Modeling particle size distribution for subaqueous soil survey applications

Abstract

Coastal environments face a growing number of challenges as a result of a changing climate (e.g., sea level rise, flooding, and erosion). In response, intertidal and subaqueous soils (SAS) are being mapped to provide a soil resource inventory for use and management decisions. An essential part of any soil resource inventory is particle size distribution (PSD) analysis. Coastal soils have elevated levels of salts and sulfides that can complicate PSD analysis, requiring time-intensive pretreatments. We tested a regression model to reduce reliance on labor-intensive methods for PSD analysis. Analysis of 257 SAS samples revealed a strong sand–silt relationship (p < 0.0001; r² = 0.975), allowing for accurate silt and clay prediction from sand content. For samples with >40% sand (70% of the 257 samples), average absolute residuals of predicted silt ranged from 0.80% to 3.58%. Randomized iterative testing (10,000 iterations) showed that as few as 50 samples of the original 257 could be used to develop a model to provide PSD data with <4% absolute error for predicting silt for samples with >40% sand. Accuracy of the model declined for samples with ≤40% sand, especially <20% sand where average absolute residuals ranged from 7% to 8%. We hypothesized that diatom skeletons disrupted the sand–silt relationship in the silt-dominated samples, which contained as many as 9% diatoms. The regression model developed in this study offers a faster, more time- and cost-effective alternative for determining PSD analysis in SAS with >40% sand, aiding large-scale soil survey efforts.