The NRL 94 GHz High-power WARLOC Radar as a Cloud Sensor

Introduction: NRL has recently set up a high-power 94 GHz radar called WARLOC at its Chesapeake Bay Detachment (CBD) on the western shore of the Chesapeake Bay. The heart of the WARLOC system is a state-of-the-art 94 GHz gyroklystron, developed by a team led by NRL.¹ The 1.8-m Cassegrain antenna has a 0.1° angular beam width and is mounted on a high-precision pedestal with a 2π steradian solid angle scanning capability. This project was primarily supported to investigate a variety of DOD-related issues, including aircraft identification, precision tracking, and counter stealth. However, the team realized early on that WARLOC could also be an unmatched cloud sensor, and this has been investigated. Radars operating at 94 GHz have been used previously for remotely sensing cloud properties,² but the much higher power and antenna gain of the WARLOC radar give it about 50 dB additional sensitivity. This allows the detection of lower reflectivity clouds at longer ranges, and permits rapid, high-resolution two- and three-dimensional imaging of clouds for the first time.

Because the radar cross section of a cloud droplet of radius r is proportional to r⁶/λ,⁴ where λ is the wavelength, millimeter-wave radars scatter much more strongly from clouds than conventional radars, which normally do not detect clouds. On the other hand, laser radars (lidars) cannot penetrate visibly opaque clouds. Thus millimeter-wave radars are well suited to cloud imaging. While the use of millimeter-waves radars is limited by atmospheric attenuation, there are propagation windows at 35 and 94 GHz; the former having less atmospheric attenuation, the latter enhancing the droplet cross section. The radar reflectivity per unit volume of a cloud is denoted by Z and measures the sixth moment of the droplet distribution. It is usually expressed in units of mm⁶/m^-3, and is often displayed on a logarithmic scale as dBZ = 10 log Z.

A variety of waveforms have been programmed into WARLOC. One is the short pulse (SP) mode, a 100-ns pulse (15-m range resolution); another is the search radar (SR) mode, which is a much longer pulse, but which is compressed on return to the same 15-m range resolution. The sensitivity is increased by the ratio of the uncompressed to compressed pulse times. Both have been used for cloud imaging, the latter being necessary for weak clouds, long range or high attenuation.

A Millimeter-Wave Cloud Image: A sector of a medium-altitude cirrus cloud was scanned at about 1 p.m. on April 9, 2002.³ On-site measurements showed the temperature at 11 a.m. was 23°C and the relative humidity was 62%. At CBD, the sky was mostly overcast, with low-altitude stratocumulus clouds. However, WARLOC obtained images of the cloud above this. Note that no ground-based laser system could have obtained these data because of the intervening lower level cloud. The radar waveform was a 100 ns, unchirped pulse, and the pulse repetition frequency (PRF) was 500 Hz. This pulse length gives a 15-m range resolution, and the 0.1° beam width gives an angular resolution of about 10 m at a range of 6 km. The cloud data were obtained with the radar pointing in an easterly direction while sweeping upward in the vertical plane between 30° and 50° to the horizontal at a rate of 0.1°/s. At this PRF and scanning rate, the radar elevation angle changes by 0.02° during a train of 100 pulses. This is much less than the 0.1° angular width of the antenna beam, so the 100-pulse train is sampling essentially the sample cloud volume element, and the results can be averaged over this number of pulses.

Figure 6(a) shows a white-on-black false-color plot of dBZ plotted as a function of distance R = (X,Y) from the radar. The wedge-shaped region is the sector sampled by the radar. The image resolution of about 15 m allows the cloud structure to be seen in extremely fine detail. A horizontal and diagonal vein is apparent in the cloud, as well as filamentary structure throughout the volume and wispy structures emerging from the top. Notice also the intricate structure of the cloud boundary on the bottom left-hand side of the figure. We are quite sure that no other sensor could image the internal structure of a visibly opaque cloud in this kind of detail.

FIGURE 6
A white-on-black false-color plot of the reflectivity factor Z of a cloud scanned on April 9, 2002.

The Cloud Correlation Function: Figure 6(b) shows the 104-pulse-averaged data for Z displayed on a linear scale. The figure shows a speckle pattern overlaid on the cloud structure, indicative of a random process, perhaps fluid turbulence, playing a role. The speckle pattern is all but imperceptible in Fig. 6(a) because of the compression of the logarithmic scale. To investigate this random process, we have calculated the spatial correlation function of Z: C(r). It is a scalar function of a two-dimensional vector r and is shown in Fig. 7. Notice that it comes to a cusp at zero point separation. The figure shows that the cloud fluctuations are correlated over about 3-4 times longer distance horizontally (x-direction) than vertically (y-direction). Also, the correlation function is quite anisotropic, even at the shortest scale lengths that WARLOC can resolve.

FIGURE 7
Correlation function for the cloud region shown in Fig. 6(b).

It is natural to think that fluid turbulence occurs in clouds. One of the most important results of modern fluid turbulence theory was obtained by Kolmogorov, who predicted that turbulence scale lengths would include an "inertial range" in which eddies would progress from larger to smaller scale lengths without dissipation. The inertial range falls between the long (outer) scale length at which power is injected, and the short (inner) scale length at which power is dissipated by viscous damping. If the turbulence is homogeneous and isotropic, Kolmogorov showed that in the inertial range the correlation function scales as r ^2/3 where r is the separation between fluid points. The one-dimensional correlation functions C(x,0) and C(0,y), for the horizontal and vertical directions, respectively, are plotted in Fig. 8. Under each plot is another curve illustrating x^2/3 (y^2/3) behavior. Clearly, the cusp behavior is characteristic of the Kolmogorov inertial range with one important caveat. The theory is derived under the assumption that the turbulence is at least locally isotropic, whereas the cloud correlation function is anisotropic down the smallest scale lengths we can resolve. Figure 8 also allows us to easily discern the outer scale length, the distance at which the measured correlation function departs from the x^2/3 dependence. It is about 1.5 km horizontally and about 0.5 km vertically. Furthermore, the horizontal correlation decreases monotonically, but the vertical correlation shows an interesting wave-like structure probably associated with the process driving the turbulence.

FIGURE 8
The solid curve in (a) shows the correlation function C(x,0), and the dashed curve shows a function proportional to x^2/3 for comparison. Similar curves for C(0,y) are shown in (b).

To summarize, the NRL WARLOC radar has demonstrated the ability to image the internal structure of visibly opaque clouds in tremendous detail. These images can provide new multi-dimensional information on cloud turbulence.

[Sponsored by ONR]