Reducing cadmium and arsenic accumulation in rice grains: The coupled effect of sulfur's biomass dilution and soil immobilization analyzed using meta-analysis and machine learning.
Journal:
The Science of the total environment
PMID:
39490837
Abstract
The biogeochemical cycling of sulfur (S) in paddy soil influences cadmium (Cd) and arsenic (As) migration. However, the impact of S application on Cd and As within the soil-rice system has not been fully explored. This study aimed to examine the effect of S application on Cd/As soil-rice system dynamics by conducting an extensive meta-analysis of 322 sets of observational data from 46 publications, which were published between 2004 and 2023. Furthermore, a machine learning model was only used to forecast the potential influence of S on Cd within the soil-rice system rather than the influence of As due to the limited data samples. The results indicated that the basal application of S mainly reduced the accumulation of Cd and As in the grains [Cd: 29.00 % (28.48 % to 29.52 %); As: 38.31 % (37.79 % to 38.85 %)] by the coupling effect of promoting rice growth 40.87 % (40.61 %-41.14 %) and reducing the soil bioavailable Cd/As by 18.20 % (18.05 % to 18.36 %)/19.59 % (19.44 % to 19.75 %). However, the efficacy of actual field farmland remediation is often suboptimal because the actual soil physical and chemical properties frequently do not meet the ideal conditions [pH: 6.5-7.5, Total S: < 200 mg/kg, soil organic matter (SOM): 30-40 g/kg, Total Fe: 20-30 g/kg] that are required to mitigate Cd and As accumulation in rice grains. Notably, the random forest machine learning model achieved an acceptable level of accuracy when compared to the excessive linear regression simulation. The model suggested that the decrease in the Cd/As accumulation in the rice was due to the soil available S content, which was primarily influenced by S application. This study provides novel insights for managers and researchers for the amelioration of Cd/As-contaminated farmland soil.