1. Introduction

The implementation of a national network of operational Doppler weather radars has resulted in greatly improved precipitation and storm monitoring over the contiguous United States (e.g., Crum et al. 1993). The development and continued support of the radar network was, and is, a joint agency program of the National Weather Service (NWS), the Federal Aviation Administration, and the Department of Defense. This program, and its component radars, was originally known as Next-Generation Weather Radar (NEXRAD). The radars currently are usually referred to as Weather Surveillance Radar-1988 Doppler (WSR-88D).

The National Research Council (NRC 1995) conducted a comprehensive assessment of the spatial coverage and detection capabilities of the new national radar network. The NRC report concluded that, “in comparison with the old network, the NEXRAD network will cover a much broader area of the contiguous United States. For the detection of specific weather phenomena—supercells, mini-supercells, macrobursts, lake-effect snow, and stratiform snow—the NEXRAD coverage is much greater than it was for the old system….” The improvements were illustrated in a number of report tables and figures (e.g., Fig. 1) showing the coverage of the new system for a variety of weather phenomena. For example, Fig. 2-5a in the NRC report (not shown here) illustrates the coverage of the new radar system for detecting macrobursts, where it was assumed that the top of the median macroburst extends to 3 km above ground level (AGL). [Fujita (1985) defined a macroburst as a large downburst with its outburst winds extending in excess of 4 km (2.5 mi).] Over the eastern United States, the coverage for this phenomena is nearly comprehensive, but it is very limited over the western United States.

The WSR-88D data have proved valuable for use and applications far beyond storm detection and warning functions of the NWS. The spectrum of users of data from the radar network is very wide and includes operational meteorologists, hydrologists, climatologists, atmospheric and hydrologic modelers, and a wide variety of researchers. It is the authors' experience that users, or potential users, of data from the radar network are often unaware of the details of the spatial coverage of the data. This brief note presents several spatial coverage maps for the WSR-88D network to heighten awareness within the user community of weaknesses and strengths of the data.

2. Background and methodology

The chart that appears most frequently in the literature (e.g., Klazura and Imy 1993; Serafin and Wilson 2000) to illustrate the coverage of the WSR-88D network is shown in Fig. 1, which is from the NRC (NRC 1995) report. This chart shows radar coverage (at 10 000 ft above the elevation of each individual radar) for the WSR-88D network and indicates nearly complete coverage for the contiguous United States. If adjacent radars are at reasonably similar elevations (e.g., the eastern half of Fig. 1), interpretation of this radar coverage chart is reasonably straightforward. The coverage of the WSR-88D network depicted in Fig. 1 for the western third of the United States appears, particularly to the casual reader or user of WSR-88D data, to be very good. Terrain elevations for the country and the locations of the WSR-88Ds are shown in Fig. 2. Because the elevations of the individual western U.S. WSR-88Ds range from very near sea level to more than 3.05 km (10 000 ft) above mean sea level (MSL), interpretation of Fig. 1 becomes a conundrum. It is nearly impossible to infer meaningful information from Fig. 1 about WSR-88D coverage for the western United States.

Westrick et al. (1999) discussed the impacts that both low-level terrain blockage and radars sited at high elevations have upon WSR-88D spatial coverage and the network's ability to detect and to monitor rainfall along the northwest coast of the United States. They processed radar and terrain elevation data in two ways to determine realistic WSR-88D coverage from northern California northward across western Washington. They first determined the coverage for the lowest beam, not blocked by 50% or more, from the surface to 8 km (see their Fig. 3) to estimate the region over which the network could reliably detect the presence of stratiform precipitation. For this detection criterion, the coverage along the northwest coast is very good.

However, they then determined coverage both at, and below, 2 km AGL and at, and below, a height 300 m below the average freezing level for winter storms at each radar site in their study region. These criteria were designed to estimate coverage for stratiform precipitation events, when data are needed below the bright band to estimate precipitation rate accurately at the surface. Their charts estimating effective radar coverage for quantitative precipitation estimation (QPE) are far more restrictive (see their Fig. 4) and indicate that only very small portions of their study area have adequate coverage for QPE.

This study employs analysis techniques essentially the same as those of Westrick et al. (1999) to generate several WSR-88D network coverage maps for the contiguous United States. The analysis process begins with the terrain-based hybrid scan files (O'Bannon 1997), which are available from the NWS WSR-88D Radar Operations Center (ROC) for each individual WSR-88D in the contiguous United States. The hybrid scan data bins, always from one of the four lowest-elevation tilts, are the WSR-88D data gathered near the ground and used by the Precipitation Processing System to generate estimates of precipitation rate (Fulton et al. 1998). A brief overview of the terrain-based process, as described by O'Bannon (1997), used by the ROC to identify the hybrid scan bins follows.

Data taken at the lowest radar elevation angle above the terrain are identified as the hybrid scan bins, with two exceptions: 1) the bottom of the radar beam must clear the terrain by at least 500 ft (∼150 m) and 2) the radar beam cannot be blocked by 50%, or more, at ranges beyond an intervening terrain obstruction. These constraints are partially illustrated in Fig. 3. The heavy, solid line shows the bottom of the beams,¹ and their elevation angles are also indicated, defined as the hybrid scan as a function of range, with a mountain present.

Near the radar [which is at location (0, 0)], the fourth tilt (3.35°) is defined as the hybrid scan bin because the heights of lower tilts are within 150 m of the ground. At farther ranges, the lowest elevation angle data (0.5°) are defined as the hybrid scan bins, once the bottom of this lowest beam clears the terrain by more than 150 m. At ranges between approximately 80 and 130 km the mountain's terrain causes the second tilt (1.45°) and then the third tilt (2.4°) data to be defined as the hybrid scan bins. Beyond the mountain the second tilt (1.45° beam) remains more than 150 m above the terrain and thus is defined as the hybrid scan bin at all longer ranges. If the mountain peak were higher and blocked up to 49% of the second tilt radar beam, the definition of the hybrid bins would remain essentially the same, jumping to the third tilt right over the mountain but then becoming the partially blocked second tilt at all longer ranges. If the mountain peak were still higher and blocked 50% or more of the second tilt radar beam, the third tilt beam would remain the hybrid tilt beam at all ranges beyond the mountain peak.

Note that the NRC coverage map shown in Fig. 1 was defined and computed using the height of the bottoms of the radar beams. Other coverage maps for phenomena detection in the NRC report considered the height of the center of the beam. It is more realistic to assume that useful data are obtained only if at least one-half of the beam samples the feature of interest [as per NRC (1995), O'Bannon (1997), and Westrick et al. (1999)]. Thus, coverage at any given level in this study means that at least one-half of the radar beam volume is at and below that level. Radar coverage has been computed out to ranges of 230 km. All coverage data have been computed starting with the ROC-defined hybrid scan data bin. Additional data from elevation angles above the hybrid tilt were then considered.

The U.S. Geological Survey's “GTOPO30” terrain data, along with the height of the center of the beam, have been used to determine beam heights above the terrain. These elevation data, having approximately 900-m horizontal resolution, have been resampled onto a 2750-m Cartesian grid covering the contiguous United States. Maximum terrain elevations for very steep, mountainous terrain have been damped by as much as 15% (i.e., about 450 m). The uncertainties due to terrain smoothing lead to slightly optimistic coverage computations over complex terrain. The computed coverage AGL is estimated to have uncertainties due to terrain smoothing and use of the hybrid scan bin (which causes slightly pessimistic calculations) of no more than 300 m over the most complex terrain features of the West.

Last, it has been assumed that the radars are scanning in the mode that provides maximum coverage in the vertical. Gaps in coverage near the radar that result if the WSR-88D is operating in other scanning strategies [refer to NOAA (1991)] have not been addressed. If a radar is operating in volume coverage pattern-21 [VCP-21, 11 elevation tilts every 6 min; again see NOAA (1991)] there are very significant gaps, or blind spots, in the coverage at high altitudes. For example, echoes at 10 km above the elevation of a radar operating in VCP-21 cannot be detected within an annulus from 60- to 95-km range. Details concerning WSR-88D scanning strategies and resultant coverage gaps can be found in Howard et al. (1997).

3. WSR-88D radar network coverage maps

The first two coverage maps illustrate coverage above mean sea level. Coverages are presented here both for MSL heights and also for AGL heights. Operational products from the WSR-88Ds are computed in one or the other of these coordinate systems. For example the clear-air winds and echo-top products are presented to the user in MSL, whereas all storm cell parameters are presented in AGL, with ground level being the elevation of the radar.

Radar coverage at 3 km MSL (i.e., near the 700-hPa level) for the entire United States is depicted in Fig. 4a. Because 3 km (9836 ft) is very close to 10 000 ft MSL, this figure is similar to the NRC's Fig. 1 for regions in which radars are located near sea level. However, for the entire United States, the coverage patterns are strikingly different than those shown in Fig. 1. There is extremely limited WSR-88D coverage at 3 km MSL over the West. There are also a number of regions in the central United States for which there is no coverage at 3 km MSL because of both the radar elevations and use here of the height of the center of the beam. Oklahoma and the southeastern United States have large areas of overlapping coverage, that is, extremely comprehensive sampling, by two–four radars.

When coverage at 5 km MSL (i.e., nearly at 500 hPa) is considered, (Fig. 4b) the area having radar surveillance is considerably improved relative to 3 km. However, there are still large data voids, particularly over the interior of the West. Because the data at this elevation are, for most locales, well above the ground surface, the implication is that the utility of the radar data for estimating precipitation at the surface is highly compromised across the West, as emphasized by Reynolds (1995), Westrick et al. (1999), and Gourley et al. (2002). In contrast, there are many locales in the eastern half of the country in which as many as four–seven adjacent radars simultaneously scan the 5-km MSL level.

The WSR-88D network coverage at 2 and 1 km above the ground surface is shown in Figs. 5a and 5b, respectively. Once again, there are very limited radar data available within 2 km of the surface across almost all of the West, the notable exception being portions of California. There are also some fairly large regions in the central United States that do not have radar coverage within 2 km AGL. For winter storms in the West, and also frequently in the rest of the country, precipitation tends to be either stratiform snow or rainfall in environments with low freezing levels. In these situations, the presence of an elevated bright band limits the range and heights at which useful reflectivity data (i.e., data taken below the height of the bright band) can be gathered. Thus, Figs. 5a and 5b illustrate that coverage for which useful QPE can be generated is very limited for cold-season, stratiform precipitation events [again as pointed out by Westrick et al. (1999) and Gourley et al. (2002)]. Indeed, Fig. 5a is similar to Westrick et al.'s analysis for stratiform QPE (cf. their Fig. 4b), but it has been computed for the entire contiguous United States.

Radar data are needed close to the surface if the user is to make reliable inferences about whether significant weather events are actually occurring at the ground, for example, heavy rains, hail, tornadoes, and strong winds. Dunn and Vasiloff (2001) have discussed problems with interpretation of the data from an elevated WSR-88D during the unusually intense Salt Lake City tornado from August of 1999. The direct detection of downbursts with the WSR-88D radars requires that data be gathered within about 1 km of the surface because of the shallow character of the outflow winds (e.g., Elmore and McCarthy 1992). Downbursts are the most common severe convective weather phenomena over much of the West during summer. Except for extremely limited areas, the occurrence of downburst winds cannot be detected directly in the WSR-88D velocity fields in the West and also over most of the contiguous United States.

Figures 5c shows that coverage at 3 km AGL is limited for much of the West. Note that Fig. 5c is essentially the same as the NRC chart (their Fig. 2–5a) for macroburst detection coverage. It is clear that forecasters in the West must use data from high levels to infer the weather threat at ground level. For summer thunderstorms, the use of reflectivity data from high above the surface to estimate precipitation is fraught with uncertainties in the West because of high cloud bases and low mean relative humidities within subcloud boundary layers. There are also areas scattered across the central United States for which there are no data available within 3 km of the surface, complicating the use of the data to determine QPE and to identify severe weather. The fairly large gaps in coverage at 1 and 2 km AGL over the eastern two-thirds of the United States (Figs. 5a,b) indicate that forecasters across the country must make operational decisions for many regions based up radar data obtained well above the ground.

4. Summary

The chart often used to depict the coverage of the national Doppler weather radar network (Fig. 1) presents an optimistic picture. The coverage charts presented here illustrate that the effective radar coverage at low levels remains very limited, especially in the West. Radar data to support the warning mission of the NWS are very limited below 2 km AGL over much of the contiguous United States. Over the eastern portion of the country, coverage is nearly complete at 3 km AGL; however, there are some limited areas in the severe storm belt that do not have radar coverage within 3 km of the ground. In contrast, the south-central plains and the Southeast are characterized by coverage by multiple radars, sometimes as many as seven different radars, scanning at the same levels, indicating that these two regions provide excellent opportunities for applied research.

We have computed many more radar coverage charts than can be presented in this note. At the time of writing, these charts were available online (http://www.nssl.noaa.gov/∼jzhang/radcover.html), and they include coverage maps for Alaska and Hawaii and many major metropolitan areas, national parks, and recreation areas.