At present, there are a limited number of measurement techniques available for the assessment of wind in the middle and upper atmosphere. Balloon-borne instruments can make in situ wind and temperature measurements in the troposphere and lower stratosphere (surface to balloon bursting height at approximately 35 km) . Radars are capable of measuring winds below 20 km and again in the mesosphere between 75 and 100 km due to the presence of favorable scattering targets but are unable to measure in the intervening region. Wind measurements relying on the release of tracers are also possible in the upper middle atmosphere using rockets . Unfortunately, these high-quality measurements are expensive and infrequent. Microwave radiometers are capable of making low-resolution wind measurements at five pressure levels between 30 and 80 km . In the upper mesosphere and lower thermosphere (UMLT), metal resonance lidars can measure temperature and wind in the so-called metal layer . Finally, there were several past satellite instruments which measured wind at various levels. In the stratosphere and UMLT missions like WINDII , TIDI , HRDI , and more recently ICON used passive remote sensing to measure horizontal winds. In the troposphere and lower stratosphere, the AEOLUS mission used active remote sensing and the double-edged Doppler–Rayleigh technique to measure line-of-sight winds .

The pressing need to measure middle-atmospheric winds, which has motivated the construction of previous Doppler–Rayleigh wind lidars, has been discussed in the literature . There are currently four other operational stations in the world: (1) L'Observatoire de Haute Provence (OHP) in France (44° N, 6° E), (2) the Arctic Lidar Observatory for Middle Atmosphere Research (ALOMAR) in Norway (69° N, 16° E), (3) Observatoire de Physique de l’Atmosphere de la Réunion (OPAR) on Réunion Island (21° S, 55° E), and (4) two mobile wind lidars in China developed in the Hefei region (approx. 32° N, 117° E). In addition to these permanent stations, short-lived Doppler–Rayleigh lidars have been built at the University of Michigan , Arecibo Observatory , and NASA Goddard . A Doppler wind lidar using aerosol scattering was recently described by . To provide a comprehensive overview of the current “state of the art”, we will give a detailed review of the technical developments of each station and summarize the key scientific results that have been obtained using Doppler–Rayleigh lidars.

We describe in this section the transmitter of our RMR wind–temperature lidar, namely the pulsed power laser and the beam guiding on the laser bench, the seeding system, and the laser pulse spectrometer (LPS) for frequency monitoring of the pulsed laser. The central part of the transmitter is an Innolas SpitLight EVO IV diode-pumped Nd:YAG with an externally stabilized seed laser (see below). The power laser is triggered at 100 Hz by the central trigger controller (see Appendix ). We have developed a cavity control that adjusts the Piezo-coupled end mirror of the laser for buildup-time reduction (BUTR). The laser contains a second harmonic generator to get light at 532 nm wavelength with a pulse energy of ∼ 500 mJ. Only this second harmonic is used, while the remaining IR light is dumped.

Doppler–Rayleigh wind and daylight temperature measurements require a high degree of stabilization of the power laser frequency. We achieve this by externally seeding the power laser with a Coherent Prometheus 100 laser. The seed laser is tuned to the line 1109 in the iodine absorption spectrum using an I2 vapor cell, i.e., to 532.112 nm (air) (532.260 nm, vac.). This cell is heated to 45 °C. We split some of the frequency-doubled seed light and send 70 % through the iodine cell and into a photodiode and 30 % directly to a photodiode. By calculating the ratio of the signal from the two photodiodes, we measure the absorption of the iodine gas at a given wavelength. Stabilization is achieved using built-in cavity control of the seed laser. The I2 line 1109 has been selected for optimal wind sensitivity within the wavelength range that is covered by the Nd:YAG laser. See Fig.  in Sect.  below for the measured spectrum of the seeder iodine cell.

The path of the pulsed laser is displayed in Fig. . Two cascaded wedge plates (BP) pick ∼ 0.2 % of the pulsed laser for frequency measurements in the LPS (see below). Next, the galvanometer scanner separates the laser pulse by pulse into two different paths. Through this, we achieve two virtual 50 Hz subsystems (called A and B) for wind measurements in two directions. The light along path A then encounters mirror A and is sent to beam-widening telescope A (BWT-A), the same with path B. The BWT is a custom-designed 10× telescope. The outgoing pulse is measured by photodiodes (PDs), generating a trigger pulse which is then sent to the central trigger controller. Finally, the light encounters the first of three beam-guiding mirrors (BGM-1A/B). This mirror is used to steer the beam into the telescope hall. Technical details of the optics can be found in Appendix .

The frequency offset between the power laser and the seed laser as well as the associated jitter between the two frequencies are major sources of systematic uncertainties in wind measurement. In the retrieval of horizontal wind components, a frequency offset of only a few megahertz can result in a significant wind bias (3.75 MHz ≈ 1 m s-1). To measure this effect and correct for the laser offset, 0.01 % of the power laser light is diverted on the transmitter table into a fringe imaging etalon. By measuring the separation of fringe peaks between the seed light and power laser light, we record the frequency offset for each laser pulse. Details of the laser pulse spectrometer (LPS) are described in a later publication (see Part 2 of this study currently in preparation). Figure  shows a histogram of the offset for ∼ 2.5 h of measurement on 5 April 2023. The distribution is nearly symmetric; i.e., negative and positive deviations from the mean cancel each other in the wind processing algorithm (assuming constant pulse energy and a known atmospheric temperature). The frequency offset is measured to be taken into account in the processing chain.

The RMR wind–temperature lidar is set up as a bi-static system, with the laser being emitted off the optical axes of the telescopes. This implies that we do not achieve overlap between the laser and the telescope's field of view (FOV) in the lowest part of the atmosphere. We aimed for a system where at least one telescope can be set vertically as well as tilted in different cardinal directions (e.g., north and west). In combination with a mono-static system, this would require complex beam-steering optics. We therefore decided on the more cost-effective bi-static setup, taking into account that full overlap of laser and FOV is achieved above ∼ 30 km. In the following, we describe the beam-guiding, beam stabilization, and telescope setup in detail.

The two beams are directed by BGM1-A/B vertically up into the telescope room, where they are reflected by BGM2-A/B horizontally towards BGM3-A/B mounted on poles beside the telescopes. Both BGM2 mirrors are motorized for automatic alignment of the beam if, e.g., the mirrors' BGM3-A or BGM3-B is moved to another viewing direction of the lidar. BGM3-A and BGM3-B direct the laser light into the atmosphere. The distance between the BGM3-A and BGM3-B and the optical axes of the tilted telescopes is about 1 m. BGM3-A and BGM3-B are both equipped with motors for fine alignment of the beam (see Appendix  for details). Path A can be directed either northward or westward at 65° elevation or higher, including zenith pointing. This requires steering the BGM3-A not only with the fine-alignment motors but additionally with a combination of a 360° motorized azimuthal mount and a ±10° goniometer. BGM3-B is fixed at 90° azimuth (eastward) and 65° elevation because of mechanical restrictions by the building. This combination of optical mounts allows for a full range of motion and beam pointing, which can be controlled in real time by software (see below).

The receiver assembly comprises two 70 cm Newtonian telescopes with motorized mounts allowing for tilting off-zenith and azimuthal rotation. We use laser-synchronized cameras mounted along the telescope's optical axis to ensure that the laser is correctly situated inside the telescope's field of view. For this, 10 % of the received light is guided into the camera and 90 % into the fiber for the detector. The basic principle of this setup is described in and used, e.g., in .

Monitoring the position of the laser beam in the telescope FOV allows us to stabilize our system alignment against thermal drifts, which can occur throughout the night, and it reduces the variability in signal strength by constantly optimizing the alignment. In practice, this is done by stabilizing the beam position every 5 s to the pixel position which corresponds to the best signal quality in each camera (Fig. a).

The value of the stabilization system can be seen in Fig. . Panel (a) shows a measure of lidar signal at 50 km before the introduction of our beam stabilization system. We use black arrows to indicate the common slow misalignment that lidar systems can experience as the system thermally shifts throughout the night. These misalignments require an operator to occasionally optimize the signal by making a small adjustment to the beam pointing (marked yellow). Panel (b) shows the same signal metric from a night when the beam stabilization system was active. The stabilization automatically compensates for beam drift and variability in the signal, with the result being a measurement with more consistent signal quality throughout the night.

With the stabilization camera having a comparatively small FOV (∼ 0.15° or 2.5 mrad), the laser may drift off the camera between two soundings. We therefore attached another camera to the telescope, having ∼ 3.5° FOV for coarse alignment of the laser beam (see Appendix ). Secondly, we can use this camera image of the laser beam against the background stars to determine the exact beam-pointing direction and elevation angle of our laser beam and telescope (Fig. , lower right). In practice, this is done by aligning the camera so that the laser tip is centered and uploading a still frame from the camera to an astrometry website (https://nova.astrometry.net/, last access: 2 June 2022). The website then gives the precise viewing direction based on the identified star constellation. The well-defined pointing information is crucial for avoiding systematic errors in the estimation of the zonal and meridional wind components.

The logical schematic for the detector bench is shown in Fig. . Light enters the bench via three fiber-optic cables; two fibers come from the telescopes (FIB-A and FIB-B), and one brings light from the seed laser (FIB-S) for use in determining the efficiency calibration between the detector channels. The lidar signals from the telescopes are 50 Hz interspersed so a second galvanometer scanner mirror, similar to the one on the transmitter bench and using the same 50 Hz trigger, is used to direct the pulses down the correct optical path. Both beams are collimated (COL), and then the dichroic mirror (D-608) separates the nitrogen Raman signal (608 nm). The Raman signal is sent through a 608 nm interference filter (IF-608) and is focused onto a single photon counting module (VVR). The Doppler–Rayleigh signal passes through a 532 nm interference filter (IF-532) and is split with a 30:70 beam splitter (BS-1). The reflected 30 % is sent to a 95:5 beam splitter (BS-2) where it is transmitted onto the high-gain detector channel (VH). The 5 % reflected light is focused onto the low-gain detector channel (VL). The remaining 70 % of the light at BS-1 is directed through the iodine vapor cell (IOD) and is split by another 95:5 beam splitter (BS-2) before being directed to a high-gain channel (VDH) and a low-gain channel (VDL). All these detection channels are equipped with avalanche photodiodes (APDs) for high quantum efficiency.

The detected signals are recorded by an FPGA-based counting system with dedicated pulse counting software for up to 16 channels. The FPGA clock rate of 100 MHz allows for a physical range resolution of 1.5 m. Up to three photons per bin can be recorded. An IAP-built signal-conditioning board is used to adapt the signal level of the APD pulses to the FPGA input level and to discriminate electronic noise. Backscatter profiles for each laser pulse and each detection channel are transmitted to a PC via a USB-3 connection. The PC runs a Python code for real-time data acquisition and storage. The overall system is called LISA, LIdar Singleshot Acquisition, and runs automatically without operator action. All operator-based actions are performed with external software if needed for, e.g., alignment purposes (see Sect. ).

Besides the backscattered light of the pulsed laser, we feed the seed light via the fiber FIB-S (see Fig. ) into the detector to get an independent calibration metric for the system . The seed light follows nearly the same optical path as the lidar signal onto each APD. It is passed through a mechanical chopper (CHP) so that it appears in the detector only at a range of approximately 300 km and above. The light passes through some collimating optics (COL) and is reflected by a mirror through and off of the backside of a dichroic filter (D-608).

Knowledge about the exact iodine absorption spectrum is crucial for the wind calculation from the VDH / VH ratio (or VD / VL for the low channels). Figure  shows the modeled absorption spectrum for gaseous molecular iodine at 40 °C (blue) produced using IodineSpec software from the Institute of Quantum Optics at Leibniz University in Hanover, Germany. The I2 line 1109 has been selected for optimal wind sensitivity within the wavelength range that is covered by the Nd:YAG laser. The measured spectra are shown as magenta (SCS, seeder cell spectrometer) and green (detector) lines in Fig. . The SCS cell serves only to stabilize the seeder, with the stabilization point marked by the black star. By construction, the laser frequency optimally fits the absorption cell in the detection bench. The transmission of the cell depends on the Doppler shift of the backscattered light and is therefore a direct measure of the wind speed . The red line in Fig.  shows the solar spectrum. The selected I2 line is close to the center of a weak Fraunhofer line, which reduces the background by approximately 30 % for the daylight-capable RMR-T lidar and planned future daylight wind measurements.

Our wind lidar is designed as a modular system with independent units for triggering, laser control, beam steering, and safety. The modular design has allowed us to save time, as well as work and maintain subsystems in parallel. The subunits share different labs and are connected to the local network. All subunits are connected to a software back end that is responsible for the overall control of the system. All operator functions are bundled in a single front end, also allowing the operation of the lidar with less experienced personnel.

The Message Queuing Telemetry Transport (MQTT) protocol forms the basis for our lidar system's networking concept. All of our computers, lasers, optoelectronics, triggers, safety systems, interlocks, detectors, data acquisition, and control software are networked by “subscribe and publish” under a common framework. This system has proven efficient and useful for coordinating multiple hardware devices (with disparate proprietary software elements) into a single, cohesive lidar system. Compared to that, a simple TCPIP socket connection is less versatile and always requires distinct relations between the sender and receiver of the information, which reduces the flexibility during development and operation. The status messages are typically broadcasted with 0.5–1 Hz cadence. This is fast enough for the overall control of the lidar, even if the different devices work and communicate at a much higher cadence (up to 100 Hz).

The Kühlungsborn Lidar Automation Software (KLAUS) is the heart (i.e., back end) of the lidar operation software. It is a multi-thread state machine that is coded in Python with two main sections. One section (the actual state machine) does the sequential startup and shutdown of the different subunits. The states are connected by actions like powering on the different hardware parts, activating the laser safety devices, starting the laser, or initializing the beam steering. The action requests are broadcasted to the subunits by MQTT messages, and the successful execution is messaged the same way, setting the system to the next state. State machines are one possible concept to formalize the different tasks of the lidar measurement. They allow a modular concept and retain a high degree of flexibility. Figure  describes the KLAUS concept in more detail.

The system reaches its final startup state by opening the laser shutter. The sounding, i.e., data acquisition, starts automatically as soon as the trigger signal from the power laser is available. The two laser beams are steered by the beam stabilization software to their respective targets on the cameras attached to the telescopes once the backscattered light has been detected by the camera (see Fig. a). The sounding state contains a check of the most important system parameters, which is the second main section of the script. If needed, the KLAUS script shuts down the whole lidar automatically. The system check comprises different countdown timers that are put on hold as long as the particular criterion is fulfilled. If not, the timer counts down until either the criterion is met again or the time has expired. The former resets the timer, and the latter initiates an automated complete shutdown of the lidar. Checked criteria include the actual and 30 min mean signal level, the actuality of the data, actions of the laser safety system (see below), watchdog messages of all subsystems, laser conditions, and cooling water flow. For example, the lidar is shut down in the morning twilight if the signal is too low for 30 min or if the secondary laser cooling water does not flow for 10 s. The operator can always take over and override the automated system checks.

Some critical parts of the lidar receive redundant control by the particular sub-software, independent of the KLAUS script. For example, a rain detector which is connected to the hatch control sends a direct signal to close the hatch in case of rain. The laser firmware stops the laser automatically if the primary cooling water flow or laser head temperatures are out of limits. The KLAUS script reacts to such events with a countdown of, e.g., the signal timer and initiates the shutdown of the whole lidar after a defined period. The operator is always notified about the state of the systems and the shutdown, mainly because of safety concerns: a Class IV laser operating in a shared lab or office building and propagating into unrestricted airspace. Notifications about shutdown (“regular” because of persistent cloud coverage or “irregular” because of technical problems) are automatically sent to the operator by a Telegram messenger bot included in the KLAUS Python script. Further redundant safety measures cover some interlock software that shuts down the power laser, closes the hatch of the telescope room, and notifies the operator automatically in case of a lost connection to the KLAUS script (watchdog).

While KLAUS is a flexible Python script for lidar control, it does not comprise a graphical user interface (GUI). KLAUS solely receives and sends information via either MQTT or, in special cases, TCPIP socket connections. The GUI is developed with the open-source development tool Node-RED. This browser-based tool can run on any PC connected to the same MQTT broker but is for practical reasons installed on the same PC as KLAUS. The GUI offers a comfortable interface between the human operator and the KLAUS automation software (see Fig. ). The Node-RED front end receives status information as well as text messages and displays them in the GUI. Different buttons in the browser display send commands for, e.g., stepwise or complete, start or shutdown of the lidar. All information between KLAUS and the Node-RED front end is exchanged via the MQTT broker. The GUI additionally displays information about the lidar signal quality that is received via MQTT from the LISA counting system. Furthermore, different webcam photos provide good visual feedback to the operator. Overall, the KLAUS script allows either a manual, GUI-guided operation or a semiautomated operation, where the script takes over control after starting the lidar manually.

Some other subsystems are worth describing that belong to the lidar operation package. As mentioned above, these make use of different coding languages, operating systems (Linux and Microsoft Windows), and hardware platforms (e.g., Raspberry Pi computers). A so-called laser warning system (LWS) informs the personnel in the lab or office building about the laser operation via displays on all doors with laser access. At the same time, these doors are watched by the LWS and the laser shutter is automatically closed for safety reasons if one of the doors is opened. A receiver for Automatic Dependent Surveillance-Broadcast (ADS-B) airplane transponder signals is included in the LWS. The laser shutter is closed if an aircraft gets close to the beam in the sky and opened again afterwards. Because ADS-B is not mandatory for aircraft, four modified Furuno M1835 radars watch the airspace above the institute in different directions. In the case of air traffic within an angle of ∼ 20° around the laser beams, the laser shutter is automatically closed. Besides these systems for laser safety, we installed some hardware for housekeeping purposes (monitoring of temperatures of air, water, laser housing, and cooling water flow). The electric roof was installed in 1996 without a computer interface. We developed a computer interface for remote operation to include it in our KLAUS concept. As mentioned above, a rain sensor acts directly on the hatch controller and triggers the hatch to close in case of precipitation.

The daylight-capable temperature lidar (RMR-T) has used a previous version of the KLAUS software since the autumn of 2019. Recently, in May 2023, we updated the RMR-T lidar with a new laser, laser controller, and APD controller. Both lidars now have about the same technical level. Therefore, we can now operate RMR-T with the same KLAUS version as RMR-WT. Both lidars can be operated in parallel or independently.