| Soil moisture is an important physical variable influencing the land-atmosphere interactions, which controls the water and energy exchange among the soil-vegetation-atmosphere interfaces by changing the surface heat capacity, affecting the surface albedo, runoff and evaporation, and adjusting the partitioning ratio of the sense heat flux and the latent heat flux, and thus has a significant impact on the weather and climate system, so the prediction of soil moisture becomes very important for the numerical forecast. The in situ soil moisture observation is accurate, but limited in spatial resolution. Remote sensing can provide the global observations, but limited to the top few centimeters. Land surface model can simulate a wide range of soil moisture, but influenced by physical process defects itself, unknown sub-grid scale physical process, inaccurate atmospheric driving data and input parameters, the simulated soil moisture is not accurate. Land surface data assimilation system could provide soil moisture profile estimates with dynamic consistence at the scales required in time and space dimensions by integrating different spatial and temporal resolution observations into the model, but the estimates still contain errors in the presence of model systematic biases. So how to deal with systematic biases including time correlation error in the process of data assimilation should be taken into consideration. Therefore, the thesis will focus on the investigation of systematic model errors related with soil moisture estimation in the land data assimilation system. The main contents and conclusions are as following:Firstly, this paper considers the impacts of time correlated model errors on soil moisture estimation with the ensemble square root filter (EnSRF). Specifically, the model errors are regarded as a whole by adding time correlated error, then explore the feasibility of the model error correction while estimating soil moisture. Results show that the estimated soil moisture is no longer accurate with EnSRF in the presence of systematic biases, thus the correction to model error is needed; the soil moisture estimates are close to the true values while the time correlated model errors are updated.Secondly, investigates the impacts of constant bias as well as the bias with the form of sine function on the soil moisture estimation. Results show that the estimated soil moisture is no longer reliable with the ensemble kalman filter (EnKF) in the presence of biases above. Especially when the bias is large, the difference between the soil moisture estimates and true values is very large, but SepKF(Separate bias Kalman Filter) can provide good soil moisture esimates with the bias corrected at the same time.Finally, discusses how to estimate the soil water content synchronously while assimilating surface soil moisture to update soil moisture profile. For this, an empirical method is provided to estimate the soil water content while the soil moisture profile is updated. Results show that the precision of both soil moisture and soil water content estimates is low if not adopting the above method. The result will be better with the synchronous estimation scheme proposed in this paper than without updating.In summary, when the systematic bias exists, the estimates with EnKF will generally be influenced, particularly when systemic bias is large, the model error correction should be considered while updating soil moisture profile. |
