Bayesian Tobit Model for Handling Left-Censored in Water Quality Assessment: A Case Study of Sunda Strait, Indonesia

As an archipelagic nation, Indonesia relies heavily on coastal activities that may affect marine water quality, yet studies addressing this issue remain limited, particularly in the presence of left-censored data. This study aims to evaluate an appropriate method for handling left-censored data in water quality assessment using the CCME-WQI, with a case study in the Sunda Strait. A Bayesian Tobit model was applied to account for left-censored observations and integrated with the CCME-WQI framework. For comparison, conventional substitution methods and exclude left-censored were also used. The performance of approaches was assessed based on their ability to produce reliable water quality index estimates. The results indicate that the Bayesian Tobit model provides more robust estimates than substitution methods, as it incorporates uncertainty through credible intervals and reduces potential bias. The estimated water quality index ranged from 83.8 to 92.1, classifying the water quality as “good.” In conclusion, the Bayesian Tobit model is a more reliable approach for handling left-censored environmental data and improving water quality assessment. This method is particularly relevant for routine monitoring and can be extended to other fields with similar data characteristics.