Strategically placed Doppler lidars (DLs) offer insights into flow processes that are not observable with meteorological towers. For this study we use intersecting range height indicator (RHI) scans of scanning DLs to create four virtual towers. The measurements were performed during the Perdigão experiment, which set out to study atmospheric flows in complex terrain and to collect a high-quality dataset for the validation of meso- and microscale models. Here we focus on a period of 6 weeks from 1 May 2017 through 15 June 2017. During this Intensive Observation Period (IOP) data of six intersecting RHI scans are used to calculate wind speeds at four virtual towers located along the valley at Perdigão with a temporal resolution of 15 min. While meteorological towers were only up to 100 m tall, the virtual towers cover heights from 50 to 600 m above the valley floor. Thus, they give additional insights into the complex interactions between the flow inside the valley and higher up across the ridges. Along with the wind speed and direction, uncertainties of the virtual-tower retrieval were analyzed. A case study of a nighttime stable boundary layer flow with wave features in the valley is presented to illustrate the usefulness of the virtual towers in analyzing the spatially complex flow over the ridges during the Perdigão campaign. This study shows that, despite having uncoordinated scans, the retrieved virtual towers add value in observing flow in and above the valley. Additionally, the results show the virtual towers can more accurately capture the flow in areas where the assumptions for more traditional DL scan strategies break down.
Scanning Doppler lidar (DL) systems have proven to be useful in many different sectors of atmospheric study. They have been used for boundary layer meteorology
By applying assumptions to the flow, one can use these different scan strategies to derive the two-dimensional (2D) and three-dimensional (3D) wind from a single DL. The simplest techniques to derive the wind speed and direction are the velocity azimuth display (VAD) technique
In complex terrain, retrievals using multiple DLs with beams that intersect in the same volume of air may be needed to accurately measure the 3D wind field. By combining multiple different radial wind vectors, one can solve the 3D transformation matrix to get the 3D wind vector without applying any assumptions to the flow. There are multiple ways that this can be done. For example, coplanar RHI scans were used in the Terrain-induced Rotor Experiment to study rotors caused by mountains
Multi-Doppler measurements can augment more traditional observation strategies and are highly adaptable to an experiment's science objectives, but they are not perfect. Though some precision is lost due to volumetric averaging, the results are generally accurate to within 0.2 m s
This study assesses how well traditional methods and 2D and 3D multi-Doppler measurements perform in a complex setting. Additionally, it provides an analysis of terrain-induced uncertainties compared to both multi-Doppler measurements and traditional single-Doppler wind estimation techniques. One challenge was however the fact that all of the techniques applied have limitations, and a true reference dataset for quantifying the measurement errors did not exist. This is further addressed in Sect.
The data presented in this study were collected during the Perdigão field campaign during the spring and early summer of 2017, which is one of multiple experiments conducted in order to build the New European Wind Atlas (NEWA)
To validate numerical models, detailed measurements of the flow at multiple scales are required
Map of DLs, meteorological towers, and virtual towers at the Perdigão site. The color fill corresponds to the height of the terrain above sea level, red stars are virtual-tower locations, blue triangles are DL locations, and green diamonds are instrumented meteorological towers. Tower 9 is a 60
The Perdigão experiment is one of the measurement campaigns that took place for NEWA
Additionally, there is a single 2
Leading up to the Intense Observation Period (IOP) in the spring of 2017, a meteorological tower had been operating on the SW ridge for a few years. This was used to construct a climatology of the wind directions over the ridge. According to the climatology, winds often were found to be directed perpendicular to the two ridges
For this study, multiple different DL configurations were jointly analyzed to retrieve the 3D wind vector in the valley. In total, six DLs from three different institutions were combined to retrieve virtual towers multiple times per hour. Figure
Characteristics of the RHI scans performed by each DL used in this study and the virtual towers they were used for.
The University of Oklahoma (OU) deployed the Collaborative Lower Atmospheric Mobile Profiling System
The University of Colorado (CU) operated a Leosphere V1 Windcube Profiling DL at the Orange Grove site, colocated with the CLAMPS DL. The CU DL measures the wind speed and direction using the DBS technique
The German Aerospace Center (DLR) contributed three Leosphere Windcube 200S scanning DLs upgraded with the Technical University of Denmark’s (DTU) WindScanner software. Two of these DLs performed continuous RHI scans in the cross-valley direction, which resulted in an RHI approximately every 30 s. One of these DLs was located up on top of the NE ridge (DLR no. 1) and performed RHI scans to the SW, capturing one horizontal component of the wind. The other DL (DLR no. 2) was located on the slope of the NE ridge and performed RHIs to the SW as well.
In addition to the DLR DLs, DTU operated eight DLs of the same kind on top of the ridges. Six of these were configured to do coplanar scans inside the valley so the horizontal wind in the plane and the vertical velocity could be retrieved. These DLs also operated in a continuous scan mode and produced a new RHI every 24 s.
The RHIs from DLR no. 1 and DLR no. 2 overlapped in a coplanar fashion, so by combining these DLs with the CLAMPS DL scans, it is possible to retrieve the three-dimensional wind field in the form of a virtual tower where the three planes intersect. Due to its positioning, DLR no. 2 was able to capture more of the vertical component of the wind in the location of the virtual tower, which allowed the retrieval of the three-dimensional wind vector.
Regarding the DTU DLs, only the DLs from each coplanar cross section that reached deeper into the valley were used. This resulted in three possible 2D horizontal wind retrievals using WS2, WS5, and WS6. The 2D retrievals assume there is no vertical velocity and thus only provide the along- and cross-valley wind components.
In total, four virtual towers distributed along the valley are retrieved every 15 min when the CLAMPS DL performed its along-valley RHI. The virtual towers typically cover heights from 50 to 600 m above the valley floor depending on the minimum and maximum height of RHI intersection. It should be noted that different to previous virtual-tower retrievals
When combining data from the uncoordinated DL scans in the virtual-tower retrievals, several analysis steps were necessary. First, data from each DL were converted to a common coordinate system. Once the
Temporal resolution of the virtual towers is limited to the time resolution of the CLAMPS RHI. Since the scans were not coordinated to sample the same volume of space simultaneously, a time window needed to be determined.
Possible uncertainties contained in the retrieval are analyzed using an idealized scheme derived from the methods discussed in
Uncertainty from the various virtual towers. Solid lines are uncertainty in along-valley direction, dashed lines are uncertainty in the cross-valley direction, and dotted lines are uncertainty in
Since we had only one site where all three components of the wind vector could be retrieved but had multiple 2D virtual-tower sites where the horizontal wind vectors were retrieved assuming that the vertical wind velocity is negligible, it is important to quantify the errors introduced in the 2D retrievals. This is particularly important in complex terrain where large vertical velocities are often present. To do this, a theoretical idealized setup was constructed (Fig.
Idealized setup to examine the errors associated with neglecting the vertical velocity and only producing 2D virtual towers. The setup is meant to closely mimic the positions of the CLAMPS DL, DLR no. 1, and DLR no. 2 used in the 3D virtual tower. DL3 is varied along
Next, we selected different 3D wind speeds and directions at the intersection of all the DL beams and calculated the radial velocities that are observed by each DL by rearranging Eq. (
Figure
Wind speed and direction errors associated with neglecting the vertical velocity in 2D virtual towers. Results are for
As mentioned in Sect.
We focused the analysis on three cases: first we selected a case with quasi-2D flow and small vertical velocities, next we chose a quasi-2D flow case with large vertical velocities, and last we selected a fully 3D case with complex flow interactions that are associated with strong variability of the flow along the valley. The selected days from the Perdigão IOP were subjectively identified based on the analysis of DL scans and stability profiles. Additionally, data availability was taken into account.
For the first case, we would expect that the VTs and single-Doppler retrievals agree well, even though the sites vary along the valley. We argue that in this case, the observed differences between the retrieval methods are primarily caused by measurement errors and uncertainties in the retrieval methods and less by flow variability; i.e., the comparison allows us to assess the measurement errors introduced by different retrieval approaches. The second case, with stronger vertical motions while still being quasi-2D, allowed us to analyze the amount of error vertical motion introduces into the 2D retrievals for a realistic scenario. The observed differences can then also be compared with the results from the systematic study described in Sect.
Time–height plots of three nights observed by the CLAMPS system. The top row
For our simplest case, we selected a day with wind speeds that exceeded 7–10 m s
During this period, a shortwave trough was off the coast of Portugal. Winds at 500 mbar were approximately 10 m s
By starting with a relatively simple case, it is possible to directly compare the retrieval methods with minimal violations to the underlying assumptions. Unfortunately, VT4 and VT1 did not meet the retrieval criteria because the CLAMPS DL became slightly out of sync with WS5 and WS6, and the time difference between scans did not meet the retrieval criteria. However, for this application, having VT2 and VT3 will suffice since they are the closest to each other spatially (Fig.
Profiles of horizontal wind speed
Similar to the previous case, we again targeted a time period with wind speeds that exceeded 7–10 m s
For this case, consistent differences between the 2D and 3D retrievals can be noted where there are vertical velocities present in the profile. Based on Fig.
Time series of wind speeds
Same as Fig.
While the two previous cases were selected with the intent of assessing the retrieval methods by limiting the analysis to quasi-2D flows, the final case was chosen to illustrate how useful the virtual towers can be for measuring the spatial variability of features along the valley. A moderately complex case, 8 May 2017, was selected. During this time, there was a low-pressure system off the coast of the Iberian Peninsula and a ridge over Perdigão. Aloft, winds over the IOP site were from the southwest at 15 m s
Similar to Fig.
The complexity of the flow for 7–8 May is shown in Fig.
As mentioned previously, the traditional DL technique for retrieving the wind speed and direction with a single DL is the VAD or the DBS technique. However, in the overnight hours of the Perdigão campaign, the assumption of horizontal homogeneity is often violated due to flow phenomena created by the terrain. This is expected to be particularly prominent at the lowest levels of the VAD or DBS profiles near the surface. We will now examine data from 8 May (Sect.
Scatter plots of the wind speeds
Figure
Around 100
The vertical velocities from VT3 and the CLAMPS DL vertical stare also agree within the uncertainty VT3 above ridge height. Around the same level as the wind speeds and directions, the stare and the 3D virtual tower start to diverge. The differences could be due to each measurement representing a different time. The tower data are 5 min averages, while the CLAMPS DL stare data and the CLAMPS/DLR virtual tower are instantaneous measurements at slightly different locations, so some differences are to be expected.
Breakdowns in the single-Doppler DL retrievals can be observed by closely examining Figs.
In summary, Fig.
Same as Fig.
One of the main reasons Vale do Cobrão was chosen for the experiment was the quasi-two-dimensional nature of the ridges; they were thought to be the best way to represents a series of periodic rolling hills and that flow perpendicular to the ridges could also be considered quasi-2D. While this assumption is valid during some nights with higher wind speeds, periods with lower wind speeds tend to have more spatial heterogeneity. In order to visualize this heterogeneity, it is best to think of wind components in the cross-valley and along-valley sense.
Analysis of the virtual towers coupled with the cross-valley RHIs from the DLR and DTU DLs often shows that the flow can not be considered entirely 2D, especially at lower wind speeds. In Fig.
Looking at the along-valley component of the flow, winds are very calm and near zero within the valley. There does appear to be a weak jet just above the ridge at 550
Analysis of a different time period during the night of 8 May further shows the complexity of the flows in the terrain. During this night, winds veered rather quickly aloft and weakened closer to the surface (Fig.
Multi-Doppler analyses are a useful tool for understanding and quantifying wind characteristics in complex terrain. Though the scans used for the virtual towers were uncoordinated, they can be useful for diagnosing flow conditions in and above the valley. The virtual towers help fill the gap in wind speed measurements inside the valley above the height of the physical towers (100
Though the virtual towers are well suited to study the complex flows observed during the IOP, they are not without limitations. The uncertainty in the radial velocities needed to be propagated through the retrieval. Due to the positioning of the DLs used for the virtual towers, this meant that uncertainty in the horizontal wind retrieval was larger with increased height. However, vertical velocity retrievals on the single 3D virtual tower became more certain with height as a larger component of the vertical velocity was observed in the radial velocities at the higher elevation angles. Additionally, the 2D virtual towers made the assumption that there was no vertical velocity, which is often violated in this terrain. Due to this, they are prone to errors, particularly in wind direction. For example, with vertical velocities of 1 m s
There are some uncertainties in the virtual towers that can not be accounted for. Since the scans were uncoordinated, the beams very rarely were observing the same point at the same time. For this reason, cases with less steady flow may have more errors. The cases chosen for this study were more steady and quasi-2D than many nights during the campaign, which made them ideal for testing the different profiling techniques. Other differences may be caused by flow inhomogeneities that may occur along the valley. Though we tried to minimize this through careful case selection, the flow was still complex, and the assumption of quasi-2D during the cases presented in Sect.
In terms of wind direction, despite the uncertainties in the 2D retrievals, the virtual towers agree better with the meteorological towers situated inside the valley than the VAD/DBS scans at those levels. However, in the analyzed cases wind speeds at these levels were quite small, so a more detailed intercomparison between all the different methods of wind estimation is needed. This could be done with a DL simulator that is able to use fine-scale model output from the valley to mimic DL radial velocities and noise. These could then be fed into the virtual-tower retrieval and compared directly to the model output.
Combining various types of profiler data with meteorological towers on the ridge allows wind speeds to be sampled as cross-valley flow enters, goes through, and exits the valley and gives a more complete picture of the spatial evolution of features in complex terrain. They also allow a more comprehensive validation dataset for numerical models in the future. While simulated RHI scans from numerical models can be compared to the observed RHIs, only the combination of multiple DLs allows the retrieval of the 3D wind vector. These data could one day be fed into a model-based wind retrieval using advanced data assimilation methods to estimate the full 3D wind field in and around the valley to gain more insight into the governing physics of the flow.
Data of all instruments that were used in this study are stored on three mirrored servers owned by DTU, University of Porto, and the NCAR Earth Observing Laboratory (EOL), respectively. The data are publicly available through dedicated web portals of the University of Porto (
TMB, PK, NW, and RM were in charge of the conceptualization and methodology; TMB performed the formal analysis; PK, NW, and RM were in charge of the resources; TMB wrote the original draft; PK, NW, and RM reviewed and edited the original draft; TMB provided the visualizations; PK provided supervision; PK and NW acquired funding.
The authors declare that they have no conflict of interest.
This article is part of the special issue “Flow in complex terrain: the Perdigão campaigns (ACP/WES/AMT inter-journal SI)”. It is not associated with a conference.
We would like to thank José Palma and José Caros Matos for their work to make this experiment a success. We would also like to thank Matt Carney and Edward Creegan for their hard work in getting CLAMPS running in Perdigão despite many initial challenges, and we thank the entire Perdigão team for their collaborative spirit before, during, and after the campaign. Lastly, we appreciate the hospitality of the people in the municipality of Alvaiade.
This research has been supported by the Division of Atmospheric and Geospace Sciences (grant no. 1565539) and the Federal Ministry of Economy and Energy on the basis of a resolution of the German Bundestag under the contract numbers 0325518 and 0325936A.
This paper was edited by Jose Laginha Palma and reviewed by two anonymous referees.