Over-ocean rainfall retrieval from multisensor data of the Tropical Rainfall Measuring Mission. Part II: Algorithm implementation

The objective of this paper is to establish a computationally efficient alg orithm making use of the combination of Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) and precipitation radar (PR) observations. T o set up the TMI algorithm, the retrieval databases developed in Part I ser ved as input for different inversion techniques: multistage regressions and neural networks as well as Bayesian estimators. It was found that both Bay esian and neural network techniques performed equally well against PR estim ates if all TMI channels were used. However, not using the 85.5-GHz channel s produced consistently better results. This confirms the conclusions from Part I. Generally, regressions performed worse; thus they seem less suited for general application due to the insufficient representation of the nonli nearities of the TB-rain rate relation. It is concluded that the databases represent the most sensitive part of rainfall algorithm development. Sensor combination was carried out by gridding PR estimates of rain liquid water content to 27 km x 44 km horizontal resolution at the center of gravi ty of the TMI 10.65-GHz channel weighting function. A liquid water dependen t database collects common samples over the narrow swath covered by both TM I and PR. Average calibration functions are calculated, dynamically updated along the satellite track, and applied to the full TMI swath. The behavior of the calibration function was relatively stable. The TMI estimates showe d a slight underestimation of rainfall at low rain liquid water contents (< 0.1 g m(-3)) as well as at very high rainfall intensities (>0.8 g m(-3)) an d excellent agreement in between. The biases were found to not depend on be am filling with a strong correlation to rain liquid water for stratiform cl ouds that may point to melting layer effects. The remaining standard deviations between instantaneous TMI and PR estimate s after calibration may be treated as a total retrieval error, assuming the PR estimates are unbiased. The error characteristics showed a rather const ant absolute error of <0.05 g m(-3) for rain liquid water contents <0.1 g m (-3). Above, the error increases to 0.6 g m(-3) for amounts up to 1 g m(-3) . In terms of relative errors, this corresponds to a sharp decrease from >1 00% to 35% between 0.05 and 0.5 g m(-3). The database ambiguity, that is, t he standard deviation of near-surface rain liquid water contents with the s ame radiometric signature, provides a means to estimate the contribution fr om the simulations to this error. In the range where brightness temperature s respond most sensitively to rainwater contents, almost the entire error o riginates from the ambiguity of signatures. At very low and very high rain rates (<0.05 and >0.7 g m(-3)) at least half of the total error is explaine d by the inversion process.