Knowledge-guided methodology reducing computational effort in the heuristic calibration of a river model

Accurate calibration is essential for reliable river-flow simulation and decision-making. Exhaustive parameter exploration incurs high computational costs and limits scalability in hydrodynamic river systems. This work introduces the Agile Adjustment in Successive Steps (A2S2) methodology, a hypothesis-driven calibration framework based on spatial and temporal continuity. Spatial continuity assumes adjacent river sections tend to exhibit similar hydraulic behavior, allowing parameter values selected for one section to initialize calibration of the next. Temporal continuity assumes that best-fitting parameter configurations tend to be similar across consecutive adjustment intervals; therefore, the configuration selected for the preceding interval initializes the search for the current interval and is reevaluated using current-interval data. This spatial and temporal reuse reduces simulator executions while maintaining low calibration error levels. The methodology was evaluated using the EZEIZA V hydrodynamic simulator developed by Argentina’s National Water Institute (INA) for Paraná River simulations. Results showed a maximum reduction of approximately 57% in annual mean squared error (MSE) relative to the baseline configuration, observed in 1995, and an approximately 86% reduction in estimated simulator executions relative to exhaustive evaluation. A2S2 was compared with a two-stage Optimization via Simulation (OvS) approach, highlighting trade-offs between localized parameter reuse and broader parameter-space exploration.