The health of terrestrial waters is of utmost importance for the sustenance of aquatic ecosystems, and dissolved oxygen (DO) represents one of the key indicators used to evaluate water health. In this study, the objective was to predict DO levels in the Danube river, using data on flow velocity and temperature, and to investigate intricate relationships among these variables. Firstly, a Random Forest regression algorithm was deployed to predict DO levels using flow velocity and temperature as predictors, and SHapley Additive exPlanations (SHAP) values were calculated. Then, an autocorrelation function was integrated into the Random Forest regression algorithm, incorporating DO data from previous time steps. Autocorrelation and partial autocorrelation functions, together with partial and cross-correlation analyses were also performed. Altogether, DO levels predicted, using flow velocity and temperature, resulted in model with R2 equal to 0.88 (MSE = 0.59; RMSE = 0.77), and SHAP values highlighted that temperature exerted a more pronounced influence on DO levels than flow velocity. When DO data from previous time steps were used, this enhanced model achieved an R2 equal to 0.97 (MSE = 0.11; RMSE = 0.33) and DO from preceding time steps had the most substantial influence on the predictive accuracy of this model. Autocorrelation and partial autocorrelation function plots for DO, underlined its strong correlation with adjacent time steps values. Partial and cross-correlation analyses revealed that temperature exhibited a strong negative correlation with DO. In conclusion, this research provides valuable insights into the dynamics governing terrestrial water health, emphasizing the indispensable role of temperature in modulating DO levels.