ALEX17 Diurnal Cycles Benchmark: a large domain in complex terrain

This benchmark proposes a 4-days episode for flow modellers to simulate the interaction of mountain-valley wind with mesoscale forcing and thermal stratification

Status

The benchmark was launched on November 4th 2019 and is open for participation within the IEA-Wind Task 31 Wakebench. We expect to receive your simulations by March 2020 to discuss the results in the next Task 31 annual meeting next to the Torque 2020 conference.

A recording of the kick-off webinar is available below:

The presentation slide-pack (with some corrections) is published in Zenodo

Participants can formally register by filling out this questionnaire

Background

The ALEX17 campaign was the last one of a series of experiments executed in the New European Wind Atlas (NEWA) project. It is based near Pamplona, Spain, with the general objective of characterizing the wind conditions in the vicinity of CENER’s test site in complex terrain at the Alaiz mountain range. Additional background information about the campaign and associated references to published data and reports is available in a previous blog post:

The Alaiz Experiment (ALEX17): revealing mountain-valley large scale flow patterns

Managed by

Javier Sanz Rodrigo / Pedro Santos

Data Provider

Pedro Santos / Roberto Chávez Arroyo

Scope and Objectives

This benchmark is intended for flow models that can reproduce wind conditions at microscale, relevant for wind turbine siting and energy yield assessment, with realistic large-scale forcing characterized by mesoscale modelling. Therefore, the benchmark is designed with transient meso-to-micro modelling in mind. Nevertheless, steady-state modellers are also welcomed to participate. To this end, we have identified situations where the flow is quasi-steady at different levels of thermal stratification.

The ALEX17 intensive campaign, where all the instruments were operational, lasted 4 months. During this period numerous interesting phenomena were captured like gravity waves, hydraulic jumps, etc., but we would like to focus first on testing models in a simpler environment to make sure models can deal with more fundamental challenges, such as:

• How to set up an efficient meshing strategy for a large microscale domain while capturing all relevant topographic features;
• How to deal with surface boundary conditions in the presence of thermal heating/cooling, roughness changes and heterogeneous forest canopies;
• How to parameterize turbulence models;
• How to couple mesoscale and microscale.

Test Case

Figure 1 shows the elevation map and the layout of the ALEX17 instrumentation and Z-transect positions that will be used in the benchmark. The 10-km long Z-transect was built from 2D (red lines )and 3D (blue lines) Doppler scanning lidar measurements at a constant height of 125 m a.g.l.

Figure 1: ALEX17 Elevation map with instruments and Z-transect positions that take part in the benchmark. WS — WindScanners; M (green) — 80m cup masts; M (orange) — 80m sonic masts; MP —118m reference mast at test site. Measurements of dual-doppler (red lines) and triple-doppler (blue line).

The proposed 96h period spans from Sep 30th 2018 00Z to Oct4th 2018 00Z

The selected episode corresponds to a period when the synoptic conditions were relatively uniform to produce northerly winds for four consecutive days. Figure 2 presents the time series of wind speed, direction and temperature at the reference MP5 position. During these days, the wind direction was constant during the entire period at the mountain top, with winds varying from 4 to 15 m/s. The last daily cycle has the strongest thermal stratification, with larger temperature gradients as well as wind shear/veer.

Figure 2: Time series of wind speed, wind direction and temperature at the mountain top (MP5).

The wind regime at the valley floor is illustrated with the time series measured at M7 location (figure 3). It’s clear that valley winds are weaker and have larger variations. At the same location the evolution of net radiation (Rn) and sensible heat flux (Hs) is shown by the bottom plot, measured from an energy balance station. The second daily cycle has stronger and more stationary wind conditions, but is also the one without clear sky conditions.

Figure 3: Time series of wind speed, wind direction, net radiation and sensible heat flux at the valley floor (M7).

The differences in wind regime between the valley floor and mountain top have to be captured by the microscale models

The animation below shows 10min snapshots from two lidar RHI scans at the Transect (blue) Line. Positive radial wind speeds represent the northerly flow, from left to right in the figure. The results highlight two major flow patterns, with periods when windward flow blockage at the mountain is present and others with an additional recirculation zone in the lee-side of the north ridge. The mesoscale (WRF) results don’t show wind turning and recirculation effects measured at the site (see Webinar slides), hence the coupling with the microscale models are necessary.

Figure 4: Animation of RHI Scans along the Transect Line. The colorbar represents radial wind speeds with positive values for notherly winds (flow from left to right).

Recirculation zones, windward flow blockage and Dual-doppler synchronized measurements can be directly compared with model results when lidar scans are available.

The Dual-doppler wind reconstruction of synchronized WindScanner measurements are also available for the selected period. Santos et al. (2019) shows an animation of the wind flow on the mountain top and at the North ridge, which will also be used for validation.

Validation Data

Three sets of measurements will be used for validation of the submitted results. The meteorological mast measurements are going to be readily available for the modelers to check and calibrate their results, whereas the lidar measurements will be used for validation after submission (blind test).

1. Vertical profiles at “mast” positions (readily available to set up your modeling strategy): UTM coordinates of all masts and 10min averages of wind data at MP5, cup masts and sonic masts (NetCDF4).
2. Wind vector (125m a.g.l.) along the “Z-transect” (validation after submission): UTM coordinates of 167 validation points.
3. Vertical RHI plane across the valley along the blue segment of the Z-transect (validation after submission): A Plane that connects WS3 [617846.796; 4732496.246] km and WS5 [617305.590; 4729848.751] km and follow the dimensions provided on RHI scans.

All validation and input data is available inside a B2Drop input folder, to be shared with registered participants

Input Data

The input data for ALEX17 benchmark is divided into static and dynamical datasets, where the former consists of land surface information valid for all the benchmarks, and the latter is the outcome from the mesoscale simulations that can be used as boundary conditions, forcing (or both depending on the modeling procedure) for every benchmark.

• Static Datasets contains orography and land cover models, more specifically:

1. Two raster-type files containing the Digital Elevation and Digital Surface Models (DEM and DSM) data of the ALEX17 domain. The models were built by TRACASA through lidar airborne scans taken during years 2011 and 2012 and updated by photogrammetry with orthophotos taken in year 2014.

Figure 5: ALEX17 DEM (left) and DSM-DEM (right).

1. A raster-type file which contains the approximate annual average aerodynamic roughness length (z0) in meters. The map was created through visual estimation of the roughness length and from Corine Land Cover (CLC) 2006 data. The roughness values derived from the Land Cover data were based on the relation between CLC and the aerodynamic roughness length applied in the Finnish wind atlas
2. The last data source is based on the Shuttle Radar Topography Mission (SRTM) Digital Elevation Model dataset with ~80m resolution covering a larger area (~200x200km) around the experimental domain.

The metadata of the static datasets are published here: https://doi.org/10.11583/DTU.8143775.v1

• Mesoscale Dataset contains results from the mesoscale simulation carried out with the Weather Research and Forecasting (WRF) model for the period of the northern case benchmark. It consists on the WRF fields around the ALEX17 domain that can be relevant to those modellers that plan to simulate the benchmark case via some offline model-chain (mesoscale-microscale coupling) method. The provided netCDF files contain:

1. Virtual time series from WRF innermost domain (d03, 3km), from nearest grid point, at north domain boundary (1.5751ºW, 42.778ºN); at M7 (1.578ºW, 42.725ºN) and at MP5 (1.5578ºW, 42.695ºN).
2. Tendencies, i.e. contributions from different terms of momentum and energy budget from WRF d02 (9km). These fields are spatially averaged in a window of 45x45km for all vertical levels. Hence, it can be considered a single-column input (time, height) but the 3D tendencies can also be provided. The file has the following quantities: Ug, Vg, UADV, VADV, POT_ADV, POT, TSK, PSFC and HFX

The simulation is performed with the WRF model version 3.8.1 (Skamarock et al., 2018), mostly following the configuration derived from the sensitivity analysis carried out by the mesoscale group of the NEWA project (Hahmann et al. 2018). The ERA-5 reanalysis data together with OSTIA SST and GLDAS/NOAH land surface were used as forcing, initial and boundary conditions for the WRF simulation. For the selected period, three nested domains centered at the M7 location were created with horizontal grid spacing of 27(d01), 9(d02) and 3(d03) km; with respectively 110, 100 and 88 grid points laterally, and 61 vertical levels up to a height of 50 hPa at the top. The runs are carried out by running sets of 8-day simulations with 24 hours of spin-up while performing spectral nudging in domain 1(d01). Outputs are stored and provided in 5-min periods. Other details about the WRF configuration can be seen at Chavez-Arroyo et al., 2018.

To improve model intercomparison we recomend all microscale modellers to use these mesoscale inputs. Participants with multiscale models using online coupling can use their own mesoscale model but they should use ERA5 input data to keep the same global input data in all simulations. If needed, the raw wrfout files, WRF namelists and config files can be shared upon request.

Output Data

The selected quantities of interest are the 3D wind vector components (U,V,W), potential temperature (Th) and turbulent kinetic energy (TKE) at the locations and with the formats specified below, with simID representing your simulation identifier (which will be provided).

1. simID_masts.nc: Vertical profiles extending 1km from the surface at the met mast locations (see inputs/masts.csv for UTM coordinates);
2. simID_Ztransect.nc: Vertical profiles along the Z-transect extending 1 km in the vertical (see inputs/Ztransect.csv for UTM coordinates);
3. simID_box.nc: Horizontal planes at 125m and 40m a.g.l. with horizontal resolution of 100 m, of the “box” [612–622; 4726–4736] km.

The temporal resolution must be 10 min averages, stamped at the beginning of the period.

Simulation results must be submitted in the specified format and can follow the example results from the WRF runs provided for guidance.

The example results, from WRF runs, are provided in netCDF4 and uploaded to the outputs folder under the ID “alex17_00a”. A template python script is posted to assist the output data compilation, with appropriate naming convention and metadata.

Results can be submitted via a B2Drop upload-only folder, to be shared with registered participants

When you submit your output data please follow the naming convention and choose a two-digit user ID followed by a letter to differentiate several submissions from the same user.

Additional Remarks

Steady-state microscale modelers: You can target intervals where conditions are more stationary. Here we shall focus on how the ensemble-averaged flow depend on different levels of thermal stratification. The following 2-hour periods (so called “events”) have been selected (Table 1):

• Neutral 2018/10/1 18Z — 2018/10/1 20Z: evening transition after a windy day result in rather vertical wind profile at the hill top with low turbulence intensity. This seems like a canonical case for neutral models to target first.
• Stable 2018/10/2 3Z — 2018/10/2 5Z: Large speed-ups between the valley and Alaiz hilltop develop with 27º wind direction change and moderate wind shear at the hilltop.
• Unstable 2018/10/2 12Z — 2018/10/2 14Z: Windy and high-turbulent conditions develop negative wind shear at Alaiz hilltop.
• Very Stable 2018/10/3 4Z — 2018/10/3 6Z: This is a extreme case that microscale models are not expected to capture, with more than 90º wind direction change between valley and hilltop to develop large speed-ups and wind shear at the hilltop.

Table 1: Wind conditions at M7 (valley) and MP5 (hilltop) for the 4 “quasi-steady” events. Wind speed, direction and turbulence intensity at 80 m, wind shear is the power law between 40 and 80 m and stability is z/L at 10 m.

When submitting your data please use the same format that was defined for transient models and fill in the corresponding timestamps with your data. You may want to run several simulations to account for the variability withing the 2-hour events. During the assessment, we will compute mean and std of the data within each 2-hr event.

Unsteady meso-micro modelers: You are welcome to run the four days of the episode, but if your set up and/or computational resources does not allow for this, you can concentrate on the diurnal cycle from 2018/10/1 18Z to 2018/10/2 18Z to cover the first three events discussed before. Allow a few hours of spin-up before the period to have fully developed turbulence already at the beginning of the cycle.

Participants can focus on a 24h period if computational resources are limited.

All participants can choose to have their submitted results anonymized during the stage of comparison of model runs. Please inform the benchmark manager if so in the sign-up questionnaire.

RANS modelers: Following the conclusions of Ivanell et al (2018) the following constant values should be employed when using the Sogachev et al (2012) turbulence model:

κ = 0.4, Cd=0.2, Cε1 = 1.176, Cε2 = 1.920, σk = 1, σε = 1.238 and Cμ = 0.033

After an initial submission we could try to calibrate turbulence models based on the flux measurements at the valley and/or the hill-top.

Schedule

• 4 November 2019: Kick-off Webinar
• 2020-Q2: Submission of results
• 2020-Q3: Discuss results and draft paper
• 2020-Q4: paper submission (WESC 2021?)

Communication between participants is done through a Slack channel.

Acknowledgements

The ALEX17 experiment was conducted with the support from the NEWA project (FP7-ENERGY.2013.10.1.2, European Commission’s grant agreement number 618122). The benchmark is run under the umbrella of the IEA-Wind Task 31 “Wakebench” with additional support from H2020-Marinet2 to make the benchmark a pilot on the generation of FAIR data using of a virtual research environment.

References

Chavez-Arroyo, R., Irigoyen-Indave, A., Sanz-Rodrigo, J., & Fernandes, P. (2018). Analysis and validation of Weather Research and Forecasting model tendencies for meso-to-microscale modelling of the atmospheric boundary layer. Journal of Physics: Conference Series, (1037), 1–12. doi :10.1088/1742-6596/1037/7/072012

Ivanell, S. Arnqvist, J., Avila, M., Cavar, D., Chavez-Arroyo, R. A., Olivares-Espinosa, H., Peralta, C., Adib, J. and Witha, B. (2018). Microscale model comparison (benchmark) at the moderate complex forested site Ryningsnäs. Wind Energ. Sci., 3, 929–946, doi:10.5194/wes-3-929-2018

Hahmann A, Witha B, Sile T, Dörenkaemper M, Söderberg S, Navarro J, Leroy G, Folch A, Garcia-Bustamante E and Gonzalez-Rouco F 2018 Geophysical Research Abstracts (Vienna, Austria)

Santos, P. Borbón, F. Mann, J. Cantero, E. Vasiljević, N. Sanz Rodrigo, J. … Cuxart, J. (2019, June). Multi scanning lidar measurements for resource assessment: a case study in complex terrain. WESC 2019. doi:10.5281/zenodo.3358598

Sanz Rodrigo, J., Churchfield, M. and Kosović, B. (2017) A methodology for the design and testing of atmospheric boundary layer models for wind energy applications. Wind Energ. Sci. 2: 1-20, doi:10.5194/wes-2-1-2017

Sogachev, A., Kelly, M. and Leclerc, M.Y. (2012). Consistent Two-Equation Closure Modelling for Atmospheric Research: Buoyancy and Vegetation Implementations. Boundary-Layer Meteorol 145: 307-327, doi:10.1007/s10546-012-9726-5

Skamarock W C, Klemp J B, Gill D O, Barker D M, Duda M G, Wang W and Powers J G 2008 A Description of the Advanced Research WRF Version 3 Tech. Rep. June National Center for Atmospheric Research Boulder, CO