Airflow and Precipitation Properties within the Stratiform Region of Tropical Storm Gabrielle during Landfall

a. Analysis techniques

Four different methods were used for obtaining vertical air motion: 1) reflectivity–fall speed relationships, 2) extended velocity azimuth display (EVAD), 3) quasi-velocity azimuth display (Q-VAD), and 4) Sans Air Motion (SAM) model (discussed in section 2b). The EVAD, Q-VAD, and the divergence theorem were used to determine horizontal divergence. Vertical air motion (w) is estimated using the reflectivity–fall speed relations for rain (Ulbrich and Chilson 1994) and for snow (Atlas et al. 1973):

View Expanded

and

View Expanded

where Z_rain and Z_snow are radar reflectivity factors (mm⁶ m⁻³) for rain and snow, respectively; w is the vertical air motion; and W_dop is the mean Doppler velocity at vertical incidence measured by the wind profiler. Both w and W_dop are negative downward. The term (ρ₀/ρ)^0.4 is the atmospheric density correction (Beard 1985), where ρ is air density parameterized by ρ(z) = ρ₀ exp(−z/9.58) and z is altitude in kilometers. The two relationships were applied to pure snow and rain regions separated by the snow and rain levels defined at the maximum curvature in the Z profile. For the intermediate layer between the snow and rain levels, w was acquired by performing a linear interpolation between the values within the rain and overlying snow levels.

The EVAD provides profiles of w, terminal fall velocity, and horizontal divergence by processing data at horizontal rings in all elevation cones per layer (Matejka and Srivastava 1991). The EVAD performs a linear regression using the equation

View Expanded

where div is horizontal divergence, W_dop is Doppler velocity, α is the elevation angle, r is the radius of the VAD circle, and a_o is the zeroth harmonic coefficient as a function of the radius and elevation angle. Two unknowns, div (intercept) and W_dop (slope), per layer are determined by the linear regression between the x and y terms (i.e., y = aX + b) in (3). After a div profile is obtained from the regression, a w profile is produced by performing a variational integration of the mass continuity equation. One important assumption is that the wind field over the analysis domain is horizontally linear. Also divergence and terminal fall velocity for a given layer are considered as homogeneous over the maximum horizontal range of the rings.

The Q-VAD computes the wind gradient using two opposite beams of the 915-MHz profiler. The vertical air motion, assumed to be identical over all five beams at the same time, is derived from downward integration of the anelastic continuity equation starting with EVAD vertical air motion as the top boundary condition (w_top):

View Expanded

where div is the sum of Δu/Δx and Δυ/Δy, the east–west and north–south gradients of the winds divided by a horizontal distance between the off-vertical range gates of the two beams. Divergence estimates become increasingly noisy at low levels because the horizontal distance between the opposite beams decreases with decreasing height. Thus, div and w from the Q-VAD technique are not displayed below 2 km AGL.

The divergence theorem was utilized to estimate div using the highest elevation angle (44.2°) of the SMART-R. This technique has been used to diagnose diabatic div profiles in tropical mesoscale convective systems (MCSs) using aircraft radar measurements (Mapes and Houze 1993, 1995). This highest elevation angle was selected to most closely match the sampling volume of the 915-MHz profiler. The horizontal divergence over an area A is equal to the closed line integral of the normal component of Doppler radial velocity (V_r) shown by

View Expanded

where V_h is a horizontal component of V_r and dl is a length element along the circular boundary. The horizontal component (V_h) of Doppler radial velocity is calculated by

View Expanded

where V_f is the terminal fall speed expressed with positive values for upward motion to match the vertical air motion sign convention.

b. The SAM model—The nonzero vertical air motion

The RSDs and their integral parameters such as rain rate and liquid water content can be retrieved from the Doppler velocity power spectrum of wind profiler radars. In doing this, w, turbulence, noise level, and deviation of Rayleigh scattering need to be taken into account. In this study, the SAM model (Williams 2002) is used to retrieve the RSDs below the melting layer in the stratiform rainband. The retrieved RSDs are approximated by the gamma size distribution of the following form:

View Expanded

where N(D) is the number concentration of particles which diameters range between D and D + dD. Here N₀ (m⁻³ mm⁻¹), μ, and Λ (mm⁻¹) are the intercept, shape, and slope parameters, respectively, of the size distribution. The SAM model retrieves the hydrometeor spectrum by comparing the observed Doppler spectrum with many modeled spectra with the final retrieved spectrum minimizing a least squares difference cost function. The modeled Doppler velocity spectrum is formed by convolving the clear-air spectrum with the modeled hydrometeor spectrum. The shape of modeled hydrometeor spectrum is varied by changing the mean mass-weighted diameter (D_m) and μ, and the clear-air spectrum by changing the clear-air mean vertical air velocity (w) and spectral broadening (σ_air). The rainfall rate is calculated from the three parameters {N₀, D_m [or Λ using Λ = (4 + μ)/D_m], and μ} of the gamma RSD by using

View Expanded

To account for different vertical air motions, the SAM model uses a two-step process. The first step estimates the best hydrometeor spectrum (N₀, D_m, and μ) and spectral broadening (σ_air) for a fixed vertical air velocity (w) by minimizing the squared difference between the modeled and observed spectra. The first step is then repeated for many different vertical air motions. Values of w are acquired by shifting the observed spectrum along the velocity axis in one spectral bin increment (Δυ = 0.169 m s⁻¹) removing the effects of updrafts or downdrafts on the observed spectrum. As shifting the observed spectrum by a spectral bin increment, the best-fit hydrometeor spectrum with the final estimate of w, as a multiplication of the increment, is determined with the minimum cost function between the observed spectrum and the modeled spectrum. Shifting the observed spectrum to account for updrafts decreases rainfall rates, whereas shifting the observed spectrum to account for downdrafts increases rainfall rates (Rajopadhyaya et al. 1998).

a. Storm evolution and environment

Gabrielle was first classified as a tropical depression at 1800 UTC 11 September 2001 and attained a tropical storm status at 1200 UTC 13 September over the western Gulf of Mexico (Lawrence and Blake 2002; Molinari et al. 2006). Gabrielle then moved northeastward after 0000 UTC 14 September. A continuous intensification was observed for the next 12 h, particularly between 0600 and 1200 UTC, the time of landfall (Molinari et al. 2006). The MIPS, SMART-R, and TBW radar sampled an extensive stratiform rainband between 0430 and 0700 UTC. The storm circulation center was located southwest (upshear) of center of the stratiform rainband. During this period Gabrielle became more organized, as the deep convection (bottom part of Fig. 1) intensified near the circulation center. The arc-shaped stratiform rainband (Fig. 1) appeared at around 0400 UTC and persisted until 1100 UTC. This rainband exhibited an arc length of ∼250 km and a mean width of ∼60 km.

A pronounced brightband signature was observed by the 915-MHz profiler and the TBW radar between 0500 and 0650 UTC. In Fig. 2, vertical sections of reflectivity factor and radial velocity at 0601 and 0656 UTC from the TBW radar, located 70 km north of the MIPS, show an extensive bright band near 4 km AGL. The increase in magnitudes of radial velocity, and vertical gradients in radial velocity near and above 6 km AGL reveal large vertical wind shear within the ice region at 0656 UTC, consistent with the findings of Molinari et al. (2006). Fig. 2e depicts a cyclonic flow of about 15 m s⁻¹ magnitude along the rainband in the dual Doppler analysis at 3.5 km AGL.

Two soundings were taken at 0000 and 0600 UTC from Tampa Bay (KTBW) within the stratiform precipitation (Fig. 3). At 0000 UTC, the 650–950-hPa layer is nearly saturated. The 0600 UTC sounding released near the leading edge of the rainband shows a slightly subsaturated layer (RH = 85%) centered near 850 mb. This relatively unstable layer is between 770 and 850 hPa where mesoscale downdrafts are likely maintained by evaporational cooling. The extended saturated layer down to 770 hPa (2.2 km) at 0600 UTC could be ascribed to heat and moisture transport, supported by increased winds from the southeast direction (see the wind barbs on the right).

At 0600 UTC, it should be noted that an increase in temperature and wind speed occurred in the 500–800-hPa layer for the 6-h period. Cooling within and below the melting layer is probably counteracted by heat and moisture transport from the south and southeast at these levels. Comparison with the 0000 UTC sounding reveals that cooling at low levels was associated with the offshore easterly flow. Knupp et al. (2006) noted that an extensive region of cool air at the surface preceded landfall and that the surface cold air exerted large effects on the ABL, and apparently on storm intensity changes. This cooling and its association with rainfall evaporation, mesoscale downdrafts, and precipitation properties is a primary motivation for this study.

b. Overview of 915-MHz profiler measurements

In Fig. 4a, the 915-MHz profiler-derived reflectivity factor (Z) patterns show a prominent and continuous brightband signature whose height oscillated by ∼200 m from a mean height of 4.3 km. Rainfall rates up to 23 mm h⁻¹ and low-level Z maxima approaching 40 dBZ were measured within this rainband. Within the melting layer near 4.2 km AGL, Z values exceeded 50 dBZ. These observations indicate that this was a relatively intense stratiform rainband. In Fig. 4b, the larger W_dop regions reaching about 10 m s⁻¹ located particularly just below the stronger bright band, were well correlated with the regions of higher Z streaks around 0550, 0610, 0625, and 0635 UTC. A positive correlation between W_dop and Z and an increase in their magnitudes over several kilometers above the melting layer, particularly around 0550, 0625, and 0635 UTC, indicates a substantial increase in snow aggregation close to 0°C level.

In Fig. 4c, spectral widths show a notable contrast between the rain and ice regions except for thin layers of large spectral width near the bright band. The large spectral widths exceed about 4.5 m s⁻¹ and are centered near 4.7 km, which appears to be a consequence of wide distributions in fall speeds of snow aggregates. The narrow region of small spectral widths less than 2 m s⁻¹ near the level of maximum Z (4.0–4.5 km) between 0535 and 0645 UTC suggests much less breakup and more uniform distributions of fall speeds relative to that at the top of the melting layer. The w patterns and precipitation characteristics of this rainband are examined in greater detail in the following sections.

a. Melting layer and brightband signatures

In Fig. 7a, the contoured frequency by altitude diagram (CFAD; Yuter and Houze 1995) of almost constant Z in the rain layer is shown with a strong bright band above during the 0530–0648 UTC period. Particularly, the Z CFAD shows a slight increase below 2 km AGL. The almost constant Z profiles are often measured in tropical stratiform regions (Cifelli et al. 2000; Schumacher and Houze 2003; Tokay et al. 1999) as a result of the compensation between breakup of large drops and continuous collection of small drops, which maintain a constant Z (Huggel et al. 1996). Figures 7a,b show a narrow distribution in both Z and W_dop from the top of the melting layer (4.8 km) to about 5.5 km AGL, suggesting that aggregation, accretion, and deposition processes at these levels compose equilibrium in snow size spectra. The CFAD of W_dop shows relatively high frequencies centered near 2 m s⁻¹ over a 1-km depth above the melting layer, implying that the mean fall speeds are independent of particle size and shape. However, as snow aggregates start melting, their size and fall velocity distributions become wider because the probability of collision rapidly increases at the same time. It is noted that an increase in the vertical gradients of W_dop is smaller in the upper part of melting layer than in the lower part because of a slower increase in melted fraction of snow in the upper part.

Spectral widths for snowflakes are narrower than those for raindrops since the range of fall speeds of snowflakes is smaller. But this range of fall speeds rapidly broadens as snow particles aggregate near the 0°C level where collision efficiency is relatively high. Large spectral widths near the top of the melting layer (Fig. 7c) are attributed to broader distributions of fall speeds of melting snow aggregates and perhaps to small-scale turbulence triggered by melting-driven cooling. However, an uncertainty still exists in the spectral width values obtained from the 915-MHz profiler observations as the spectra above the melting layer are more asymmetric with smaller peak values than those below the melting layer. One or both edges of the spectrum may be too steep, the spectral peak may be off-center, or the number of spectral peaks may be greater than 1 (Janssen and Van der Spek 1985). The increase in vertical wind shear (not shown) may have played a role in very asymmetric shapes (including some noise) of spectra near the top of the melting layer.

The CFADs for spectral widths and wind speeds (Figs. 7c,d) show a slow increase from 5 km AGL to near the 0°C level. The heights of high spectral widths and large horizontal winds are almost identical near the top of the melting layer, suggesting that the increase in spectral widths is partially due to increased horizontal wind speed and wind shear.

The smallest spectral width values near the level of maximum Z (i.e., 4.4 km in Fig. 7a) suggest a minimum in breakup and a maximum in collection efficiency. Right below the levels of the Z maxima, the spectral widths rapidly increase, indicating that collisional breakup becomes more significant since the range of particle fall speeds in the lower half of the melting layer is, on average, larger due to partly or completely melted particles than in the upper half (Barthazy et al. 1998). Higher collision efficiency due to a larger range of fall speeds may produce more efficient breakup in the lower half of the melting layer. The second peak in the CFAD for spectral widths near 3.8 km may be attributed to increased instability associated with cooling (i.e., increased lapse rate of air temperature) as well as mixed distributions of raindrops and partially melted particles near the bottom of the layer.

b. Melting layer parameters

Melting layer microphysics is quantitatively analyzed by using various melting layer parameters defined at the top and bottom of the melting layer. The top of the melting layer is defined as a height of the maximum curvature in the Z profile (Drummond et al. 1996). The snow reflectivity expressed in equivalent reflectivity factor (Z_esnow) is an averaged Z value (in order to reduce the noise) over two successive range gates above the top of the melting layer (Fig. 8). The bottom of the melting layer is also determined at the height of the maximum curvature in the Z profile. The rain reflectivity also expressed in equivalent reflectivity factor (Z_erain) is taken as an averaged value over three successive range gates below the bottom of the melting layer. The basic eight parameters, H_top, H_bottom, H_peak, Z_esnow, Z_epeak, Z_erain, W_snow, and W_rain, and their extended parameters, γ, ΔZ = Z_epeak − Z_erain, and ΔH = H_top − H_bottom, are estimated during the 0530–0648 UTC period when the Z profiles had distinct curvatures. The parameter γ is defined as the ratio of the products of Z and W_dop at the top (H_top) and bottom (H_bottom) of the melting layer (Drummond et al. 1996) by

View Expanded

where Z_esnow and Z_erain are the equivalent reflectivity factor (mm⁶ m⁻³) and W_snow and W_rain are the mean Doppler velocity of snow and rain at the top and bottom of the melting layer, respectively, which are shown in Fig. 9a. We assume a 1:1 relationship of the snowflake mass entering the melting layer and the raindrop mass leaving it, implying an absence of aggregation and breakup in the melting layer. Therefore, the deviation of the ratio of the two masses from unity indicates aggregation or breakup in the melting layer.

Aggregation- and breakup-dominant periods are determined by using smaller γ values and larger γ values, respectively. Drummond et al. (1996) suggested a threshold of γ = 0.23 to separate aggregation and breakup-dominate periods. Additional reflectivity constraints of Z_epeak > 49 dBZ and Z_erain > 38.3 dBZ are used for reliably determining aggregation-dominant periods since Z_epeak and Z_erain tend to be proportional to the degree of aggregation. In this study, the period (0600–0608 UTC, a time when H_peak descends as shown in Fig. 9a) is identified in order to examine the relationship of the descending bright band with the microphysics below the melting layer. In Fig. 9b, breakup is suggested over a limited time period between 0612 and 0618 UTC when γ values are largest, and aggregation is suggested for the other times. It is shown that the dominance of breakup or aggregation is solely determined by the Z ratio at the two levels. For breakup-dominant periods, the Z ratio is relatively higher due to smaller Z_erain than for aggregation-dominant periods when smaller γ values are due to higher Z_erain. It is found that the Doppler velocity ratio is inversely proportional to the Z ratio during aggregation-dominant periods, indicating that snow aggregates are probably denser, more rimed, and even larger in size.

c. Quantitative analysis

In Fig. 10a, the scatterplot of Z_epeak with brightband height (H_peak) and thickness (ΔH) shows that ΔH is proportional to Z_epeak (Fabry and Zawadzki 1995; Klaassen 1988), but H_peak is inversely proportional to Z_epeak. During breakup-dominant periods, ΔH is especially small. The inverse relationship is probably due to melting-driven cooling because the downward propagation of cooled air would lower the brightband height as indicated in Durden et al. (1997). Fig. 10a shows that the average ΔH for aggregation-dominant periods is about 30% larger than that for breakup due to melting of large snow particles at a long fall distance.

Figure 10b shows a slightly positive correlation between ΔZ and Z_epeak, and a negative correlation between ΔZ and Z_erain. Both Z_epeak and Z_erain are larger and less scattered for aggregation-dominant periods than those for breakup. Since all ΔZ values (9.5–15 dB) in Fig. 10b are greater than 7 dB, the threshold that Huggel et al. (1996) used following Fabry and Zawadzki (1995) to discriminate weak and well-defined bright bands, the bright band within this TC is classified as strong and well-defined throughout the 80-min analysis period. For breakup-dominant periods, the average ΔZ value is 2 dB larger than that for aggregation periods. This is due to a large reduction of Z_erain at the bottom of the melting layer, which contributes more to an increase in ΔZ, indicating active breakup in the lower half of the melting layer during breakup-dominant periods. In this study, ΔZ for breakup-dominant periods is larger (ranging from 11 to 15 dB) than aggregation-dominant periods (ranging from 10 to 13 dB). However, the large ΔZ does not necessarily indicate the dominance of breakup because both periods have ΔZ, which includes the range of 11–13 dB. Thus, the parameter ΔZ alone cannot be used for diagnosing the aggregation or breakup dominance, especially for the strong brightband case. On the whole, the average ΔZ and ΔH during the descending brightband period are between those during aggregation- and breakup-dominant periods, implying that the microphysical properties would be between the two distinguishable periods.

In Fig. 11a, the rain rate (R) retrieved from the SAM shows positive correlations with both the Z_epeak and ΔH, particularly during aggregation-dominant periods. That is, R is proportional to brightband intensity. Since brightband intensity is highly dependent on the characteristics of the largest particles (Stewart et al. 1984), this is attributed to stronger precipitation produced by melting of larger snow aggregates over a longer fall distance (Fabry and Zawadzki 1995). In Fig. 11b, overall R shows little dependence on H_peak, indicating that for a given H_peak, there can be a wide range of R. Thus, a higher H_peak does not always imply a higher R (or larger ΔH). However, it is shown that the larger R during aggregation-dominant periods corresponded to lower H_peak, relative to other R–H_peak pairs. A primary cause for this is melting-driven cooling, which plays a role in lowering H_peak (Durden et al. 1997) and producing negative buoyancy for downdrafts at rather small temporal and spatial scales. It seems that the large-scale variation in H_peak is probably more associated with other physical factors like temperature change, vertical air motion, and so on, besides melting-driven cooling.

d. Raindrop size distributions

Figure 12 shows the time–height sections of Z, R, D_m, and N₀. As proportional to Z, it is evident that R is higher during aggregation-dominant periods when the bright band was strong. Overall, D_m tends to be inversely proportional to R above 2 km AGL, and it is proportional to R below 2 km, particularly during aggregation-dominant periods. The R is positively correlated with N₀, indicating that the increase in R is related to narrow spectra having a large number of small raindrops, thereby increasing the RSD slope and decreasing D_m. The notable increase in N₀ particularly below 2 km AGL is attributed to raindrop breakup. (Evaporation would be small, given the nearly saturated sounding.) During aggregation-dominant periods, Z profiles are nearly uniform or decrease with height, indicating that collision–coalescence produces moderate to large raindrops and balances or even dominates over breakup (around 0540 and 0640 UTC). It appears that breakup is more dominant than collision–coalescence when R increases as D_m decreases, but collision–coalescence is more dominant than breakup when R increases as D_m also increases, intermittently shown below 2 km AGL. During the descending brightband period, both Z and R are smaller than those during aggregation-dominant periods. During this period, the slight increase in Z with decreasing height, along with a slightly larger D_m and smaller N₀ indicates a relative increase in the number of moderate to large drops. Relatively stronger mesoscale downdrafts were measured below 2 km AGL during this period. Evaporation may decrease the number of smaller drops the most and collection of large and midsize drops may increase. Both the sporadic increase in R and D_m below 2 km AGL probably indicates the possibility of raindrop growth within the mesoscale downdrafts. On the other hand, during breakup-dominant periods, the smallest Z and D_m are shown among the sampling periods.

In Fig. 13, the correlation between R from the SAM and Z–R relation (Z = 367R^1.3; Tokay and Short 1996) shows better agreement when vertical air motions are included. The increase in R for a given Z in Fig. 13a is due to the fact that corrected downdrafts shift the spectrum to smaller fall speed and smaller drop diameter range. In Fig. 13a, the R derived from the Z–R relation exceeds the SAM-retrieved R values. We interpret this as raindrops in this case are, on average, larger than those for a given Z. The inference of larger than average raindrops is consistent with the stronger bright band. In Fig. 13b, the overall underestimation in SAM-retrieved R is due to the absence of downward air motions, which were estimated below the melting layer.

Figure 14 shows the composite RSDs retrieved from the SAM model at 3472 m AGL during three periods. We chose this height closest to the bottom of the melting layer in order to minimize uncertainties or errors by horizontal advection of particles. Each RSD is an average of the five successive RSDs. The composite RSD during aggregation-dominant periods shows larger number concentrations at almost all drop sizes, corresponding to high reflectivity streaks shown in Fig. 12a. During breakup-dominant periods, the RSD shows a larger number of small drops and smaller number of large drops. This may be evidence of small raindrops resulting from melting of smaller snow particles produced by more active breakup in the melting layer. The RSD during the descending bright band shows smaller number concentrations at the small to midsize drop range when compared with those during the two other periods. The D_m is a little larger during this period than during aggregation-dominant periods. The contributions of the number concentration of large drops on Z and D_m are relatively larger than at least those during breakup-dominant periods when D_m was smallest.

Since these composite RSDs were obtained near the bottom of the melting layer, the properties of snow size distributions near the top may be deduced from the RSDs. For aggregation-dominant periods, as expected, large snow aggregates were present just above the melting layer as aggregation dominates breakup within the melting layer, increasing the brightband intensity. Small and dense rimed particles may be present as well in terms of large number concentration of small drops in Fig. 14 and large Doppler velocity ratio in Fig. 9b. During breakup-dominant periods, snow aggregates are expected to be relatively smaller and less dense. The smallest Z_erain suggests substantial breakup in the lower part of the melting layer during these periods. During the descending brightband period, presumably less efficient breakup within the melting layer and larger D_m indicates that on average, large and dense hydrometeors are probably present in terms of small γ values in Fig. 9b. However, their sizes seem to be smaller than those during aggregation-dominant periods since their Z_erain values are about 3 dB less.

a. Kinematics

Similar to this study, mesoscale up- and downdrafts above and below the melting layer in stratiform regions of hurricanes have been observed by Marks and Houze (1987) and Black et al. (1996). The mean range of w₉₁₅ (from −1 to 1.5 m s⁻¹) in this study is larger than the mean w range from −0.3 to 1.3 m s⁻¹ from Marks and Houze (1987) and the mean w range from −0.5 to 1.0 m s⁻¹ in Fig. 8 of Black et al. (1996) in stratiform regions. Particularly, the mean downdraft velocity of −1 m s⁻¹ at low levels in this study is relatively large. It is likely that the strong downdrafts below the melting layer are overestimated by an underestimation of fall speeds of hydrometeors.

A convergence–divergence couplet within the melting layer has been documented in a limited number of papers (Cifelli et al. 1996; Mapes and Houze 1993, 1995; Steiner et al. 2003; Szeto et al. 1988; Szeto and Cho 1994). However, finding the convergence–divergence couplet within such an intense stratiform rainband of TC is unprecedented to our knowledge. Szeto and Stewart (1997) indicated that melting-driven cooling can produce mesoscale circulations that enhance horizontal convergence around the melting layer. In this study, the convergence–divergence couplets both in the div profiles from both the Q-VAD and divergence theorem are probably indicative of cooling-induced circulations. (Although the couplet is not as obvious in the divergence theorem profile, the trend in strong convergence to weak convergence near the 4-km height is similar to that indicated in the Q-VAD profile.) The convergence above and divergence below the levels of Z maxima within the melting layer implies indirect finescale circulations of airflow as a coupling of the upper and lower half of the melting layer.

b. Melting layer studies

Huggel et al. (1996) used a ΔZ threshold of 7 dB to determine weak or well-defined bright bands and showed a prominent negative correlation between ΔZ and N₀ particularly for the R range of 1–10 mm h⁻¹. However, in this study, all ΔZ values (9.5–15 dB) were greater than 7 dB and an overlapping range of ΔZ was found for both the aggregation- and breakup-dominant periods. The average ΔZ for breakup-dominant periods was 2 dB greater than that for aggregation-dominant periods. But the average Z_epeak for aggregation was rather higher than that for breakup (ΔZ = Z_epeak − Z_erain), suggesting that there are larger numbers of large particles (i.e., larger D_m, see Fig. 14) during aggregation-dominant periods. Therefore, one should exercise caution in interpreting the inverse relationship between ΔZ and N₀ (or Λ), proposed by Huggel et al. (1996), if the bright band is strong with moderate precipitation. This is because a large ΔZ does not always indicate a large number of large particles (i.e., small N₀) as we found that the average ΔZ was larger during breakup-dominant periods when D_m is comparatively smaller.

c. Precipitation properties regarding brightband intensity

During aggregation-dominant periods, the increase in number concentrations at all drop sizes was shown in Fig. 14. Such wide distributions have dependence on snow properties (size and habit) near the top of the melting layer. Yuter et al. (2006) indicated that the standard deviation of fall speeds of wet (i.e., partially melted) snow aggregates is larger than that of dry snow aggregates. The collision efficiency for aggregation-dominant periods would be higher because of the larger standard deviations of fall speeds of wet snow aggregates. A higher R during aggregation-dominant periods (Fig. 12b) as proportional to Z implies higher collection efficiency since R increases as the number concentration of large drops increases (Hu 1995; Low and List 1982). Considering that the largest factor to the decrease in Z below the melting layer is raindrop breakup (Stewart et al. 1984), a slow increase in Z and D_m with decreasing height suggests that moderate to large raindrops are still produced while breakup proceeds, even within the mesoscale downdrafts that were present below the melting layer throughout the analysis period. The almost constant or slightly increasing Z CFAD (Fig. 7a) indicates the raindrop growth by collection (particularly below 2 km AGL), suggesting relatively less breakup and evaporation in this particular stratiform rainband. During the descending brightband period (i.e., strong downdraft period), the collection efficiency is probably not as large as that during aggregation-dominant period because of smaller Z_erain and smaller numbers of small and large drops. However, relative to aggregation-dominant periods, evaporation associated with the strong downdrafts may come to play a more important role in decreasing R near the surface during the descending brightband period.

It is found that the effects of mesoscale downdrafts cannot be neglected for interpreting all dynamics and microphysics in and below the melting layer, particularly for intense stratiform rainbands of TCs as in this case study. Therefore, the accurate estimation of w is more significantly required for this case. Even though the w variation would be larger with shorter periods (about 80 min) and the Z–V_f relations for snow and rain might not work well just below and above the melting layer, w from the profiler and the SAM showed good agreement. The RSD retrievals from the nonzero w assumption gave rise to less errors or biases than from the zero w assumption.

Errors in determining the top and bottom heights and the parameter values (e.g., γ) based on these levels can lead to false interpretation. Since the determination of snow and rain levels using the reflectivity curvatures can never be perfect, a secondary constraint such as the Doppler velocity would be auxiliary to make the levels be more accurate. Also, one should keep in mind that an increase of breakup in this study is a relative increase over aggregation within the melting layer, even during breakup-dominant periods.