Online vegetation parameter estimation using passive microwave observations for soil moisture estimation

Vegetation affects the ability to estimate soil moisture from passive microwave observations by attenuating the surface soil moisture signal. To use radiobrightness observations in land data assimilation a vegetation opacity parameter is required as input to a radiative transfer model, which maps surface soil moisture to radiobrightness temperature. Ancillary vegetation data options, for example visible/NIR remote sensing data or monthly climatology, are external and uncertain. Other approaches which rely on dual incidence angles are not feasible for the proposed NASA SMAP (Soil Moisture Active Passive) mission which has a fixed incidence angle instrument.

We propose to estimate the vegetation opacity parameter online using an ensemble Kalman filter. The parameter is part of the observation operator, the radiative transfer model. A state augmentation approach is used where the vegetation parameter is added to the soil moisture state vector. The filter consists of a two-layer soil hydrology model and the radiative transfer model and is tested using simulated passive microwave observations. We study both time-invariant and time-varying parameter cases. The novel time-varying parameter case yields satisfactory parameter estimation results. The time-varying vegetation opacity case is derived from field leaf area index observations from a field site of corn crop during a growing season. Adding small variance mean zero Gaussian noise is required in the time-varying case to converge to the true parameter, consistent with theory. Vegetation information could be extracted from passive microwave observations using the data assimilation system itself. This online approach differs from physical methods which require dual frequency, dual polarization or dual incidence angle observations.