Abstract
A technique is presented whereby the noise level of satellite measurements of the atmosphere and earth can be estimated. The technique analyzes a spatial array of data measured by a satellite instrument. A minimum of about 200 satellite measurements is required, preferably in a regular pattern. Statistical structure analysis is used to describe a combination of the mean gradient and noise in the data. The noise level is then estimated by separating out the gradient information and leaving only the noise. Results are presented for four satellite sounding instruments, and effective blackbody or brightness temperature noise levels were compared to prelaunch specifications or inflight calibrations for each instrument. Comparisons showed that in the absence of cloud-contaminated measurements (in the case of infrared data) and away from the highly variable ground surface, the noise level of various satellite instruments can be obtained without the need for calibration data. The noise levels imply how much spatial averaging is possible, without smearing the detected geophysical gradient, and how much is necessary, to meet the absolute signal accuracy requirements for the intended use of the satellite measurements.
