Comparative evaluation of Tobit and Free Ion Activity Models for optimizing genotype-specific permissible arsenic...

Arsenic (As) contamination in soil and irrigation water poses a significant threat to food safety, particularly in rice-growing regions of Gangetic West Bengal, where As-enriched groundwater is widely used for irrigation. Rice cultivation under flooded, anaerobic conditions promotes As mobilization through reductive dissolution of Fe/Mn hydroxides and the conversion of arsenate [As(V)] to the more mobile arsenite [As(III)], which is readily taken up by rice roots via silicic acid transporters. Although inorganic As species dominate, small amounts of methylated As (MMA, DMA) from microbial transformations may also contribute to bioavailability. However, scientifically validated threshold values for arsenic in both soil and irrigation water remain poorly defined for rice-based agroecosystems. This study quantified the permissible limits of As in soil and irrigation water for safe Boro rice cultivation using a combination of controlled pot experiments and predictive modelling. Soils from five arsenic-affected districts of Gangetic West Bengal were spiked with graded As concentrations and cultivated with rice, followed by analysis of As in soil, irrigation water, and rice grains. Two modelling approaches were employed: Tobit regression, which accounts for censored dietary risk thresholds, and the solubility-based Free Ion Activity Model (FIAM), which integrates soil pH and organic carbon effects on As mobility. The novelty of this study lies in the comparative application of a censored regression model and a mechanistic solubility-based model to derive regulatory thresholds for arsenic in rice cultivation systems. The Tobit model predicted higher permissible limits, suggesting a maximum soil Olsen extractable As content of 4.00 mg kg^-1 under As-free irrigation water, while FIAM explained 88.09% of the variation in grain As. The Tobit approach provided more practical and flexible regulatory limits across irrigation scenarios, whereas FIAM offered a conservative, process-based assessment driven by soil chemical properties. Together, the two models complement each other by combining empirical threshold prediction with mechanistic risk evaluation. These findings provide a mechanistic and statistical framework for defining safe As thresholds, supporting evidence-based management and policy development for sustainable rice cultivation in As-affected agroecosystems.