An Assessment of MWR Retrievals of Liquid Water Path Using Clear-sky Data
Marchand, R.T. and Ackerman, T.P., Pacific Northwest National Laboratory
Twelfth Atmospheric Radiation Measurement (ARM) Science Team Meeting
The ARM radiometers make measurements at two frequencies. These two measurements are used to infer two quantities, the total column vapor and liquid water. Both microwave emission frequencies respond to both the liquid and vapor, but one channel is more sensitive to vapor (23.8 GHz) and the other is more sensitive to liquid water (31.4 GHz). Under clear sky conditions we can use one measurement to solve for the water vapor and determine if the forward calculation from a radiative transfer model in the other channel matches the measured value. Any difference between the calculation and the measurement directly reflects errors that are introduced into the liquid retrieval if a cloud had been present. Using the above approach, we have studied over a dozen carefully identified clear sky periods from the Nauru, Barrow, and Southern Great Plains ARM sites. The ARM microwave radiometer data were examined during these periods using three microwave absorption models MonoRTM, Liebe 87, and Rosenkranz 98. From these case studies this poster shows that clear sky differences between the model and measured brightness temperature in the 31.4 GHz channel can be as large as +/- 1.5 K, which is equivalent to about +/- 30 g/m2 in liquid water path. The forward model calculations require profiles of atmospheric temperature and pressure with altitude, as well as, the partitioning of water vapor with altitude. In the model calculations, radiosonde data is used to provide this information and it is observed that choosing different nearby-in-time sonde data could make as much as a 1 K difference in the 31.4 GHz brightness temperatures (~ 20 g/m2). It is also clear that interpolating the sonde data in time will not solve this problem. At the Barrow site, the Rosekranz 98 model agrees with the measurements better than the other two models (which are in close agreement), but the reverse is true at Nauru. The Rosenkraz 98 model and the other two models differ by about 1 K, which implies a potential bias of ~20 g/m2. Half the difference between Rosekranz 98 and other models is from the O2/N2 contribution and half is the water contribution (some in line and some in continuum). It is important to realize that even if the uncertainties in the model and atmospheric profile data could be eliminated, that the microwave radiometer calibration is only good to ~ 0.3 to 0.5 K under clear-sky conditions and ~ 0.6 to 0.8 K under cloudy-sky conditions. This means that for this instrument there will always be at least a 10 to 15 g/m2 uncertainty.
Note: This is the poster abstract presented at the meeting; an extended version was not provided by the author(s).
