Studies On Improving Land Surface Model Simulations Via Assimilating Land Data Products From Polar-Orbiting Satellites

In the background of global warming, drought frequency, duration and intensity are all increasing in the past several decades. Land surface model simulations have be successfully used in drought monitoring, and can be improved by assimilating land surface products generated from polar orbit satellite platforms. In this dissertation, the Land Information System (LIS6.2), polar-orbiting satellite observations and in-situ measurements of land surface state variables are employed to investigate the improvements of the land surface model simulations by using Moderate Resolution Inmaging Spectroradiometer (MODIS) Collection 5 (MODIS-C5) land cover, weekly green vegetation fraction (GVF) and monthly albedo products. In addition, it is demonstrated that how to obtain the optimal ensemble size of the ensemble Kalman Filter (EnKF) in a sequential soil moisture data assimilation system using mathematical derivation and numerical experiment for validation. Studies in this dissertation also include enhancing model skill by assimilating SMOPS-Blended soil moisture (SM) into land surface model and how to quality control satellite SM data in a data assimilation system. Main results of these studies are as follows.1. Residual errors of the ensemble Kalman Filter (EnKF) in sequential soil moisture data assimilation can be sharply reduced with the ensemble size N increasing when it is small. However, the ensemble size impact on assimilation effectiveness decreases quickly after the ensemble size is larger than 12. In theory, the contributions of EnKF can increase infinitely with ensemble size increasing, which leads to the residual error can be reduced infinitely. But the maximum efficiency of the EnKF for assimilating soil moisture (SM) observations can be reached when the ensemble size is set as 12. The finding with the mathematical derivation does not depend on the model or observational data used in the assimilation although the numerical experiment for the validation is based on soil moisture data assimilation using the Noah land surface model.2. Both of 1992-1993 AVHRR and 2007-2010 MODIS land cover maps have abilities to reflect the land cover accuratelly in the time periods when the satellite data were acquired. The differences between the two land cover products have large influences on land surface model soil moisture simulations. Land cover changing into grassland may increase root zone soil moisture (RZSM,15%), while the change from grassland and primeval forest areas to cropland or reforest areas reduce RZSM (-20%).3. The weekly Advanced Very High Resolution Radiometer (AVHRR) GVF/monthly MODIS albedo data of an individual year and their multi-year average could be very different. These differences are found to have large influences on Noah model simulations. Assimilation cases using current year weekly GVF can significantly improve model surface soil layer SM under dry conditions (9.3%). They could improve 60-100 cm layer SM (11.4%) and surface net long wave radiation (LWnet,9%) simulations too. The significant improvements of weekly GVF on model soil temperature (0.8 K) are found for both surface and root zone soil layers, but which are insignificantly degraded by using monthly albedo. Both near real time GVF and albedo have insignificant impacts on model surface net short wave radiation (S Wnet), which leads to using both of dynamic data performs obviously improvement on model SWnet.4. The SMOPS-Blended satellite soil moisture data product has higher spatial coverage on high-(84.2%), middle-(90%) and low-latitude (79.2%) areas. Spatial patterns of surface layer soil moisture climatologies of SMOPS-Blended retrievals and Noah model simulations agree well on global domain. But there are important differences between model and satellite SM in North America and North Eurasia. Improvements of Noah model simulated soil moisture (11%-25%) and soil temperature (6%-26%) through assimilating SMOPS-blended soil moisture data areas are found for different GVF conditions while the best performance of those simulations is obtained for areas with intermedian GVF values. Temporal correlations between in situ measurements and model simulations of SWnet/LWnet are higher with assimilating SMOPS-Blended soil moisture product than without the benefit of data assimilation.5. The benefit of assimilating satellite data is found to be subjective satellite data quality. Assimilating quality controlled satellite data has assimilation better effect on model simulations. Based on in situ measurements of the surface and root zone soil moisture, the time serie of CONUS domain-averaged RMSE of Noah LSM simulations from assimilating quality-controlled satellite soil moisture data shows the best performances while all three assimilation cases demonstrated simulation performance improvements.The general conclusions from the above studies are the following:land surface model simulations can be significant improved by using MODIS-C5 land cover, current year weekly AVHRR GVF and monthly MODIS albedo and assimilating SMOPS Blended soil moisture data. The ensemble size of ensemble Kalman filter should be set as 12 to obtain the maximum assimilating efficiency in a sequential data assimilation system. Quality controlled satellite data may enhance the benefit of the assimilation of SMOPS product into land surface models.