土壤盐分
含水量
土壤水分
土壤科学
降水
盐度
压实
结晶
冰点
环境科学
材料科学
矿物学
地质学
岩土工程
化学
热力学
气象学
物理
有机化学
海洋学
作者
Liyang Wang,Pengcheng Wang,Jingyu Liu,Jiankun Liu,Weihang Chen,Qianli Zhang,Tengfei Wang
标识
DOI:10.1016/j.coldregions.2023.103804
摘要
The volumetric expansion of saline soils under cooling could pose a threat to infrastructure constructed on these soils. This study examines the crystallization deformation characteristics of coarse-grained sulfate saline soils subject to stepwise cooling from 30 °C to −15 °C. Soil columns are prepared in plexiglass buckets with different fine contents, percent compaction, moisture contents, and sodium sulfate content. Cumulative heave and temperature evolution are measured when the specimens are maintained in a temperature-controlled environmental chamber (freezing in three dimensions). Upon cooling, salt and ice crystals grow in soil pores, leading to volumetric expansion. A model is proposed to describe the precipitation-induced deformation varying with temperature based on random forest (RF). The RF-based model reveals that the initial salt-to-water ratio exerts the most significant influence on the accumulated deformation, followed by moisture content, soil salinity, percent compaction, and soil gradation. The initial precipitation and freezing points (response variables) are expressed as two separate functions of soil properties (predictor variables) using the multivariate adaptive regression splines (MARS). The MARS-based models demonstrate that the precipitation points are mainly affected by the initial salt-to-water ratio and soil salinity; all factors contribute to the freezing point prediction, given the combined action of salt and ice crystallization. A threshold soil salinity of 33.33% existed in the evolution of precipitation points, while a threshold fine content of 35% and a threshold soil salinity of 39.68% were observed in freezing point predictions. The interaction mechanism of salt and ice crystallizations led to incredible difficulty capturing freezing points rather than precipitation points and predicting the sample heave. Although the evaluation metrics (RMSE, MAPE, and R2) confirm that the RF- and MARS-based models achieve the regression tasks well, more efforts are needed toward the second phase transition phenomenon and its implications for explaining thermally induced saline soil response. The findings may guide the utilization and management of land resources in saline areas.
科研通智能强力驱动
Strongly Powered by AbleSci AI