examples of improvements in seafloor discrimination using multispectral backscatter data are presented and suggestions for standardized field procedures for the data collection are put forward.