Albin J. Gasiewski
School of Electrical and Computer Engineering
Georgia Institute of Technology

I. Introduction

A relatively new area in Earth remote sensing concerns the measurement and interpretation of the fully-polarized microwave emission signature from geophysical surfaces and volumes. In essence, fully polarized microwave radiometry is an extension of the well-known techniques used to measure and interpret the vertically and horizontally polarized microwave brightness temperatures (Tv and Th). These two quantities are the first two Stokes’ parameters. Under specific conditions of observation, however, the third Stokes’ parameter (TU = Re) also conveys geophysical information and can be used for remote sensing. Collectively, the four modified Stokes parameters (or Stokes’ vector) provide a complete characterization of a Gaussian-random electromagnetic field.

It is the purpose of this paper to discuss the state of technology with regard to the measurement and application of the full Stokes’ vector in passive remote sensing. In order to report new developments, the paper will refrain from discussing mature applications of the first two modified Stokes’ parameters. Many such sensing and retrieval studies have been reported in the literature.

II. Application Area: Remote Sensing of Ocean Surface Winds

The wind-driven ocean is decidedly anisotropic, with the wave crests and troughs oriented primarily in the direction of the prevailing wind. Both gravity and capillary wave direction are correlated with wind direction; capillary waves are more closely related to local wind direction while gravity waves are related to the average wind direction over a long fetch. The anisotropic nature of this surface gives rise to azimuthal variations in its bistatic scattering function. One result of these variations is that the upwelling Stokes’ vector exhibits related azimuthal variations of up to a few degrees Kelvin in amplitude. These brightness variations are due both to harmonic variations in the emissivity (or equivalently, the reflectivity) and to harmonic variations in the elevation angle of the surface scattering lobes. The extent of the latter contribution depends on the magnitude of the first derivative of the downwelling radiation field with respect to elevation angle. In general, the amplitude of the azimuthal harmonics are dependent on wind speed, angle of observation, and frequency. However, azimuthal variations in the upwelling brightness have been observed from X- to W-bands.

The aforementioned anisotropic brightness signatures in Tv and Th could potentially facilitate the retrieval of ocean surface wind direction from space using purely passive means. However, the contribution from the relatively large average brightness temperatures and tends to make accurate measurements of subtle azimuthal brightness variations difficult. Moreover, the Tv and Th variations are co-phased, resulting in a four-fold ambiguity in the retrieved wind direction. Here, measurements of the third Stokes’ parameter TU are expected to facilitate wave direction measurement by virtue of the zero-mean nature of the azimuthal TU signature along with the quadrature phasing of the dominant TU variation with respect to Tv and Th. It is worthwhile to note that no useful information on wave direction has been observed to be carried by the fourth Stokes’ parameter TV =Im .

III. Current Status of Instrumentation

Although both vertical and horizontally polarized brightness temperatures have been used for Earth remote sensing of the atmosphere, ocean surface, and cryosphere for almost two decades, the utility of the third parameter has only recently been shown. An important issue, however, is the optimal architecture for a polarimetric microwave radiometer. A simple measurement of TU can be made using a dual-channel orthogonal-linear polarized radiometer with an IF correlator. This method is preferable for satellite applications since integration time is maximized. Two receivers, however, are required and the orthogonal-mode signals must be kept phase coherent.

The IF correlator can either be an adding type (add then-detect) or multiplying type. If an adding correlator is used then the detected signals must be subtracted in software to remove Tv and Th, and thus to recover TU or TV. If a multiplying correlator is used, then no software differencing is required. The multiplying correlator can be of the analog type; however, reduced-bit digital correlators yield nearly the same accuracy as analog correlators. By switching a 90° phase shift into one of the two orthogonal-channel signal paths the correlating channel can be made to selectively measure either TU or TV. A 90° quadrature hybrid placed in the IF stage will permit measurement of both TU and TV simultaneously.

Another method employs three looks of a single-channel linearly-polarized radiometer, rotated in polarization by 45° for each look, along with an appropriate subtraction of the resulting brightness temperatures. Either mechanical or ferromagnetic rotation of the radiometer’s polarization can be employed. While these schemes (or variants thereof) have been demonstrated to be useful, the optimal configuration for satellite application remains an open issue.

Using any of the above architectures requires accurate absolute polarimetric calibration. The three-look mechanically-rotated radiometer can be calibrated using conventional hot-and-cold blackbody targets. A more accurate means of calibration, applicable for any of the above architectures, utilizes a polarized calibration standard. The polarized standard can be constructed using two targets (hot and cold) along with a polarization-splitting wire grid. Mechanical rotation of the polarized standard around a polarimetric radiometers feedhorn axis allows all twelve parameters (nine gain coefficients and three offsets) of a three-channel (e.g., Tv, Th, and TU) radiometer to be identified. The required viewing interval of a polarized standard depends on the stability of the radiometric hardware.