Application of targeted observation in a model's physical parameters for the simulation and forecast of heat flux with a land surface model

Abstract

The targeted observation is that the additional observation is implemented in important regions to improve a numerical forecast of a weather or climate event. These additional observations are aimed to reduce the uncertainties of the initial state with the data-assimilation system. In the study, the targeted observation strategy is attempted in the uncertainties of the model's physical parameters to improve the numerical simulation and forecast skills of surface turbulent fluxes (sensible heat flux—SH and latent heat flux—LH). First, an approach for determining important parameter combinations based on conditional nonlinear optimal perturbation related to parameters (CNOP-P) is employed to choose sensitive parameter combinations for the SH and LH. The characteristic of the new approach is to judge the sensitivity of the parameter combination and consider the nonlinear effect of the combination. The sensitivity of the parameter combination using the new approach is compared with that using the traditional one-at-a-time (OAT) approach. The latter method ranks parameters one by one according to the sensitivity of each parameter and it fails to judge the sensitivity of the parameter combination. Within the Common Land Surface Model (CoLM), numerical results show differences between the sensitivity of the parameter combinations using the new approach and the top-ranked sensitive parameter using the OAT approach in dry and wet regions of China for the SH and LH. Second, the benefits of forecast skills of the SH and LH are evaluated by reducing the parameter error within the sensitive parameter combination. It is found that the prediction skills of the SH and LH are improved in China. For example, elevated extents of the SH (84.3%) and LH (78.3%) calculated from the sensitive parameter combination using the new approach are greater than those of the SH (57.2%) and LH (48.5%) using the OAT approach in semi-arid regions. This result suggests that it is practicable to apply the targeted observation to reduce the model physical parameters to improve the simulation skill. The new approach for determining parameter combinations based on the CNOP-P is a promising method with which to explore the problems of the targeted observation to the model physical parameters.