Optimal Calibration of Radiometers Using System Identification Techniques

Microwave radiometers are valuable for obtaining geophysical information such as the ocean surface salinity, stratiform precipitation, water vapor, temperature, soil moisture, vegetation cover, and others [1]. These environmental monitoring applications of passive microwave remote sensing systems require precision measurements of brightness temperatures over a range of more than 300K with absolute accuracy as good as 0.1K or better depending on the application. Therefore, developing an accurate and practical internal radiometer/sensor calibration technique becomes necessary and critical to future radiometer application success. The classical two-load calibration technique is in terms of a predominantly linear dependence between brightness temperature (T) and detected voltage (v) [2]. This simple relationship , b mT v where m is the receiver gain and b is the offset, is often adequate to describe the aggregate system response of the instrument. Subsequently, the classical two-load calibration technique is widely applied in many radiometer calibration applications with great success in the past thirty years [2-6]. However, due to the following reasons the classical two-load technique is inadequate to account for variations in system response that occur in the path from the antenna and calibration references to the receiver. First, there can potentially be unwanted thermal emission sources in between the calibration references and the receiver, thus producing additional unaccounted emission. Second, the calibration references may themselves be imperfect blackbodies, thus reflecting radiation from other unknown sources [7, 8]. They may also exhibit instabilities that require characterization or redundancy [9]. Third, there are in general, unknown emissions from losses in the antenna itself. In calibration reference design, the parametric optimization of target geometry has been discussed with respect to electromagnetic and thermal analysis [7, 8] including trough-to-tip temperature profile, wave scattering and absorption from the target surface, target shapes, and coating thickness. Long-term calibration stability can be achieved with frequent recalibration of the noise diodes using external calibration techniques [9]. In our radiometer studies, we have found that radiometer front-end components produce significant thermal noise contribution with the following characteristics: 1) these thermal noise contributions are very hard to be determined to desired levels of precision, 2) their emission changes subtly when components are removed and re-installed, 3) they exhibit emission variations with thermal drift. Further, the gain and offset of a radiometer are not constants but vary with time and ambient temperature. To this end we find that in radiometer calibration there are two time scales and two classes of stability: radiometers commonly exhibit random fluctuations in gain (therefore in offset) that occur on short time scales (~0.001-100 sec.) and fluctuations in noise temperature that are the result of relatively slow but measurable thermal variations need to be considered. Therefore, instead of describing radiometer calibration in terms of a simple linear dependence between brightness temperature and detected voltage, in our study a radiometer is treated as linear instrument with a number of unknown non-ideal front-end parameters in addition to radiometer systematic gain and offset. Based on the above assumption a general expression for the response of a radiometer to antenna temperature, with additional input data being the measured front-end components temperatures is defined to characterize the non-idealities: