Smith, E. A., J. E. Lamm, R. Adler, J. Alishouse, K. Aonashi, E. Barrett, P. Bauer, W. Berg, A. Chang, R. Ferraro, J. Ferriday, S. Goodman, N. Grody, C. Kidd, D. Kniveton, C. Kummerow, G. Liu, F. Marzano, A. Mugnai, W. Olson, G. Petty, A. Shibata, R. Spencer, F. Wentz, T. Wilheit, E. Zipser, 1998: Results of WetNet PIP-2 Project. J. Atmos. Sci., 55, 1483-1536.


The second WetNet Precipitation Intercomparison Project (PIP-2) evaluates the performance of 20 satellite precipitation retrieval algorithms, implemented for application with Special Sensor Microwave/Imager (SSM/I) passive microwave (PMW) measurements and run for a set of rainfall case studies at full resolution-instantaneous space-timescales. The cases are drawn from over the globe during all seasons, for a period of 7 yr, over a 60°N-17°S latitude range. Ground-based data were used for the intercomparisons, principally based on radar measurements but also including rain gauge measurements. The goals of PIP-2 are 1) to improve performance and accuracy of different SSM/I algorithms at full resolution-instantaneous scales by seeking a better understanding of the relationship between microphysical signatures in the PMW measurements and physical laws employed in the algorithms; 2) to evaluate the pros and cons of individual algorithms and their subsystems in order to seek optimal "front-end" combined algorithms; and 3) to demonstrate that PMW algorithms generate acceptable instantaneous rain estimates.

It is found that the bias uncertainty of many current PMW algorithms is on the order of ±30%. This level is below that of the radar and rain gauge data specially collected for the study, so that it is not possible to objectively select a best algorithm based on the ground data validation approach. By decomposing the intercomparisons into effects due to rain detection (screening) and effects due to brightness temperature-rain rate conversion, differences among the algorithms are partitioned by rain area and rain intensity. For ocean, the screening differences mainly affect the light rain rates, which do not contribute significantly to area-averaged rain rates. The major sources of differences in mean rain rates between individual algorithms stem from differences in how intense rain rates are calculated and the maximum rain rate allowed by a given algorithm. The general method of solution is not necessarily the determining factor in creating systematic rain-rate differences among groups of algorithms, as we find that the severity of the screen is the dominant factor in producing systematic group differences among land algorithms, while the input channel selection is the dominant factor in producing systematic group differences among ocean algorithms. The significance of these issues are examined through what is called "fan map" analysis.

The paper concludes with a discussion on the role of intercomparison projects in seeking improvements to algorithms, and a suggestion on why moving beyond the "ground truth" validation approach by use of a calibration-quality forward model would be a step forward in seeking objective evaluation of individual algorithm performance and optimal algorithm design.