The Hornamossen diurnal-cycle benchmark for flow modeling in forested and moderately complex terrain
The test challenges the models to predict the wind field during a full diurnal cycle using mesoscale input data versus traditional methods based on onsite measurements and idealized boundary conditions.
Status
The benchmark is now open for anyone to participate in. The final date for submitting results is 31:st of January 2019.
Background
The benchmark was initiated within the NEWA (New European Wind Atlas) project to provide a test for model accuracy in complex topography and heterogeneous forests. The choice of challenging the modellers to estimate the wind field during a full diurnal cycle comes from the increasing importance of stratification as turbines grow larger, and besides covering a wide range of atmospheric stabilities, diurnal cycles also include some natural transient effects.
Managed By
Stefan Ivanell/Johan Arnqvist
Scope
The benchmark is open to anyone interested in participating, but is specifically aimed at studying models for potential use in the making of the NEWA.
Objectives
The task is to model the flow in complex forested terrain and provide wind-profile estimates at a number of validation locations. The motivation is to study how well models can capture the differences in wind profiles in different locations of a typical wind farm, during different atmospheric conditions. Hence, the objective is very similar to the objective of a normal wind resource siting process.
There are two cases, one with easterly and one with westerly flow. Both cases cover diurnal cycles in relatively clear sky conditions and hence include both stable, unstable and neutral atmospheric stratification as well as transitions between different conditions. Validation is made both on the diurnal average as well as in different stratifications. The cases have been selected based on the criteria of stationary geostrophic wind speed and direction in order to simplify the modeling process as well as barotropic conditions in order to simplify the evaluation of model performance.
The feat of modelling a full diurnal cycle is challenging, so the benchmark will also accept submissions that are only modelling stationary conditions, neutrally stratified or otherwise.
Data accessibility
The benchmark is open to anyone willing to participate. The results of the benchmark will be compiled into a scientific paper in which the models will be shown by their name.
Input data
The surface data consists of a 10x10 m resolution data set, covering 40x40 km and containing elevation height, forest height and Plant Area Density (PAD) provided in netCDF format. The vertical resolution of PAD is 1 m. The vertically integrated PAD, the PAI, is also included for simplicity. A flag for water surface is also included, with the value 1 for water and 0 otherwise.
Input data for the boundary conditions of the models is provided in netCDF format and comes from a 3x3 km resolution run by the mesoscale model WRF based on a similar configuration used for the production of the NEWA wind atlas. These include wind speed, heat fluxes, temperatures, tendencies, geostrophic wind speed and boundary layer height for 72 hours. Since the validation cases are relatively stationary, it is sufficient to provide output data during 24 h.
The coordinate system of the input data is in SWEREF99TM, which is a coordinate format with longitude, latitude and latitude measured as distance in meter from a reference point.
The input data can be downloaded from
https://b2drop.eudat.eu/s/q3ewCwpiroFMGKd
Validation data
Velocity and TKE values will be normalized with respect to the tower measurement at 100 m. The validation dataset includes measurements during four full diurnal cycles in each case. The use of validation measurements from several diurnal cycles enables estimation of the confidence of the measurements as well as it provides some redundancy for instrument down time.
Model runs
There are two cases. Winds coming from the east, and winds coming from the west. Both runs should cover time series from at least one full diurnal cycle. If that is not possible to calculate, stationary runs are also interesting, such as a weighted average between runs in different atmospheric stratifications or, in the simplest case, one stationary run for a single atmospheric state. In that case, suitable geostrophic wind speeds and directions are 12.2 m/s, 115° (east case) and 11.7 m/s, 264° (west case)
Following the conclusions of Ivanell et al (2018) the following constant values should be used 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
Output data
For efficient handling of the benchmark it is preferable that all the results are submitted in a standardized format. Hence, in the folder for downloading the input files there is a matlab script that shows how to create the output variables data set in netCDF, and there is also an example netCDF file that you can examine for the structure, should you not be using matlab.
Output quantities and dimensions:
- dimensions: time (t), position (x,y,z)
- time-height: U, V velocity components, , where U is the east-west component, positive eastward, V is the south-north component, positive northwards, turbulent kinetic energy (TKE), and if possible potential temperature (Th).
- Planes at 100,150 and 200 m above local ground height of wind components U and V, as well as TKE. These planes should cover as many of the validation points as possible, which in most cases is expected to be more or less the size of the inner modelling domain. The planes should contain one hour block averages starting and ending with the regular clock hours.
If other boundary conditions are used than the ones provided by the input, please also provide these.
To homogenize the output data please consider these indications:
- time series should be provided based on 10-min or 1-hr block averages
- Vertical profiles should be provided in the whole boundary layer.
- Please follow the same format as the input .nc file
- Naming convention for variables can be found at http://cfconventions.org/Data/cf-standard-names/50/build/cf-standard-name-table.html and also in the input files.
After the simulations has been performed, please upload the data to
https://b2drop.eudat.eu/s/EiSYyx2bPAdLqNo
You will not be able to see or download any other files in that folder, it is simply a space for uploading the results.
Remarks
The east case starts at 15:00 local time, whereas the west case starts 00:00 local time. Both cases does however include input data for 72 hours.
The meso scale model results have been compared to the measurements of the campaign to ensure that they agree to such an extent that permits validation of the model results by the measurements. The reason why measurements are not directly given as input data is the desire to test the entire NEWA modelling chain. As long as the wind directions, wind speeds, and fluxes agree reasonably between the measurements and the meso scale model, the differences by using either as input data has been deemed minor. Due to the blind nature of the benchmark a detailed comparison of the meso scale model data and the measurement will only be published after the benchmark has closed.
References
Sogachev, A., Kelly, M. & Leclerc, M.Y. (2012). Consistent Two-Equation Closure Modelling for Atmospheric Research: Buoyancy and Vegetation Implementations. Boundary-Layer Meteorol (2012) 145: 307. https://doi.org/10.1007/s10546-012-9726-5
Ivanell, S. and Arnqvist, J. and Avila, M. and Cavar, D. and Chavez-Arroyo, R. A. and Olivares-Espinosa, H. and Peralta, C. and Adib, J. and Witha, B. (2018). Microscale model comparison (benchmark) at the moderate complex forested site Ryningsnäs. Wind Energy Science Discussions 2018 DOI: 10.5194/wes-2018–20
