Theoretical and empirical scale dependency of Z‐R relationships: Evidence, impacts, and correction

Abstract

[1] Estimation of rainfall intensities from radar measurements relies to a large extent on power-laws relationships between rain rates R and radar reflectivities Z, i.e., Z = a*R^b. These relationships are generally applied unawarely of the scale, which is questionable since the nonlinearity of these relations could lead to undesirable discrepancies when combined with scale aggregation. Since the parameters (a,b) are expectedly related with drop size distribution (DSD) properties, they are often derived at disdrometer scale, not at radar scale, which could lead to errors at the latter. We propose to investigate the statistical behavior of Z-R relationships across scales both on theoretical and empirical sides. Theoretically, it is shown that claimed multifractal properties of rainfall processes could constrain the parameters (a,b) such that the exponent b would be scale independent but the prefactor a would be growing as a (slow) power law of time or space scale. In the empirical part (which may be read independently of theoretical considerations), high-resolution disdrometer (Dual-Beam Spectropluviometer) data of rain rates and reflectivity factors are considered at various integration times comprised in the range 15 s -64 min. A variety of regression techniques is applied on Z-R scatterplots at all these time scales, establishing empirical evidence of a behavior coherent with theoretical considerations: a grows as a 0.1 power law of scale while b decreases more slightly. The properties of a are suggested to be closely linked to inhomogeneities in the DSDs since extensions of Z-R relationships involving (here, strongly nonconstant) normalization parameters of the DSDs seem to be more robust across scales. The scale dependence of simple Z = a*R^b relationships is advocated to be a possible source of overestimation of rainfall intensities or accumulations. Several ways for correcting such scaling biases (which can reach >15-20% in terms of relative error) are suggested. Such corrections could be useful in some practical cases where Z-R scale biases are significant, which is especially expected for convective rainfall.Citation: Verrier, S., L. Barthe`s, and C. Mallet (2013), Theoretical and empirical scale dependency of Z-R relationships:

Data-driven clustering of rain events: microphysics information derived from macro-scale observations

“…Rain rates lower than T 0 are thus set to zero. This conventional threshold is also chosen to ensure coherency with previous studies (Verrier et al, 2013;Llasat et al, 2001). In the present study, we worked with two data sets recorded during the period between July 2008 and July 2014, at the Site Instrumental de Recherche par Télédétection Atmosphérique (SIRTA 1 ) in Palaiseau, France.…”

“…The notion of an SOM, introduced by Kohonen (1982Kohonen ( , 2001, makes use of a popular clustering and visualization algorithm. SOM is a neural network algorithm based on unsupervised learning, derived from the technique of competitive learning (Kohonen, 1982(Kohonen, , 2001Vesanto and Alhoniemi, 2000). It may be considered as a nonlinear generalization, which has many advantages over the conventional feature extraction techniques such as empirical orthogonal functions (EOF) or PCA (e.g., Liu et al, 2006).…”

Two months of disdrometer data in the Paris area

“…This approach has been implemented by numerous authors (Jaffrain and Berne, 2012a;Leinonen et al, 2012;Ryzhkov et al , 2005;Verrier et al, 2013;Gires et al, 2015). It should be mentioned that it relies on strong unrealistic hypothesis, notably the homogeneity of the DSD within a radar bin (see Gires et al, 2017b for a discussion of the limitations of this approach).…”

Two months of disdrometer data in the Paris area

“…Using external information and assumptions about raindrop scattering properties, it is also possible to estimate, from the DSD, equivalent local pointwise radar quantities and hence study and possibly help improve radar rainfall retrieval algorithms. This approach has been implemented by numerous authors (Jaffrain and Berne, 2012a;Leinonen et al, 2012;Ryzhkov et al, 2005;Verrier et al, 2013;Gires et al, 2015). It should be mentioned that it relies on a strong unrealistic hypothesis, notably the homogeneity of the DSD within a radar bin (see Gires et al, 2018, for a discussion of the limitations of this approach).…”