247

Task 247

GPM Algorithm Development

Principal Investigator(s):

J. Munchak

Sponsor(s):

A. Hou

Last Updated:

October 26, 2012 15:26:13


Description of Problem

The general objective is to develop a general framework and state-of-the-art algorithms to advance precipitation observations from space using combined information from active and passive microwave sensors. In particular, two areas of investigation have been selected: 1) determining the limits of detection of snowfall from passive-only microwave sensors; and 2) analyzing the sensitivity of the combined radar-radiometer algorithm to non-precipitation parameters such as cloud water and water vapor.

Scientific Objectives and Approach

For the first area of study (limits of snowfall detection), an observational approach is being employed using coincident CloudSat and AMSU-B overpasses. The CloudSat radar is an active sensor with greater sensitivity than the GPM Dual-frequency precipitation radar (DPR), although it lacks the cross-track scanning ability. AMSU-B observes similar frequencies to the GPM microwave imager (GMI). Taking advantage of A-Train sensors such as the AIRS, AMSR-E, and MODIS instruments on the Aqua satellite, emissivities are calculated under clear-sky conditions and stored in a database along with surface parameters available from AMSR-E products. These emissivities are then used as input to a radiative transfer model along with AIRS temperature and water vapor profiles to simulate microwave radiances for all matched CloudSat/AMSU-B footprints. In the presence of precipitation-sized ice particles, it is expected that the simulated radiances will be significantly higher than observed due to the (intentional) absence of ice scattering in the forward model. The output of this procedure may then be analyzed to determine the minimum signal, in terms of CloudSat reflectivity, that can be reliably discerned from non-precipitating scenes by each AMSU-B channel or combination thereof.

In the second area of study (sensitivity of combined algorithm), data from field experiments and cloud-resolving models will be used as input to radiative transfer models and a 1D DPR algorithm (Grecu et al. 2011). The current proposed combined precipitation algorithm for GPM involves two steps: 1) Deconvolution of the GMI brightness temperatures (Tbs) to DPR resolution, and 2) Filtering of candidate profiles given the deconvolved Tbs. This research will mainly focus on the information content present in the Tbs in the second step, particularly regarding non-precipitation parameters such as surface emissivity, cloud water, and water vapor profiles. The outcome of this research can be used to guide future field experiments to better determine a priori estimates of these parameters, or, in the case of those that are well-determined by the combined algorithm, to incorporate those into GPM output products.

Accomplishments

At the present time some preliminary results from the CloudSat/AMSU-B thresholds of snowfall detection are available. The top panels of the figure below show, by channel, the detected fraction of CloudSat reflectivity profiles. These show that the channels near the 183.1 GHz water vapor absorption line are most sensitive to snow, with increases in detected fraction beginning near -5 dBZ (corresponding to a snowfall rate of about 0.1 mm/hr liquid equivalent according to Kulie and Bennartz (2009)) and over 50% of profiles with reflectivity greater than 10 dBZ are detected when the simulated brightness temperature is at least 3 K greater than observed. The ground-sensitive channels (89 and 150 GHz), meanwhile, do not respond to reflectivities below about 10 dBZ, corresponding to a snowfall rate of 0.5 mm/hr liquid equivalent. The bottom panels show, in a similar manner, the false alarm percentage. The 183±1 GHz channel has the lowest false alarm percentage, with rates of less than 50% for profiles with reflectivity up to -5 dBZ when a brightness temperature difference of less than about -6K is observed. Similar sensitivities to the maximum reflectivity in the column and the column-integrated reflectivity have also been calculated (not shown). These results are preliminary and it is anticipated that further refinements to the surface emissivity database and use of physical emissivity models may result in higher detection and lower false alarm percentages.

Other Publications and Conferences

Munchak, S.J. G. S. Jackson, B. J. Johnson, 2011: Thresholds of Passive Microwave Snowfall Detection Determined Using CloudSat and AMSU-B. European Geosciences Union General Assembly, Vienna, Austria, April 4-8, 2011 (abstract accepted)

Task Figures


Fig. 1 – Top panels: Percentage of CloudSat profiles with surface reflectivity greater than value on the x-axis corresponding to an AMSU-B brightness temperature colder than simulated non-precipitating value by the amount shown on the y-axis. Bottom panels: Fraction of AMSU-B footprints with brightness temperature difference less than y-axis value corresponding to CloudSat surface reflectivity less than value on x-axis.