Widespread sampling bias in atmospheric data

Abstract Air’s compressibility practically precludes unbiased atmospheric sampling because warmer air, occupying more specific volume at equal pressure, is more likely to be measured. Weighted statistics are required to compensate for the effects of sampling bias on statistical moments of state and flow variables. These effects occur from the microscale to the global scale, at which uncorrected sampling bias causes underestimation of both global warming and its statistical certainty.