This study will validate the S-band dual-polarization Doppler radar (S-Pol) radar refractivity retrieval using measurements from the International H2O Project conducted in the southern Great Plains in May–June 2002. The range of refractivity measurements during this project extended out to 40–60 km from the radar. Comparisons between the radar refractivity field and fixed and mobile mesonet refractivity values within the S-Pol refractivity domain show a strong correlation. Comparisons between the radar refractivity field and low-flying aircraft also show high correlations. Thus, the radar refractivity retrieval provides a good representation of low-level atmospheric refractivity. Numerous instruments that profile the temperature and moisture are also compared with the refractivity field. Radiosonde measurements, Atmospheric Emitted Radiance Interferometers, and a vertical-pointing Raman lidar show good agreement, especially at low levels. Under most daytime summertime conditions, radar refractivity measurements are representative of an ∼250-m-deep layer. Analyses are also performed on the utility of refractivity for short-term forecasting applications. It is found that the refractivity field may detect low-level boundaries prior to the more traditional radar reflectivity and Doppler velocity fields showing their existence. Data from two days on which convection initiated within S-Pol refractivity range suggest that the refractivity field may exhibit some potential utility in forecasting convection initiation. This study suggests that unprecedented advances in mapping near-surface water vapor and subsequent improvements in predicting convective storms could result from implementing the radar refractivity retrieval on the national network of operational radars.
Accurate, horizontally distributed, high-resolution moisture measurements are critical, yet largely unavailable, for many atmospheric science applications (e.g., Weckwerth et al. 1999). Such applications include boundary layer studies, convection initiation and evolution, nowcasting, and quantitative precipitation forecasting (QPF). In fact, several national study groups (National Research Council 1998; Emanuel et al. 1995; Dabberdt and Schlatter 1996) have suggested that the primary factor limiting the prediction of convective precipitation is measurement uncertainty in the high-resolution distribution of water vapor. Radiosondes, the traditional means of obtaining water vapor measurements, are insufficient because they provide vertical profile information at widely distributed locations, are typically only available 2 times per day, and have significant errors and biases (e.g., Soden and Lanzante 1996; Guichard et al. 2000; Wang et al. 2002; Revercomb et al. 2003; Turner et al. 2003). Additionally, there is a general absence of operational ground-based water vapor remote sensing systems, and many satellite techniques cannot obtain high-vertical-resolution water vapor measurements with high accuracy in the lower troposphere. This water vapor measurement uncertainty was a primary justification for the International H2O Project (IHOP_2002; Weckwerth et al. 2004) and for the effort described here to test the radar refractivity technique to map the low-level distribution of water vapor.
Radar refractivity retrieval algorithms show great promise in estimating the surface layer moisture field (Fabry et al. 1997; Fabry 2004). The refractivity signal is derived from an apparent fluctuation in the range of fixed ground targets from the radar. The slow variations in target ranges are caused by differences in the time taken for the radar waves to travel to the target and back. The velocities of the radar waves are modified by the index of refraction in the intervening atmosphere. This results in a change in the phase of the fixed targets that can be measured by radar:
where ϕ is the phase of a given target at range r, f is the radar frequency, and c is the speed of light in vacuum. The phase delay difference between two times t0 and t1 is due to the path-integrated atmospheric refractivity (N; dimensionless N units) variations that are caused by changes in pressure (P; hPa), temperature (T; K) and water vapor pressure (e; hPa), following Bean and Dutton (1966):
A reference time period t0 with homogeneous and constant refractivity conditions across the refractivity domain is first identified to provide a base field of phase differences between targets. Then, if numerous targets are aligned, the many overlapping paths of integrated N can be decomposed into shorter paths that can be as short as a range gate in regions where ground targets are numerous. This process is repeated for all azimuths to build a field of refractivity. The quality and the resolution of the refractivity field are, hence, strongly dependent on the density of fixed ground targets in the area considered. Each pixel of the S-band dual-polarization Doppler radar (S-Pol) refractivity retrieval is smoothed using a 4-km pyramidal-like smoothing function, which places the greatest weight on targets nearest to the pixel location. If there are numerous ground targets in the area, this smoothing function allows for the possibility of a refractivity resolution better than 4 km. For refractivity field examples, please see Fig. 1.
Typical refractivity values during IHOP_2002 were 270–300 N units. At a temperature of 18°C a change in 1 N unit can be caused by either a temperature change of 1°C or a much smaller dewpoint temperature change of 0.2°C (e = 0.2 hPa). Thus, in the summer season with warm temperatures, variations in refractivity are primarily caused by changes in water vapor content, and the refractivity field may be used as a proxy for the low-level moisture field. This will also be illustrated with the data in this paper.
During IHOP_2002 a field of refractivity measurements was routinely obtained from the National Center for Atmospheric Research (NCAR) S-Pol (Lutz et al. 1995) in the Oklahoma Panhandle. The nominal retrieval range was approximately 40 km, but it extended 60 km toward the northwest because of the more numerous ground targets and more optimal slope of the land. Useful ground targets include power and roadside poles, buildings, and towers. Note the clear appearance of a 1-mi (1.62 km) grid pattern, identifying the poles along roads in the ground clutter map (Fig. 2). Because these structures are typically less than 20 m high, the refractivity field is a measurement within the surface layer, but the actual depth that the radar refractivity represents will be investigated.
The objectives of this study are to (i) assess the accuracy of the radar refractivity retrieval, (ii) evaluate the horizontal and vertical representativeness of the radar refractivity retrieval, and (iii) assess its potential for forecasting applications. Section 2 describes the intercomparisons with fixed and mobile surface stations, aircraft in situ measurements, Atmospheric Emitted Radiance Interferometer (AERI) profiles, soundings, and scanning Raman lidar (SRL) measurements. Section 3 discusses some of the potential nowcasting applications of the real-time refractivity retrieval. A summary is given in section 4.
The radar refractivity retrieval was compared with several different IHOP_2002 datasets to assess the accuracy of the retrieval algorithm and determine the horizontal and vertical scales over which the retrieval is representative. These datasets include fixed and mobile surface stations, aircraft in situ measurements, mobile AERI, soundings, and SRL measurements.
Nine surface stations were within the coverage of the radar refractivity measurements and were used for their validation. Figure 3 shows the comparison of refractivity measurements by radar with the values computed from the Integrated Sounding System’s (ISS) surface station data at the Homestead profiling site 16 km east of S-Pol. Although by calibration the measurements were set to match each other on the afternoon of 14 May, there appears to be a slight time-dependent difference between the two datasets (standard deviation of 3 N units, or about 0.5°C in dewpoint temperature) despite their excellent correlation. Identical results were found with the data from the other surface stations.
Possible causes for biases between the radar and surface stations were explored. Errors may occur locally if the refractivity retrieval performs poorly because of unsuitable ground targets (swaying vegetation, central pivot irrigation systems, and other nonstationary targets). Errors may also occur over the whole domain if the vertical profile of refractivity changes significantly (Fabry 2004), which is the case at night (Gossard et al. 1999). Ideally the ground targets should be uniformly distributed—both radially and azimuthally—and at the same level. If the latter condition is met, each target pair will provide a perfect estimate of N. Otherwise, each estimate will be noisy, and the estimates from many pairs must be combined to obtain a more accurate measurement of N. Larger measurement errors will occur in regions scarce in ground targets, when these ground targets are at very different heights, and when propagation conditions differ greatly from those at the calibration time t0. Increased smoothing of the refractivity fields (to reduce local biases) and stratification of the data by the time of day showed that while these sources of biases were present, they only accounted for a small fraction of the observed difference between the radar and station measurements of refractivity.
The dependence of the error with meteorological parameters was then investigated. The bias between radar and a surface station showed a strong relationship with relative humidity (Fig. 4), which was similar for all of the stations. Radar refractivity comparisons with radiosonde measurements from the same site at the lowest available level, of which half were below 2 m AGL and the other half were between 2 and 10 m AGL, showed no such bias. It is possible that there is a height dependence to the bias with the surface station measurements at 2 m and the radar refractivity measurements from 10–20-m-high targets. This may not have been apparent in the sounding comparison because of the use of some 10-m data points.
To examine this potential height dependence further, Monin–Obukhov similarity theory was used to simulate expected temperature and specific humidity profiles, and, therefore, refractivity profiles (e.g., Kaimal and Finnigan 1994). The three Integrated Surface Flux Facilities (ISFFs) near S-Pol were used because flux measurements are required for the calculations. Only data between 1800 and 1830 UTC on clear-sky days were used because Monin–Obukhov similarity theory works best for convective conditions. The average refractivity difference between the surface and 10 m and between the surface and 20 m were used as proxies for the radar refractivity estimates for those two layers.
The simulated relationship between the refractivity difference and relative humidity (Fig. 5a) corresponds well to the observations in Fig. 4. Both figures show that the difference is greater at larger relative humidities. Latent heat flux is also strongly related to the vertical refractivity difference (Fig. 5b). These relationships result primarily from three factors. First, the specific humidity near the surface decreases rapidly with height at a rate that is directly proportional to the surface moisture flux (which is, in turn, proportional to the latent heat flux). This results in a decrease in the refractive index with height. Second, for the conditions of IHOP_2002, higher moisture fluxes were correlated with higher relative humidities. Third, the refractivity difference is affected by the temperature decrease with height, which is proportional to the sensible heat flux and, thus, is usually reduced at higher latent heat fluxes (and relative humidities). Note that the refractivity difference is larger and has a greater slope for the greater height separation between the measurements. Thus, it appears that the difference between the station and radar estimates of refractivity results from the difference in altitude between the station (2 m) and radar (between 0 and 10–20 m) measurements.
To investigate the correspondence between radar and surface station data at shorter time scales, data from a couple of IHOP_2002 days are examined in more detail. On 22 May 2002 there was interesting temporal and spatial variability in the refractivity field. Figure 1 illustrates these refractivity variations. In the morning hours, at 1503 UTC [1003 central daylight time (CDT)] the refractivity field and surface station measurements are fairly uniform. By 2031 UTC there is substantial variation across the domain with a nearly north–south refractivity gradient forming right through the radar (Fig. 1b). This is likely depicting the dryline because the surface station moisture gradient is also pronounced across the refractivity gradient. By 2257 UTC (Fig. 1c), the refractivity field varies by ∼40 N units across the region sampled. This corresponds to a mixing ratio gradient of about 6.6 g kg−1 across the domain. The north–south dryline of Fig. 1b has moved eastward and another gradient (a secondary dryline) has formed, running northeast–southwest through the radar. This double-dryline structure is also apparent in the surface dewpoint temperature values, showing three distinct air masses. In Fig. 1d, approximately 1 h later, two well-defined gradients are apparent in the refractivity field—one dryline running north–south ∼10 km east of S-Pol, and a secondary dryline running northeast–southwest ∼15 km west of S-Pol.
The radar refractivity comparisons with several surface stations (locations shown in Fig. 1a) show excellent agreement, both in magnitude and gradient, on this day that exhibited large variations. The refractivity data from S-Pol show high correlation (r = 0.91–0.99) with the refractivity computed from the individual surface stations (Fig. 6). Note the excellent correspondence even at times of strong gradients in the field, but, just as in Fig. 3 the surface station refractivities typically show a slight positive bias. Figure 7 shows refractivity comparisons on 14 June 2002 (refractivity images not shown), which had less spatial and temporal variability than those on 22 May 2002, but also exhibited very high correlations (r = 0.69–0.99). On this more humid day, especially before 1400 UTC (0900 CDT), the surface refractivity values show larger differences with the radar refractivity values than were seen on 22 May 2002.
Beyond the issue of biases, these fixed surface station data are also useful to confirm the result of Fabry and Creese (1999) that, in summer conditions, refractivity is most correlated with mixing ratio. Figures 6 and 7 also show the temperature and mixing ratio traces for 22 May and 14 June 2002, respectively. Apparent in the figures is an inverse relationship between the S-Pol refractivity and temperature, and a more pronounced direct relationship between the S-Pol refractivity and mixing ratio. Even small variations in moisture directly affect the radar refractivity values. It takes larger temperature variations to have a similar impact on radar refractivity. This is a more empirical illustration of the strong relationship between radar refractivity and surface moisture as was discussed theoretically in the introduction.
Radar refractivity comparisons with the mobile mesonets (MM; Straka et al. 1996) are useful to assess the representative horizontal resolution of the radar refractivity field. Correlations between S-Pol refractivity and refractivity calculated from the MMs are high even though the spatial resolution is significantly different. S-Pol refractivity values obtained from the time within the range of the MM times are used. Thus, within one MM leg, multiple refractivity times may be utilized. Mobile mesonets record data every 1 s, which corresponds to ∼20 m resolution, assuming an average speed of 45 mi h−1 (20 m s−1). Recall that if there are numerous ground targets in the area, the pyramidal smoothing function allows for the possibility of refractivity resolution better than 4 km.
Figures 1b and 1d show refractivity maps with overlays of MMs and aircraft tracks (to be discussed in the following section). The gradients in S-Pol refractivity are often significantly smoothed as compared with the corresponding gradients in MM refractivity (Fig. 8). For example, in Figs. 8a and 8b, the S-Pol refractivity gradient is smoothed out over a 4-km distance, while the corresponding MM5 and MM1 both detect a sharp gradient (11 N units in less than 0.5 km). Later in the day, two MM tracks are compared with S-Pol. MM9 covers a uniform refractivity region; thus, it was deemed inappropriate to calculate a correlation coefficient for this intercomparison (Fig. 8c). Both MM9 and S-Pol show uniform values for most of the track, while MM9 indicates a more pronounced and rapid gradient at the eastern end of the track. In Fig. 8d, S-Pol’s refractivity does not show structures smaller than 2-km scale, while MM1 again detects a gradient tighter than 0.5 km. Because the spatial resolution of radar refractivity is somewhat dependent upon the number and spacing of ground targets, it is probable that some of the variability of horizontal resolution is location dependent; that is, resolution is better at a location surrounded by more ground targets than at a location surrounded by fewer ground targets. Another important factor is the smoothing of the phase data to remove noise in the refractivity retrieval. Because this value is set to 4 km, it is likely that features with scales of less than 4 km cannot be resolved well.
Aircraft in situ
The aircraft in situ measurements provide another means of comparing the radar refractivity field with high-resolution spatial data. Similar to the mobile mesonet observations, the aircraft measurements are obtained at a higher resolution (1 Hz; 90–120 m) than the radar refractivity retrieval (4 km). Thus, they are also used to illustrate the difference in scales between the in situ measurements and refractivity retrieval. Additionally, they are used to address the vertical depth over which the refractivity retrieval is representative.
Aircraft radar refractivity intercomparisons along the tracks plotted in Figs. 1b and 1d are shown in Fig. 9. When variations in refractivity occur, the radar refractivity identifies the overall trend of the refractivity gradient, but the aircraft-measured refractivity gradient is much sharper than the radar refractivity gradient, particularly at low levels. Figure 9c illustrates the high correlation (r = 0.89) between the University of Wyoming King Air (UWKA) refractivity measurements at 168 m AGL or 5% of the convective boundary layer (CBL) depth (i.e., 0.05zi) and the radar refractivity retrieval. While the UWKA samples an abrupt gradient in refractivity in less than 1 km, the radar retrieval technique observes the gradient smoothed over ∼10 km. At higher levels within a well-mixed CBL (Fig. 9a) the correlation between the P-3 and radar refractivity is still quite high (0.98). Note that the P-3 track (1225 m MSL; 363 m AGL; 0.1zi) illustrated in Fig. 1d is above a uniform refractivity region, and this is also shown in the radar refractivity trace (Fig. 9b); thus, the correlation coefficient was not calculated.
The aircraft data are also used to examine the vertical depth over which the refractivity retrieval is relevant. Although the radar beam is seeing only 10–20-m-tall towers and poles as ground targets, the refractivity retrieval may be representative of a deeper layer in some circumstances. Unfortunately, stacked flight legs are available primarily during times when the CBL is well mixed. For example, the UWKA flew stacked flight legs on 22 May 2002, flying approximately the same track at different heights (not shown). One set of stacked tracks within the CBL (164, 358, 673, and 1475 m AGL, or 0.05zi, 0.1zi, 0.2zi, and 0.4zi) from 2158 to 2246 UTC had good agreement between the aircraft legs and radar refractivity (correlations of 0.94, 0.95, 0.91, and 0.94, respectively). The second stacked track from 2322 to 2352 UTC, with legs at 167, 837, and 1494 m AGL (0.05zi, 0.2zi, and 0.4zi, respectively), also showed high correlations between radar and aircraft measurements, even with increasing height (0.89, 0.95, and 0.88, respectively). Thus, in a well-mixed CBL, S-Pol refractivity is representative of at least one-half of the depth of the CBL.
Profiling instruments are more readily used to assess the vertical depth over which the refractivity measurements are representative. The AERIBAGO, an AERI (Feltz et al. 2003) deployed within a recreational vehicle, was located 16 km east of S-Pol at the Homestead profiling site. AERIs measure downwelling infrared radiances and use them to retrieve temperature and moisture profiles, thus, allowing for the calculation of refractivity profiles. Figure 10 shows the trace of radar refractivity at the AERI location along with refractivity calculated from AERI profiles at average heights of 44 and 446 m AGL for 1 week of the project (7–13 June 2002). Similar results were observed throughout the entire project, but only 1 week is shown for enhanced clarity in the figure. While the magnitude of the refractivity should decrease with increasing height because of the decrease in pressure, the correlation should not be adversely affected due to the altitude variation alone. The different vertical levels of the AERIBAGO data show a difference in the pattern above ∼250 m AGL with the 446 m AGL AERI refractivities clearly not depicting the variations as well. The layer between 220 and 355 m seems to be a transition from high correlation values to lower ones (0.89 at 220 m, 0.87 at 265 m, 0.83 at 310 m, and 0.80 at 355 m AGL).
Figure 11 shows the radar–AERI refractivity composites at various low levels to assess the diurnal and height-related refractivity features. This figure illustrates that the refractivity bias between radar and various AERI levels is strongest at night and steadily improves throughout the day as the CBL becomes well mixed. It also shows that the strongest relationship between the radar and AERI occurs at low levels. Even during the night the 44 m AGL AERI measurements correspond well with S-Pol refractivity. A couple of hours after sunrise the low-level AERI line exhibits a somewhat different, but still strong, relationship with radar refractivity. This strong relationship with uniform bias commences at 1300 UTC (0800 CDT) and continues into the night until 0400 UTC (2300 CDT). At 88 m AGL the separation between the AERI and radar refractivity lines is smallest during the daytime, but it begins later (1900 UTC) and continues only until 2300 UTC when the trend between the two lines somewhat diverges. At 132 m AGL, the strongest relationship occurs at 1900–2200 UTC. Above 132 m AGL the time of strong correspondence also only occurs for 3 h—from 1900 to 2200 UTC. Only the lowest level (i.e., 44 m AGL) shows a strong relationship past 2300 UTC (1800 CDT). The same results were obtained when the AERI measurements were pressure corrected to ground level.
In summary, at low levels there is a strong daytime relationship between radar refractivity and AERI refractivity measurements from 1 h after sunrise until 2 h after sunset. As the height increases, the strong relationship begins later in the morning as the CBL becomes well mixed, and drops off by 2300 UTC as the CBL begins to decay.
Radiosondes were also compared with S-Pol radar refractivity. Serial ascents of radiosondes from the same locations within the refractivity range were obtained on 14 June 2002. Figure 12 shows comparisons of the S-Pol refractivity and radiosonde refractivity from the ISS location at several altitudes. Similar to the AERI results, the soundings also suggest that the refractivity is representative of the lowest 200–250 m of the atmosphere.
The SRL (Whiteman and Melfi 1999) obtained vertical profiles of mixing ratio. These data were qualitatively compared with the refractivity field because refractivity cannot be calculated from SRL alone. Figure 13 shows a time–height display of mixing ratios from the SRL on 22–23 May 2002. Note the strong correspondence, particularly in gradient and trend, between the SRL mixing ratio, surface station mixing ratio, and S-Pol refractivity.
Potential utility of refractivity for short-term forecasting
This section will evaluate the potential use of the refractivity field for making short-term forecasts of convective development. Three cases showed features within the S-Pol refractivity range and will be used to illustrate whether the refractivity shows good potential for making improved forecasts.
On 22 May 2002 two boundaries formed within the radar refractivity range. At 2022 UTC (left panels of Fig. 14), the reflectivity field showed fairly uniform values (Fig. 14a) while the velocity field indicated a strong south-southwest flow of ∼18 m s−1 with no obvious convergence lines (Fig. 14c). The refractivity field, however, showed a tightening gradient oriented north from S-Pol and along the 195° radial from S-Pol. Because this gradient was persistent and strengthened over the previous couple of hours, IHOP_2002 aircraft and mobile ground vehicles were directed toward the feature. By 2309 UTC (right panels of Fig. 14) the reflectivity field indicated two fine lines representing low-level convergence zones (Fig. 14b). There was no apparent velocity convergence associated with the fine lines (Fig. 14d). The refractivity field (Fig. 14f) showed strong gradients corresponding to the fine-line locations. The surface stations also suggest a hint of a double boundary structure with driest air to the west, intermediate air between the two boundaries, and the moistest air to the east of the eastern boundary. This case clearly shows that the refractivity field has great potential to identify boundaries that can often be precursors to convective development (Wilson and Schreiber 1986).
Furthermore, the refractivity field may even identify these boundaries prior to the boundaries appearing in the more traditional radar fields depicting reflectivity fine lines and velocity convergence zones. One possible explanation for the refractivity field showing these boundaries prior to the reflectivity or velocity fields is that the latter are cumulative in nature. They require the concentration of insects to accumulate within the convergence zones before they will be detected by the radar (e.g., Achtemeier 1991). The refractivity gradients, on the other hand, are nearly instantaneous in that the moisture and/or temperature gradients define the boundary.
The importance of the early detection of boundaries to nowcasting the onset of convection cannot be underestimated because most current nowcasting systems rely heavily on monitoring the location and behavior of boundaries as illustrated in radar fine-line signals (Wilson et al. 1998). In addition, assimilation of these measurements may also improve the ability to predict storm systems using numerical models (e.g., Montmerle et al. 2002).
On 10 June 2002, although there were no obvious boundaries in the reflectivity field, a persistent northeast–southwest refractivity gradient existed ∼40 km northwest of S-Pol (Fig. 15). This was also apparent as a broad moisture gradient and a suggestion of weak surface convergence in the surface station measurements. A couple hours after the formation of the refractivity gradient, convection developed at the ∼40 km range along the gradient (Fig. 15d). Note that the locations of the first echoes aloft correspond well to the locations of the tighter refractivity gradient. Small cells continued to develop along the gradient for the next couple of hours (Fig. 15f). The individual echoes had lifetimes of 20–30 min and did not become large storms. This shows, however, that a region of a persistent refractivity gradient may indicate a low-level weak boundary, which can sometimes be a precursor to convective development.
At 2347 UTC 12 June 2002 a northeast–southwest boundary with undulations was apparent running east of and ∼15 km southeast of S-Pol. This boundary was observed in both reflectivity (Fig. 16a) and velocity fields (not shown). The refractivity field showed relatively cool, moist air north of the boundary and warm, dry air to the south of the boundary. This was observed by the surface stations, which also indicated low-level convergence associated with the east–west portion of the boundary east of S-Pol. Additionally, a gust front was approaching the refractivity domain from the south. Approximately 15 min later the northeast–southwest reflectivity boundary remained in approximately the same location (Fig. 16c), but high-refractivity air extended southward. This is apparent in the cooler, moister green colors extending farther south to the 120° radial out to a range of 35 km (Fig. 16d). The gust front from the south propagated to within 40 km of S-Pol (Fig. 16c), with strong low-level convergence in the Doppler velocity field (not shown). By 0021 UTC the two boundaries had nearly collided, with an echo forming at 40 km east-southeast of S-Pol (Fig. 16e). Several cells were initiated in the same area and merged together to form a large, long-lived storm with reflectivities over 65 dBZ. Although the initial echo occurred outside of the refractivity coverage, the increase in refractivity was widespread enough to infer that it also occurred beneath the new storm. It appears that the combination of the increasing moisture associated with the northeast–southwest boundary and the enhanced convergence brought by the southern gust front allowed for convection initiation. Because there were cirrus clouds in this area, the visible satellite imagery was not useful for monitoring the convective development. Thus, the radar and surface stations were the only real-time observational tools available to forecasters. In this example, the refractivity field played an important role in monitoring the situation for convective development.
Summary and operational implications
The radar refractivity retrieval during IHOP_2002 provides an excellent representation of the low-level moisture distribution. This was illustrated by the pronounced correspondence between refractivity and surface station moisture gradients. The overall validity of radar refractivity was demonstrated by the high correlations with refractivity derived from surface measurements, aircraft, profiling instruments, and radiosondes.
There is a slight bias between radar refractivity and some surface station refractivity measurements. Differences in altitude between these two measurements likely caused the bias. Although no high-resolution profiling measurements were available to address this hypothesis empirically, an examination of surface station measurements and refractivity profiles, simulated using Monin–Obukhov similarity theory, was applied. It was shown that the bias between the station and radar estimates of refractivity results from the difference in altitude between the station (2 m) and radar (between 0 and 10–20 m) measurements.
The vertical depth represented by radar refractivity appears to be the low levels of the CBL, typically below 200–250 m AGL. It represents even greater depths when the CBL is well mixed. This was illustrated with composites of AERI data, which showed greater bias at night with a stable nocturnal boundary layer. During the afternoons when the CBL was well mixed, there was less bias. The apparent horizontal resolution of radar refractivity varies as well. It appears to be as good as 2 km at times, but at other times can be worse than 4 km. Although the initial phase differences are smoothed over 4 km, a pyramidal function is used; thus, better resolution measurements are possible if the region is densely populated with ground targets.
Although the ideal spatial water vapor accuracy, resolution, and range for sampling preconvective processes (e.g., 0.4 g kg−1; 100 m in the horizontal and vertical directions every 10 min out to 15 km; Weckwerth et al. 1999) may not all be achieved with this refractivity retrieval technique, it still shows great promise as a potential nowcasting and forecasting tool. It is clear that one technology is insufficient to satisfy all requirements for water vapor measurements, but this technique uniquely provides a low-level 2D map—admittedly a small map—of moisture measurements. It may identify boundary layer convergence zones prior to their appearance in the more traditional reflectivity and Doppler velocity fields. Additionally, it shows some promise as an aid in forecasting convective development. This potential application requires further examination, because IHOP_2002 did not collect numerous cases of convection initiation within S-Pol’s limited refractivity range.
The findings herein suggest that implementing the radar refractivity retrieval on the national network of operational radars may provide unprecedented improvements in the ability to map the distribution of water vapor. Experience with S-Pol indicates that the radar refractivity estimates can likely be obtained from the operational National Weather Service Weather Surveillance Radar-1988 Doppler (WSR-88D) S-band radars and Federal Aviation Administration Terminal Doppler Weather C-band Radars (TDWRs) with only the addition of new software. The 10-cm S-Pol radar was the ideal radar for further evaluating the refractivity technique, because the technique had previously been run on McGill’s S-band radar (Fabry et al. 1997). Some of the potential difficulties of operating the refractivity retrieval at shorter wavelengths, including the 5-cm TDWRs, include (i) less ground clutter, (ii) more difficulty in obtaining frequency stability, and (iii) more severe phase aliasing and measurement noise. With a field project planned for the summer of 2006 [Refractivity Experiment For H2O Research And Collaborative Operational Technology Transfer (REFRACTT)], several refractivity retrieval–enabled radars (including operational radars) will be positioned in a network to provide refractivity coverage over a larger area than was available during IHOP_2002. This upcoming project should help to determine whether refractivity can be successfully implemented on operational radars and whether it can actually play a crucial role in improving convective storm forecasting. Ongoing refractivity research also includes the assimilation of radar refractivity into numerical models to further evaluate the utility of refractivity in making short-term forecasts of convective development.
Much of this research was funded through the NCAR/U.S. Weather Research Program (USWRP). Some of this work is based upon research supported by the National Science Foundation under Grant 0208651. Author ML’s work on the vertical profiles of temperature and mixing ratio was partially funded by GLOBE through NASA Grant NCC5-735.
Erik Rasmussen and Conrad Ziegler led the data collection of the mobile mesonets in IHOP_2002, which is greatly appreciated. The University of Wyoming King Air group also collected an excellent dataset used in this study. Belay Demoz provided the scanning Raman lidar data. Thanks are given to Jenny Sun who reviewed an early version of this paper. Comments from three anonymous reviewers are also appreciated.
Corresponding author address: Tammy M. Weckwerth, Earth Observing Laboratory, National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307-3000. firstname.lastname@example.org
* The National Center for Atmospheric Research is sponsored by the National Science Foundation.