L-band Soil Moisture Retrieval at Different Spatial Scales: From Satellite to Paddock Scale

R. Panciera, J.P. Walker, E. Kim, J.D. Kalma, J. Hacker and O. Merlin

Power Point Presentation

The National Airborne Field Experiment 2005 (NAFE’05) provides a unique wealth of multi-resolution L-band observations over a variety of land covers and wetness conditions. This data set is very suitable to address a number of issues related to near-surface soil moisture mapping from L-band passive microwave sensors onboard satellite missions and their use in a range of applications such as weather prediction, water resources management and irrigation practices. For most applications, the spatial scale at which predictions are needed is finer than the typical observation scale of such sensors (~50km). The physical meaning of the spatially averaged measurements provided by the sensor over such a large area, as well as the mismatch between the “observation scale” and the “prediction scale” is yet to be properly understood.

Current retrievals techniques of near-surface soil moisture from L-band (1.4 GHz) observations are based on inversion of radiative transfer models, and have been developed making use of tower-mounted radiometers, with resolutions of the order of 10’s of meters. Synthetic studies show that there are strong non-linearities involved in such models with respect to the parameters characterizing the land surface. These stem the question whether the same model can be applied at multiple spatial scales, and in particular at satellite footprint scale, where the land surface heterogeneity within the radiometer field of view is expected to be significant.

In this study multiple resolution, independent L-band passive microwave observations from NAFE’05 were used to verify the transferability of current soil moisture retrieval models across scales ranging from 10’s meters to 10’s kilometers. Top 5cm near-surface soil moisture ground measurements were correlated with airborne L-band observations at scales ranging from 1km to 62.5m. This relationship was investigated for different land cover and wetness conditions. The same observations were then used to retrieve near-surface soil moisture, and the accuracy of the retrieval was compared across scales and related to the heterogeneity of the land surface features within each pixel.