How the roughness rose simplifies roughness maps

[doha]

Hi all,

I have been investigating how large roughness maps are processed on WAsP, and have found the following two paper which describe some aspects of how the roughness rose is generated:
https://wes.copernicus.org/articles/6/1379/2021/wes-6-1379-2021.html
https://wes.copernicus.org/articles/3/353/2018/wes-3-353-2018.pdf

I now understand that WAsP simplifies large roughness maps with many roughness changes into a roughness rose, and that each sector of this rose has a maximum of 10 roughness changes by default. I have also observed that a great amount of simplification is involved, as I have personally tested large roughness maps with more than 10 roughness changes in a given sector and seen resulting roughness roses with at most 5 changes in any sector.

I have not found a detailed description of how the roughness changes used on the roughness rose are chosen or how WAsP simplifies or averages several roughness changes on the original roughness map into one roughness value on the corresponding section of the roughness rose. Does such a description exist anywhere?

 

[Rogier]

Hi Doha, the papers you mention are indeed the right references. This paragraph in the paper 1 you mention probably explains it best:

Quote

To account for roughness changes, WAsP calculates sector-wise speedup factors for a certain point. Because the effect of a roughness change on the wind speed in a sector is distance dependent, with nearby areas having a higher impact, the distance to each z₀ area is multiplied with an exponential weighting function as described in Floors et al. (2018). From these weighted values z_0w, the ones that explain most of the variance of z_0w are stored for further processing (up to a maximum of n_max). This is done for computational efficiency, so that equations that take into account the effect of internal boundary layers can be used. These equations are given in Sect. 8.3 of Troen and Petersen (1989). The output of these equations are sector-wise speed-up factors, which are used to remove the effect that microscale features may have on the wind observations. Apart from the speed-up factors, z_0w is also used to compute a geostrophic (sometimes referred to as an effective or mesoscale) roughness length z₀G.

The original zooming grid is just a grid where the grid cells increase in size further away from the site:

Quote

The distance to the first radial segment r₀ in the zooming grid is defined by the user (default r₀=25 m), and each next segment has a grid spacing that is 5 % larger than the previous one

The roughness lengths in these cells are then multiplied with a exponentially weighted distance. From those transformed cell values the most significant ones are found with a simple algorithm where you just loop over all roughness changes until you have explained enough of the initial variance in all transformed roughness changes in a certain sector.

[doha]

Hi Rogier, thanks very much for your response. I am interested in learning more about how the most significant cells are chosen, in order to inform my roughness map creation process. Is this selection process described in more detail anywhere?