Monitoring the ecological status of water bodies is crucial to guarantee human health and economic development. However, monitoring is often deficient in developing regions due to high installation and maintenance costs, thus it is frequently supported by water quality models, whose results are themselves affected by the lack of detailed input data. A possible solution is to use probabilistic models that consider the inherent uncertainty of the different inputs. In this research, we extended a simple water quality model (QUAL-UFMG, based on Qual2E) through Monte Carlo simulations to generate probabilistic results and applied it to a representative case study in Brazil. Results showed that, depending on the distribution of probabilities and variability of parameters adopted, the outcome of a non-deterministic modelling approach may differ significantly from a deterministic one regarding compliance with water quality standards. Moreover, the probabilistic strategy is more scientifically transparent and robust, as it explicitly communicates the uncertainty in both the measured data and modelling results. We conclude that a probabilistic approach is particularly useful in regions with a low data availability such as developing countries, as uncertainties are high due to insufficient monitoring, and the risk to human health is elevated due to a low prevalence of sanitation.

HIGHLIGHTS

• A deterministic water quality model is expanded through Monte Carlo simulations.

• Outcomes of deterministic and probabilistic approaches to water quality modelling are compared for a case study in Brazil.

• There may be significant differences in modelling results regarding compliance to water quality standards depending on the approach taken; this is key in data scarce regions with deficient monitoring networks.