Analysis of microwave radiometer noise

Noise usually stands for any disturbance that tends to obscure a desired signal: in passive remote sensing, the desired signal is the thermal noise radiation from the geophysical scene of interest. Disturbing noise includes the receiver intrinsic noise. Calibration of radiometers has involved the noise power quantification between the noise of interest, and the receiver internal noise. With the emergence of extrinsic radio frequency interferences and the increasing interest in the third and fourth Stokes parameters, noise power separation and therefore calibration has been even more challenging.; In this dissertation, noise in microwave radiometers was investigated to facilitate calibration. This study evolves naturally from a detailed analog block wise perspective, to an intermediate analog/digital calibration scheme, to a global system approach where only the output time series is needed. We found that noise power distribution within the functional blocks of an analog adding polarimater can be described using the relationship between power waves and transfer function and experimental characterization of the functional blocks. This blockwise approach yields an assessment tool describing the actual noise power distribution within the system, and doubles as a calibration technique using only external black body looks for applications where a 1-2K accuracy is sufficient. Applied to a structured senescent corn canopy, a non-zero polarimetric signal was found along the azimuth direction.; For digital correlating polarimeters, where the analog functional blocks are not as distinct as in the analog case, we developed an equivalent Stokes vector model (ESVM) that maps the polarimeter switch position, noise diode configuration and external look, to a 'fictitious' external equivalent Stokes vector when the polarimeter is in a neutral configuration. The ESVM was used as an input to the linear calibration model, to solve for the gain and offset parameters. We found that an internal correlated noise source combined with independent reference switch control can replace an external polarimetric target provided equivalent Stokes temperatures have sufficient differences in all Stokes vector elements.; Digital radiometers with high speed and high resolution analog to digital converters make the output time series available for study. Autoregressive spectral estimation was applied to the time series obtained at the output of the ADC of a radiometer receiver, and fitting WLS regression to a generalized system spectral model shows that the equivalent receiver noise temperature (Trec) has a non-flat spectral characteristics, and that a radiometer system colors white noise input.