Improving the Initial Conditions of Hydrological Model with Reanalysis
Soil Moisture Data
Abstract
The initial conditions (e.g., soil moisture content) of the hydrological
model, which is usually obtained from the warm-up of the hydrological
modeling, significantly impact the simulation efficiency. However,
spending the valuable data in warm-up instead of calibration and
validation is luxurious. In order to improve hydrological simulation
efficiency in the case of no warm-up phase, this paper proposes a
methodology to fill the gap via improving the initial conditions of the
hydrological model using an alternative global soil moisture dataset.
Specifically, three soil moisture (SM) variables of the initial
conditions from the Block-wise use of the TOPMODEL (BTOP) model and
EAR5-Land reanalysis data were adopted and conducted correlation
analysis. Several traditional curve-fitting functions and the
state-of-art technical, long-short term memory (LSTM), were applied to
develop the relationship between BTOP and EAR5-Land SM variables in the
Fuji and Shinano River Basin, Japan. Furthermore, four configured
hydrological simulations evaluated the benefits of the proposed
methodology for improving the initial conditions. As a result, LSTM
outperforms the traditional curve-fitting method in constructing the
relationship between variables in time and space. Moreover, the
hydrological simulation cases using the initial conditions related to
the SM from the ERA5-land performs better than the case without the
warm-up phase, and the simulated discharge process approaches the
“optimal” case with the warm-up phase. It is confirmed that the
proposed methodology helps improve the initial conditions of the
hydrological model using reanalysis soil moisture data.