Quantifying the effect of emissivity on the relations between AVHRR split window temperature difference and atmospheric precipitable water over land surfaces

The effect of emissivity on the relations between the NOAA-9 AVHRR infrared split window temperature difference (ΔT = difference of brightness of temperatures for 10.1–11.5 μm and 11.2–12.6 μm bands) and atmospheric precipitable water (W), namely, ΔT = α + βW and W = α′ + β′ ΔT, is quantified in terms of red reflectance (0.55-0.72 μm) observed by the AVHRR. Two years (1987 and 1988) of AVHRR and radiosonde observations at 44 globally distributed land surface locations have been used in the analysis. Predictions of a radiative transfer model are tested using these observations, which indicate possible discrepancies between the predicted and the observed results. Estimates of 10.1–11.5 μm band emissivity have been made by interpreting the intercept of ΔT−W regression according to the theoretical predictions, and these emissivities were found to be linearly related to the red reflectance. Precipitable water estimated from the observed ΔT and red reflectance (by the AVHRR) for the two years at 11 locations over South American and Europe when compared with the observed precipitable water (from radiosonde) gave a mean absolute error of 5 mm and a bias of 4 mm, with the explained variance of 79%. Some inherent uncertainties in the data used in the present analysis have been noted.