Sie sind auf Seite 1von 22

892

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

Testing a Procedure for Automatic Classication of Hydrometeor Types


DUSAN S. ZRNIC
NOAA/National Severe Storms Laboratory, Norman, Oklahoma

ALEXANDER RYZHKOV
Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, Oklahoma

JERRY STRAKA
School of Meteorology, University of Oklahoma, Norman, Oklahoma

YIDI LIU
Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, Oklahoma

J. VIVEKANANDAN
National Center for Atmospheric Research, Boulder, Colorado (Manuscript received 18 April 2000, in nal form 7 September 2000) ABSTRACT Examples of automatic interpretation of polarimetric measurements made with an algorithm that classies precipitation, from an Oklahoma squall line and a Florida airmass storm are presented. Developed in this paper are sensitivity tests of this algorithm to various polarimetric variables. The tests are done subjectively by comparing the elds of hydrometeors obtained using the full set of available polarimetric variables with a diminished set whereby some variables have been left out. An objective way to test the sensitivity of the algorithm to variables and rank their utility is also devised. The test involves denition of a measure, which is the number of data points classied into a category using subsets of available variables. Ratios of various measures (similar to probabilities) dene the percentage of occurrence of a class. By comparing these percentages for cases in which some variables are excluded to those whereby all are included, a relative merit can be assigned to the variables. Results of this objective sensitivity study reveal the following: the reectivity factor and differential reectivity combined have the strongest discriminating power. Inclusion of the temperature prole helps eliminate a substantial number of spurious errors. Although the absence of temperature information degrades the scheme, it appears that the resultant elds are generally coherent and not far off from the elds obtained by adding temperature to the suite of polarimetric variables.

1. Introduction Knowing what precipitation type is reaching the ground is a fundamental prerequisite for accurate determination of amount. Thus, for quantitative precipitation estimation (QPE), rst a correct classication needs to be made so that appropriate semiempirical relations can be chosen to estimate the corresponding rates and/or accumulations. Because of sensitivity to hydrometeor concentration, shape, orientation, dielectric constant, and size, polarimetric variables have emerged as

Corresponding authors address: Dusan S. Zrnic, NOAA/National Severe Storms Laboratory, 1313 Halley Cr., Norman, OK 73069. E-mail: zrnic@nssl.noaa.gov

leading discriminators of precipitation type (Zrnic and Ryzhkov 1999). Very early in the development of differential polarization measurements, it became apparent that these could be used to determine the presence of hail and possibly gauge its size (Seliga and Bringi 1978). Later, Hall et al. (1980) proposed a table of relations between the reectivity factor and differential reectivity to classify hydrometeors; noteworthy are the descriptive (nonquantitative) entries in the table that, it turns out, are very suitable for building fuzzy classication rules. A more detailed table can be found in Doviak and Zrnic (1993); it served as a starting point for some decisiontree type classication schemes (e.g., Lopez and Aubagnac 1997; Carey and Rutledge 1996), which were used to discriminate among rain, graupel, and hail. A

2001 American Meteorological Society

JUNE 2001

ZRNIC ET AL.

893

similar decision type rule has been advanced and applied to data collected with a 5-cm wavelength radar by Holler et al. (1994). Building on the table in Doviak and Zrnic (1993), Straka and Zrnic (1993) developed a scheme based on fuzzy logic principles in which boundaries between classes were allowed to overlap, but the weighting (membership) functions were of a pulse type (i.e., values of 0 and 1). Subsequently, Straka (1996) used a variety of overlapping weighting functions with smooth transitions in one and two dimensions; this scheme became the basis of several subsequent attempts for classication of hydrometeors (Zrnic and Ryzhkov 1999; Vivekanandan et al. 1999). A further step in the development is reported by Liu and Chandrasekhar (2000) who used a feedback scheme to adjust the weights of the fuzzy classier; they also give a comprehensive set of values for the one-dimensional membership functions. Meanwhile Straka et al. (2000) have presented an extensive set of relations in the form of tables and twodimensional graphs that delineate regions, in the space of Z h (the reectivity factor for the horizontally polarized wave) and any polarimetric variable, where specic hydrometeor signatures reside. These graphs (or variants thereof ) have been used by the authors to build membership functions for fuzzy classication schemes (Straka 1996; Ryzhkov and Zrnic 1999), one of which has been implemented on the NCARs S-Pol radar (Vivekanandan et al. 1999). The purpose of this paper is to present details of the algorithm not found elsewhere and a methodology for sensitivity analysis to the various polarimetric variables used for classication. Verication, let alone simple comparisons of classication algorithms, is a daunting task because in situ observations are too few and often not coincidental with the radar measurements. Moreover, the radar sampling volumes are several orders of magnitude larger than the typical particle probes whose images contain uncertainties as well. Because there are other simpler means to develop and evaluate the algorithm, no comparison with in situ probes is attempted herein. Rather, we rely on spatial continuity, height above ground, and comparison with conceptual models to qualify the algorithms performance. An important thrust of our paper is a procedure we developed to compare the merits of the polarimetric variables for classifying various hydrometeors. This procedure and selfconsistency checks based on intuition, precipitation physics, and conceptual models can be carried out before in situ comparisons are attempted. In section 2 of the paper we briey describe the algorithm. Then follows a description of data that are used by the algorithm. In section 3 we present the performance of the algorithm on the data from the Cimarron (Zahrai and Zrnic 1993) and the S-Pol polarimetric radars (Lutz et al. 1997). Ramication for operational radars are then mentioned, and relative merits of various

polarimetric variables to isolate hydrometeor types are discussed. 2. Algorithm The classication algorithm described herein belongs to the fuzzy logic family. Much has been written about this method (Mendel 1995), and several articles with applications to meteorological problems have appeared (Straka 1996; Cornman et al. 1998; Vivekanandan et al. 1999; Liu and Chandrasekar 2000). Nonetheless, none of the jargon used in that discipline is needed to understand its principles. In the following description, we utilize the key nomenclature, dene twodimensional membership functions, and rely on analogy with probability density functions for reasons that will be apparent shortly. In its essence a classier assigns (maps) an observed point X in the multiparameter space to a class j (which refers to the bulk hydrometeor type such as rain, hail, graupel, snow, etc.). In this paper, the point X has for coordinates the following six variables: 1) the reectivity factor for horizontally polarized waves Z h , 2) the differential reectivity ZDR , 3) the specic differential phase KDP , 4) the cross-correlation coefcient between the horizontally and vertically polarized (copolar) waves hv (0), 5) the linear depolarization ratio LDR , and 6) the environmental temperature T. The crux of a hydrometeor classication scheme is to partition this six-dimensional space of observed variables into subsets such that each can be associated with a specic hydrometeor type ( j ). The classical statistical decision theory solves, in principle, this type of a problem as follows. It starts with the probability densities P j (X ) for each class. Then, the integral

P j (X ) dX

(1)

Vj

over a subset (here six-dimensional volume V j ), corresponding to the likely class j, gives the probability of correct classication, whereas the integral over the complement of Vj is the probability of misclassication of the specic class j. The choice of the volume V j amounts to identication of a boundary delineating the class; misclassication occurs because of the ambiguity at and in the vicinity of the boundaries between classes. By considering consequences of either decision (the point does or does not belong to a class), one can partition the space of the polarimetric variables (universe in the nomenclature of set theory) to optimize the outcome. Normally an accepted false alarm rate for misclassication would be adopted, and by trial and error, the boundary would be changed as long as the probability

894

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

of correct classication increases while the probability of misclassication is kept at or below the desired level. This is analogous to the dilemma facing weather forecasters who are about to issue warnings. Over warning increases the probability of detection at the expense of more false alarms. A proper balance between these two decisions is often a subjective call. In most cases of practical signicance, and certainly for hydrometeor classication, the probabilities of classes are not available, and it is unlikely that these would become known in the near future. Therefore, simplications are sought such that reasonable classication can be made. An obvious solution is to decouple the variables, consider each individually, and then make a nal judgment on the basis of weighted averages. This is the simplest form of a fuzzy classier (Mendel 1995). One-dimensional weighting functions W j (Y i ), somewhat similar to probabilities, are dened in a way that mimics ones expectation (Straka and Zrnic 1993; Vivekanandan et al. 1999; Liu and Chandrasekhar 2000). Here, Y i stands for any one of the polarimetric variables. An intuitive yet good practice is to restrict the maxima of W j s to be 1 and locate these at values for which the variable almost always corresponds to the jth class. Further from these maxima, the weighting function should decrease to reach zero in regions where the variable is unlikely to be associated with the jth class. To distinguish between reliable and less reliable variables, Y i , one can multiply the W j (Y i ) with a multiplicative factor A i less or equal to one. In this manner, relative changes of condence in various variables can be easily accounted for. Thus, the classication based on this scheme involves sums

A W (Y ) S , A
M i1 i j i j M i1 i

(2)

where M is the number of variables, and the natural assignment is the one that maximizes S j . But, there is more to come. Because the numerical values of S j allow quantitative assignment of the condence in the classied bulk hydrometeors (in the resolution volume), if the weights represent the truth well, then the values (in %) of S j can be interpreted as the condence in outcome of the classication procedure. That is, the higher the max (S j ) is, and the larger the difference between the max (S j ) and the next to max (S j ), the more likely it is that the classication is correct. A short discussion concerning maximization of the sum (2), as opposed to some other combination of the weights, is in order. For example, Liu and Chandrasekar (2000) maximize the product of weights and report considerable skill of their classier. Simple considerations suggest that with the sum, the probability of correct classication should be larger. That is, if due to noise,

one of the weights (for a specic class) is zero, the rest still can contribute signicantly to bring the datum into the correct category. On the other hand, the percent of false (absurd) classications should be smaller in the product maximization scheme. That is, if one variable is considerably out of range for a given class, its very small weight will suppress that class. Another difference between the two procedures concerns the way that condence in the variables is expressed. In the maximization of sums, the multiplicative factors A i primarily determine the condence, whereas the width of the weighting functions has a secondary role. In the maximization of products, the width of the weighting functions determines the condence in the variables; increase of the width lowers the condence in the variable. Because comparisons between the two procedures have yet to be made, further research and tests are required to evolve an optimum scheme. Because of high dimensionality and unknown shape of the partition boundaries, a reasonable approach is to deal with projections to smaller dimensions. At the moment, this is a sound practical approach as much has been learned about how values of individual variables relate to hydrometeor types (Herzegh and Jameson 1992; Doviak and Zrnic 1993; Holler et al. 1994; Straka et al. 2000). Although simple, the use of one-dimensional weighting functions gives a less tight partition than what is possible by capitalizing on the dependencies (correlations) between the variables. If two polarimetric variables (corresponding to a specic class of hydrometeor types) are related, their scattergram has a trend as opposed to a centrally symmetric shape for uncorrelated variables; see, for example, the Z h , ZDR scattergrams (Leitao and Watson 1984), the Z h , KDP scattergrams (Balakrishnan and Zrnic 1990a; Ryzhkov et al. 1997), and the Z h , hv scattergrams (Balakrishnan and Zrnic 1990b). Thus, a more precise classication would be achieved if the class is delineated with a boundary that follows a xed percentage contour in the scattergram of the two variables for a specic hydrometeor type. Straka et al. (2000) suggest how to partition pairs of variables (two-dimensional subspaces) into regions of dominant hydrometeor types. According to several investigators, discrimination in two-dimensional (2D) regions produced very promising results (Straka 1996; Carey and Rutledge 1998; Lopez and Aubagnac 1997; Zrnic and Ryzhkov 1999; Vivekanandan et al. 1999; Liu and Chandrasekar 2000). Simulations (Scarchilli et al. 1996) and observations allow denitions of regions in the subspace of three or more variables (triplet Z, ZDR , KDP in the cited reference) whereby hydrometeors of one type prevail. Nonetheless, direct partition (i.e., creation of boundaries) in larger than two-dimensional subspaces is quite complex and, to our knowledge, has not yet been attempted. Moreover, two-dimensional partitioning is readily visualized, and therefore, the connection between changes in the boundaries and the corresponding effects on classication be-

JUNE 2001

ZRNIC ET AL.

895

FIG. 1. Two-dimensional weighting function for moderate rain, WMR (Z h , ZDR ), over the space Z h , ZDR .

comes apparent. This is important for the evolution of the algorithm, which requires numerous iterations, and this visual-cognitive feedback can lead to a rapid convergence toward satisfactory performance. Accordingly, we have dened 2D weighting functions W j (Z h , Y i ) where Z h is the reectivity factor for horizontally polarized waves, and Y i is one of the other ve variables. Boundaries of hydrometeor classes presented by Straka et al. (2000) are used to dene the values of the weighting functions. In regions identifying a specic hydrometeor type, the corresponding weighting function is set to 1; from the boundaries and outward, the values decrease linearly with distance. To construct the 2D weighting functions, the analogy with probability densities is very helpful. For example, the 2D weighting function can be expressed as a product of a one-dimensional weighting function (analogous to a priory probability) with a conditional weighting function (analogous to marginal probability density function), W j (Z h , ZDR ) W j (Z h )W j (ZDR | Z h ). (3) As an example, the shape of the weighting function for the class moderate rain (Straka et al. 2000) is plotted in Fig. 1. Here, WMR (Z h ) 1 for Z h between 35 and 45 dBz; from these boundaries, the weighting function decreases linearly (with the slope of 0.2 dB 1 ) so that if Z h is displaced by 5 dB from the boundary, it reaches 0. The conditional weighting functions are also trapezoidal except the coordinates dening each trape-

zoid (Fig. 1) depend on the value of Z h ; the decrease of WMR (ZDR | Z h ) from 1 to 0 occurs over the interval of 0.3 dB in ZDR . In a similar way, all other 2D weighting functions are dened on, altogether, ve pairs of variables. All the a priory and conditional Ws are trapezoids whose top vertices are prescribed by the diagrams in Straka et al. (2000) in a similar manner, as explained in the previous paragraph. The weighting functions overlap and intersect at the value of . Some minor deviations from these diagrams have evolved in the course of this work. The increments in the polarimetric variables over which various weighting functions change from 1 to 0 (or vice versa) are listed in Table 1. These were obtained in a subjective manner from experience, examination of scattergrams, and consideration of statistical errors in estimates of the polarimetric variables. Tests of the algorithm demonstrated that the zero degree isotherm of the environmental temperature was often above the one inferred by the location of the melting layer (which is lowered by the downdraft). This can produce inconsistent weighting functions. To avoid such contradictions, we modify the environmental temperature prole so that its zero degree height coincides with the minimum of hv . That is, in cases where the vertical cross section of hv shows a well-dened melting zone (Zrnic et al. 1993), a vertical prole through the melting zone is constructed, and the location of its minimum is found. In this way, the heights of the melting layer bottom obtained from the temperature and polarimetric measurements are forced to coincide. But through convective cores, the melting zone is lifted in the updraft and lowered in the downdraft and, thus, offset from the imposed zero T height. Nonetheless, the polarimetric variables have such strong signatures in the convective cell that their weighting functions overwhelm the temperatures, and hence, the offset does not affect the outcome of classication. We emphasis that it is not necessary to have an actual temperature prole. A standard

TABLE 1. Increment of polarimetric variables over which the weighting function changes from 1 to 0. Increment of polarimetric variables Z h 5 dB ZDR 0.3 dB ZDR ZDR ZDR KDP KDP KDP 0.1 dB 0.25 dB 0.2 dB 0.23 0.023Z h 6.6 104 Z2 h ( km1 for Z h in dBZ ) 0.2 km1 0.1 km1 Hydrometeors to which weighting applies All Light, moderate, heavy rain, large drop rain, graupel/small hail, and wet snow Rain/hail mixture Hail Dry snow, ice crystals Light, moderate, heavy rain, large drop rain, and rain/hail mixture Hail, wet snow Graupel/small hail, horizontally, and vertically oriented ice crystals Dry snow Light, moderate, heavy rain, large drop rain, rain/hail mixture, graupel/small hail, hail, dry snow, horizontally, and vertically oriented ice crystals Wet snow All

KDP 0.05 km1 hv 0.02 hv 0.05 LDR 2 dB

896

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

TABLE 2. The transition points for the weighting trapezoidal functions of temperature. T1 (C) T 2 (C) T 3 (C) T 4 (C) 5 10 60 3 15 60 40 0 0 50 0 0 35 15 NA NA 1 7 20 0 15 NA NA 1 10 40 20 40 Hydrometeors to which weighting applies Light, moderate, heavy rain Large drop rain Dry snow, horizontally, and vertically oriented ice crystals Wet snow Rain/hail mixture Graupel/small hail Hail

TABLE 3. Values of thresholds and types of suppressed hydrometeors. Variables ZDR T T T Thresholds 0 dB 20C or 10C 15C 15C Suppressed hydrometeors Light, moderate, heavy rain, and rain with large drops Wet snow Rain/hail mixture Light, moderate, heavy rain, and rain with large drops

atmosphere prole sufces if it is modied to have the zero isotherm at a heigh determined by the radar. The weighting functions of temperature are rooted in physical principles and are also trapezoidal, except the ones for rain are open ended at temperatures larger than 0C. We have tabulated the transition points of the weighting trapezoidal functions in Table 2, whereas Fig. 2 relates the tabulated values to the typical trapezoidal weights. Currently, there are 11 hydrometeor classes that build on the synthesis in Straka et al. (2000). The hydrometeor classes are 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 11) light rain (LR 5 mm h 1 ), moderate rain (MR: 5 to 30 mm h 1 ), heavy rain (HR 30 mm h 1 ), rain dominated by large drops (LD), rain/hail mixture (R/H), graupel and/or small hail (GSH), hail (HA), dry snow (DS), wet snow (WS), horizontally oriented ice crystals (ICH), and vertically oriented ice crystals (ICV).

Throughout this paper (in the gures and the text), the classes are numbered as in the above list. There is also a nonhydrometeor class that absorbs biological scatterers, ground clutter, and other strong point scatterers as well as noise. This class is excluded from the displays. The real-time version of this algorithm implemented on the S-Pol radar (Vivekanandan et al. 1999) differs in some categories. Drizzle, cloud drops, supercooled

droplets, irregular ice crystals, and insects are additional classes of scatterers, but there is no distinction between horizontally and vertically oriented ice crystals, nor is there a category for rain with large drops. Vertical orientation of crystals is caused by strong electric elds and their presence might be a precursor to the onset of lightning. The extra hydrometeor categories for the SPol radar may be possible because it measures LDR , which is not available on the Cimarron radar. In addition to the described weighting functions, there are thresholds on differential reectivity and temperature such that the condence (sum S j ) in a hydrometeor class j is set to zero if the thresholds are satised. Table 3 lists the thresholds and the affected hydrometeors. It is self evident that these thresholds prevent absurd classication. Finally, the multipliers A i representing the importance attached to the various polarimetric variables are listed in Table 4. Experience combined with trial and error lead us to adopt these values. Reasons related to hardware on the Cimarron radar cause bias in the hv such that the values are lower and noisier than expected, and therefore, A is set to 0.3 as opposed to 0.5 for the SPol radar. It is of utmost importance to present valid data to this or any other automated algorithm. To that end, we utilize a median lter on Z h , a running average to smooth the ZDR , hv , and LDR , and a least-squares t (Ryzhkov and Zrnic 1996) to obtain specic differential phase. Furthermore, we discard KDP in and adjacent to regions where its values are below 0.5 km 1 . Reectivity factor and differential reectivity are corrected for attenuation following the procedure suggested by Bringi et al. (1990). 3. Sensitivity tests Objective comparison of polarimetric schemes for classication could be done, in principle, if a large set of comprehensive in situ observations were available.
TABLE 4. Multipliers (Ai) of the weighting functions. A Zh A ZDR 1 1 A KDP 0.8 0.8 Ahv 0.3 0.5 A LDR NA 0.5 AT 0.4 0.4

FIG. 2. Sketch of a trapezoidal weighting function for temperature. The points T1 , T 2 , T 3 , and T 4 serve to dene the numerical values in Table 2.

Cimarron data S-Pol data

1 1

JUNE 2001

ZRNIC ET AL.

897

FIG. 3. Vertical cross sections of the elds of (a) reectivity factor (color contours are in dB Z as indicated on the color bar), (b) differential reectivity (contours are in dB as indicated on the color bar), (c) cross-correlation coefcient (contours are indicated on the color bar), and (d) specic differential phase (contours km 1 are indicated on the color bar). Data were obtained on 6 Jun 1996, at 2316 UTC with the Cimarron polarimetric radar at the azimuth of 148.2; range from Cimarron radar is in kilometers, and the grid cell is 10 5 km.

Currently, there are no platforms to provide such observations over large volumes and at rates compatible with radar scanning times; nonetheless, for the purpose of this paper, such comparisons are premature. For preliminary comparisons and crude adjustments of the weighting (membership) functions, there is a much simpler procedure that we have adopted. It consists of visual identication of the obvious gross errors and determination of the responsible variables. Then, the boundaries and/or membership functions built on these variables are changed until the gross errors are reduced or disappear. This requires care so that the ripple effect on the previously correct classication is minimized, whereas the reduction in erroneous classes is maximized. The reectivity factor and environmental sounding are available to all those with access to ubiquitous nonpolarimetric radars. Hence, it is pertinent to determine how these two variables fair in comparison with the full set of polarimetric variables. In addition, it is useful to know which polarimetric variable in combination with the reectivity factor has the most discriminatory power. To illustrate the effects of temperature, we present results of classication whereby use only of the pair Z h , T is contrasted with application of the other variables. Similarly, we consider the other pairs individually and in suitable combinations. We have chosen a dataset obtained with the Cimarron polarimetric radar from a hail storm and a dataset obtained with the S-Pol radar from an airmass storm in Florida to demonstrate the effects variables have on the classication outcome. Besides presenting images of

classied hydrometeor elds, we also make a somewhat subjective but nonetheless quantitative ranking of the variables according to their relative importance. a. Hail storm observed with the Cimarron radar 1) VISUAL
ANALYSIS OF CLASSIFIED FIELDS

The polarimetric variables available on the Cimarron radar are reectivity factor Z h , differential reectivity ZDR , specic differential phase KDP , and cross-correlation coefcient hv . Vertical cross sections of the elds of these variables through a convective storm reveal some well-dened hydrometeor signatures as well as artifacts (Fig. 3). For example, the high reectivity core aloft (Fig. 3a) likely contains hail. The positive ZDR near ground indicates rain, and the column of positive ZDR at 80 km (Fig. 3b) contains large drops in relatively low concentration. Positive ZDR above the storm top is due to mismatch of antenna sidelobes for vertical and horizontal polarizations; these data are excluded from the forthcoming presentations. The KDP shows concentration of rain between 70 and 80 km, and there is a column of KDP that overlaps the ZDR column (Fig. 3d). The censored values aloft correspond to KDP 0.5 km 1 , which are likely due to artifacts (Ryzhkov and Zrnic 1998). The eld of correlation is least reliable, and there is no indication of the melting layer in this storm (Fig. 3c). An abrupt increase of ZDR in the lower part of precipitation in this case hints at the onset of melting of graupel and snow. Because the strong vertical gradient of ZDR at its highest point above ground (at the

898

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

FIG. 4. Vertical cross section of classied hydrometeors for the data from Fig. 3. The letters below the color bar correspond to the hydrometeor categories as follows: LR light rain, MR moderate rain, HR heavy rain, LD rain with large drops, R/H rain/hail mixture, GSH graupel small hail, HA hail, DS dry snow, WS wet snow, IH horizontally oriented ice crystals, and IV vertically oriented ice crystals. Classication was made using all available polarimetric variables (Z h , ZDR , KDP , hv ) and the environmental temperature T.

range of 40 km) almost coincides with the 0C isotherm of the sounding, no adjustment was made to the environmental temperature. The algorithm was applied to the elds of polarimetric variables (Fig. 3), and it created the hydrometeor classes represented in Fig. 4. We consider this eld to be the standard against which elds obtained with less than the full set of variables can be gauged. Although we do not have independent verication about the quality of the eld in Fig. 4, we submit that the comparison can reveal some useful information. Things like relative merit of polarimetric variables and the average condence (Mendel 1995) in various classes of hydrometeors can be quantied.

Figures 5ad contain the elds of hydrometeors, each of which was obtained from a single pair of variables (one variable in any of the pairs is always Z h ); comparison with Fig. 4 reveals the relative signicance of the selected pair. This signicance will be quantied after a brief qualitative description of salient features attributed to the various pairs. The primary variables responsible for locating the onset of rain are the Z h and ZDR , as can be seen in Fig. 5a. Further, the heavy rain and rain with large drops at 50 km as well as aloft at 80 km are also identied from the Z h , ZDR pair. The light rain in the anvil is erroneous and is a consequence of inability to distinguish between light rain and dry snow in the two-dimensional subspace Z h , ZDR . The areas of light rain and dry snow overlap in a large region of the Z h , ZDR subspace (see Figs. 2a and 2b in Straka et al. 2000). With further research, it might be possible to decrease the overlapping area. The pair Z h , KDP identies some horizontally oriented ice crystals, but at the top of the cloud it misclassies crystals into light rain. Otherwise, graupel and rain/hail mixture are the two dominant categories. Both extend into the region where their presence (in this case) is physically forbidden; for rain/hail that is throughout a large upper part of storm, and for graupel it is the northwest part near ground. These errors are a consequence of inadequate information in the Z h , KDP pair. Sporadic data void regions (black patches within the eld) are where the KDP is excessively negative. The pair Z h , T (Fig. 5d) produces a smooth eld in which the vertical stratication of hydrometeors is controlled by the temperature prole. Note how the phase

FIG. 5. Vertical cross sections of classied hydrometeors for the data from Fig. 3 except only a pair of variables is used for each eld as follows, (a) the pair Z h , ZDR ; (b) the pair Z h , KDP ; (c) the pair Z h , hv ; and (d) the pair Z h , T.

JUNE 2001

ZRNIC ET AL.

899

FIG. 6. Same as in Fig. 4 except a specic pair has not been applied to the classication scheme, but other pairs each of which includes Z h have been applied. (a) The pair Z h , ZDR has not been used. (b) The pair Z h , KDP , has not been used. (c) The pair Z h , hv has not been used. (d) The pair Z h , T has not been used.

transition at zero temperature is at the constant height of about 4 km; it is followed in height with graupel/ small hail, and dry snow is at the top. This general pattern is modulated in the horizontal direction by the inuence of the reectivity factor. Although pleasing to the eye and somewhat physically satisfying, the classication misses important details seen in Fig. 4. The pair Z h , hv produces the least reliable classication due to extensive overlap of discrimination regions (Straka et al. 2000). The classes close to the ground are credible mainly due to the inuence of Z h . Data-void regions are where the values of the correlation coefcient are below 0.3. Figures 5ad indicate that none of the pairs by themselves can create wet snow in the melting layer; its presence is revealed only if the algorithm is applied to all the polarimetric variables (Fig. 4). The pairs Z h , ZDR , and Z h , T produce elds that closely match what is expected as function of height in storm cells. Other works (Wakimoto and Bringi 1988) and our experience suggest that the boundary between frozen and liquid precipitation at low levels is faithfully delineated using the Z h , ZDR pair. For further comparison and analysis, it is worthwhile to examine classied elds obtained from the sets of the polarimetric variables in which one polarimetric variable is absent (Fig. 6). Absence of ZDR (Fig. 6a) produces a eld in which the strong inuence of KDP and T is seen, whereas the absence of KDP (Fig. 6b) replicates fairly well the eld in Fig. 4; neither one of these elds contain wet snow in the melting layer. Only if Z h , ZDR , and KDP are simultaneously present do the weights add

up to produce wet snow in the melting layer (Figs. 6c and d). Absence of hv (Fig. 6c) has a minor effect on the classication outcome; however, it increases the thickness of the wet snow region. The value of temperature in eliminating physically impossible classes is seen in Fig. 6d. These are the sporadic areas of wet snow at heights above 7 km, spots of rain/hail mixture at similar heights, patches of light rain at the edges of storm, some ice crystals close to the ground, and wet snow in the upper part of the storm. Comparison of Figs. 6ad clearly shows that ZDR is the most important polarimetric variable. In its absence, the classied elds clash with those expected from conceptual models and physical considerations. 2) RELATIVE
IMPORTANCE OF THE VARIABLES

Next, we will quantify the discussion concerning relative contribution of variables to hydrometeor classes in Figs. 46. This we do, as in the previous section, by examining classication with single pairs versus the rest of variables. The reader is referred to appendix A for explanation of the measure that we assign to the eld of classied hydrometeors. Briey, the measure is a number of data points that are classied to the same category of hydrometeors. With appropriate normalization of measures, we obtain various percentages (akin probabilities) of classication, which are entered in tables. Thus, from the tables one can objectively establish relative merits of variables used in the classication scheme. Two tables (5 and 6) are presented and dis-

900

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY TABLE 5. Percent of correctly classied hydrometeors (P A)/(A) using the pair alone; Oklahoma storm. Hydrometeor categories

VOLUME 18

With ZDR KDP hv T

LR 96 62 70 98

MR 77 40 94 100

HR 81 26 83 100

LD 100 0 0 0

R/H 69 95 90 26

GSH 91 65 0 83

HA 95 24 0 78

DS 8 17 0 64

WS 0 1 15 0

ICH 9 0 0 0

ICV 0 98 0 0

Tot 70 51 25 68

70%* 30%* 6 2 4 5 4 6 7 5

* 70% and 30% indicate the number of classes for which these conditions hold.

cussed in detail herein; several others are in appendix B. We will often make reference to appendix A (Fig. A1 and Table A1) in the following text. Thus, (A), (P), and (R) stand for the number of data in a class obtained from all the available variables, a single pair of variables, and the rest of the variables (i.e., excluding the polarimetric variable that was part of the pair), respectively, as explained in appendix A. The notation in Table A1 claries the meaning of various measures presented here and in appendix B. The total number of classied data points (i.e., resolution volumes) for this storm was 14050. We start with Table 5 to illustrate values obtained for the Oklahoma hailstorm data. The entries are percentages with respect to the case in which all the variables are used for classication (see also Table A1). Thus, the large value indicates that a particular pair contributes signicantly to the classied category, whereas very small value indicates irrelevance of the variable. If a column contains one large value and other very small values, then the variable corresponding to the large value is almost exclusively responsible for the classication to the corresponding category. In the twelfth column are the percentages for the whole data eld (i.e., for all the classes combined). A glance at the Table 5 reveals that Z h , ZDR pair is the most signicant of the pairs. In column 14 listed is the number of hydrometeor classes for which the percent of correct classication is larger than 70. (For Gaussian distributions, this is almost equal to the probability within one standard deviation.) If we rank the variables according to the number of hydrometeor classes for which the pair produces 70% of agreement with that produced by the complete set, the order would be ZDR , T, hv , KDP (6 categories 70% for ZDR , 5 for T, 4 for hv , and 2 for KDP ). But if the rank is according to the number of categories for which the percent is 30%, the ranking would be slightly altered to ZDR , T, KDP , hv ; in this ranking the variable with the smallest number of categories is ranked highest because it fails to classify the smallest number of categories. (The other variables fail to classify more categories.) Ranking according to the percentage for all classes (column 12) is ZDR , T, KDP , hv . It is signicant that Z h , T pair is ranked second, although the multiplier for its weighting function is 0.4

(Table 4); only the multiplier A hv has a smaller weight of 0.3. This strong inuence of temperature is partly due to its independence of the radar variables (which inherently are somewhat related). Furthermore, the temperature information inuences a large number of data above the melting layer and in weak reectivity regions where the classication is often ambiguous. These factors and the strong effect of Z h (in the horizontal direction) are likely the cause for such a high importance of the Z h , T pair. Even though the Z h , KDP pair has a 0.8 weight multiplier, for classication it is of secondary importance, similar to the Z h , hv pair. This is because for the majority of hydrometeors, the boundaries in the Z h , KDP space overwhelmingly overlap. Further in this case, there is a large region of KDP close to 0, which is ambiguous. But the specic differential phase offers advantages for rainfall measurement (Zrnic and Ryzhkov 1996), which are not included in our ranking. Further examination of the table reveals features in data that are a direct consequence of the classication scheme, like the large drop category is detected exclusively with the Z h , ZDR pair. Dry snow, wet snow, and vertically oriented ice crystals are not at all classied by this pair. Vertically oriented ice crystals are exclusively detected with the Z h , KDP pair, but in this particular case, KDP misses large drops and horizontally oriented crystals. Another aspect of the scheme and the variables inuence can be gleaned by normalizing with respect to the number of points obtained if a particular pair (P) or lack thereof (i.e., the rest R of the pairs) is used in the classication. For that reason, we present Table 6. The values in Table 6 correspond to the categories that are wrongly classied (analogous to false alarm) by the use of the pair. That is for each category, the measures are normalized to the total number of points classied into that category. For example, 71% of the data are erroneously classied as light rain out of all data classied as LR from the Z h , ZDR pair; these are mostly in the anvil (Fig. 5a). The missing entries indicate that there were no such categories identied by the pair [i.e., the corresponding (P) is 0]. This lack of identication is a deciency routed in the fact that the particular hydrometeor signature in the polarimetric pair is week or nonexistent. From Table 6, it becomes immediately apparent what pairs misclassify which categories. So in

JUNE 2001

ZRNIC ET AL.

901

TABLE 6. Percent of wrongly classied data (per category) P P A/(P) using a pair of variables out of all those classied using the same pair; Oklahoma storm. Hydrometeor categories With ZDR KDP hv T LR 71 59 75 23 MR 24 3 72 40 HR 49 32 54 66 LD 6 R/H 2 56 65 26 GSH 14 33 28 HA 3 18 6 DS 13 43 44 WS 93 98 ICH 84 100 ICV 33 Tot 30 41 71 32 70% 30% 2 2 3 0 6 2 0 4

addition to the Z h , ZDR pair, signicant misclassication of LR is caused by the Z h , KDP pair and also by the Z h , hv pair. Substantial mis-classiers of other categories are Z h , KDP for R/H, WS, and ICH; and Z h , hv for LR, MR, R/H, and WS. Use of combined variables eliminates these misclassications. Nonetheless, we have no independent conrmation that the combined use of all the variables produces a correct eld of hydrometeors. Hence as stated earlier, the values in this and other tables are only meant for relative comparisons. As far as the number of categories that can be identied (that is, minimum of missing entries), the ranking would be KDP , ZDR , T, hv . The total percentages imply the following ranking: ZDR , T, KDP , hv . Other aspects of merits of different polarimetric variables can be seen in Tables B1 to B6 in appendix B. This is where the ranking of variables is justied and explained. The summary of the rankings according to the columns labeled total (Table 7) indicates that the Z h , ZDR pair (note that Z h is always used) is overall most effective in the classication scheme. The strong importance of temperature, in spite of its relatively low weight (0.4 in Table 4), suggests that ambiguities in polarimetric signatures and statistical uncertainty might be signicant spoilers. It remains to be seen if the addition of spatial lters and quantitative use of weighting functions can reduce the importance of temperature. (This is pertinent for cases where T is not available.) As expected, because of the 0.3 multiplier, the hv contributes least to our current classication scheme. b. Ordinary storm observed with the S-Pol radar During August and September 1998, the S-Pol radar was located near Melbourne, Florida, to support the TexTABLE 7. Relative rank of the variables; Oklahoma storm. Summaries from Tables 5, 6, and B1 to B5 5 B1 B2 B3 B4 6 B5 Rank in order from left to right ZDR, T, KDP, hv T, ZDR, KDP, hv ZDR, T, KDP, hv T, ZDR * , K* DP, hv ZDR, T, KDP, hv ZDR, T, KDP, hv T, ZDR, KDP, hv

as-Florida Under ying experiment (TEFLUN). Because the S-Pol radar measures LDR and produces better quality hv than the Cimarron, this set allows an augmented analysis and a more extensive comparison of the utility of the various variables. Also, comparison of two different storm types adds value to the analysis. 1) VISUAL
ANALYSIS OF CLASSIFIED FIELDS

( P ( R ( R ( P ( R ( P ( R

A )/(A ) A )/(A ) P A )/(A ) A R P A )/(A ) A R P A )/(A ) P A )/(P ) R A )/(R )

* ZDR and KDP are ranked the same.

The elds of the polarimetric variables (Fig. 7) indicate the storm is well developed with peak reectivities of about 55 dBZ below the melting layer. The transition from ice to rain is seen in the ZDR as an abrupt increase between 3.5 and 4 km; it is also apparent in the LDR eld as a sharp decrease to less than 26 dB. The bright band in the hv is well dened and discernable in few places even within the storm core. The height of the minimum of the bright band was used to shift down by 1.1 km the temperature prole of the environmental sounding (0C isotherm from 4.7 to 3.6 km). Both ZDR , and KDP have visible columns that coincide, and KDP indicates heavy rain (100 mm h 1 ) between 20 and 23 km from the radar. Note the LDR cap (Jameson et al. 1996; Hubbert et al. 1998) at a range of 20 km that coincides with the top of the ZDR column. It indicates freezing of the supercooled drops lifted by the updraft. Adjacent to it (at farther range) is a vertical band of LDR minima that suggest the presence of graupel. Classication from the whole suite of variables and from single pairs produces the results in Fig. 8. Graupel/ small hail is located in the frozen region of the three distinct cells (Fig. 8a). Dry snow lls most of the rest of the cloud above the melting layer, and at the top of the cloud, some horizontally and even vertically oriented ice crystal are indicated. The sporadic presence of vertically oriented crystals suggests that these are likely artifacts. Although the horizontally oriented crystals are expected at cloud tops, the detection herein could be fortuitous; this is because the crystals also appear at the bottom of the anvil in the region of weak reectivity where they do not belong. Heavy rain and R/H mixture are interleaved below the main core, and there is a small patch of hail in the core just above the melting zone. Noteworthy is the presence of snow in the bright band, and a sliver below it is rain composed of large drops. Examination of the ZDR eld (Fig. 7b) indicates its

902

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

FIG. 7. Vertical cross section of the elds of (a) reectivity factor, (b) differential reectivity, (c) correlation coefcient, (d) specic differential phase, (e) temperature, and (f ) linear depolarization ratio. Range from the radar is in kilometers, and the grid cell size is 10 5 km. Data were obtained on 14 Aug 1998, in Florida with the S-Pol radar.

strong inuence on the moderate, heavy, and large drop rain (Fig. 8b). The wet snow is correctly located, and its chief identier is the hv (Fig. 8f). Hydrometeors within the major cell (1727 km from the radar) are almost equally well identied with the ZDR , Z h pair as with the KDP , Z h pair (Fig. 8c), except the eld obtained from the latter one is noisier and has a larger area of heavy rain. The phase of precipitation in the two weak cells is correctly separated with the ZDR , Z h pair but is not separable with the KDP , Z h pair. Note that the ZDR , Z h pair fails to discriminate between light rain and dry snow in a large region above the freezing level; neither does it detect vertically or horizontally oriented ice crystals. Clearly KDP , Z h identies the horizontally and vertically oriented ice crystals. Note the area of vertical crystals just above the cell at 40 km (Fig. 8c). Although in the composite classication (Fig. 8a) this area is absent, it might have been wrongly eliminated; similar

signatures above growing storms have been previously noted and attributed to aligned crystals (Zrnic and Ryzhkov 1999; Caylor and Chandrasekar 1996). Identication based on the T, Z h pair is too smooth, generally credible, but lacks details of wet snow or rain/ hail mixture. We stress again that in this particular case, the minimum of hv was used to place the zero degree isotherm at a correct altitude. If the LDR , Z h pair is the only discriminant, the eld of hydrometeors shows structure, but with the exception of some light and moderate rain below the melting layer and some snow above, the types are mostly wrong. The eld produced by the hv , Z h is surprisingly smooth (Fig. 8f). The strength of this pair is in identifying the wet snow and moderate to heavy rain below the melting layer. As expected, it is not possible to discriminate between rain and ice forms above the melting layer. By examining the effects of omission of variables

JUNE 2001

ZRNIC ET AL.

903

FIG. 8. Vertical cross sections of classied hydrometeors for the data in Fig. 7. (a) All available variables were used to classify this eld. (b) Only the pair Z h , ZDR was used. (c) Only the pair Z h , KDP was used. (d) Only the pair Z h , T was used. (e) Only the pair Z h , LDR was used. (f ) Only the pair Z h , hv was used.

from the classication process, one can obtain another feel for the relative merits among the variables (Fig. 9). Absence of ZDR caused misclassication in the rain region where some LR and MR are assigned to GSH. This would degrade rainfall measurement. Also in the region of frozen hydrometeors, there are substantial sporadic areas of horizontally and vertically oriented ice crystals (although the area above the two-week cells at 40 km could contain vertically oriented crystals). For most of the hydrometeor categories, the absence of KDP has little or no effect (Fig. 9c), except in the core of the strongest cell; there, below the melting layer, GSH occupies larger area, the one touching the ground looks suspicious. Otherwise the elds are smoother and, thus, more appealing than in Fig. 9a! Perhaps KDP should be used only in regions where its signatures are strong (moderate or larger rain, rain/hail mixture, etc.). It is very satisfying

that the absence of T does not change the basic patterns of the classied elds (Fig. 9d). As expected, the eld looks noisier; sporadic dry snow appears below the melting layer, and speckles of light rain are seen in the region of snow. The absence of either LDR or hv has a similar effect. Minor differences are in the depth of regions with oriented crystals at the top (larger in the absence of LDR ) and in the area of rain/hail mixture (slightly larger in the absence of hv ). The presented images indicate the signicance of Z h , ZDR in convective cores, the importance of hv for wet snow, and the temperature for resolving ambiguities above and below the melting zone. It is signicant that the absence of temperature information leaves generally consistent elds. Nonetheless, LDR or T are needed to discriminate between dry snow and light rain, as can be seen in Fig. 10. Absence of LDR and T (Fig. 10a)

904

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

FIG. 9. Same as in Fig. 8 except a specic pair has not been applied to the classication scheme, but other pairs each of which includes Z h have been applied. (a) All available variables were used to classify this eld. (b) The pair Z h , ZDR has not been used. (c) The pair Z h , KDP , has not been used. (d) The pair Z h , T has not been used. (e) The pair Z h , LDR has not been used. (f ) The pair Z h , hv has not been used.

produces a light rain throughout most of the upper part of the storm where the ice phase should dominate. Inclusion of LDR at the expense of hv and T (Fig. 10b) restores the snow region above the melting layer. This nding is signicant for applications to winter storms in which the spatial distribution of temperature is not available, and hence, LDR might be the most important discriminator between rain and snow. In the absence of temperature, hydrometeor classes appear noisier but, nonetheless, have coherent structure and are at the expected location. This attests to the strength of the polarimetric variables. Thus, we believe that in similar cases, additional processing with spatial lters would reduce the incoherent speckles. 2) RELATIVE IMPORTANCE OF THE VARIABLES Some 14 812 data values were classied (between 10 and 50 km from the radar), and the smallest number

in a class was 20 for hail; the next smallest was 220 for wet snow. Because of insufcient sample size, we will refrain from quantitative interpretation of data classied as hail. The percentage of hydrometeors classied using a pair of variables that agree with those obtained from the full set is tabulated in Table 8. The ranking according to column 13 would be rst ZDR and T, second hv , and third LDR and KDP , but according to column 14, the ranking would be ZDR and T, KDP , hv , LDR . For all classes combined, the temperature leads and is followed by ZDR , KDP , LDR , and hv . The 25% lead of T over ZDR is somewhat unexpected considering that in the Oklahoma storm, temperature was a close second (Table 5). But in the Florida event, a large portion of frozen precipitation has weak reectivity whereby classication is often ambiguous (Fig. 9d is an example where the T is excluded), unlike the Oklahoma storm, which has highly

JUNE 2001

ZRNIC ET AL.

905

FIG. 10. Same as in Fig. 8 except two variables have not been applied to the classication scheme. (a) The missing variables are LDR and T. (b) The missing variables are hv and T.

reecting upper part (see how few obviously wrong classes are in the frozen part in Fig. 6d). Perhaps the reason is that the 0C isotherm is set to the correct height, which is constant for the Florida storm. The equivalents to false alarms, that is, percentages of data wrongly classied by the pair, are in Table 9. From this table and according to the totals, the order of variables is T, ZDR , KDP , LDR , and hv . It is noteworthy that the three less useful variables KDP , LDR , and hv each has a large false classication for a specic hydrometeor type. Here, KDP misclassies 100% of hail, LDR mis-classies 97% of the large drops, and hv misclassies 85% of the large drop rain. Relative merits of other variables are discussed and tabulated in appendix C, and the rank of variables according to the metrics in these tables is listed in Table 10. Comparison of the rows of this table with the corresponding rows of Table 7 brings out the following curious fact. The rst variable in Table 7 it always sec-

ond in Table 10 and vice versa. Moreover, according to all the criteria in Table 10, the hv is ranked consistently last. Because of immense practical implications of a simultaneous scheme proposed for the WSR-88D (Doviak et al. 2000), we further focus on the relative merits of LDR versus hv . Current plans are to simultaneously transmit and receive the H and V polarizations, and this precludes measurement of LDR (unless it is done on alternate scans). At least in this Florida storm, LDR does not seem to add signicantly more to the classication, as can be seen by the fact that omission of either LDR or hv (with the use of temperature) has almost the same effect (95% and 96% agreement in column 12 of Table C1). This we have veried in an alternate manner by comparing the percentage of agreement for classication, which omits LDR with the one that omits hv (93% of classied data agree). A more important question (to which we have no answer) asks, Can the addition of LDR be the only way to detect a signicant precipitation type (like icing conditions)? We do have, however, an objective way to determine the potential utility of LDR in special meteorological conditions such as transition between rain and snow in the horizontal direction. Because surface temperature measurements are sparse, the horizontal T eld might not be available, in which case, the polarimetric variables must stand on their own. Figure 10 suggests that in the absence of temperature, LDR should be a much better discriminator of dry snow than hv . This assertion is quantied in Table 11, where the percent of correctly classied hydrometeors is presented for the case without T and LDR , as well as for the case without T and hv . If LDR is lacking, only 36% of dry snow is correctly identied. In the absence of hv (but presence of LDR ) 88% of the dry snow is still correctly classied. Although these results are for vertical cross sections, they apply equally well to horizontal elds of the classication variables. 4. Discussion Both the specics of our classication scheme and the particular dataset to which it is applied bear on the deduced merits of the variables. Presently, we have not yet evolved the boundaries to the point of diminished return, nor do we have optimum weighting functions.

TABLE 8. Percent of correctly classied hydrometeors (P A)/(A) using the pair alone; Florida storm. Hydrometeor categories With ZDR KDP LDR hv T LR 94 59 69 96 99 MR 96 54 74 99 98 HR 81 74 25 99 99 LD 85 0 52 0 0 R/H 94 21 76 1 0 GSH 94 64 20 0 74 HA 83 0 0 0 94 DS 18 19 29 0 85 WS 0 10 0 98 0 ICH 2 49 0 85 8 ICV 0 84 0 0 0 Tot 50 38 33 25 75 70% 30% 7 2 2 5 7 4 5 7 6 4

906

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

TABLE 9. Percent of wrongly classied data (per category) (P P A)/(P) using a pair of variables out of all those classied using the same pair; Florida storm. Hydrometeor categories With ZDR KDP LDR hv T LR 80 79 49 85 25 MR 5 14 18 63 9 HR 6 55 19 63 45 LD 33 97 R/H 32 73 50 60 GSH 16 39 54 13 HA 0 100 76 DS 4 9 8 6 WS 63 57 ICH 35 21 20 29 ICV 93 Tot 49 60 64 75 14 70% 30% 1 4 1 1 1 5 3 3 1 4

Therefore, we cannot isolate the two effects. Nonetheless, comparison of the relative performance of classication for the two storms can, in some instances, point out to what extent data affect the classier. Although the two storms are convective, they have distinct characteristics that might induce substantially different responses of the classier. Note that the percentages of identied hydrometeor categories out of the total for the two storms differ by a factor of 2 in ve categories (Table 12); these are R/H, HA, DS, ICH, and ICV. The contrast in hail is good and expected considering that the Oklahoma storm was a cell in a squall line, and the Florida storm was an airmass storm. This is also the reason that the Florida storm has substantially more dry snow; the decit of the Oklahoma storm in dry snow is made up by an increase in graupel-small hail. It is gratifying that all four rain categories are comparable (Table 12), this implies consistency in detecting rain. Similar conclusions apply to the wet snow. As far as the rain and wet snow are concerned, the algorithm is not sensitive to the type of convective storm or the radar. The algorithms relative performance on the Oklahoma and Florida data is similar. Comparison of entries in Tables 5, 6, B16 with corresponding Tables 8, 9, C16 indicates that on the average, the difference between likewise entries in the tables exceeds 30% in only 15% of the total number of entries. Thus, in these cases, the relative merits of the polarimetric variables are invariant to differences in the type of storm and environment, the extra LDR available on the S-Pol radar, and the quality of hv data. Comparison of Tables 7 and 10 reveals that for all
TABLE 10. Relative rank of the variables; Florida storm. Summaries from Tables 8, 9, 10 and C1 to C5 8 C1 C2 C3 C4 9 C5 Rank in order from left to right T, ZDR, KDP, LDR, hv ZDR, T, KDP, LDR, hv T, ZDR, KDP, LDR, hv ZDR, T, KDP, LDR * , * hv T, ZDR, KDP, LDR, hv T, ZDR, KDP, LDR, hv ZDR, T, KDP, LDR, hv

( P ( R ( R ( P ( R ( P ( R

A )/(A ) A )/(A ) P A )/(A ) A R P A )/(A ) A R P A )/(A ) P A )/(P ) R A )/(R )

rows except the seventh the leading two variables (either ZDR or T ) are opposite. It is signicant that the LDR and hv rank similarly behind the rest of the variables, and the LDR is ahead of hv . For several reasons, the polarimetric scheme contemplated for upgrades of the WSR88D is one with simultaneous transmission and reception of H and V waves (Doviak et al. 2000). In that mode the LDR is not available, but if needed, it could be obtained in alternate volume scans that transmit H and receive H and V. Our analysis supports the notion that LDR might not be essential for precipitation identication in convective storms. But where temperature is not available, such as when it changes through zero on the ground, like in winter storms, LDR might offer advantages not found in the other variables. This was deduced by excluding the temperature and comparing classications in the presence and absence of LDR and hv . A very important aspect for future examination is ranking classication according to the value associated with the knowledge of a specic hydrometeor type. Such ranking might not even be unique; rather, it would be a function of the utility in a given situation. The value of detecting hail in convective storms is much higher than detecting snow aloft. But in a snow storm, the principal utility would be to detect the freezing zone and accurately measure the amounts. Pragmatic reasons suggest that in a short run, tuning the algorithm for a storm type might result in a faster evolution toward better performance. We realize that our analysis incorporates partly a circular argument. That is, we have given weights (which depend on the condence in the variables, Table 4) to the variables, and the results of the analysis are generally consistent with these weights. Yet, specics differ. For example, the Z h , ZDR , which both have a weight of one, come in as a strong rst in most comparisons. But T, which has a weight of 0.4, is a very close second (behind ZDR ). Two reasons could explain its importance: one, T has information independent of the polarimetric variables, and two, it is paired with Z, which has very strong discriminatory power. 5. Conclusions Polarimetric measurements from an Oklahoma squall line and a Florida airmass storm were automatically

* LDR and hv are, for this row, ranked the same.

JUNE 2001

ZRNIC ET AL. TABLE 11. Percent of hydrometeors correctly classied without T and either LDR or hv ; Florida storm.

907

Without LDR hv

Hydrometeor categories LR 75 72 MR 86 85 HR 92 89 LD 51 71 R/H 79 91 GSH 95 97 HA 78 83 DS 36 88 WS 95 79 ICH 97 45 ICV 86 86 Tot 63 85 70% 9 11

interpreted. The automatic interpretation is in the form of an algorithm that classies hydrometeors. The principles behind the algorithm are so simple that a prerequisite for understanding is high school algebra. Because the equations are linear sums, it would seem that there should be no difculty in analyzing and improving its performance. Yet this is far from the truth. It is the multidimensionality that obfuscates the progress here. The partitioning boundaries are many, and they overlap; furthermore, there are numerous combinations of values of the variables that can cause maxima in the condence factors. Therefore, the analysis is involved and the evolution tedious. It will just take much time to develop a ne version of the algorithm. At this early stage of algorithm development, in situ verication seems premature; thus, we have opted to access its attributes by examining self consistency of data, spatial continuity, and compliance with conceptual models. Our rst goal was to test sensitivity of the algorithm to the various polarimetric variables. This was done subjectively by comparing the elds of hydrometeors obtained using the full set of available polarimetric variables with a diminished set whereby some variables have been left out. Comparisons reveal the following. The reectivity factor and differential reectivity combined have the strongest discriminating power. Inclusion of the temperature prole helps eliminate a substantial number of spurious errors. Although the absence of temperature information degrades the scheme, it appears that the resultant elds are generally coherent and not far off from the elds obtained by adding T to the suite of polarimetric variables. An objective way to test the sensitivity of the algorithm to variables and rank their utility was devised. It involves denition of a measure (appendix A), which turns out to be a number of data points classied into a category in the presence or absence of variables. Ratios of various measures (similar to probabilities) dene the percentage of occurrence of a class. By comparing these percentages for cases in which some variables are excluded to those where all are included, a relative merit can be assigned to the variables.

Acknowledgments. Partial funding for this research was provided by the National Science Foundation Grants ATM-9120009, ATM-9311911, EAR-9512145, and ATM-9617318, the National Severe Storms Laboratory, the Cooperative Institute for Mesoscale Meteorological Studies, the Graduate College of the University of Oklahoma (Dr. E. C. Smith), and the Federal Aviation Administration. Mike Schmidt and Richard Wahkinney have maintained and calibrated the Cimarron radar, Joan OBannon drafted Figs. 2 and A1, whereas Chris Curtis produced Fig. 1, respectively. Finally a perceptive reviewer pointed out an inconsistency with the weighting functions for temperature and prompted us to correct it. APPENDIX A Metrics for Quantitative Comparisons Denitions of metrics for quantitative comparisons of the effects of various polarimetric variables on the outcome of the classication scheme are presented herein. To aid the reader, we start with a conceptual diagram (Fig. A1) of the eld of an identied class of hydrometeors. One can think of it as a horizontal or vertical cross section in which only one class of hydrometeors is shown. The set A corresponds to the data points (here from radar resolution volumes), which have been classied as one hydrometeor type (e.g., hail) using all (A) the available variables; this set is a standard for comparisons, and the corresponding classication will be called correct. The set P corresponds to the same hydrometeor class (hail) except that the data have been classied using a single pair (P), and the set R corresponds to the same class (hail), but the data have been classied using the rest (R) of the variables (all except the pair that was used to generate P). For the sake of simplicity in the sequel, the sets P, R, and A refer to any single hydrometeor type or to all the classied hydrometeors. Contrast this to the cumbersome PLD (Z h , ZDR ), which could otherwise signify the set of large drops classied using the pair Z h , ZDR . From the context

TABLE 12. Percent of hydrometeors classied in various categories for the Oklahoma (OK) data and the Florida (FL) data. LR OK FL 8.7 10.5 MR 6.8 5.7 HR 4.3 3.4 LD 1.9 2.2 RH 9.9 2.9 GSH 27.3 19.1 HA 19.1 0.1 DS 12.5 48.1 WS 1.1 1.5 ICH 0.7 4.8 ICV 7.7 1.6

908

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

and in the tables, it should be clear to which hydrometeor classes the shorthand notation refers. Next, we prove that the partitions (i.e., subsets) in Fig. A1 are the only ones possible. That is, for the classication procedure based on Eqs. (2) and (3), agreement of the pair with the rest implies that the use of all the variables will also agree. In other words, the seven subsets including the null category in Fig. A1 are the only ones possible. Without loss of generality, assume that the pair is Z h , ZDR , and the classied category by the pair and by the rest is the same, say snow(s). Thus, the weighting function for the pair satises W s (Z h , ZDR ) W j (Z h , ZDR ), j s. (A1) Similarly, the sum of weights for the rest of variables satises

A W (Z , Y ) A W (Z , Y ),
i s h i i j h i Yi Yi

(A2)

FIG. A1. Sketch of elds of any one classied hydrometeor type. A (all) is obtained using all the available variables. P (pair) is obtained using one pair of variables of which the rst variable is always Z h . R (rest) is obtained using all the variables except the second variable of the pair P.

j s, and the summation is over all the variables Y i except ZDR . Normalization by the A i is omitted from (A2) as it has no bearing on the proof. Now, multiply both sides of (A1) with A ZDR and add the right side of (A1) to the right side of (A2) and the left side of (A1) to the left side of (A2). This operation does not change the inequality, neither does normalization by A i , which is actually used when all the variables are present. This completes the proof. The readers might recognize that the sets P, R, A belong to a Borel eld (e.g., Papoulis 1965) and so do the various intersections and unions of these sets. Next, we discuss assignment of a measure to these sets, from which some quantitative comparison of performance will ensue. Briey, a measure on a Borel eld is an assignment of a number (E ) for every element E that belong to the eld such that (E ) 0. For our purpose, the measure will be the number of data points in a classied category of hydrometers. (Each data point corresponds to a single resolution volume.) That is, we count the number of classied data points (in a specic hydrometeor category) for which use of all variables,

TABLE A1. Metrics used for relative comparisons between variables.*

(P A )/(A ) (R A )/(A ) (R P A )/(A ) (P A R P A )/(A ) (R A R P A )/(A ) (P P A )/(P ) (R R A )/(R ) (A R A P A R P A )/(A )

% of correctly classied hydrometeors using the pair alone % of correctly classied hydrometeors using the set without one variable % of correctly classied hydrometeors for which use of a pair and absence of a variable produces the same class % of correctly classied hydrometeors that would be misclassied in the absence of a variable % of correctly classied hydrometeors that would be missed in the absence of the remaining variables % of wrongly classied hydrometeors using a pair of the variables out of all those classied using the same pair % of wrongly classied hydrometeors using the rest of the variables out of all those classied using the rest of the variables % of hydrometeors that can be correctly classied only if all the variables are used

* The intersection operation takes precedent over algebraic operations.

JUNE 2001

ZRNIC ET AL.

909

use of the single pair, and use of the remaining variables (without the single pair) agree or disagree. Then, a suitable combination and normalization of these measures can reveal the relative importance of the variables. Thus in our notation, we dene metrics as ratios of measures (completely analogous to probabilities); for example,

of variables in the classication scheme. For example, the points for which use of a pair, use of the rest of the variables, and use of all the variables agree are depicted in Fig. A1c (45 staples); on the same gure is the area (45 staples) corresponding to points (categories) that can be obtained only if all the variables are used. APPENDIX B Relative Comparison of Variables Utility Oklahoma Storm Tables with various metrics for relative comparison of the variables utility in the classication scheme are listed herein. In tables with measures on P and subsets of P, the second member of the relevant pair is listed in the vertical column dening the rows. (The rst member of the pair is always the reectivity factor Z h .) Thus, the variable is either part of the pair used exclusively for classication, or it is the one that is excluded from the classication; from the table titles and appropriate label of the dening column, it should be clear which one it is. The indicated class categories are as listed in the text and the gures. The meaning of various measures presented in the forthcoming tables is explained in appendix A. Similar to Table 4, Table B1 presents percentages of

(P A)/(A),
indicates the % of points (correctly classied) in a class (obtained with a specic pair), which agrees with the points obtained if all the variables are used. On the graph (Fig. A1b), this is the stapled (45) area divided with the sum of the two stapled areas (depicted with 45 and 45 lines), and the value thus computed is analogous to a probability of detection. A quantity similar to a false alarm ratio is

(P P A)/(P);
it indicates the % of points in a class (obtained using a specic pair), which does not contribute to the nal classication (that uses all the variables). Its graphical depiction (Fig. A1b) is the ratio of the stapled (vertical) area to the sum of the stapled areas (vertical and 45). Seven meaningful metrics of this type are described in Table A1. A quick look at this table, at Fig. A1, and similar gures can help physical understanding and visualization of comparisons between the relative values

TABLE B1. Percent of correctly classied hydrometeors (R A)/(A) using one less variable; the variable not used is indicated to the left of the table. Without ZDR KDP hv T Hydrometeor categories LR 95 99 98 66 MR 62 81 98 42 HR 33 85 98 86 LD 0 100 100 71 R/H 96 67 74 99 GSH 94 95 96 79 HA 82 98 96 98 DS 61 98 100 34 WS 36 18 92 86 ICH 0 100 100 39 ICV 100 0 100 94 Tot 80 85 95 77 70% 30% 5 8 11 7 2 2 0 1

TABLE B2. Percent of correctly classied hydrometeors (R P A)/(A) for which the pair and its absence (i.e., the rest of variables) produce the same class. Hydrometeor categories LR ZDR KDP hv T 91 61 68 65 MR 39 21 92 42 HR 14 4 82 86 LD 0 0 0 0 R/H 60 62 64 26 GSH 85 62 0 63 HA 78 22 0 76 DS 5 17 0 24 WS 0 0 7 0 ICH 0 0 0 0 ICV 0 0 0 0 Tot 56 37 22 49 70% 30% 3 0 2 2 6 8 7 6

TABLE B3. Percent of correctly classied hydrometeors (P A R P A)/(A) that would be misclassied in the absence of the variable. Hydrometeor categories LR ZDR KDP hv T 5 1 2 34 MR 38 19 2 58 HR 67 15 2 14 LD 100 0 0 0 R/H 4 33 26 0 GSH 5 2 0 21 HA 18 2 0 2 DS 3 0 0 40 WS 0 1 8 0 ICH 9 0 0 0 ICV 0 98 0 0 Total 14 14 3 19

910

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

TABLE B4. Percent of correctly classied hydrometeors (R A R P A)/(A) that would be missed in the absence of the remaining variables. Hydrometeor categories LR ZDR KDP hv T 4 38 30 2 MR 23 60 6 0 HR 19 74 17 0 LD 0 100 100 71 R/H 36 5 9 74 GSH 8 33 96 17 HA 5 76 96 22 DS 56 81 100 11 WS 36 18 85 86 ICH 0 100 100 39 ICV 100 0 100 94 Total 24 48 73 27

TABLE B5. Percent of wrongly classied data (R R A)/(R) using the rest of the variables out of all those classied using the rest of the variables. Without ZDR KDP hv T Hydrometeor categories LR 19 3 0 60 MR 11 18 1 8 HR 39 46 37 0 LD 100 4 0 3 R/H 34 2 0 3 GSH 25 8 3 6 HA 2 4 0 1 DS 8 38 1 20 WS 2 0 57 92 ICH 17 0 81 ICV 18 0 7 Tot 20 15 5 24 70% 1 2 30% 7 8 10 8

TABLE B6. Percent of hydrometeors (A P A R A P A R)/(A) that can be correctly classied only if all the variables are used. Hydrometeor categories With ZDR KDP hv T LR 0 0 0 0 MR 0 0 0 0 HR 0 0 0 0 LD 0 0 0 29 R/H 0 0 0 1 GSH 1 2 4 0 HA 0 3 4 0 DS 36 2 0 26 WS 64 82 0 14 ICH 91 0 0 61 ICV 0 2 0 6 Total 6 2 2 5

classied precipitation, except it is for the case whereby a variable has been excluded. Large percentages indicate that the variable is relatively insignicant; that is, without it classication is still successful; small numbers indicate the opposite. Ranking of the variables according to the number of classes for which the percent is 70% is ZDR , T, KDP , hv ; smaller numbers in column 13 mean increased importance of the excluded variable because less classes are correctly identied by the rest of the variables. Ranking of the variables according to the number of classes for which the percent is 30% is ZDR , KDP , T, hv , and according to the percentage in the total (column 13), the ranking would be T, ZDR , KDP , hv . To determine how one pair and the rest of the variables contribute to the various categories, we present in Table B2 the percentages of correctly classied data for which the pair and the absence of the variables produces the same class (see Table A1 for denition). The table can loosely be interpreted as the information carried by the pair that is also present in the rest of the pairs. This means that the pair is worth as much as the rest; hence, large numbers imply higher ranking of the pair. Thus in this data, a good portion of light rain, graupel/small hail, and hail would be classied the same with or without the Z h , ZDR pair; that is for these categories, this pair is as good as the rest of the pairs. The information about

heavy rain (86% of data in Table B2) and hail (76%) carried by the pair Z h , T is also very high. The Z h , hv pair is almost sufcient for LR, MR, and HR classication. This occasional redundancy of a pair is partly caused by the common presence of Z h in all pairs. Small percentages in B2 indicate that the main contributor is either the variable or the rest but not both. According to the total, the variables (understood to be always paired with Z h ) would rank as ZDR , T, KDP , hv . Counting the data points in sets described by rows 4 and 5 of Table A1 produces Tables B3 and B4 (for clarity, see also Fig. A1). Thus, in B3 is the percentage of data that is missed in the absence of a variable. In B4 is the percentage of data that is missed in the absence of the rest of the variables. According to Table B3, HR and LD are primarily classied by the Z h , ZDR pair, the signicant contributor to MR is the pair Z h , T, and vertically oriented ice crystals are due to Z h , KDP . In the overall hydrometeor categories, the percent of contribution by any single pair is 19% or less, and according to the totals, the ranking is T, ZDR , and KDP second, and hv last. Table B4 is almost, but not quite complementary to B3, and from column 12, the rank of variables is ZDR , T, KDP , hv . Table B5 presents somewhat complementary information to Table 6. It indicates the percent of mis-clas-

JUNE 2001

ZRNIC ET AL.

911

sied data for cases whereby a specic variable is excluded from the scheme. Small percentages indicate the excluded variable is not very relevant for a classied category (i.e., the lack of it does not cause signicant mis-classication). Large values signify substantial misclassication due to the missing variable. Hence, this criteria would rank the variables as T, ZDR , KDP , hv . Note again the ZDR and T lead because the rst is needed to identify LD and the second to identify WS. The missing entries are where (R) 0; hence, the omitted variables prevent classication in the corresponding category. Finally, Table B6 presents the percentage of correctly identied hydrometeors for which all the variables are needed. These percentages are relative with respect to the pair and its absence. The mis-classied data is because the pair assigns it one class; the rest assigns it another class, but use of all the variables assigns it a class different from either of these two. That is the combined use of the pair, and the rest overrides either one. For example, consider dry snow (DS) and the pair Z h , ZDR ; if this pair and the rest were used separately, then 36% of dry snow would be missed. Note the high percent of wet snow obtained by combining Z h , KDP , and the rest which could not be obtained by either one. As

expected the, categories where percentages are relatively high are those for which there is signicant overlap in the boundaries (i.e., DS, WS, ICH). APPENDIX C Relative Comparison of Variables Utility Florida Storm In Table C1 are the percentages of classied hydrometeors that agree if one variable is excluded from the set with those obtained from all the variables. Large percentages go with relatively unimportant variables. Thus, the leading variable according to the number of categories that can be classied (column 13) would be ZDR ; the rest are of about equal importance. This same ranking ensues if the total % of the eld is a benchmark (column 12). The information carried by a pair of variables that is also present in the rest of the variables is summarized in Table C2. Here, the large percentages imply that the pair has relatively high importance. A surprisingly strong inuence of T is seen in both the total percentage (column 12) and the number of categories (column 13). This is perhaps due to the generally weaker polarimetric

TABLE C1. Percent of correctly classied hydrometeors (R A)/(A) using one less variable; the variable not used is indicated to the left of the table. Hydrometeor categories Without ZDR KDP LDR hv T LR 67 100 93 97 81 MR 61 98 97 99 96 HR 66 78 96 97 91 LD 33 99 75 99 75 R/H 51 82 80 100 89 GSH 91 95 97 98 98 HA 16 100 83 83 83 DS 64 98 96 100 88 WS 62 88 88 82 100 ICH 95 72 100 48 97 ICV 97 0 100 100 85 Tot 70 94 95 96 90 70% 3 10 11 10 11 30% 1 1 0 0 0

TABLE C2. Percent of correctly classied hydrometeors (R P A)/(A) for which the pair and its absence (i.e., the rest of variables) produce the same category. Hydrometeor categories LR ZDR KDP LDR hv T 63 58 63 92 80 MR 58 53 71 98 94 HR 47 52 20 95 91 LD 19 0 27 0 0 R/H 44 17 56 1 0 GSH 88 61 18 0 73 HA 0 0 0 0 78 DS 11 19 28 0 76 WS 0 7 0 81 0 ICH 2 23 0 34 8 ICV 0 0 0 0 0 Tot 35 33 30 21 67 70% 30% 1 0 1 4 6 6 7 8 6 6

TABLE C3. Percent of correctly classied hydrometeors (P A R P A)/(A) that would be mis-classied in the absence of the variable. Hydrometeor categories LR ZDR KDP LDR hv T 32 0 6 3 19 MR 38 2 3 1 4 HR 34 22 4 3 9 LD 66 0 25 0 0 R/H 49 4 20 0 0 GSH 6 4 2 0 2 HA 83 0 0 0 17 DS 7 0 1 0 10 WS 0 3 0 18 0 ICH 0 26 0 51 0 ICV 0 84 0 0 0 Total 14 5 3 3 8

912

JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY

VOLUME 18

TABLE C4. Percent of correctly classied hydrometeors (R A R P A)/(A) that would be missed in the absence of the remaining variables. Hydrometeor categories LR ZDR KDP LDR hv T 4 41 30 4 0 MR 2 46 26 1 2 HR 19 26 75 1 1 LD 14 99 48 99 75 R/H 6 65 24 98 89 GSH 3 34 79 98 26 HA 5 100 83 83 6 DS 53 79 68 100 13 WS 62 81 88 2 100 ICH 93 49 100 14 89 ICV 97 0 100 100 85 Total 35 60 64 75 23

TABLE C5. Percent of wrongly classied data (R R A)/(R) using the rest of the variables out of all those classied using the rest of the variables. Hydrometeor categories Without ZDR KDP LDR hv T LR 9 2 5 1 30 MR 3 4 1 0 1 HR 48 5 22 0 0 LD 82 4 30 6 0 R/H 45 8 8 13 9 GSH 23 9 0 0 15 HA 0 22 0 0 6 DS 1 6 0 1 3 WS 17 5 19 31 16 ICH 9 11 9 0 17 ICV 91 54 0 25 Total 30 6 5 1 10 70% 2 0 0 0 0 30% 7 10 10 11 11

TABLE C6. Percent of hydrometeors (A P A R A P A R)/(A) that can be correctly classied only if all the variables are used. Hydrometeor categories With ZDR KDP LDR hv T LR 2 0 1 0 0 MR 1 0 0 0 0 HR 0 0 0 0 0 LD 1 1 0 1 25 R/H 0 14 0 0 11 GSH 3 2 2 2 0 HA 11 0 17 17 0 DS 29 1 3 0 2 WS 38 9 12 0 1 ICH 5 1 0 1 3 ICV 3 16 0 0 15 Total 16 2 2 0 2

signatures in the Florida storm compared to the Oklahoma one. The strength of a pair for classication can be gauged by the percent of data that would be missed in the absence of it (Table C3). The dominant pair for hail detection is Z h , ZDR ; in its absence, 83% of the hail would not be detected. (The importance of this miss is diminished by the fact that sample size for hail is very small.) The order of importance according to the total percentage is ZDR , T, KDP , and the last spot is shared by LDR and hv . The importance of T is seen in Table C4 whereby the percent of correctly classied data in the absence of a variable is indicated. Small percentages mean that the rest of the variables contribute relatively little to the class. Accordingly, we rank in order of importance T, ZDR , KDP , LDR , and hv . Percentages of wrongly classied hydrometeors due to the absence of one variable are in Table C5. From the total percent of misclassied data (Table C5), we note that the absence of ZDR has the largest effect on the total: 30% of the misses (column 12). This is followed by T, then LDR , KDP , and hv . The last Table, C6, indicates the percent of classied hydrometeors for which combination of the pair and the

rest is needed to accomplish classication. The results are similar to the ones in Table B6. There are few differences such as in wet snow for (Z h , KDP ) and horizontally oriented crystals (Z h , ZDR ), which have a much lower percentage for the Florida storm.
REFERENCES Balakrishnan, N., and D. S. Zrnic, 1990a: Estimation of rain and hail rates in mixed phase precipitation. J. Atmos. Sci., 47, 565583. , and , 1990b: Use of polarization to characterize precipitation and discriminate large hail. J. Atmos. Sci., 47, 15251540. Bringi, V. N., V. Chandrasekar, N. Balakrishnan, and D. S. Zrnic, 1990: An examination of propagation effects in rainfall on radar measurements at microwave frequencies. J. Atmos. Oceanic Technol., 7, 829840. Carey, L. D., and S. A. Rutledge, 1996: A multiparameter radar case study of the microphysical an kinematic evolution of a lightning producing storm. Meteor. Atmos. Phys., 59, 3366. Caylor, I. J., and V. Chandrasekar, 1996: Time-varying ice crystal orientation in thunderstorms observed with multiparameter radar. IEEE Trans. Geosci. Remote Sens., 34, 847858. Cornman, L. B., R. B. Goodtich, C. S. Morse, and W. L. Ecklund, 1998: A Fuzzy logic method for improved moment estimation from Doppler spectra. J. Atmos. Oceanic Technol., 15, 1287 1305. Doviak, R. J., and D. S. Zrnic, 1993: Doppler Radar and Weather Observations. Academic Press, 562 pp.

JUNE 2001

ZRNIC ET AL.

913

, V. Bringi, A. Ryzhkov, A. Zahrai, and D. S. Zrnic, 2000: Considerations for polarimetric upgrades of operational WSR88D radars. J. Atmos. Oceanic Technol., 17, 257278. Hall, M. P. M., S. M. Cherry, J. W. F. Goddard, and G. R. Kennedy, 1980: Raindrop sizes and rainfall rate measured by dual-polarization radar. Nature, 285, 195198. Herzegh, P. H., and A. R. Jameson, 1992: Observing precipitation through dual-polarization radar measurements. Bull. Amer. Meteor. Soc., 73, 13651374. Holler, H., V. N. Bringi, J. Hubbert, M. Hagen, and P. F. Meischner, 1994: Life cycle and precipitation formation in a hybrid-type hailstorm revealed by polarimetric and Doppler radar measurements. J. Atmos. Sci., 51, 25002522. Hubbert, J. V., V. N. Bringi, and L. D. Carey, 1998: CSU-CHILL polarimetric radar measurements from a severe hail storm in Eastern Colorado. J. Appl. Meteor., 37, 749775. Jameson, A. R., M. J. Murhpy, and E. P. Krider, 1996: Multipleparameter radar observations of isolated Florida thunderstorms during the onset of electrication. J. Appl. Meteor., 35, 343 354. Leitao, M. J., and P. A. Watson, 1984: Application of dual linearly polarized radar data for prediction of microwave path attenuation at 1030 GHz. Radio Sci., 19, 209221. Liu, H., and V. Chandrasekar, 2000: Classication of hydrometeor type based on multiparameter radar measurements: Development of a neuro-fuzzy system and in situ verication. J. Atmos. Oceanic Technol., 17, 140164. Lopez, R. E., and J. P. Aubagnac, 1997: The lightning activity of a hailstorm as a function of changes in its microphysical characteristics inferred from polarimetric radar observations. J. Geophys. Res., 102, 16 79916 813. Lutz, J., B. Rilling, J. Wilson, T. Weckwerth, and J. Vivekanandan, 1997: S-Pol after three operational deployment, technical performance, siting experiences, and some data examples. Preprints, 28th Conf. on Radar Meteorology, Austin, TX, Amer. Meteor. Soc., 286295. Mandel, J., 1995: Fuzzy logic systems for engineering: A tutorial. Proc. IEEE, 83, 345377. Papoulis, A., 1965: Probability, Random Variables and Stochastic Processes. McGraw-Hill Co., 583 pp. Ryzhkov, A., and D. S. Zrnic, 1996: Assessment of rainfall mea-

surement that uses specic differential phase. J. Appl. Meteor., 35, 20802090. , and , 1998: Beamwidth effects on the differential phase measurements of rain. J. Atmos. Oceanic Technol., 15, 624634. , , and D. Atlas, 1997: Polarimetrically tuned R(Z) relations and comparison of radar rainfall methods. J. Appl. Meteor., 36, 341349. Scarchilli, G., E. Gorgucci, and V. Chandrasekar, 1996: Self-consistency of polarization diversity measurement of rainfall. IEEE Trans. Geosci. Remote Sens., 30, 2226. Seliga, T. A., and V. N. Bringi, 1978: Differential reectivity and differential phase shift: Applications in radar meteorology. Radio Sci., 13, 271275. Straka, J. M., 1996: Hydrometeor elds in a supercell storm as deduced from dual-polarization radar. Preprints, 18th Conf. on Severe Local Storms, San Francisco, CA, Amer. Meteor. Soc., 551 554. , and D. S. Zrnic, 1993: An algorithm to deduce hydrometeor types and contents from multiparameter radar data. Preprints, 26th Int. Conf. on Radar Meteorology, Norman, OK, Amer. Meteor. Soc., 513516. , , and A. V. Ryzhkov, 2000: Bulk hydrometeor classication and quantication using polarimetric radar data: Synthesis of relations. J. Appl. Meteor., 39, 13411372. Vivekanandan, J., D. S. Zrnic, S. M. Ellis, R. Oye, A. V. Ryzhkov, and J. Straka, 1999: Cloud microphysics retrieval using S-band dual-polarization radar measurements. Bull. Amer. Meteor. Soc., 80, 381388. Wakimoto, R. M., and V. N. Bringi, 1988: Dual-polarization observations of microbursts associlated with intense convection: The 20 July storm during the MIST project. Mon. Wea. Rev., 116, 15211539. Zahrai, A., and D. S. Zrnic, 1993: The 10-cm wavelength polarimetric radar at NOAAs National Severe Storms Laboratory. J. Atmos. Oceanic Technol., 10, 649662. Zrnic, D. S., and A. Ryzhkov, 1996: Advantages of rain measurements using specic differential phase. J. Atmos. Oceanic Technol., 13, 454464. , and , 1999: Polarimetry for weather surveillance radars. Bull. Amer. Meteor. Soc., 80, 389406. , N. Balakrishnan, C. L. Ziegler, V. N. Bringi, K. Aydin, and T. Matejka, 1993: Polarimetric signatures in the stratiform region of a mesoscale convective system. J. Appl. Meteor., 32, 678 693.

Das könnte Ihnen auch gefallen