Lidar
Report on Regional OSSE
ETL Contribution
Modeling Observations from the Wind-finding Lidarby
Michael Hardesty, Barry Rye, and Graham Feingold
25 January 20021. Space-based Lidar Modeling
a) Instrument concept
An important factor in simulating the performance of a space-based Doppler lidar is specification of the instrument operating parameters. A wide range of potential choices could have been used for this study. Probably the most important decision was the specification of the basic instrument type, i.e., deciding whether the space-based instrument simulated should be modeled as a heterodyne (coherent) or direct (incoherent) Doppler lidar. The two types of instruments have fundamental differences in both their design and their performance. Coherent lidar is typically more sensitive and provides better wind measurements at aerosol levels consistent with the boundary layer and lower troposphere, as well as from atmospheric ice and water clouds. However, in regions where aerosol loading is low, such as in the free troposphere or over the oceans, coherent lidar systems based on currently available technology often lack the sensitivity to obtain wind measurements with sufficient precision for use in numerical weather forecast models.
Direct-detection Doppler lidars, although less sensitive than coherent lidars when aerosol levels are high, have the advantage of being able to obtain wind measurements even when no aerosol is present. Under low aerosol conditions, direct detection lidars are able to detect and process the return from atmospheric molecules to obtain usable wind measurements. Thus, a direct detection instrument deployed in space can, in principle, measure winds within any region that is not blocked by clouds. This attribute makes a direct detection lidar, if sufficiently sensitive, well-suited to the problem of maximizing measurements over the widest possible region. This research effort will help to determine if the objective of maximizing coverage of the estimates is, in fact, optimal for improving weather prediction. An alternative strategy, for example, might be to obtain highly accurate and dense measurements over targeted regions, such as regions of cyclogenesis, jet streams, or hurricanes. For some alternative observing strategies, a coherent Doppler lidar may be the preferred instrument.
For the regional observing system simulation experiments described here, we assumed that the instrument to be modeled is a direct detection Doppler lidar employing a Fabry Perot interferometer as a spectrum analysis element. This configuration is often called a "fringe imaging" system. By assuming a fringe-imaging configuration, we match the simulated space-based system to ground-based prototype instruments actually being developed and demonstrated under a NOAA-funded grant at the University of New Hampshire.
b) Scanning
Because a Doppler lidar measures the radial component of the wind, some type of scanning telescope is required if the full horizontal wind vector is to be calculated from the lidar return. Although single component space-based lidars have been proposed ( the European Space Agency Atmospheric Dynamics Mission expects to incorporate a fixed, rather than a scanning telescope), it is generally acknowledged that a scanning instrument capable of measuring wind components at multiple viewing angles will provide significantly higher impact for forecasts. The cost for a scanning telescope is increased weight and complexity, due to precision motion and pointing subsystems that need to be incorporated on the spacecraft.
Another parameter to be specified is the pointing angle of the lidar beam relative to the spacecraft nadir. Selection of the pointing angle is a tradeoff between better coverage and larger components of the horizontal wind (larger angles) and increased sensitivity for a given set of parameters (smaller angles). Most concepts for space-based lidars assume nadir angles ranging from about 30 degrees to about 45 degrees.
For this simulation, we assumed that the scanning is designed to provide two looks at approximately the same volume from two different angles. Such a scan pattern is implemented by directing the lidar beam forward at a fixed angle, then holding the beam stationary as the returns from several shots are averaged to increase sensitivity. After the satellite has passed the volume of interest, the beam is directed back toward the same volume, and again fixed for several laser shots. The number of regions that can be interrogated in this fore-and-aft manner depends on the needed dwell time to obtain a useful measurement; this in turn is a function of the instrument sensitivity. Scanning is discussed in more detain in the next section, in which the implementation of the instrument simulation model is described
c) Orbital altitude
Like the nadir angle of the telescope, selection of the orbital altitude entails a trade off between coverage and sensitivity. Designing an instrument to operate at the NPOESS altitude of 833 km would have the advantage of potentially being a candidate for deployment on the NPOESS satellite with a variety of other sensors. However, given the current state of demonstrated technology available for space-based lidar wind measurements, we felt that assuming a lower orbit, on the order of 450 km, would be more realistic. A low orbit also matches the orbital parameters assumed in a recent system concept proposed by industry as the baseline for a potential wind measuring instrument.
d) Sensitivity
Sensitivity of the lidar system is a key factor in assessing the potential impact of a spacebased wind measuring instrument. Several parameters determine the sensitivity of the instrument, including laser pulse energy, laser pulse repetition frequency, receiving telescope diameter, optical efficiency of the lidar system, number of pulses averaged, and range to the target. The key parameter is the total number of backscattered photons actually detected and processed by the lidar receiver. For direct detection lidar, on the order to 100,000 photons must be detected to obtain a radial velocity estimate approaching one m s-1 accuracy. Because current laser and telescope technology does not enable instruments to be constructed with enough sensitivity to obtain the requisite number of photons on a single shot, multiple pulse averaging is assumed. For our simulations, when actual lidar parameters are assimilated, we plan to assume a 20-W laser transmitter operating in the ultraviolet spectral region at 355 nm, a 1-meter telescope, and 5-s averaging. This technology is probably somewhat beyond the capabilities of current technology.
Perhaps the most uncertain component in the instrument simulation is the optical efficiency of the receiver. A recent patent (Hays, 2000) describes a novel technique for increasing the efficiency of the optical receiver by "recycling" reflected photons back into the receiver. If the predicted improvement in efficiency is validated, requirements on laser power and telescope size will be significantly reduced. The photon recycling technique was first implemented in a prototype system deployed by the University of New Hampshire. During the fall of 2000, an intercomparison campaign was carried out to characterize efficiency of the prototype system for instrument simulation. The intercomparison is described in more detail in Section 4. Sensitivity data from this campaign are still being analyzed.
Incorporating realistic values for each of the system parameters is critical to the validity of the instrument simulation. This is particularly true in the case of direct detection space-based lidar, since very little actual field data have been available to assess the actual versus predicted performance of prototype instruments. Our strategy for specifying system sensivity is to examine the impact of sensitivity variation by simulating, in turn, ideal, optimistic, and realistic instruments.
e) Effects of clouds
Clouds must be considered when simulating perfomance of a space-based lidar for wind measurements because they affect system performance in a variety of ways. High, tenuous ice clouds can enhance performance by providing high backscatter with low attenuation. However, water clouds are generally opaque to lidar radiation, resulting in a very high signal at the edge of the cloud, but blocking the beam from regions below the cloud.
To realistically model the optical effects of clouds, information on cloud phase, distribution of particle sizes, and shape is required. Because the regional nature run only provides ice and liquid water content (rather than more detailed size distribution information), we developed a simple relationship between these parameters and the backscatter and extinction coefficients for optical propagation. This is described in Section (3).
The small-scale variability of cloud fields is not well simulated in global and even regional scale observing system experiments. For example, in this study ice and water are specified over a 10-km grid, which is much larger than the typical scale of cloud field variability. Initially, we will simulate the lidar using these unrealistically large cloud fields. Eventually, we hope to apply a realistic randomness to the cloud fields produced by the nature run and evaluate the effects on lidar performance. We are collaborating with scientists at Colorado State University in this effort, which is discussed in Section (3).
f) Instrument Simulation Strategy
In developing an instrument modeling strategy for the regional OSSE, we elected to simulate an "ideal " instrument for the initial study, and to employ a very simplistic representation of cloud effects. A primary reason for this was to duplicate the strategy of the global OSSEs being performed at the National Centers for Environmental Prediction, which also initially looked at the effect of new, ideal wind observations as provided by the space-based lidar. For our initial simulation, the ideal lidar provides observations with no error. We simply average the winds available from the nature run across the appropriate horizontal distance and provide that average for data assimilation into the RUC. Thus, the limit of the maximum benefit provided by the lidar is obtained. In subsequent simulations, we will include more realistic assessment of both lidar operating parameters and cloud effects, to better assess the impact of a realizable system.
2. Instrument Model
a) Simulation concept
The interaction between the lidar instrument and the atmospheric model takes place in three stages. First, the satellite orbit and its scan pattern is established, and the space/time coordinates of the observation points are established. The atmospheric conditions (wind velocity, aerosol content, cloud conditions) at these points and on the intervening path between them and the satellite are established. From these data, the performance of the lidar instrument can be inferred, and statistical properties (the line-of-sight, LOS, mean wind speed, instrument precision) of the lidar measurements can be computed. By combining this with the representativeness or sampling error of the measurement, which characterizes the variability of the atmosphere, simulated, appropriately randomized, lidar LOS velocity estimates are generated. These estimates are assimilated in a forecast model so that the incremental effect of the measurements can be assessed. ETL is responsible for writing the software to calculate the satellite orbit, modeling the lidar instrument, and describing its error characteristics. Modeling the atmospheric conditions and assimilating the measurements in a forecast model are FSL tasks.
Data are exchanged between the two laboratories in the standard format of netCDF files. The ETL software has been written by A. Belmonte and incorporated in a MATLAB file called WindLab.
b) Simulation parameters
It was decided at the outset to assume a low satellite orbit (altitude 450 km) in order to emphasize the impact of lidar wind velocity estimates. Initially, a simple circular scan of the lidar beam was assumed at a nadir angle of either 30 or 45 degrees. Each circular scan is interrupted by eight equally spaced pauses, or ‘stares,’ per period to make observations. The pulse energy of the transmitted beam was assumed to be insufficient to produce a Doppler estimate with any useful degree of precision from a single shot, so that within each stare, it was assumed that five shots would be transmitted at 1 second intervals (in practice a lidar might fire more shots than this, but this would be at the expense of the energy in each pulse, so that the overall time of about 5 seconds to complete a useful observation is believed to be approximately correct). During each stare, the beam at ground level travels forward a distance of about 35 km, and the observations made over this distance are averaged together. Later, a more sophisticated scan was developed, in which a given area is viewed twice, once when the satellite is approaching and once when it is receding (Fig. 1). This four-point bi-perspective scan gave rise to closely spaced stare-pairs at different LOS. Although observations from the two stares are not precisely co-located or simultaneous, they have been found in global OSSE’s to provide better forecasts, presumably because the two horizontal components of the wind can be inferred. Of course, this is achieved at the expense of coverage.