MORPHOMETRIC ANALYSIS: ASSESSING ERROR IN DIGITIZATION AND AUTOMATED COLLECTION OF LARGE QUANTITIES OF DATA

The collection of data for morphometric analyses inevitably incorporates some amount of researcher error related to such factors as specimen orientation, lighting, scaling, and digitization. Here, we focus on resolving two methodological issues in the collection of data in the venerid bivalve genus Chione: 1) consistency of specimen orientation and 2) reduction of error in landmark/outline capture.

Standardization of specimen orientation is critical to avoiding error in the translation of a three dimensional specimen into two dimensional space. For morphometric analyses of bivalves, standard protocols orient the plane of the commissure parallel to the plane of the photographic surface. This orientation can be difficult to replicate when positioning a specimen using a sand box or wedges of modeling clay, as has often been done in previous studies. As an alternative, positioning specimens with the commissural plane directly on the glass of a flat bed digital scanner ensures consistency of shell orientation and may provide a rapid, high quality method with which to acquire high-resolution digital images.

Morphometric analyses of bivalve shells may involve collection of landmark (discrete homologous points) or outline data. Error in outline traces produced by even the most consistent human hand likely outstrips most other sources of error. We investigated techniques for automating the outline-tracing procedure by means of digital image processing, in which the specimen outline is acquired by an algorithm that detects significant differences in values of adjacent pixels. A threshold value of difference between the specimen and background pixel values serves as an effective discriminator to extract the specimen outline. These methods show great promise for creating a consistent outline that not only greatly improves the reliability of the data, but also dramatically increases the speed with which additional data can be acquired.

Additionally, we determined the degree of error introduced when digitization was performed manually by students with varying levels of experience. A methodology that maintains precision and accuracy for a variety of data collectors, while maximizing the rate at which specimens are processed, will greatly promote the acquisition of robust morphometric datasets.