The purpose of the SMOS data processor is to convert the raw data downloaded from the satellite into calibrated microwave brightness temperature maps at the top of the atmosphere which, using suitable algorithms and ancillary data, can ultimately be used to product useful geophysical measurements such as global soil moisture and ocean salinity data products.
To this end, the SMOS Level 1 processor is a vital element in the space segment to ground processor chain which forms the minimum configuration needed to produce meaningful results (Zundo).
The processing of raw data to brightness temperature is not direct and the raw data needs to undergo a complex sequence of multi-step software processing (summarised in Fig. 9) in order to obtain brightness temperature first (Level 1 processing) and SM and OS later (Level 2 processing). Implementation of this processing is performed in the DPGS by two separate multi-node computer facilities named respectively the Level 1 Processor and the Level 2 Processor.
Regarding the Level 1 processing stage, the following three steps have been identified:
1) Level 0 processed to Calibrated Visibilities (Level 1a)
2) Image reconstruction i.e. Calibrated Visibilities to Brightness temperatures
3) Brightness temperature to geo-located map (Level 1c).
Level 2 processing relies on co-located brightness temperature measurements at different angles. In general, LEO satellite measurements taken at different times are never co-located due to satellite motion so at each moment a different patch of surface is sensed making it necessary to interpolate in space with a corresponding loss of accuracy. MIRAS, however, is a synthetic aperture radiometer so each of the sensing beams, which have an approximate width of 2.6 deg., is created mathematically by combining the data measured by each of the array's receiver-baselines during the process of image reconstruction. The directions in which each beam, at any time, is pointed can therefore be mathematically changed resulting in effect in a virtual "steerable" sensor with a resolution at nadir corresponding to a pixel size of 30 km, the - 3dB beam-width and its intersection with the ground.
This feature is highly valuable and has been exploited by SMOS in creating a fixed Discrete Global Grid (DGG). The SMOS DGG has been selected in a way as to offer the most uniform sampling of the earth's surface at a resolution of 15 km, twice the actual value at nadir.
The DGG, based on a hexagonal geometry, partitions the Earth's surface into approximately 2.6 million cells, it does not present any preferential directions or symmetry and is as accurate at middle and high latitudes as it is at the equator. An example of the hexagonal sampling for a SMOS data set is shown in Fig. 10.
The value of brightness temperature in an area sensed by MIRAS at each snapshot can then be computed at each DGG point in the instantaneous field-of-view (FOV) and a Level 1c product built consisting of pole-to-pole swaths of fixed pixel records each listing the number of attached measurements (more near the centre, less toward the edges) and their value. No interpolation is needed and each pixel can be processed independently from any other by the Level 2 processors using only data associated with that pixel in Level 1c, producing considerable computational advantage.
Since the Level 1c defined in this way consists of a list of pixel and brightness temperature values (although variable in number), it can be easily plotted and accessed by user applications and there is no need to "search" in time for data related to each ground pixel.
It is to be noted that L-Band signals undergo rotation while propagating through the atmosphere due to the presence of the ionosphere and the earth's magnetic field so that the values measured by MIRAS at the Top-Of-Atmosphere (TOA) need to be corrected for on-ground use. This correction depends among others on the varying geophysical input like total electron content (TEC) which is not known exactly at the moment of processing. In order to avoid permanently changing the brightness temperature value with a value that is not accurate, the correction is computed but not applied and stored in the data product independently. In this way the user can easily compute the brightness temperature at the Earth's surface using the pre-computed correction or apply a better one if known.
Fig. 10 SMOS Level-1c simulated data for the coast of Portugal on DGG
Was this article helpful?