A cloud filtering method for microwave upper tropospheric humidity measurements
1Lulea Technical University, Dept. of Space Science, Kiruna, Sweden
2IUP, University of Bremen, Bremen, Germany
3Satellite Applications, Met Office, Exeter, UK
4RSMAS, University of Miami, USA
5Dept. of Radio and Space Science, Chalmers University of Technology, Gothenburg, Sweden
Abstract. The paper presents a cloud filtering method for upper tropospheric humidity (UTH) measurements at 183.31±1.00 GHz. The method uses two criteria: a viewing angle dependent threshold on the brightness temperature at 183.31±1.00 GHz, and a threshold on the brightness temperature difference between another channel and 183.31±1.00 GHz. Two different alternatives, using 183.31±3.00 GHz or 183.31±7.00 GHz as the other channel, are studied. The robustness of this cloud filtering method is demonstrated by a mid-latitudes winter case study.
The paper then studies different biases on UTH climatologies. Clouds are associated with high humidity, therefore the possible dry bias introduced by cloud filtering is discussed and compared to the wet biases introduced by the clouds radiative effect if no filtering is done. This is done by means of a case study, and by means of a stochastic cloud database with representative statistics for midlatitude conditions.
Both studied filter alternatives perform nearly equally well, but the alternative using 183.31±3.00 GHz as other channel is preferable, because that channel is less likely to see the Earth's surface than the one at 183.31±7.00 GHz.
The consistent result of all case studies and for both filter alternatives is that both cloud wet bias and cloud filtering dry bias are modest for microwave data. The recommended strategy is to use the cloud filtered data as an estimate for the true all-sky UTH value, but retain the unfiltered data to have an estimate of the cloud induced uncertainty.
The focus of the paper is on midlatitude data, since atmospheric data to test the filter for that case were readily available. The filter is expected to be applicable also to subtropical and tropical data, but should be further validated with case studies similar to the one presented here for those cases.