Balance between detail and map size

[conor.coady]

WAsP is limited to 10 roughness changes. The recommended size of a roughness map however, I believe is 20km by 20 km. In any case, with online sources of roughness, we often end up using polygons for things like forestry or water bodies, inside a background roughness. Polygons next to one another, which is the more accessible form of online mapping, gives twice as many roughness changes as single lines of roughness change (as intended in WAsP). In short with any sort of detail 10 roughness changes are arrived at very quickly and often quite near to the wind farm we are modelling. Are there any plans to increase the number of roughness changes WAsP can model, or are we better off using a smaller roughness map with the added detail. What is the guidance and where is the balance? 

[Rogier]

These are some good questions and the exact behaviour depends also on WAsP version, as the roughness model has had some changes (mostly version Release 2022-A-1 as described here: https://www.wasp.dk/download/wasp12_releasenotes)

If you are using WAsP from that version onwards it generates a zooming grid as a first step of processing a vector map. This is called spider grid analysis in this paper:

https://wes.copernicus.org/articles/6/1379/2021/wes-6-1379-2021.html

you can read more about how it works. It is probably most useful to look at Fig 2 there, which shows the spider grid analysis with the roughness lengths in all cells (2b) and the reduced number of roughness changes that is indeed 10 as you describe (2c). This is just to say that if there is a very small gap between polygons it will be aggregated anyway within a cell and it will not necessarily count as a change, because first a log averaged z0 weighted by landcover class areas in a cell is calculated.

I copy the relevant part about the roughness reduction algorithm from this paper:

https://wes.copernicus.org/articles/3/353/2018/wes-3-353-2018.pdf

Quote

For computational efficiency, WAsP filters the roughness changes to include only those that have the most significant impact on the wind speed. This is carried out by creating a distance weighted roughness length (ln(z0w)), which is found by multiplying log-transformed roughness values with the exponentially weighted distance from the mast to the roughness changes xk, xˆk = xd 1 − exp −xk xd , (4) where xd is a decay length, which is currently set to 10 km. Then the n items in the array with all transformed roughnesses and distances are found by fitting a step function that stops when either the maximum number of allowed steps nmax is reached or the residual variance RMSmax is below a specified threshold. More details about the algorithm can be found in Sect. 8.3 in Troen and Petersen (1989). The default values for nmax and RMSmax are 10 and 0.3, respectively.

So in summary the both the distance to the roughness change and the change itself are important for the algorithm that selects the most significant roughness changes. It does not filter from the site outwards and stops when it has found 10 as you seem to think. It looks at all roughness cells simultaneously and also takes into account the distance to change. 

To some extent it can be problematic to have high-resolution maps of forested sites, because the few pixels with low roughness will lower the aggregated roughness quite a lot for WAsP (see Fig. 8 of the last mentioned paper), while in reality these pixels might even increase the effective roughness of a area due to additional form drag. For those cases it might make sense to use a higher roughness length then you would expect to alleviate this problem. But this is a different issue from which you are talking about, I think.

 

[conor.coady]

Thanks Rogier. This certainly shines a light on the subject. It would be great to get more of this info into the WAsP Help doc to better describe the limitations of WAsP, but appreciating the new developments. 

[conor.coady]

Rogier, it would also be good to understand what gaps I can keep between for example forestry polygons, as this is often how they come in parcels of forestry with forestry access tracks between. I also feel the industry in general is using only 0.5 roughness length for forestry, as opposed to a roughness length relating to different tree heights. In some ways this also relates and depends on what displacement height is used? What is your view on these two approaches?

[Rogier]

I think you can just add the gaps as you want, the significance algorithm is quite robust. Regarding roughness for forest I would for sure not use a standard roughness of 0.5 m. That is generally too low in Europe, where a typical forest has h>10. The rule with z0=h*0.1 will give you usually a much better estimation. The displacement height is automatically calculated since WAsP Release 2022-A-1, if you specifiy them in a .gml file (you can use treeheight *(2/3), see paper I linked before).

There is a video about using displacement heights and creating a GML file here:

https://panopto.dtu.dk/Panopto/Pages/Viewer.aspx?id=0394a399-d5c2-4a71-9954-b0d800dbcb7f