Retrieving stratocumulus drizzle parameters using Doppler radar and lidar E WAN J. O’C ONNOR , ROBIN J. H OGAN AND A NTHONY J. I LLINGWORTH Department of Meteorology, University of Reading, United Kingdom Submitted to J. Appl. Meteorol., July 2003 ABSTRACT Stratocumulus is one of the most common cloud types globally with a profound effect on the Earth’s radiation budget, and the drizzle process is fundamental in understanding the evolution of these boundary-layer clouds. In this paper a combination of 94-GHz Doppler radar and backscatter lidar is used to investigate the microphysical properties of drizzle falling below the base of stratocumulus clouds. The ratio of the radar to lidar backscatter power is proportional to the fourth power of mean size so potentially can provide an accurate size estimate. Information about the shape of the drop size distribution is then inferred from the Doppler spectral width. The algorithm estimates vertical profiles of drizzle parameters such as liquid water content, liquid water flux and vertical air velocity, assuming that the drizzle size spectrum may be represented by a gamma distribution. The depletion timescale of cloud liquid water through the drizzle process can be estimated when the liquid water path of the cloud is available from microwave radiometers and our observations suggest that this timescale varies from a few days in light drizzle to a few hours in strong drizzle events. We have used both radar and lidar observations from Chilbolton, and aircraft size spectra taken during the Atlantic Stratocumulus Transition Experiment, to derive the following power law relationship between liquid water flux (LWF) in g m 2 s 1 and radar reflectivity (Z ) in mm6 m 3 : LWF 0 0093 Z 0 69 . This is valid for frequencies up to 94 GHz and therefore would allow a forthcoming spaceborne radar to measure liquid water flux around the globe to within a factor of two for values of Z above 20 dBZ . 1. Introduction bution from the vertical velocity of the air. Wakasugi et al. (1986) proposed that clear air Doppler radar provided the necessary air velocity information and Gossard (1988), using a 915 MHz wind profiler, also demonstrated that it is possible to separate the clear air return from the drizzle return and thus obtain the vertical air velocity directly. Millimeter-wave cloud radars have the necessary sensitivity and narrow beamwidth to measure the Doppler spectrum of the smaller drizzle droplets of interest but are no longer sensitive to the clear air return, so a method of estimating the vertical air velocity accurately by some other means is required as the terminal velocity of the drizzle droplets is comparable to the expected updrafts and downdrafts. The airborne Wyoming Cloud Radar aboard the University of Wyoming King Air uses the aircraft inertial navigation system to attempt to correct for the vertical air velocity (Vali et al., 1998; Galloway et al., 1999; French et al., 2000) but this is of course not possible from the ground. In addition to measuring the mean velocity and spectral width, Doppler radars are able to measure the full spectrum of velocities within the radar sample volume. It is therefore possible, in principle, to separate the cloud droplet component of the spectrum (droplets less than 50 m in diameter) from the drizzle component (Gossard et al., 1997; Babb et al., 1999). The terminal fall velocity of these cloud droplets is typically only a few cm s 1 , so that they can be considered as tracers of the air motion and the air velocity can be inferred. However, due to the sixth power dependence of the radar return on droplet diameter, the echo from the cloud mode tends to be much smaller than that of the drizzle, and only a small amount of turbulence is sufficient to smear the Doppler spectrum Boundary-layer clouds are one of the most significant components of the shortwave radiation budget of the Earth (Ramanathan et al., 1989; Harrison et al., 1990) and the accurate representation of such clouds in numerical models is crucial for understanding climate (Slingo and Slingo, 1991; Jones et al., 1994). Observations (Miller et al., 1998; Albrecht, 1989) and modelling studies (Albrecht, 1993; Wood, 2000) have shown that drizzle is important principally because it is involved in determining the cloud lifetime and evolution. The drizzle process may also have implications for the radiative properties of such clouds (Feingold et al., 1996, 1997) through alteration of the cloud droplet spectra. For the purposes of this paper, we define drizzle as water droplets greater than 50 m in diameter (e.g. French et al., 2000) which may or may not evaporate before reaching the surface, rather than the 200 m in the American Meteorological Society glossary (1959). Previous ground-based Doppler radar techniques for measuring rain and drizzle droplets have attempted to exploit the direct relationship between terminal fall velocity and droplet size. By assuming a model for the droplet size distribution, such as lognormal or gamma, the radar reflectivity factor, mean Doppler velocity and Doppler spectral width can be used to estimate the number concentration and mean size of the drops (Atlas et al., 1973; Frisch et al., 1995). However, an ambiguity arises because the measured mean Doppler velocity has a significant contri- Corresponding author address: Department of Meteorology, Earley Gate, PO Box 243, Reading RG6 6BB, United Kingdom. E-mail: [email protected]. 1 6250 Hz, yielding a folding velocity of 5 m s 1 . The first three moments of the Doppler spectrum are calculated from the average of thirty 256-point FFTs (fast fourier transforms) which gives a temporal resolution of 1.25 s. The estimated sensitivity is 50 dBZ at 1 km. It has been calibrated to within 1.5 dB by comparison with the 3 GHz radar at Chilbolton which itself has been calibrated to better than 0.5 dB using the non-independence of its polarimetric parameters (Goddard et al., 1994). The radar used in the second case study was the zenithpointing 35 GHz Rabelais radar, on loan from the University of Toulouse. It is of the conventional pulsed type with a pulse width of 0.33 s, a beamwidth of 0.4 , a PRF of 3125 Hz, has a folding velocity of 6 m s 1 and was sampled every 75 m. It has been calibrated to within 1.5 dB by comparison with the 3 GHz radar at Chilbolton in the same manner as the 94 GHz Galileo radar and the estimated sensitivity is 42 dBZ at 1 km. Situated close to the radar during both case studies was a zenith-pointing Vaisala CT75K ceilometer consisting of an InGaAs diode laser operating at 905 nm with a divergence of 0.75 mrad and a field of view of 0.66 mrad (both half angle). It is a fully automated system which produces averaged profiles every 30 s with a range resolution of 30 m. Calibration of the lidar to within 5% is achieved using the technique described by O’Connor et al. (2004). so that the narrow cloud peak is lost in the larger drizzle mode. In this paper we combine the Doppler radar measurements with those of a backscatter lidar to retrieve the crucial drizzle parameters. The lidar backscatter signal is approximately proportional to the second power of the diameter so the ratio of the radar to lidar backscatter power is a very sensitive function of mean size (Intrieri et al., 1993). Once the size is known, concentration and higher moments of the distribution can be derived from the observed radar reflectivity, which depends on the assumed size distribution. However, the technique may only be applied in the drizzle below cloud base as the lidar beam is strongly attenuated as soon as it penetrates the cloud. In drizzle there is minimal attenuation of the lidar and radar signals and cloud base is always well defined by the lidar, but cannot be detected using the radar echo which is dominated by the drizzle droplets. The inferred drizzle droplet concentration and mean size are refined further by using the Doppler spectral width to infer the shape of the droplet size distribution. We also correct the observed Doppler spectral width for turbulence by calculating the standard deviation of the measured mean velocities and assuming a 5 3 power law for the vertical velocity spectrum. We then calculate bulk parameters such as drizzle liquid water content and liquid water flux, and if total liquid water path is available from microwave radiometers, the timescale for the depletion of cloud water by drizzle may be estimated. The absolute value of the mean Doppler velocity is not used in the retrieval of the drizzle droplet size distribution but we are able to calculate the theoretical Doppler velocity that would be measured in still air for the derived size distribution. The difference between this and the actual Doppler velocity therefore yields the air vertical velocity. A useful validation for the technique is that over a few hours the mean of this inferred air velocity should be zero providing there are no topographical effects. The instrumentation used in this paper is described in section 2. In section 3, the theoretical basis for the technique is explained in detail. The parameters that are available from the radar and lidar are presented, as are any necessary assumptions. The expected accurracy and possible shortcomings of the technique are discussed in section 4. In sections 5 and 6 results are presented from two case studies undertaken at Chilbolton and a relationship between radar reflectivity and liquid water flux is proposed in section 7. 3. Theory a. Measured parameters A Doppler radar commonly measures the first three moments of the Doppler spectrum which, in principle, contain the information required to derive the parameters of a three-parameter droplet size distribution (Frisch et al., 1995). However, although the radar reflectivity is directly related to the droplet size distribution alone, the Doppler velocity and Doppler spectral width may have significant contributions from the air motion as well. The radar reflectivity factor for spherical liquid water droplets at frequency f , is defined as Zf K f T 2 2 K f 0 n D D 6 f D d D (1) 0 where K f T 2 is the dielectric factor of liquid water at temperature T, K f 0 2 is the dielectric factor of liquid water at 0 C, n D d D is the number concentration of water droplets with diameters between D and D d D, and f D is the Mie/Rayleigh backscatter ratio. The ratio of dielectric factors present in (1) ensures that radars of different wavelengths will all measure the same Z for a 0 C cloud containing Rayleigh-scattering liquid water droplets. The dielectric constant varies with temperature at 94 and 35 GHz and is calculated using the empirical formula given by Liebe et al. (1989). 2. Instrumentation The radar used in the first case study was the zenithpointing 94 GHz Galileo radar located at Chilbolton in Southern England. It is of the conventional pulsed type with a pulse width of 0.5 s, a beamwidth of 0.5 and is operated with a range resolution of 60 m and a PRF of 2 d d (2) n D D 6 D f D d D n D D 6 f D d D 0 0 log (vertical velocity spectral energy) The mean Doppler velocity, , measured by a zenithpointing Doppler radar is the sum of the vertical air motion, , and the mean Z -weighted droplet terminal fall velocity, d : (3) In this paper we adopt the convention that velocity is positive away from the radar. Beard (1976) provided semik k k empirical formulae for calculating the terminal velocity l s λ log (horizontal wavenumber) of individual water droplets, D . The radar Doppler spectral width, , is the Z - F IG . 1: Theoretical turbulent spectrum plotted on a log-log scale where weighted standard deviation of velocities within the pulse the dark shaded area is the turbulent energy affecting the radar in 1 secvolume. In the absence of turbulence a distribution of ond and the light grey area is the turbulent energy measured over 30 droplets will have an intrinsic Doppler variance due to the seconds. range of terminal velocities given by where a is the universal Kolmogorov constant, & is the D! " d 2 n D D 6 f D d D 0 2 (4) dissipation rate and k is the wavenumber which can be d 6 n D D D d D related to a length scale, L, by k 2() L. The turbulent f 0 contribution to the spectral width is then Broadening of the spectrum can occur if the droplets expek* rience additional random motion due to turbulence and for S k dk (8) t2 a vertically pointing radar with finite beamwidth there is ks a contribution from the component of the horizontal wind 3 along the beam. If it is assumed that the sources of spec(9) a & 2' 3 k+ 2' 3 ks 2' 3 2 tral broadening are independent of one another, the ob2' 3 3a & served spectral variance, 2 , is the sum of the variances L s 2' 3 L + 2' 3 (10) 2 2 ( from each source (Doviak and Zrnić, 1993) such that 2 d 2 # b 2 $ t 2 where k+ 2(, L + is the smallest scale probed by the Doppler radar (L + is half the radar wavelength) and ks 2(, L s corresponding to the scattering volume dimension which also includes large eddies travelling through the sampling volume within the dwell time (1 second) for the radar. For a beamwidth of 0 5 and no wind, L s is about mm and the impact of L + 9 m at 1 km whereas L + is 1.6 appears negligible. The cut-off for the turbulent kinetic energy spectrum in the viscous sub-range may be at larger scales than the smallest scale probed by the Doppler radar but, even if the cut off is at 10 cm, the second term in (10) is only 5% of the first term and L + can be ignored. The difficulty with this technique using the observed Doppler width, , for a one second dwell is that we cannot separate the d and t components in (5). We now follow Bouniol et al. (2003) and consider a new parameter, . - 2 , which is the variance of the individual mean Doppler velocities measured each second, computed over 30 seconds, and show that . - 2 is dominated by turbulence and can be used to estimate the t 2 component of the one second Doppler variance 2 in (5) which can be subtracted so that d 2 can be derived. Figure 1 displays the theoretical turbulent energy spectrum as a function of wavenumber and the two in- (5) where b 2 is the contribution due to finite beamwidth and t 2 is the contribution from air turbulence. We are interested in d 2 so we need to estimate b 2 and t 2 and remove them. The value of b , assuming a circularly symmetric Gaussian pattern, is given by (Doviak and Zrnić, 1993) b U% 4 ln 2 (6) where % is the one-way, half-power beamwidth of the radar antenna in radians and U is the horizontal wind. For a typical wind speed (U 10 m s 1 ) in the boundary layer, b 0 032 m s 1 for the 94 GHz Galileo radar. Kollias et al. (2001) estimated the turbulence in fairweather cumuli by assuming that t 2 in (5) was the dominant contribution. We follow the method of Bouniol et al. (2003), where it is assumed that turbulence is a homogeneous and isotropic process of energy dissipation. The Kolmogorov hypothesis then states that the statistical representation of the turbulent energy spectrum S k is given by S k a & 2' 3 k 5' 3 (7) 3 0.5 0.4 Ratio of σt 2 to σ−v 2 The equations (7-15) are based on the assumption that the length scales of turbulent eddies being probed by the Doppler radar lie within the inertial subrange of the turbulence spectrum and that - 2 is dominated by turbulence rather than any coherent fluctuations in droplet terminal velocity. These assumptions have been shown to be valid by Bouniol et al. (2003) who observed that, in drizzle, the value of & derived from - 2 is independent of the integration time. The lidar extinction coefficient, 4 (in m 1 ), is defined as ( (16) n D D 2 d D 4 2 0 0.5 km 1 km 2 km 3 km 0.45 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0 10 20 30 40 −1 Horizontal wind speed, U (m s ) 50 where it is assumed that the lidar wavelength is small compared to the particle size and the geometric optics apF IG . 2: Theoretical ratio of / t 2 to /12 0 2 as a function of horizontal wind proximation can be applied. The relationship between 4 speed for a beamwidth of 0 53 at various altitudes. and the lidar backscatter coefficient, 5 (in m 1 sr 1 ), is described by tegrals that relate to - 2 and t 2 , where we assume the (17) 4 S5! turbulent contribution dominates . - 2 and is given by where S (in sr) is termed the lidar ratio and varies with ks wavelength and droplet size. Raman lidars (Ansmann (11) et al., 1992) and high spectral resolution lidars (Grund S k dk - 2 kl and Eloranta, 1990) can measure 4 directly but most li2' 3 3a & 2' 3 2' 3 dars measure only the attenuated backscatter coefficient Ll Ls (12) 2 2( 5,6 which is related to the true backscatter coefficient, 5 , where kl 2(, L l and relates to the large eddies trav- by z elling through the sampling volume during the averaging 5 6 z 57 z exp 2 48 z 6 dz 6 (18) 0 time. If we take the ratio of (10) to (12) we have Generally, the attenuation in drizzle is small (4 9 L s 2' 3 t2 1 0 5 km ) and therefore 4 can be retrieved from (17) (13) (18) without experiencing instability, providing S can . - 2 L l 2' 3 L s 2' 3 and be estimated with sufficient accuracy. The estimation of The length scale is given by, S using Mie theory is described below. % L U t 2z sin (14) b. Drizzle parameters 2 We now have three independent measurables, Z , where t is the observation time and z is the height in m. and 4 from which to derive the three parameters (NW , For an averaging time of 30 seconds the second term in D0 and ) describing the droplet size distribution, n D , (14) can often be ignored and L l U t. For L s , t is 1 which we assume can be represented by a normalised second and the correction for the beamwidth is necessary gamma distribution of the form for low U. : D [3 67 ] D The theoretical ratio of t 2 to - 2 is displayed in Fig. 2 n D NW f exp and shows the effect of the beamwidth correction at low D0 D0 wind speeds. It also shows that for low altitudes and hori(19) zontal winds greater than about 10 m s 1 the ratio is close where NW is the concentration normalised so that the liqto 0.14 so that for the Galileo radar at a height of 1 km the uid water content is independent of , D0 is the median contribution of turbulence to 2 on a 1 second timescale equivolumetric diameter, represents the shape of the is estimated as distribution and (15) t 2 0 14 - 2 6 3 67 4 f (20) ; The horizontal winds from the Met Office mesoscale 3 674 4 model are generally accurate to 1-2 m s 1 (Panagi et al., ; 2001) so, in this study, we have used the model winds and where denotes the gamma function. For a value of Fig. 2 to provide a more precise estimate of the turbulent 0, (19) reduces to the familiar inverse-exponential contribution. distribution. 4 be determined accurately using Mie scattering theory by including the factor f D from (1) in (21). For the purposes of this paper we define 6 to describe this departure from pure Rayleigh scattering in terms of distributions of droplet sizes rather than for single droplets (i.e. Z Mie 6 Z Rayleigh ). Figure 3 shows theoretical values of 6 at 94 and 35 GHz as a function of D0 and . For D0 9 200 m, 6B D0 1 to within 2% at 10 C for all , but at larger sizes Mie scattering effects become significant and, for certain sizes, Mie scattering is stronger than the Rayleigh scattering assumption due to the influence of the resonance region described by van de Hulst (1957). The temperature profile obtained from observations or an operational forecast model is used to account for the small temperature dependence of 6 . 1.4 Mie−to−Rayleigh ratio, γ ’ 1.2 1 0.8 0.6 0.4 µ= 0 µ= 2 µ= 5 µ = 10 0.2 0 10 100 Median volume diameter, D0 (µm) 1000 µ= 0 µ= 2 µ= 5 µ = 10 d −1 Intrinsic Doppler spectral width, σ (ms ) F IG . 3: Theoretical Mie-to-Rayleigh ratio, <>= , at 94 GHz (black) and 35 GHz (grey) for gamma distributions of droplet sizes with different values of ? at 103 C. If we initially assume Rayleigh scattering so that D f 1 in (1) and that, in the absence of attenuation, the lidar extinction coefficient can be obtained from the observed lidar backscatter, then we can take the ratio of Z to 4 defined in terms of (19) to remove the dependence on NW f and obtain as a first approximation Z 4 ( 2 0 D 6@ 0 D 2@ : : exp [3A 67@ D0 exp [3A 67@ D0 : : ]D dD ]D dD (21) 1 0.1 0.01 10 100 Median volume diameter, D (µm) 1000 0 F IG . 4: Intrinsic Doppler spectral width, / d , at 94 GHz for gamma distributions of droplet sizes with different values of ? at 103 C. µ= 0 µ= 2 µ= 5 µ = 10 d −1 Intrinsic Doppler spectral width, σ (ms ) where the slight variation of the dielectric factor with temperature in (1) has been accounted for by using the temperature profile obtained from observations (radiosondes) or numerical operational forecast models. Integration over all sizes yields ; Z 2 7 1 (22) D0 4 ; 4 ( 3 3 67 4 and the potential accuracy of the technique is illustrated by the fourth power dependence of the radar/lidar ratio on D0 , which results in the relative error in retrieved size being much less than the error in the input parameters Z and 4 , and any small errors due to the truncation ratio of the gamma function (Ulbrich and Atlas, 1998) will be similarly reduced to negligible levels. At high frequencies the Rayleigh scattering assumption can only strictly be applied when dealing with individual cloud and drizzle droplets smaller than around 300 m at 94 GHz and 1 mm at 35 GHz. Since we are considering size distributions whose droplets are spherical (droplets are not significantly aspherical until they are a few millimetres in diameter) and have a well defined refractive index, their backscattering properties can 1 0.1 0.01 10 100 Median volume diameter, D0 (µm) 1000 F IG . 5: Intrinsic Doppler spectral width, / d , at 35 GHz for gamma distributions of droplet sizes with different values of ? at 103 C. Figures 4 and 5 display theoretical values of 5 d as a where S and 6 are functions of and D0 ( 6 also varies very slightly with temperature), and are implemented as look-up tables. A first estimate of D0 can be found by assuming a value of 0 in (23). This estimate can then be refined iteratively by comparing the observed spectral width (corrected for turbulence) with that calculated using (4), adjusting to agree with observations and recomputing D0 until convergence. Once D0 and are established, NW is derived from observed Z . Now that we have derived the three parameters, NW , D0 and that define n D , d can be calculated independently of the mean Doppler velocity, using (3), and the vertical air motion can be inferred. The drizzle liquid water content (LWCd and the drizzle liquid water flux (LWFd ) are defined as follows; ( n D D 3 d D (24) LWCd F l 6 0 ( LWFd (25) n D D 3 D d D F l 6 0 function of median volume diameter for various values of at 94 and 35 GHz. These show the smooth transition from the Stokes regime, where d C D 2 for the terminal fall velocity of droplets 9 60 m, to a more linear regime. Mie scattering effects are again noticeable when D0 D 200 m. It can be seen that d varies by about a factor of two between 0 and 10, allowing retrieval of from after it has been corrected for turbulence and any horizontal winds using (6). The relationship between lidar extinction and backscatter in stratocumulus clouds and drizzle can also be derived using Mie theory. The value of the lidar backscatter coefficient oscillates wildly with size for individual droplets but once a realistic spectrum of droplet sizes is present these oscillations are smoothed out (O’Connor et al., 2004). Figure 6 displays the 905 nm extinction-to-backscatter ratio, S, as a function of D0 and . Our computations show that the change in visible refractive index is negligible for the range of temperatures expected. For droplet median diameters ranging from Extinction−to−backscatter ratio, S (sr) where F l is the density of liquid water. If we can estimate how much liquid water is in the cloud and the rate at which drizzle is leaving the cloud, then the timescale for cloud liquid water depletion by drizzle is given by 100 G LWP LWFd (26) 10 where LWP is the total liquid water path (i.e. cloud plus drizzle) obtained from coincident radiometers and LWFd is the liquid water flux immediately below cloud base. µ= 0 µ= 2 µ= 5 µ = 10 1 10 100 Median volume diameter, D0 (µm) 4. Error analysis 1000 In this section we estimate the error in the retrieval of D0 , drizzle liquid water content, drizzle liquid water flux and vertical air velocity. We first assume that the shape of the droplet size distribution (i.e. the value of in Eq. 22) is known and the droplets are Rayleigh scattering with respect to the radar. If we integrate the appropriately weighted gamma function over the drop size distribution then we have F IG . 6: Theoretical lidar ratio, S, at 905 nm as a function of D0 for gamma distributions of droplet sizes with different values of ? . 10 to 50 m the value of S D0 is almost constant with a value of 18 8 sr. Typical droplet diameters for stratocumulus with significant liquid water content (Miles et al., 2000) lie between 8 and 20 m. It is this feature of stratocumulus clouds that can be used to calibrate the lidar (O’Connor et al., 2004). The median diameter of the droplet size distribution for drizzle droplets falling below the cloud (D0 D 50 m) lie in the range where S D0 can no longer be assumed constant and needs to be taken into account. Z 5 LWCd LWFd d C C C C C N W D0 7 N W D0 3 N W D0 4 N W D0 5 D0 (27) (28) (29) (30) (31) c. Algorithm Equation 22 can now be written as ; Z 2 7 S D0 6E D0 D0 4 ; 5 ( 3 3 67 4 where it has been assumed that for LWFd , the terminal velocity distribution D in (25) is adequately represented by C D. The weak functional dependence on is ad(23) dressed later. 6 The random error in Z can be expressed, in linear and is about 18%. space, as The Z -weighted droplet terminal velocity, d , is proH 1 portional to D0 according to (31) and will therefore have Z I (32) the same error characteristics as D0 with a relative error MI of 12%. If D0 is very low, then the contribution of drops where M I is the number of equivalent independent samfalling in the Stoke’s regime could lead to d C D0 2 , and ples, given by Doviak and Zrnić (1993): a consequent error in d of 24%. The mean Doppler ve1 locity is measured directly by the radar and the error 4 G d ( 2 (33) in is better than the bin width of the Doppler spectrum MI J produced by the FFT as part of the coherent processing J where is the spectral width, G d the dwell time and the algorithm. For the Galileo radar the velocity resolution radar wavelength. In this paper the radar data are averaged is on the order of 4 cm s 1 . The error in the vertical air over 30 seconds to match the temporal resolution of the li- velocity is given by H H H dar data. With a typical spectral width of 0 5 m s 1 , the 1 2 random error in Z is about 0 02 dB, or 0.5%, at 94 GHz (38) d 2 2 H 35 GHz. A larger systemand close to 0 04 dB, or 1%, at atic error may be present due to the difficulty of accurately and as is so low, the error in is the same magnitude calibrating a radar of this wavelength. We estimate the as the error in d . accuracy of the calibration to be around 1.5 dB (Hogan The Doppler spectral width contains the information et al., 2003) or 40%. about the shape of the droplet size distribution. Figures 4 The lidar is a commercial instrument and the errors and 5 show that estimating the value of to within one are difficult to determine. However, the signal to noise element of the sequence 0, 2, 5, 10, requires that d can be ratio is good at low altitudes and the lidar has been obtained to within 30%. In principle, the Doppler spectral calibrated to within 5% using the technique described width, , can be measured to much better than 30%, but by O’Connor et al. (2004) so that we estimate the error in the error in deriving d depends on the relative size of the lidar backscatter, 5 , for individual rays to be about 10%. other contributing terms to plus their associated error. From (22), if S is constant, D0 C Z K5) 1' 4 but for It was shown in section 3 that b is typically a few drizzle the factor S D0 from (23) should also be in- cm s 1 and can be neglected. We observe that typical valcluded; the slope of the curves in Fig. 6 for D0 D 100 m ues of M - in drizzle are of the order 0 1 m s 1 so from (15) indicates this can be approximated by S C D0 0A 5 so the estimated value of t is about 0 04 m s 1 . From Figs. 4 that and 5, the terminal velocity component of the Doppler 2 1 7 (34) width, d , is above 0 1 m s once D0 D 100 m, and so D0 C Z L5M is less than 15%. Consequently, the turbulent correction The fractional error in the median volume diameter is then we should be able to distinguish changes in from 0 to 2, H H H 1 2 to 5 or 5 to 10 quite easily for D0 D 100 m and obtain 2 5 2 D0 2 Z 2 (35) an indication of the value of for D0 D 50 m. The effect D0 7 Z 5 of these changes in can be derived by an appropriately weighted integration of the normalised gamma function so that if the fractional error in Z is about 40% and the (19) to give the dependence in equations (27-31), and fractional error in 5 is about 10%, the relative error in D0 reveals that these step changes in lead to a change in D0 is about 12% when is known. and LWFd of about 7% and minimal change in the liquid For the drizzle liquid water content, LWCd C NW D0 4 , water content. Hence we conclude that, for the instrument which can be written as LWCd C Z D0 3 or LWCd C errors considered, the overall error in D0 and is about Z 1' 7 5 6' 7 and the fractional error, 14%, LWCd about 10% and LWFd about 20%. These erH H H rors are valid provided that, at 94 GHz, D0 is less than 1 2 2 around 500 m, and at 35 GHz, D0 is less than around LWCd 1 Z 6 5 2 (36) 1 mm. LWCd 7 Z 5 is about 10%. Considering the drizzle liquid water flux, 5. 11 September 2001 Case Study LWFd C NW D0 5 , which can be written as LWFd C We now apply the technique described in section 3 to 2 Z D0 or LWFd C Z 3' 7 5 4' 7 , the fractional error is given data taken by the Galileo Doppler radar and the Vaisala by lidar ceilometer at Chilbolton. Figure 7 shows 3 hours of H H H 1 data taken on the morning of 11 September 2001 during 2 2 2 LWFd 1 3 Z 4 5 typical stratocumulus conditions with appreciable drizzle (37) 5 LWFd 7 Z falling beneath the base of the cloud. Panel (a) shows 7 1 dBZ Height (km) 2 (a) Chilbolton 94 GHz Galileo radar − Radar Reflectivity Factor m −1 2 10−4 (b) Chilbolton 905nm CT75K Lidar Ceilometer − Attenuated backscatter coefficient −1 1 sr Height (km) 0 −7 10 −1 −1 0 −2 0.1 −1 1 0 5:30 −1.5 1 (d) Chilbolton 94 GHz Galileo radar − Doppler Spectral Width ms Height (km) 10−6 −0.5 1 2 10−5 0 (c) Chilbolton 94 GHz Galileo radar − Mean Doppler Velocity ms Height (km) 0 2 10 0 −10 −20 −30 −40 −50 6:00 6:30 7:00 Time (UTC) 7:30 8:00 8:30 0.01 F IG . 7: Observed variables for 11 September 2001. Time-height plots of (a) radar reflectivity factor, (b) attenuated lidar backscatter, (c) radar Doppler velocity (positive away from the radar) and (d) radar Doppler spectral width. The black line in each panel indicates cloud base derived from the lidar. radar reflectivity, Z , and panel (b) coincident attenuated lidar backscatter 5,6 . The lidar shows a prominent cloud base at all times, while the radar appears unable to discriminate between drizzle in cloud and drizzle falling beneath the cloud. Cloud base remains relatively constant at 1.5 km with occasional departures to 1 km which may indicate pannus, and cloud top appears constant for long periods with a typical cloud depth of 350 m. The radar reflectivity is dependent on the sixth power of the diameter and so the larger, but far less numerous, drizzle droplets in cloud and below cloud dominate the radar return even though their liquid water content is neglible compared to that of the small but numerous cloud droplets. This also explains why, in regions where it is not drizzling (such as before 05:40 and after 08:20 UTC), cloud base is detected by the lidar but not the radar as it is not so sensitive to the small cloud droplets. Reflectivity values reach 0 dB both in cloud and below cloud while 5,6 values reach 5 N 10 5 sr 1 m 1 below cloud and jump rapidly to 5 6 D 2 N 10 4 sr 1 m 1 when penetrating the cloud. The drizzle evaporates completely before reaching the ground. The background lidar signal of 10 6 sr 1 m 1 is due to boundary layer aerosol. The radar Doppler velocity, displayed in panel (c), shows a large variation in the velocity of the drizzle, ranging from 0 5 to 2 m s 1 , and the pattern of ve locities is characterized by narrow fall streaks, indicating the stongly inhomogeneous nature of drizzle. These streaks usually coincide with the increases seen in radar reflectivity and lidar backscatter below cloud. Within the cloud the velocities are generally of a smaller magnitude and decrease towards cloud top where they are close to 0 m s 1 . Cloud droplets have terminal velocities of only a few cm s 1 and as the drizzle droplets grow a corresponding increase in terminal fall velocity is observed. The Doppler spectral width, depicted in panel (d), is also characterized by streaks which coincide with the increases in radar reflectivity and lidar backscatter below cloud. The high values seen just prior to 07:00 UTC coincide with the strong negative values in the Doppler velocity. The derived microphysical parameters are displayed in Fig. 8. Panel (a) shows D0 , the median volume diameter of the derived drizzle droplet size distribution, varying from 40 m to 250 m, similar sizes to those found by Vali et al. (1998). The value of which describes the shape of the size distribution is shown in panel (b). The derived distributions are consistently broad although 8 (a) Drizzle Median Diameter 200 1 100 0 Height (km) 50 µm Height (km) 2 2 10 8 6 4 2 0 −1 (b) Drizzle Shape Parameter 1 2 10−1 (c) Drizzle Liquid Water Content −2 10 −3 1 0 10−4 −1 2 10 (d) Drizzle Liquid Water Flux 10−2 −2 −1 1 gm s Height (km) 10−3 gm Height (km) 0 0 5:30 6:00 6:30 7:00 Time (UTC) 7:30 8:00 −3 10 −4 10 8:30 F IG . 8: Drizzle parameters derived from the radar and lidar for 11 September 2001: (a) median diameter D0 , (b) shape parameter water content and (d) liquid water flux. ? , (c) liquid (b) Estimated air velocity, w 1 0.5 5 10 15 5 10 −1 U (ms ) Velocity (ms−1) 0.5 15 Distance (km) 20 10−2 (c) Correlation −3 10 0 10−4 −0.5 10−5 5 10 15 Distance (km) 20 −1 25 −2 −1 0 0 0 LWF (g m s ) 1.5 1 (a) w (ms−1) Height (km) 2 25 F IG . 9: (a) Horizontal wind speed, U , taken from sondes at Larkhill at 05Z (solid) and 11Z (dashed) on 11 September 2001. (b) Time series of vertical air velocity, O , for a selected region (0712 to 0748 UTC) is shown with an aspect ratio of 3:1 (horizontal:vertical). Velocity is positive away from the radar. Cloud top as measured by the radar is shown by the black line. (c) Time series showing correlation of vertical air velocity, O , (black) and drizzle liquid water flux, LWFd , (red) at an altitude of 720 m for the same region as in (b). 9 there are occasions when much narrower distributions are observed (i.e higher values of ), particularly between and towards the base of the drizzle streaks. Preferential evaporation of the smaller drizzle droplets is a possible explanation. Liquid water content values (panel c) reach 0 2 g m 3 and liquid water flux values (panel d) reach 0 02 g m 2 s 1 (0 07 mm hr 1 ) which are consistent with et al. (1998) who found maximum drizzle rates of Vali 0.1-0 2 mm hr 1 . The background vertical air velocity, , can be estimated using (2) and (3) and the derived up and downdrafts reach 1 m s 1 with an error of up to 0 2 m s 1 . Radiosonde ascents at Larkhill (25 km to the west of Chilbolton) were available for 05:00 UTC and 11:00 UTC and were indicative of a decoupled cloud layer capped by a strong inversion at about 2 km and a boundary layer reaching 1 km below a transition layer which remained in place throughout the day. The mean horizontal wind measured by the radiosondes (Fig. 9a) at cloud level has been used to transform the time axis into a length scale and a section of (from 07:12 to 07:48 UTC) is shown in Fig. 9b. Wind shear is present below cloud and manifests itself by causing the drizzle streaks to fall at a significant angle to the vertical, up to 2.5 km in the horizontal for a 1 km fall in the vertical. There is a strong impression of a cellular structure which, if it extended through the whole boundary layer, had horizontal-vertical aspect ratios ranging from 1:1 to 3:1 but if confined to the cloud layer had horizontal-vertical aspect ratios of 2:1 to 6:1. The cloud top and base remain relatively constant throughout this period. A time series of and LWFd at 720 m are depicted together in Fig. 9c to show that an increase in drizzle liquid water flux is seen in the vicinity of updrafts. The mean vertical velocity during this period is 0 2 m s 1 and considering the error in this is not significantly different from zero. Thus far topography has been neglected and around Chilbolton the ground slopes up to 20 m over a distance of 1 km (a gradient of 2%). A steady horizontal airflow of 10 m s 1 could give rise to a vertical motion of 0 2 m s 1 . 6. 20 October 1998 Case Study An opportunity to estimate the drizzle depletion timescale occured during the Cloud Lidar and Radar Experiment, CLARE ’98 (ESA, 1999) , at Chilbolton in October 1998 using data from the 35 GHz Rabelais radar and Vaisala CT75K ceilometer. Estimates of LWP from microwave radiometers at 21.3, 23.8 and 31.7 GHz were provided by the Technical University of Eindhoven, Netherlands. The 35 GHz Rabelais did not measure Doppler spectral width during this period and so a constant value of 0 for n D was used when estimating D0 , LWCd and LWFd . This seems a reasonable assumption based on the values obtained on 11 September 2001 and any error in deriving values based on this assumption is not expected to significantly alter the results, as explained in section 4. For instance, if 5 when it has been assumed that 0, then LWFd is in error by only 20% whereas the depletion timescale varies over orders of magnitude. With no mean Doppler velocity available, it was not possible to estimate the vertical air velocity. Figure 10 shows two hours of data taken on the morning of 20 October 1998. Panel (a) shows radar reflectivity factor, Z , and cloud base derived from the lidar is superimposed. Panel (b) shows attenuated lidar backscatter and, as in the previous case, the cloud base is prominent in the lidar data but not in the radar data. Again, wind shear was present and affected the angle at which the drizzle fell although it does not appear to have been as strong within cloud. The cloud top remained constant at about 2.25 km while the cloud base had more variation and cloud depth ranged from 400 m to 800 m. The drizzle completely evaporated before reaching the ground. The derived microphysical parameters are shown in Fig. 11. Values of D0 (panel a) reach 300 m in the strong drizzle streaks near 08:00 UTC with LWCd values reaching 0 1 g m 3 (panel b) and LWFd values reaching 0 05 g m 2 s 1 (panel c). Values of the total column liquid water path, LWP, obtained from the microwave radiometers, range from 100300 g m 2 which are typical of stratocumulus (Greenwald et al., 1995) and distinct increases in cloud LWP correlate well with the strong drizzle streaks and associated increases in drizzle LWP. Comparison of the cloud LWP with the liquid water path of the drizzle in panel d indicates the relative partitioning of liquid water between cloud mode and drizzle mode; the drizzle LWP is often two orders of magnitude lower than the cloud LWP in light drizzle. This confirms that the drizzle LWP makes a negligible contribution to the total LWP measured by the radiometer which is dominated by the cloud. The value of G derived using (26) is shown in panel e and varies from several days for the weaker drizzle to two hours in the stronger drizzle events (the period after 07:35 UTC). The scatter in G matches the inherent variable nature of drizzle. 7. Relation between drizzle flux and radar reflectivity factor Observed values of radar reflectivity factor versus derived values of drizzle flux, LWFd , are plotted together with the line of best fit and its standard deviation, in Fig. 12 for the 94 GHz case on 11 September 2001, and in Fig. 13 for the 35 GHz case on 20 October 1998. The fits for the two cases agree quite well and we suggest that the power law relationship derived from the 11 September 2001 case; 10 LWF 9 3 N 10 6 Z 0A 69 (39) 2 dBZ Height (km) (a) 35 GHz Rabelais radar − Radar Reflectivity Factor 0 10 0 −10 −20 −30 −40 −50 −5 2 0 6:00 sr−1 m−1 Height (km) −4 10 (b) Chilbolton 905nm CT75K Lidar Ceilometer − Attenuated backscatter coefficient 6:30 7:00 Time (UTC) 7:30 8:00 10 −6 10 −7 10 2 (a) Drizzle Median Diameter 200 100 1 µm Height (km) F IG . 10: Observed variables for 20 October 1998: (a) radar reflectivity factor and (b) attenuated lidar backscatter. The black line indicates cloud base derived from the lidar. 0 50 2 −2 10 1 g m−3 Height (km) −1 10 (b) Drizzle Liquid Water Content 0 −3 10 −4 10 2 g m−2s−1 Height (km) −1 10 (c) Drizzle Liquid Water Flux 1 500 −2 LWP (gm ) 0 10−3 −4 10 (d) Liquid water path LWP cloud LWP drizzle x 20 0 3 τ (hrs) 10−2 10 (e) Depletion timescale 2 10 1 10 0 10 6:00 6:30 7:00 Time (hrs) 7:30 8:00 F IG . 11: Drizzle parameters derived from radar, lidar and microwave radiometer for 20 October 1998: (a) median diameter D0 , (b) liquid water content, (c) liquid water flux, (d) liquid water path and (e) drizzle depletion timescale. 11 −3 −3 10 10 −2 −1 log (LWF[kg m s ]) = 0.06889Z[dBZ] −5.031 10 Liquid water flux (kg m−2 s−1) Liquid water flux (kg m−2 s−1) 10 −4 −5 10 −6 10 −7 10 −8 v (m s−1) 2 −5 10 1.5 −6 10 1 −7 10 0.5 −8 10 −40 −4 10 10 −30 −20 −10 0 10 Radar reflectivity factor Z (dBZ) 20 −40 F IG . 12: Drizzle liquid water flux and radar reflectivity values derived for the 94 GHz case on 11 September 2001 with mean (solid) and P 1 standard deviation (dashed) fits to the data. 0 −30 −20 −10 0 10 Radar reflectivity factor Z (dBZ) 20 F IG . 14: Liquid water flux and radar reflectivity calculated from FSSP and 2DC size spectra measured by the Met Office C-130 during ASTEX. The shading of each point indicates the Z -weighted mean terminal velocity calculated from the spectra. −3 10 −3 log (LWF[kg m−2s−1]) = 0.07643Z[dBZ] −4.854 10 log (LWF[kg m−2s−1]) = 0.0673Z[dBZ] −4.754 10 −4 10 Liquid water flux (kg m−2 s−1) Liquid water flux (kg m−2 s−1) 10 −5 10 −6 10 −7 10 −4 10 −5 10 −6 10 −7 10 −8 10 −8 10 −40 −30 −20 −10 0 10 Radar reflectivity factor Z (dBZ) 20 −40 F IG . 13: Drizzle liquid water flux and radar reflectivity values derived for the 35 GHz case on 20 October 1998 with mean (solid) and P 1 standard deviation (dashed) fits to the data. where LWF is in kg m 2 s 1 and Z has units of mm 6 m 3 , would, from the scatter in Fig. 12, allow LWF to be measured to within a factor of two from Z alone. This is similar to the relationship derived for the 20th October 1998 case, where the shape of the size distribution was not known, and implies that the relationship is also valid at 35 GHz. It has been proposed that spaceborne radar will be able to retrieve liquid water content by using a Z -LWC relationship that also incorporates visible optical depth when available (Austin and Stephens, 2001; Stephens et al., 2002). This may be possible in drizzle-free clouds but ignores the fact that drizzle can dominate Z , especially in marine stratocumulus, while having a negligible impact on the liquid water content (Fox and Illingworth, 1997), so a Z -LWC relationship should be limited to nonprecipitating liquid water clouds (Papatsoris, 1994). A Z -LWF relationship is likewise limited to precipitating liquid water clouds, i.e. those that contain drizzle but, −30 −20 −10 0 10 Radar reflectivity factor Z (dBZ) 20 F IG . 15: Liquid water flux and radar reflectivity values calculated from FSSP and 2DC size spectra measured by the Met Office C-130 during ASTEX with thick lines indicating mean (solid) and P 1 standard deviation (dashed) fits to the data. To remove pure cloud droplet spectra, only values with a Z -weighted mean terminal velocity greater than 0 1 m s 1 are plotted and considered for the regression fit. since the presence of drizzle droplets greatly enhances the reflectivity, these are the clouds that are easily detected by a spaceborne cloud radar (Fox and Illingworth, 1997) whereas the reflectivity of non-precipitating liquid water clouds will usually be below the detection limit. No direct in situ validation was available for the radar studies in this paper, so, for the purposes of comparison, we have looked at aircraft observations of particle size spectra taken during the Atlantic Stratocumulus Transition Experiment (ASTEX) (Albrecht et al., 1995). Figure 14 shows drizzle flux, LWFd , versus radar reflectivity factor, Z , calculated from 10 second averages of the size distributions obtained by the Forward Scattering Spectrometer Probe (FSSP) and the 2D cloud probe (2DC) aboard the Met Office C-130 aircraft. It should be noted that the ASTEX data include events both in and below 12 cloud and the large scatter is due to the presence of both small cloud droplets and larger drizzle droplets. Any direct fit to the data will be biased by the high number of spectra containing cloud droplets only. However, the Z -weighted mean terminal velocity, d , calculated from 10 second averages of the size spectra using the formulae given by Beard (1976), and indicated by the shading of each point in Fig. 14, provides an objective means of separating the cloud and drizzle components so that a comparison can be made with the values of LWFd derived from the radar/lidar technique. Potentially, the velocity information in Fig. 14 could be used by a Dopplerised spaceborne radar, such as that proposed for EarthCARE (ESA, 2001) , to discriminate between drizzle and cloud. Figure 15 shows drizzle flux versus radar reflectivity factor calculated from the ASTEX data, for the drizzle component only, obtained by selecting the spectra with 0 1 m s 1 . The Z -LWFd relationship that is derived d D drizzle component of the data is relatively insenfrom the sitive to the value of d chosen as the threshold. The fit derived from the ASTEX data is reasonably consistent with those in Figs. 12 and 13 which are derived from below cloud base only. The bias of the ASTEX fit is probably due to the fact that some mixed drizzle and cloud droplet spectra have been included, and also because the aircraft probes provided a poor sample of the low concentration of the larger drizzle droplets. The 30 s averaging of the radar and lidar data corresponds to a horizontal distance of approximately 300 m, assuming a typical horizontal wind speed of 10 m s 1 , and 60 m in the vertical. The ASTEX data were averaged over 10 s, which corresponds to a horizontal distance of approximately 1 km, and was regarded as the shortest averaging period that would provide enough drizzle sized drops to form representative drizzle droplet spectra. This may account for the larger spread seen in Fig. 15, compared to Figs. 12 and 13, since longer averaging periods may have produced better spectra but would have encompassed regions with markedly different drizzle rates and concentrations. These plots show that a future spaceborne radar will be able to make much more accurate measurements of liquid water flux than liquid water content for values of Z above 20dBZ , where the radar reflectivity and liquid water flux is dominated by drizzle droplets and the liquid water content is dominated by cloud droplets. For lower values of Z , an ambiguity may arise in deriving liquid water flux or liquid water content, because both cloud and drizzle droplets can make a significant contribution to Z and liquid water flux; the ambiguity could be removed if the radar had a Doppler capability as envisaged for EarthCARE. ing radar and lidar to provide continuous measurements of drizzle. The technique only requires temporal averaging for matching the two data streams, thus 30 second temporal timescale (and smaller) is possible. This allows detection of the cellular structure and investigation of the inhomogeneities present in drizzle. An advantage of the technique is that there is no reliance on the mean Doppler velocity to obtain the droplet size distribution, in contrast to existing radar-only techniques. It has the potential to retrieve vertical profiles of D0 , drizzle LWC and drizzle LWF below cloud base to within 15-20% with the assumption that a gamma function fits the size distribution, and can also estimate the vertical wind and the shape parameter, , of the size distribution. The spatial scales of updrafts and downdrafts can also be derived and it was found that updrafts tended to coincide with the occurrence of the strongest drizzle streaks. This indicates that drizzle production may be enhanced by the evaporative cooling experienced below cloud which can have a feedback effect into stimulating the production of more drizzle. Vali et al. (1998) also observed upward transport of drizzle drops in cloud. Observations in the first case study suggest that, except at the edges of drizzle regions, the shape parameter, , tends to be close to zero and appears to confirm the findings of Ichimura et al. (1980) and Wood (2000), who both indicated that observations could be sufficiently well described by an exponential distribution. Therefore, the technique could be used by an un-Dopplerised radar assuming a fixed value of . There also appears to be a correlation between drizzle LWP and cloud LWP in the strong drizzle regions. Previous observations have stated that there is no relationship between the two and a long timeseries of data would be required to see if this is a regular occurrence. The drizzle LWP is affected by wind shear but the timescale for depletion of liquid water is derived from values of LWF taken directly below cloud base and is unaffected. The minimum value for this timescale is about 2 hours; more observations need to be made to link this timescale to other meteorological parameters. The presence of drizzle droplets enhances the reflectivity of liquid water clouds sufficiently so that they can be detected by a spaceborne cloud radar and, although it is not possible to retrieve the LWC of such clouds, a Z -LWF relationship has been shown to be robust enabling the drizzle beneath climatically important marine stratocumulus to be monitored routinely for the first time. Acknowledgements We thank the Radiocommunications Research Unit at the Rutherford Appleton Laboratory, Henri Sauvageot 8. Conclusion (University of Toulouse, France), Phil Brown (Met Office) A technique has been demonstrated that uses the in- and Suzanne Jongen (Technical University of Eindhoven, herent sensitivity of radar/lidar synergy with zenith point- Netherlands) for providing the data. The Galileo radar 13 was developed for the European Space Agency (ESA) by REFERENCES Officine Galileo, the Rutherford Appleton Laboratory and Albrecht, B. A., 1989: Aerosols, cloud microphysics, and the University of Reading, under ESTEC Contract No. fractional cloudiness. Science, 245, 1227–1230. 10568/NL/NB. This research was funded by NERC grant NER/T/S/1999/00105 and EU CloudNet contract EVK2- Albrecht, B. A., 1993: The effects of precipitation on the CT-2000-00065. The CLARE ’98 campaign was funded thermodynamic structure of trade-wind boundary layby ESA (grant 12957/98). ers. J. Geophys. Res., 98, 7327–7337. Albrecht, B. A., Bretherton, C. S., Johnson, D., Schubert, W. H., and Frisch, A. S., 1995: The Atlantic Stratocumulus Transition Experiment — ASTEX. Bull. Amer. Meteorol. Soc., 76(6), 889–904. American Meteorological Society, 1959: Glossary of Meteorology. American Meteorological Society, 45 Beacon St. Boston, MA. Ansmann, A., Wandinger, U., Riebesell, M., Weitkamp, C., and Michaelis, W., 1992: Independent measurement of extinction and backscatter profiles in cirrus clouds by using a combined Raman elastic-backscatter lidar. Appl. Opt., 31(33), 7113–7131. Atlas, D., Srivastava, R. C., and Sekon, R. S., 1973: Doppler radar characteristics of precipitation at vertical incidence. Rev. Geophys. Space Phys., 11, 1–35. Austin, R. T. and Stephens, G. L., 2001: Retrieval of stratus cloud microphysical parameters using millimetric radar and visible optical depth in preparation for CloudSat, Part I: Algorithm formulation. J. Geophys. Res., 106, 28,233–28,242. Babb, M. B., Verlinde, J., and Albrecht, B. A., 1999: Retrieval of cloud microphysical quantities from 94-GHz radar Doppler power spectra. J. Atmos. Ocean. Technol., 16, 489–503. Beard, K. V., 1976: Terminal velocity and shape of cloud and precipitation drops aloft. J. Atmos. Sci., 33, 851– 864. Bouniol, D., Illingworth, A. J., and Hogan, R. J., 2003: Deriving turbulent kinetic energy dissipation rate within clouds using ground based 94 Ghz radar. In 31st Conference on Radar Meteorology, Seattle, USA. Amer. Meteor. Soc., 193-196. Doviak, R. J. and Zrnić, D. S., 1993: Doppler radar and weather observations. Academic Press, 2nd edition. ESA (European Space Agency), 1999: International Workshop Proceedings, CLARE ’98, Cloud Lidar And Radar Experiment, ESA WPP-170. European Space Agency, ESTEC, Nordwijk, The Netherlands. ESA (European Space Agency), 2001: The Five Candidate Earth Explorer Core Missions - EarthCARE — Earth Clouds, Aerosols and Radiation Explorer, ESA 14 SP-1257(1). European Space Agency, ESTEC, Nord- Harrison, E. F., Minnis, P., Barkstrom, B. R., Rawijk, The Netherlands. manathan, V., Cess, R. D., and Gibson, G. G., 1990: Seasonal variation of cloud radiative forcing derived Feingold, G., Stevens, B., Cotton, W. R., and Frisch, A. S., from the Earth Radiation Budget Experiment. J. Geo1996: The relationship between drop in-cloud resiphys. Res., 95, 18687–18704. dence time and drizzle production in numerically simulated stratocumulus clouds. J. Atmos. Sci., 53(8), 1108– Hogan, R. J., Bouniol, D., Ladd, D. N., O’Connor, E. J., and Illingworth, A. J., 2003: Absolute calibration of 1122. 94/95-GHz radars using rain. J. Atmos. Ocean. Technol., 20(4), 572–580. Feingold, G., Boers, R., Stevens, B., and Cotton, W. R., 1997: A modeling study of the effect of drizzle on cloud optical depth and susceptibility. J. Geophys. Res. Ichimura, I., Fujiwara, M., and Yanase, T., 1980: The size distribution of cloud drops measured in small maritime — Atmos., 102(D12), 13527–13534. cumulus clouds. J. Meteorol. Soc. Jpn., 58, 403–415. Fox, N. I. and Illingworth, A. J., 1997: The retrieval Intrieri, J. M., Stephens, G. L., Eberhard, W. L., and Uttal, of stratocumulus properties by ground based radar. J. T., 1993: A method for determining cirrus cloud partiAppl. Meteorol., 36, 485–492. cle sizes using lidar and radar backscatter technique. J. Appl. Meteorol., 32, 1074–1082. French, J. R., Vali, G., and Kelly, R. D., 2000: Observations of microphysics pertaining to the development Jones, A., Roberts, D. L., and Slingo, A., 1994: A cliof drizzle in warm, shallow cumulus clouds. Q. J. R. mate model study of indirect radiative forcing by anMeteorol. Soc., 126, 415–443. thropogenic sulfate aerosols. Nature, 370, 450–453. Frisch, A. S., Fairall, C. W., and Snider, J. B., 1995: Mea- Kollias, P., Albrecht, B. A., Lhermitte, R., and surement of stratus cloud and drizzle parameters in ASSavtchenko, A., 2001: Radar observations of updrafts, TEX with a KQ -band Doppler radar and a microwave downdrafts, and turbulence in fair-weather cumuli. J. radiometer. J. Atmos. Sci., 52(16), 2788–2799. Atmos. Sci., 58, 1750–1766. Galloway, J., Pazmany, A., Mead, J., McIntosh, R. E., Liebe, H. T., Manabe, T., and Hufford, G. A., 1989: Leon, D., French, J., Haimov, S., Kelly, R., and Vali, Millimeter-wave attenuation and delay rates due to G., 1999: Coincident in situ and W-band radar meafog/cloud conditions. IEEE AP, 37, 1617–1623. surements of drop size distribution in a marine stratus cloud and drizzle. J. Atmos. Ocean. Technol., 16, 504– Miles, N. L., Verlinde, J., and Clothiaux, E. E., 2000: Cloud droplet size distributions in low-level stratiform 517. clouds. J. Atmos. Sci., 57, 295–311. Goddard, J. W. F., Tan, J., and Thurai, M., 1994: Technique for calibration of meteorological radars using dif- Miller, M. A., Jensen, M. P., and Clothiaux, E. E., 1998: Diurnal cloud and thermodynamic variations in the ferential phase. Electron. Lett., 30, 166–167. stratocumulus transition regime: A case study using in situ and remote sensors. J. Atmos. Sci., 55, 2294–2310. Gossard, E. E., 1988: Measuring drop-size distributions in clouds with a clear-air-sensing Doppler radar. J. AtO’Connor, E. J., Illingworth, A. J., and Hogan, R. J., mos. Ocean. Technol., 5, 640–649. 2004: A technique for autocalibration of cloud lidar and for inferring the lidar ratio for ice and mixed phase Gossard, E. E., Snider, J. B., Clothiaux, E. E., Martner, B., clouds. J. Atmos. Ocean. Technol., 21(5), 777–786. Gibson, J. S., and Kropfli, R. A., 1997: The potential of 8-mm radars for remotely sensing cloud drop size Panagi, P., Dicks, E., Hamer, G., and Nash, J., 2001: Predistributions. J. Atmos. Ocean. Technol., 14, 79–87. liminary results of the routine comparison of wind profiler data with the Meteorological Office Unified Model Greenwald, T. J., Stephens, G. L., Christopher, S. A., and vertical wind profiles. Phys. Chem. Earth (B) - Hydrol. Von der Haar, T. A., 1995: Observations of the global Oceans Atmos., 26, 187–191. characteristics and regional radiative effects of marine cloud liquid water. J. Climate, 8, 2928–2946. Papatsoris, A. D., 1994: Implication of super-large drops in millimeter-wave radar observations of water clouds. Grund, C. J. and Eloranta, E. W., 1990: The 27-28 OcElectron. Lett., 30(21), 1799–1800. tober 1986 FIRE IFO cirrus case-study - cloud opticalproperties determined by High Spectral Resolution Li- Ramanathan, V., Cess, R. D., Harrison, E. F., Minnis, P., dar. Mon. Weather Rev., 118(11), 2344–2355. Barkstrom, B. R., Ahmad, E., and Hartmann, D., 1989: 15 Cloud-radiative forcing and climate: Results from the Earth Radiation Budget Experiment. Science, 243, 57– 63. Slingo, A. and Slingo, J. M., 1991: Response of the National Center for Atmospheric Research community climate model to improvements in the representation of clouds. J. Geophys. Res. — Atmos., 96, 15341–15357. Stephens, G. L., Vane, D. G., Boain, R. J., Mace, G. G., Sassen, K., Wang, Z., Illingworth, A. J., O’Connor, E. J., Rossow, W. B., Durden, S. L., Miller, S. D., Austin, R. T., Benedetti, A., Mitrescu, C., and the CloudSat Science Team, 2002: The CloudSat mission and the A-train. Bull. Amer. Meteorol. Soc., 83(12), 1771–1790. Ulbrich, C. W. and Atlas, D., 1998: Rainfall microphysics and radar properties: Analysis methods for drop size spectra. J. Appl. Meteorol., 37(9), 912–913. Vali, G., Kelly, R. D., French, J., Haimov, S., and Leon, D., 1998: Finescale structure and microphysics of coastal stratus. J. Atmos. Sci., 55, 3540–3564. van de Hulst, H. C., 1957: Light scattering by small particles. Wiley and Sons. Wakasugi, K., Mizutani, A., Matsuo, M., Fukao, S., and Kato, S., 1986: A direct method for deriving dropsize distribution and vertical air velocities from VHF Doppler radar spectra. J. Atmos. Ocean. Technol., 3, 623–629. Wood, R., 2000: Parametrization of the effect of drizzle upon the droplet effective radius in stratocumulus clouds. Q. J. R. Meteorol. Soc., 126, 3309–3324. 16
© Copyright 2026 Paperzz