Evaluation of Integrated Land Simulator using site observations

Accurate representation of land processes is important because land models are used for many purposes: climate simulation, weather and flood prediction, and impact assessment studies. In order to realize high accuracy land simulations, we are developing a new framework, Integrated Land Simulator (ILS). ILS consists of multiple land components. Currently, it includes a land surface model, Minimal Advanced Treatments of Surface Interaction and Runoff (MATSIRO), a next-generation river model, CaMa-Flood, and an independent I/O component. They are coupled by a general purpose coupler Jcup, and executed using the Multiple Program-Multiple Data (MPMD) method. It can be applied to land simulations from a local scale to continental or global scales. It also includes a bounary data creation tool, which is useful for simulations with different horizontal resolutions. MATSIRO has been rewritten in Fortran 90 and separated from a climate model. In this study, we conducted off-line site experiments using ILS. We used meteorological forcing data from multiple model intercomparison projects that cover observation sites for different climate and land cover. First, reproducibility of the rewritten code was checked and confirmed that the difference was negligible. Then, model performance was evaluated using different metrics. The result shows that the correlation coefficients for latent and sensible heat fluxes are high for most sites, but there are some sites with relatively high mean bias.