Extended Data Fig. 4: Regional and local validation of DynQual_Random Forest and DynQual.
From: Climate change drives low dissolved oxygen and increased hypoxia rates in rivers worldwide
(a) Median value of the Root Mean Squared Error, normalised by the mean (nRMSE), for each region and three month period (Jan-March, April-June, July-Sept, Oct-Dec) for the process-based model DynQual (b) median nRMSE value for each region and three-month period for the hybrid DynQual_Random Forest model (c) boxplots comparing the observed dissolved oxygen concentrations (grey) during periods of hypoxia (DO < 3 mg/l) with the simulations from DynQual_Random Forest (red) and DynQual (blue). Boxplots: median, first quartile (Q1), third quartile (Q3), Q1 − 1.5IQR, Q3 + 1.5IQR where IQR is Interquartile range. (d-g) Time-series of daily DO concentrations simulated by the hybrid DynQual_Random Forest model for four stations across Asia, North America, Australia/New Zealand and South America compared to monitoring data (black).
