The influence of hydrological coupling method experimented with one-way loose coupling of TOPMODEL and NWM

Abstract:

Hydrologic models are frequently coupled together to represent complex systems. In this paper, we investigate the best location of the coupling interface when coupling a Land Surface Model with a hydrologic model: i.e., which processes should be simulated by which model? We tested how the level of dependency, type of transferred data, or modification of process steps impact the model behavior, using seven cases of one-way loose coupling of the National Water Model (NWM) and TOPMODEL. We previously found the coupling of TOPMODEL and NWM improves streamflow prediction and calibration in headwater catchments. Each test case varies in level of dependency on NWM’s land surface model (LSM; Noah-MP) and TOPMODEL is adapted accordingly; modifying process steps or structure to accommodate data input from Noah-MP, replacing input parameters with a dynamic coefficient derived from Noah-MP flux dynamics, or overriding internal states with Noah-MP outputs. The number of active TOPMODEL parameters decreases from 10 to 5 as the couplings become more reliant on LSM. All the cases enable efficient parameter calibration and outperformed the regionally calibrated NWM v1.2.2 in hourly streamflow prediction. Predictions improved as TOPMODEL was modified to utilize more LSM data, however, cases in which underground runoff or infiltration excess values from LSM directly override TOPMODEL internal states resulted in lower accuracy, and became less sensitive to two major TOPMODEL parameters (lnTe and m). The best result came from the case that took surface water flux calculated from Noah-MP as an input, removed the TOPMODEL “Rootzone” structure, and replaced vertical drainage delay parameter with a dynamic coefficient (three parameters removed from TOPMODEL). The benefits from the coupling diminish if hydrological dependency between an LSM and the coupled conceptual hydrological model increases such that internal processes in the hydrologic model are overridden and lose continuity. Minor discord in model assumptions can result in an inconsistent calculation in comparable hydrological steps, and this error accumulates through override-type coupling. Maintaining independence in each model’s operation is recommended in the coupling method.