A comparative study of model initialization for the Noah land surface model using LIS

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 19 January 2010: 5:00 PM
B212 (GWCC)
Roshan K. Shrestha, EMC, College Park, MD; and P. R. Houser

A Land surface model (LSM) needs proper initialization for proper simulation and to avoid erroneous interpretation of the model results. A properly initialized model is expected to have equilibrium land surface state to begin the model runs. Such an equilibrium state is obtained via spinning up the model. Generally, the spin-up is done either by running the model for a longer period or by looping the model runs within a chosen period repeatedly. A longer spin-up period requires longer term data which is not easily available for most of the cases. Alternatively, the looping approach allows the spin-up period run as long as it needs to reach to the preferred equilibrium state.

Because the LSM is sensitive to the model initialization, the spin up results and the time taken to reach to the equilibrium is also sensitive to the chosen method of model initialization. We have tested and compared nine different methods of initialization for the spin-up runs, which were conducted in the looping-mode using the 1-km resolution Noah LSM in the Land Information System (LIS). The experiments have used soil moisture states based on chosen methods as the surrogate to initialize the model for the spin up runs. The model outputs were compared to evaluate the discrepancies of land surface fluxes over the annual interval of the looping cycle. The experiments were conducted at 12 stations located in the SGP ARM site.

It is noted that the appropriate spin up period can vary depending upon the chosen land surface fluxes. The spin-up response is also found dependent on the spatial heterogeneity of land surface conditions. Early termination of spin-up runs could dangerously fail to adapt with the large spatial variation of land surface conditions. The model adjusts land surface states rapidly with the local land surface conditions but it needs multiple repetition cycles. The fluxes like sensible heat or latent heat are adjusted faster than the ground heat flux, which retains memory of the land surface state for considerably longer period.

The results have shown large variability despite being the experimented stations located in the same hydroclimatic region. The results indicate that the climatological average state does not necessarily return the most efficient initialization of a LSM. Spatially heterogeneous land surface state averaged over a short period is found to perform better than climatological average state in the experiment.