the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
06 Apr 2021
06 Apr 2021
Satellite-based estimation of roughness lengths and displacement heights for wind resource modelling
- 1Technical University of Denmark, DTU Wind Energy, Risø Campus, Frederiksborgvej 399, 4000 Roskilde, Denmark
- 2DHI GRAS A/S, Agern Alle 5, 2970 Hørsholm, Denmark
- 3Siemens Gamesa Renewable Energy A/S, Borupvej 16, 7330 Brande, Denmark
- 1Technical University of Denmark, DTU Wind Energy, Risø Campus, Frederiksborgvej 399, 4000 Roskilde, Denmark
- 2DHI GRAS A/S, Agern Alle 5, 2970 Hørsholm, Denmark
- 3Siemens Gamesa Renewable Energy A/S, Borupvej 16, 7330 Brande, Denmark
Abstract. Wind turbines in northern Europe are frequently placed in forests, which sets new wind resource modelling requirements. Accurate mapping of the land surface can be challenging at forested sites due to sudden transitions between patches with very different aerodynamic properties, e.g. tall trees, clearings, and lakes. Tree growth and deforestation can lead to temporal changes of the forest. Global or pan-European land cover data sets fail to resolve these forest properties, aerial lidar campaigns are costly and infrequent, and hand-digitization is labour-intensive and subjective. Here, we investigate the potential of using satellite observations to characterise the land surface in connection with wind energy flow modelling using the Wind Atlas Analysis and Application Program (WAsP). Collocated maps of the land cover, tree height, and Leaf Area Index (LAI) have been generated based on observations from the Sentinel-1 and -2 missions combined with the Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2). Three different forest canopy models are applied to convert these maps to roughness lengths and displacement heights. We introduce a modified model, which can process detailed land cover maps containing both roughness lengths and displacement heights. Extensive validation is carried out through cross-prediction analyses at ten well-instrumented sites in various landscapes. We demonstrate that using the novel satellite-based input maps leads to lower cross-prediction errors of the wind power density than land cover databases at a coarser spatial resolution. Differences in the cross-predictions resulting from the three different canopy models are minor. The satellite-based maps show cross-prediction errors close to those obtained from aerial lidar scans and hand-digitised maps. This demonstrates the value of using detailed satellite-based land cover maps for micro-scale flow modelling.
Rogier Floors et al.
Status: final response (author comments only)
-
RC1: 'Comment on wes-2021-28', Anonymous Referee #1, 25 May 2021
The comment was uploaded in the form of a supplement: https://wes.copernicus.org/preprints/wes-2021-28/wes-2021-28-RC1-supplement.pdf
-
RC2: 'Comment on wes-2021-28', Anonymous Referee #2, 13 Jun 2021
The authors investigate the ability of satellites to provide roughness length and displacement heights for wind resource assessment. They consider 3 satellite platforms which provide tree height, land cover, and leaf area index. Then, they use 3 forest modules to convert these data to roughness and displacement. They consider these satellite-derived estimates, along with other traditional ways (global land-cover maps; aerial lidar scans; manual digitalization) of obtaining these quantities, and perform power predictions at various sites using WASP. They take each sensor on a met tower and use it to predict the wind speed and power at all other sensors on that tower (vertical predictions) and other towers at the same site (horizontal predictions). A total of 10 sites are considered.
The main result is that WASP-derived power predictions obtained with satellite vs lidar land characteristics were comparable (similar error of ~10-11%), indicating that satellite data can be a more affordable alternative to costly aerial lidar campaigns over forested terrain.
Major comments
The writing quality is somewhat poor. Subsection 5.3 and Conclusion are the most well-written ones in the manuscript. The rest is certainly readable, but it takes extra effort from the reader to sort through ambiguities and incoherencies. I would expect to be able to focus on the science and results being presented while performing a review, but the writing made it difficult and I spent a substantial amount of time just trying to understand the content that was being presented. Each paragraph should have a clear message and a reason to exist. Figures should stand alone with well-labeled axes and captions, and always show the units. The reader shouldn’t have to read the manuscript in detail to understand what the figure is trying to show. Especially since there are so many acronyms for the datasets/models and foreign names for the site locations, it can become difficult to keep track of what’s what if the text is not easy to follow.
The results presented implicitly include a validation of the WASP models themselves. I suggest that the analysis be reframed to focus exclusively on the effects of what the authors are actually testing: several ways of obtaining z_0 and d. Either that, or the narrative of the paper should be expanded to include the WASP validation that is being carried out (even if WASP is being employed in a less traditional way, using only one sensor, without the ability to fit to the entire mast profile).
It feels like 90% of my time was spent reading the introduction, data, and methodology (19 pages) and then not many results were presented (less than 3 pages). Maybe that's because some results are shown in the methodology? Either way, please consider including some of the results that are not shown (as per manuscript text) and more examples that show spatial variations. It would help the interpretation of the results if the readers could see for themselves the spatial distribution of z_0 and/or d at all sites, which could be accomplished by adding a single figure with subpanels. It might also be helpful to spatially see the progression of data from satellite-derived quantities (h, land cover, lai) to z_0 and d for the best-performing and worst-performing sites. Figure 7 is also lacking discussion of some key points: e.g. I don't understand why the hand-digitized map is so bad for one of the MX sites; I don't understand why the proposed method performs worse than MODIS/GLOB/CORINE for Østerild and the two MX sites, and it's not discussed. It seems like Figure 7 can be modified and milked for a much more valuable discussion (and shown more than once, highlighting different aspects of the results, like e.g. vertical vs horizontal predictions). A key, unanswered question is: in what types of sites should one go through the trouble of deriving z_0 and d from satellites vs using the global datasets?
I would like to see the authors using WASP as they say it's usually used (page 25, lines 483-487) for the sites with two masts: Alaiz, Finland, Perdigao, Sweden, Østerild to get the best prediction possible and then evaluate the effect of different z_0 and d estimates. By using WASP the way it's usually used, won't you get results that are more interesting to your audience, and be able to tease out the real sensitivity to these lower-boundary forcing parameters?
Minor comments
A very large amount of minor comments is annotated directly onto the pdf
-
RC3: 'Comment on wes-2021-28', Anonymous Referee #3, 14 Jun 2021
The authors here present a way to represent wind resource maps using new satellite products from the Sentinel-1 and -2 and the ICESat-2 satellites. They attempt to demonstrate how these new products can be comparable, or even better than, existing best practices such as aerial lidar scan and hand-digitized maps. A large emphasis is placed on the land cover datasets and the forest roughness models used in conjunction with micro-scale flow modeling with WAsP. Ultimately, a cross-prediction is performed against a variety of sites with different roughness/forest characteristics is shown, with the new satellite products outperforming the coarser land cover datasets. Finally, it is emphasized that these new satellite products are not only more accurate for wind resource assessment, but quicker to perform and are continually updated.
In general, the authors do an okay job of presenting their work. Overall, though, it does seem that the delivery of the results is a bit muffled. For example, a great deal of the paper is spent laying out the methods of the work and introducing the different datasets/models. After this, very few pages are concerned with the results. The opportunity does present itself to show more results (especially regarding sites where WaSP performs poorly).
Specific Comments
- There are a lot of acronyms. I realized this can be hard to avoid, but I found myself constantly jumping around to recall which one meant what. This is done somewhat in Table 3. A simple table in the Appendix would prove beneficial here.
- Time frames are generally not explained in the figures, such as Figure 6. The satellite time coverage is laid out in Table 3, but it is lost in the text and not explained in the figure captions.
- Line 158: A space is needed in between “in” and “mountainous”.
- Line 182: What are the consequences of assuming that each subclass of z0 is the same as the class it inherits?
- Line 190-200: I feel this discussion of these machine learning models (Random Forest and Support Vector Regression) came out of nowhere and hardly any attention is given to the specificities of these. Please elaborate the discussion around these and why they were used in the first place.
- Line 387 (Results in general): Results are not shown for the WAsP model in complex terrain with steep slopes, since WAsP is known not to perform well in these conditions. Shouldn’t at the very least some of these results be shown? I feel we are missing some of the picture if not.
- I realize atmospheric stability is not considered in this study, but I do think that would be an interesting addition, especially since we are dealing with various forested sites, and the interactions with these sites and different stabilities could tease out further insights not previously considered. I will leave it up to the authors to include this or not.
Overall, I think there is a solid paper in here… somewhere. In essence, a greater balance between the methods and results section is needed. Reading 15-20 pages of methods for only ~5 pages of results/discussion is tough. There are some opportunities to include some more results (as explained above), and I believe this could help round out the paper, along with a restructuring of the methods.
Rogier Floors et al.
Rogier Floors et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 320 | 135 | 13 | 468 | 10 | 6 |
- HTML: 320
- PDF: 135
- XML: 13
- Total: 468
- BibTeX: 10
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1