Improving external shell volume estimation in snails using landmark-based size measurements

Shell images in digital collections and publications enable taxonomically, geographically and morphologically broad approaches to studying gastropod body size. Because traditional linear size measurements are not directly comparable across morphologically disparate taxa, we propose a transition to landmark-based size measurements, which are less sensitive to shape differences. Traditional and landmark-based size measurements were compared using two study systems: 36 specimens of diplommatinid micromolluscs from the Belau archipelago and 66 marine, freshwater and terrestrial gastropod shells from the Roosevelt Wild Life Collections (RWLC; housed at the State University of New York College of Environmental Science and Forestry, Syracuse, NY, USA). Belau diplommatinids were used to assess the effectiveness of size measurements when comparing confamilial specimens. This study system was most sensitive to methodologically introduced variance. RWLC shells were used to assess measurement effectiveness when comparing morphologically disparate taxa. A total of seven size variables measured from 2D images were compared, three variables based on interlandmark distances and four based on linear shell dimensions. Landmark-based size measurements were centroid size, a proposed new measurement dubbed size adjusted mean (SAM) and its rescaled version, SAM1/3. The linear dimension-based measurements were maximum, mean and geometric mean of shell height and width, as well as conical volume (volume of a cone with similar dimensions to specimen). Measurement accuracy was assessed by regressing external shell volume against each size measurement. For Belau diplommatinids, centroid size and maximum shell dimension most accurately predicted external shell volume. For RWLC shells, conical volume followed by SAM1/3 and geometric mean shell dimension were the most accurate predictors. Repeatability was assessed by computing % measurement error (%ME). Maximum shell dimension was most repeatable (%ME = 4.19%) for Belau diplommatinids, followed by centroid size (7.76%). Conical volume (%ME = 0.0039%) and maximum shell dimension (0.0073%) were most repeatable for RWLC shells. In both study systems, SAM1/3 was the land-mark-based size measurement most robust to missing landmarks, followed by centroid size. Based on these findings, we recommend using centroid size to compare congeneric or confamilial specimens and using conical volume to compare morphologically or taxonomically disparate specimens. However, conical volume could not be calculated when shell height or width data are unavailable. In such cases, we recommend comparing morphologically disparate shells using SAM1/3.