Incorporation of Four Dimensional Data Assimilation of Microwave Remote Sensing Observations into a Land Surface Scheme
Mahadevan Pathmathevan (1), Toshio Koike (1), Li Xin (3)
(1) Dept of Civil Engineering, University of Tokyo
(3) Cold and Arid Regions Environmental and Engineering Research Institute,Chinese Academy of Sciences, Lanzhou 730000, P. R. China
The main objective of land surface data assimilation is to produce a regular, physically consistent four dimensional representation of the land surface hydrological parameters from a heterogeneous array of in-situ and remote instruments which sample imperfectly and irregularly in space and time. The regular, physically consistent aspects of the procedure come from the use of models, and thus data assimilation is a discipline, which naturally integrates theory by models with real in-situ and satellite observations.
This paper describes the implementation and application of advanced four-dimensional data assimilation algorithm for land surface hydrological parameters. This algorithm minimizes the penalty, which is produced by remotely sensed passive microwave observations of brightness temperature or its dimensionless products such as Polarization Index (PI) and Index of Soil Wetness (ISW), and the numerically predicted results by modified simple biosphere model (Sib2) for GAME-Tibet Meso-scale experiment.
The frozen soil parameterized Sib2 model, called as a model operator, was used as the Land Surface Scheme to estimate soil moisture and temperature profiles at experimental site. These profiles can be produced with modified Sib2 scheme during the freeze-thaw cycle period also. And from the TRMM (Tropical Rainfall Measuring Mission) Microwave Instrument (TMI), the brightness temperature observations at different frequencies and different polarization were utilized in this study. In above four-dimensional data assimilation scheme the forward (radiative transfer) model, called as an observation operator, was used to estimate the surface brightness temperature from soil moisture and temperature profile.
As an important point of our assimilation algorithm, we used an annealing schedule with random generation of the state vector for minimizing the penalty function, instead of using usual ad-joint models. This scheme works very well. In our procedure, at initial stage the data assimilation algorithm was applied only to assimilate the in-situ observations at a station in Tibet, where all data is available, without applying the observation operator. Then with the smoothened results, the station's initial condition was used at nearest other station and the full assimilation scheme was applied to assimilate the soil moisture only with brightness temperature observations. The results show the estimated soil moisture profile with reasonable accuracy.
Finally, the effectiveness of brightness temperature observations from different frequency channel and different polarization, dimensionless indices and other physically based combinations were analyzed. From above conclusions the scheme is being extended to produce good initial conditions and reasonable soil moisture and temperature profiles for all grid points at the experimental site.
