Modeling Size‐number Distributions of Seeds for Use in Soil Bank Studies

Abstract

Knowledge of soil seed banks is essential to understand the dynamics of plant populations and communities and would greatly benefit from the integration of existing knowledge on ecological correlations of seed size and shape. The present study aims to establish a feasible and meaningful method to describe size-number distributions of seeds in multi-species situations. For that purpose, size-number distributions of seeds with known length, width and thickness were determined by sequential sieving. The most appropriate combination of sieves and seeds dimensions was established, and the adequacy of the power function and the Weibull model to describe size-number distributions of spherical, non-spherical, and all seeds was investigated. We found that the geometric mean of seed length, width and thickness was the most adequate size estimator, providing shape-independent measures of seeds volume directly related to sieves mesh side, and that both the power function and the Weibull model provide high quality descriptions of size-number distributions of spherical, non-spherical, and all seeds. We also found that, in spite of its slightly lower accuracy, the power function is, at this stage, a more trustworthy model to characterize size-number distributions of seeds in soil banks because in some Weibull equations the estimates of the scale parameter were not acceptable.

Seed Volume Dataset—An Ongoing Inventory of Seed Size Expressed by Volume

“…Data 2019, 4, 61 2 of 10 Therefore, we investigated [10] size-number distributions of seeds with known length, width, and thickness after being passed through a series of sieves. We found that the geometric mean of length, width, and thickness was the most adequate one-dimensional seed size estimator, providing shape-independent measures of seed volume which were directly mirrored by the sieves mesh side.…”

Self-Affinity, Self-Similarity and Disturbance of Soil Seed Banks by Tillage

“…Either before or after tillage, non-spherical seeds were more abundant than spherical seeds, their percentage ranging between 53% (in nine cores taken before tillage) and 62% (in three cores taken after harrowing). Given the adequacy of mesh side to estimate seed volume [ 30 , 39 ] seed size was calculated and found to extend almost always across three orders of magnitude, and only in one sample across four orders of magnitude.…”

“…The approach and procedures of Casco et al . [ 30 ] were generally followed including the choice of mesh side instead of mesh bisector as a surrogate for seed size. All statistics were done with Statgraphics 4.2 (STSC, Inc., Rockville, MD, USA) except Box-Cox transformations done with BIOM (Applied Biostatistics, Inc., New York, NY, USA) and lack of fit tests done with Excel ® 2010 (Microsoft Corporation).…”

“…Size-number distribution of seeds can be described either by power law or by Weibull equations [ 30 ]. However, the Weibull function [ 31 ] can reduce to the power law [ 32 ], but without the ability of the former to easily accommodate size-number distributions in which the relationship between size and number depends on and varies with size itself.…”