The merging of radiative transfer based surface soil moisture data from SMOS and AMSR-E
This paper evaluates a methodology to integrate surface soil moisture retrievals from SMOS and AMSR-E into a single, consistent dataset retrieved by the Land Parameter Retrieval Model (LPRM). In a first step, the SMOS LPRM soil moisture retrievals were used as the baseline for optimizing the interna...
Published in: | Remote Sensing of Environment |
---|---|
Main Authors: | , , , , , , , , |
Other Authors: | , , , , , , , , , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
HAL CCSD
2017
|
Subjects: | |
Online Access: | https://hal.science/hal-01595237 https://doi.org/10.1016/j.rse.2016.11.026 |
Summary: | This paper evaluates a methodology to integrate surface soil moisture retrievals from SMOS and AMSR-E into a single, consistent dataset retrieved by the Land Parameter Retrieval Model (LPRM). In a first step, the SMOS LPRM soil moisture retrievals were used as the baseline for optimizing the internal parameterization (i.e. surface roughness and single scattering albedo) of the AMSR-E LPRM retrievals. Secondly, to overcome the uniqueness of these datasets a linear scaling approach was applied resulting in a consistent soil moisture dataset. The new parameter set from the first step is similar for the two (low) frequencies of AMSR-E (i.e. C- and X-band) further improving their inter-comparability for both soil moisture and vegetation optical depth. Soil moisture retrievals from these AMSR-E frequencies were globally merged based on the availability of brightness temperatures that are free from RFI contamination (resulting in AMSR-E LPRMN). This new product was evaluated against both the SMOS LPRM product in the overlapping period (July 2010 to October 2011), as well as the standard, publicly available AMSR-E LPRM dataset (AMSR-E LPRMV3) for an almost 9 year period (January 2003 to October 2011). For the overlapping period, the AMSR-E and SMOS LPRM products show high temporal correlation coefficients (0.60 < R < 0.90) and low root mean square errors (rmse < 0.04 m3 m− 3) for NDVI values up to 0.60. Their agreement tends to drop over the well-known challenging areas such as the arctic region and tropical rainforest. A detailed evaluation over in situ sites from 5 in situ networks worldwide showed that AMSR-E LPRMN often outperforms SMOS LPRM in sparsely vegetated areas, with generally higher correlation coefficients in areas with NDVI < 0.3, and in general a lower unbiased rmse (ubrmse). In line with theoretical expectations, SMOS LPRM outperforms the AMSR-E LPRM product over the more densely vegetated areas. The newly developed AMSR-E LPRMN product was also compared against AMSR-E LPRMV3, revealing a ... |
---|