Bottom classification analysis options

A bottom points variable is created from a bottom classification on an Sv variable. Echoview offers flexible reclassification and analysis options.

Reclassify Bottom

Dataflow window > Click a Bottom points variable > Shortcut menu > Reclassify Bottom

Reclassify Bottom recalculates the clustering in the bottom points variable. The reclassification uses Cluster Dimension Selection and Bottom Class Allocation settings and reruns the randomly seeded k-Means clustering algorithm. This is why Reclassify Bottom is generally faster than the initial bottom classification.

Experimental observations suggest that the number and the quality of bottom points together with the randomness of the k-Means++ initialization can affect the number of classes and distribution of bottom points in the classes. The number of classes tends to be more stable (over multiple classifications) when there are many bottom points. Reliable bottom points satisfy the algorithm limitations for single beam incidence angle with the seabed.

Repeated use of Reclassify Bottom with unchanged algorithm settings may give you a feel for the nature of the clustering in the bottom classification. And a feel for the quality of the bottom points. For an overview and referenced discussion of Echoview bottom classification refer to Bottom classification algorithms.


  • Reclassify bottom deletes any new or renamed classes.
  • Automatic PCA automatically sends a diagnostic message to the Message dialog box
  • Automatic k-Means clustering automatically sends a diagnostic message to the Message dialog box.
  • Depth is not recalculated.

Classify bottom with different settings

Display a Sv echogram > Echogram menu > Classify bottom

Each bottom classification produces a bottom points variable that contains features and clustered bottom points. The algorithm settings used for the classification are automatically recorded in the Notes page of the Variable Properties dialog box. The classification time and intermediate details associated with "automatic" parts of the classification are sent to the Messages dialog box (these become unavailable when an Echoview session is closed).

The variable settings and the classification algorithm settings affect the bottom classification in different ways. Differently configured bottom classifications can reveal classification sensitivities to:

  • Bottom echo threshold at 1 m
  • Feature extraction interval
  • Extracted features
  • Cluster dimensions
  • Clustering algorithm
  • Clustering iterations

Automatic PCA may find up to five principal components. Follow with auto clustering using the Calinski-Harabasz algorithm to produce classes that give a general feel for the data. Then use specific component dimensions and a manual class number with increased cluster iterations to fine tune the classification. From a physical perspective, features that are likely to be strong cluster dimensions include: Roughness index, Hardness index and (first) bottom rise time. Sampled substrate information may help with the selection of class number. Bottom points graphs of features may also help in identifying trends in the data. For an overview and referenced discussion of Echoview bottom classification refer to Bottom classification algorithms.


  • Automatic PCA automatically sends a diagnostic message to the Message dialog box
  • Automatic K-Mean clustering automatically sends a diagnostic message to the Message dialog box.
  • Vertical selection > Classify Bottom can be used to classify a part of the echogram. The classification is valid within the boundaries of the selection.

Graph bottom features

The features of a bottom classification represent aspects of the physical characteristics of the first bottom echo and where available the second bottom echo. These in turn may be correlated to the physical interaction of acoustic energy with the substrate. Relationships and trends between extracted features can be observed in two dimensional graphs, where class is a unifying third variable. Echoview allows you to create multiple graphs for a bottom points variable. Observations from bottom points graphs may help you identify cluster dimensions that you can specify manually.

First echo features include Roughness index E1, First bottom length, Bottom rise time, Bottom max Sv, Bottom kurtosis, Bottom skewness and Depth.

Second echo features include Hardness index E2, and Second bottom length.

An estimation for initial cluster dimensions (for data with first and second bottom echoes), based on the physical behavior of a substrate, are: Roughness index, Hardness index and Bottom rise time or Bottom max Sv.

Anderson (2007) observes that high acoustic frequency is useful for finding out about roughness and substrate surface structure and that low acoustic frequencies are useful for penetrating the seabed.

Hamilton (2001) observes that a bottom type may be inferred from shape of the echo envelopes. Other observations made by Hamilton (2001) include:

  • Soft sediments attenuate the acoustic signal strength more than harder sediments. There may be a correlation between a lower amplitude in the first bottom echo for softer sediments.
  • Rougher surfaces provide more backscattered energy. An example of this would be that the first echo may have a lower peak and a longer tail than a smoother surface of the same composition.
  • The length and energy of the first echo tails is proportional to the acoustic roughness of the sediment surface.
  • For frequencies greater than 50 kHz, the effect of the acoustic absorption coefficient on bottom echoes becomes more important with depth.

See also: Bottom classification graph

Create, delete or modify a bottom class

Bottom points variable/Bottom points graph/Bottom points cruise track > Shortcut menu > Variable Properties > Classes page

Under a bottom points variable you can:

  • create a new bottom class
  • delete a class
  • change the color assigned to a class
  • give a class a name e.g. Mud, Sand, Mud and vegetation

Note: Delete is irreversible and will also delete the bottom points in the class.

Edit the class of a bottom point

Under a bottom points table you can change the class of a bottom point with selection from the Class list.

Ground truth with geographically matched video

Geographically synchronous video or identified substrate information from grabs or sampling can be used to ground truth classes in a bottom classification.

Auto-synchronization is available for an Sv echogram (displaying a bottom classification integram), a bottom points cruise track and a Media window.

Video data is synchronized using a time and date. Echoview can read time and date from some video file formats. For the other video formats you can create an EPJ file with a specified time and date. Video data recorded concurrently with echogram data is the easiest synchronize. For video data that is not concurrent in time, but is geographically synchronous it is possible to estimate a time and date by matching latitude and longitude on a cruise track with a time and date of a ping on the Sv echogram used for bottom classification.

See also

About media data
Using video data

Onscreen analysis

EV File Properties dialog box > Export page > Bottom > select bottom features

Calculated values for bottom features can be displayed with the results of an onscreen integration. Configuration of a bottom classification is required (beforehand), as well as the selection of bottom features from the Bottom category on the Export page of the EV File Properties dialog box.

Note: The pings intersecting the selection/region/cell are used as the Feature extraction interval. Also Analysis page Exclusion lines are ignored by bottom classification.

Export data

Dataflow window > Bottom points table/Bottom points cruise track variable > Shortcut menu > Export Data

- OR -

Table menu > Export Data
Cruise track menu > Export Data

A bottom points export to a .csv file is available to enable processing of bottom features outside Echoview.

See also

About bottom classification
Configuring a bottom classification
Bottom classification algorithms
Bottom classification - practical notes