The usefulness of Total Runoff Integrating Pathways (TRIP) for validating land surface models

In this study, the usefulness of the global river routing model Total Runoff Integrating Pathways (trip) for the validation of land surface models is discussed. An attempt is made to compare and validate the modeled runoff of two different land modules of the climate model ec-earth: the Tiled ECMWF Surface Scheme for Exchange over Land (tessel) and the revised Hydrology Tiled ECMWF Surface Scheme for Exchange over Land (htessel).
These land surface models especially differ in the generation of surface runoff, which results in high frequency variability in the modeled runoff. To validate the tessel and htessel models in terms of runoff, the modeled river discharge of the Danube and the Amazon rivers are compared with observations.
It is concluded that the usefulness of the trip scheme for the validation of the land surface models is limited by its tendency to delay the water too much and to filter out high frequency variability. For comparison, a simple method was used to estimate the river discharges without river routing. For this approach, the unrouted runoff was manipulated by applying a time shift and smoothing the data over a certain time period.
The method was applied for a couple of rivers. The quality of the obtained results are at least comparable with the trip simulations. This might suggest that the trip scheme is unnecessary complicated for its purpose.
By comparing the high frequency variability in the unrouted runoff directly with the discharge observations it is found that the tessel scheme underestimates the high frequency variability in the runoff in the Danube region. Estimates of the river discharge that were found by a simple manipulation on the htessel runoff suggest that the high frequency variability in the runoff is well modeled by the htessel scheme in the Danube
region. In the Amazon, both offline models perform equally bad and underestimate the runoff by about 60% (due to too much evaporation). A large underestimation of the runoff was also found in the Mississippi region.
An attempt has also been made to compare the behavior of the tessel and htessel land modules in an online model simulation, where the lsms were coupled to the climate model ec-earth. Online validation seems to be difficult due to uncertainties in the climate model. The climate simulation of the surface water balance is especially influenced by the modeled precipitation that drives the land surface scheme. Due to errors in the modeled precipitation it is difficult to compare the offline and online behavior of the land surface models. In most areas that were considered, the precipitation was (in various extents) un-
derestimated by the climate model. A comparison between the online and offline modeled runoff indicates that in most areas the underestimation of the precipitation by the climate model leads to a reduction in the modeled evaporation (rather than to a reduction in the runoff). In the Danube region, the correlation between the runoff and the precipitation is under-represented by the climate model. These features indicate that the atmospheric
conditions in the climate model might be different from the observed atmospheric conditions. It is not clear whether this difference should be ascribed to the effect of the land surface model on the atmosphere or to errors in the atmospheric model itself.