The concept of normalized PSD. The technique of normalisation of the PSD developed at CETP allows characterisation of any observed hydrometeor spectrum by three independent elements: (i) the liquid water content LWC (or ice water content IWC, according to the phase of the precipitation); (ii) the "mean weighted mass equivalent diameter” Dm_eq; (iii) the intrinsic shape of the PSD. For liquid particles (raindrops), Dm_eq coincides with the well known mean volume diameter Dm. For ice particles, the same concept is applied to the spectrum of melted particles. The intrinsic shape of the DSD is obtained by normalising a raindrop spectrum in diameter, by Dm, and in concentration by N0*=44/(prw). LWC/Dm4 (Testud et al., 2001). For an ice particle spectrum, the same normalisation is applied to the corresponding melted particles (thus replacing Dm by Dm_eq and LWC by IWC. The normalisation technique allows us to express the relationships between two moments of order i and j of the PSD as quasi universal expressions parameterised by N0*.
The derivation of global maps of N0* from the TRMM precipitation radar. Within the rain layer, it may be assumed that N0* is constant. With the TRMM PR, the specific attenuation A may be profiled from the apparent (attenuated) reflectivity Za , using as auxiliary observation the path integrated attenuation PIA derived from the surface echo (this is implicit in the standard 2A25 algorithm). It is shown that this algorithm implies an adjustment of N0* within the rain layer. It is thus possible to derive global maps of N0* from the TRMM PR observation. Two one month periods have been analysed. The most striking feature is probably the contrast between ocean and land. Over ocean a typical value of N0* is about 2x107m-4, while over land, it is rather 4x106m-4.
The algorithm for the TRMM precipitation radar in the ice layer. In the ice layer, because of the aggregation process, N0* varies considerably from the top of the cloud to the bottom of the ice layer (freezing level), decreasing typically of several decades. Thus any attempt to apply a unique R-Z or IWC-Z relationship is inappropriate. This paper investigates what kind of vertical model for N0* variation could be used.