the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Modelling wind farm effects in HARMONIE-AROME (cycle 43.2.2) – part 1: Implementation and evaluation
Henrik Vedel
Xiaoli Guo Larsén
Natalie E. Theeuwes
Gregor Giebel
Eigil Kaas
Abstract. With increasing number and proximity of wind farms it becomes crucial to consider wind farm effects (WFE) in the numerical weather prediction (NWP) models used to forecast power production. Furthermore, these WFE are also expected to affect other weather-related parameters at least locally. Thus, we implement the explicit wake parameterization (EWP) in the NWP model HARMONIE-AROME (hereafter HARMONIE) along-side the existing wind farm parameterization (WFP) by Fitch et al. (2012) (FITCH). We evaluate and compare the two WFPs against research flight measurements as well as against similar simulations performed with the Weather Research and Forecasting model (WRF) using case studies. The case studies include a case for WFE above a wind farm as well as two cases for WFE at hub height in the wake of farms. The results show that EWP and FITCH have been correctly implemented in HARMONIE. For the simulated cases, EWP underestimates the WFE on wind speed and strongly underestimates the effect on turbulent kinetic energy (TKE). FITCH agrees better with the observations and WFE on TKE are particularly well captured by HARMONIE-FITCH. After this successful evaluation, simulations with all wind turbines in Europe will be performed with HARMONIE and presented in the second part of this paper series.
- Preprint
(17676 KB) - Metadata XML
-
Supplement
(313 KB) - BibTeX
- EndNote
Jana Fischereit et al.
Status: final response (author comments only)
-
RC1: 'Comment on gmd-2023-63', Anonymous Referee #1, 08 Aug 2023
The article describes an evaluation of an explicit wake parameterization (EWP) in the HARMONIE-AROME mesoscale model. The authors remark that several European weather services already use the HARMONIE-AROME model. Therefore, it is vital to implement wind farm parameterization for this model to allow researchers to work with a familiar code when accounting for wind farm effects in a forecast.
Since the technical description of wind farm parameterizations in HARMONIE-AROME was already published, this article focuses more on comparing and presenting the results. The article studying one-day simulation cases also acts as a link between a technical description of wind farm parameterizations in HARMONIE-AROME and an evaluation of long-term simulation effects expected in the second part.
Considering the number of graphs presenting the results, the authors did a good job optimizing the layout and description of the figures. Except for a few concerns listed further, the captions are presented in a way that makes easy to understand the plots without looking for details in the main text.
The simulations set-ups and choices made are meticuosly documented. The concepts unique to the study are appropriately introduced first. For example, a double line plot showing the implementation of no wind farm state for the observations is introduced separately in Fig. 6; this figure is further referred to in the caption of composite figures, so a reader can quickly check what the straight dashed lines mean. The caption of Fig. 8, one of the most complex figures in the article, also helps to recall the explanation immediately without searching for it in the main text.
The evaluation puts the EWP and FITCH parameterization in HARMONIE-AROME into a broader context by comparing their performance to the same parameterizations in the WRF model (the most widely used mesoscale model) and observations. This way of comparing allows other researchers to make an informed choice when selecting a mesoscale model to work with wind farm effects.
My suggestions for the revisions would be, therefore, minor leaning to technical.
Minor revisions
1. Page 14-15, lines 240-244 and Conclusion mention a possible effect of different grids on the direct comparison of WRF and HARMONIE highlighted by wind farm placement in Fig. 5. Compared to other complications, such as a choice of TKE correction factor or artificial no-farm observations, this problem is addressed only briefly as a possible source of the differences in predicted wake effects and leaves a question if there are other differences between models that could have been caused by it. Is it possible to quantify this effect?
I like the colour coding for tables – it makes them easy to read and interpret. Nevertheless, the colour-coding may not be quite in line with the article text or cause an erroneous interpretation. It is need to be checked whether the tables communicate the main idea properly, so I'll list my concerns.
2. The colour tones in tables may be misleading in some cases. Particularly, the TKE STDE row of Page 19, Table 5: the values of 0.10 are coloured in different shades of red, while 0.08 is white. Considering the rounding to two digits, one would expect all 0.10 to be coloured the same. Otherwise, it appears as if the difference in a third digit is important but cannot be seen due to the round-up. Similar behaviour may be observed in the TKE BIAS, RMSE rows of Page 19, Table 5 and, to a lesser extent, in the TKE RMSE, STDE rows of Page 20, Table 6.
3. The colour tones also raise a question of whether the difference in the TKE STDE between 0.08 and 0.10 in Table 5 is as important as in the TKE STDE between 0.29 and 0.55 in Table 3, considering the similar shading. It is also noted in the article text that the models' agreement with observations in the wake for 15 October 2017 (Table 5) is noticeably better than above the wind farm for 14 October 2017 (Table 3). However, similar tones imply equally strong differences for the TKE BIAS, RMSE and STDE in Table 3 and in Table 5.
4. Another confusing usage of colour is the BIAS row in all colour-coded tables. While other rows set the colour map to highlight the best case regardless of its value, the BIAS row shows the absolute deviation from zero. I may be missing some common convention here. If this is the case, I would be glad for a clarification.
Technical corrections
I will try to focus on the typos which are harder to track down in the proofreading.
1. The commas are occasionally omitted for introductory phrases.
E.g., Page 2, line 49-50 does not have a comma between 'atmosphere' and 'EWP':
To parameterize the effect of wind farms on the atmosphere EWP imposes an elevated momentum sink or drag force on a control volume...
However, Page 3, lines 54-55 have a comma between 'effect' and 'EWP' in a similarly constructed sentence:
To account for this effect, EWP builds upon classical wind turbine wake theory, by assuming an exponential expansion based on an effective length scale...
In general, the punctuation for introductory phrases allows some freedom. A comma can usually be omitted for short introductory phrases and is preferable for long phrases, but its usage is the opposite here. I suggest double-checking the text and comma rules to ensure that commas are used consistently.
2. Page 3, line 78: an uncommon placement of 'not'
For the implementation of Eq. 4 not the true wind direction WD is used...
Probably, it was supposed to be
For the implementation of Eq. 4, the true wind direction WD is not used...
3. Page 6, Table 1: FITCH is not mentioned for WRF in the title row, but WRF is also used with FITCH implemented, isn't it?
4. Page 8, Fig. 3: this could be my monitor calibration, but bright and dark colours are harder to distinguish for 08 August 2017 lines than for other tracks-transect pairs. The flight track for 15 October 2017 may need to be slightly darker to avoid blending with the background.
5. Page 8, line 150: the dates are typed as 14.08.2017 and 15.08.2017. Are they supposed to be 14 October 2017 and 15 October 2017, the cases regarded in the article?
6. Page 9, Fig. 5: a typo in the caption '...the winf farms...'
7. Page 13, line 218: an uncommon placement of 'also', similar to the one mentioned in (3)
As a consequence also the exact location of the wind speed deficit is not as well captured...
Please, double-check the text for similar errors.
8. Pages 14-15: the captions of Fig. 8 and Tables 3-4 do not mention the date and height of the case shown – it requires remembering the case's specifics to compare them to similar figures/tables.
9. Page 21, lines 323-324: duplicated 'lower'
These changes happen throughout the lower lower atmosphere, e.g. also close to the surface.
10. Page 21, Section 4 and further: As noted in the Introduction, HARMONIE-AROME is shortened to just HARMONIE. However, Sections 4-5 use the full name again and occasionally switch to the shortened name. Is there any specific reason behind this?
11. Cross-referencing figures in the captions shortens the description but also forces scrolling back and forth or searching in the text to find out which case is shown in the figure. Considering the amount of information and the size of figures detaching them from the first mention in the text, it is possible to get lost. I expected to see captions similar to Fig. 11 for all figures using the same layout as Fig. 8: reference to Fig. 8 for the full description, correction for the date and height, and additional comment if needed.
However, several captions break the pattern:
- Figure 13 in the main body references Fig. A2 in the Appendix (= needs scrolling down to confirm which case is this or searching the text), which in turn references Fig. 8 but adds date and height. It would be easier to read if Fig. 13 also referenced Fig. 8 with the added date and height.
- Similarly for Fig. A1, its caption does not mention the new date and height and requires to scroll up to see the case.
12. The wind farms are referred to by name several times. Although it is not very ambiguous from the text + figures which farm is located where, it would be nice to see an overview figure of the wind farm placement.
13. Page 24, line 379: a typo in the URL Hiram.org – it should be Hirlam.org, isn't it? Also, it is not possible to access this site for an external user. The page only asks for a login and password, unlike the EWP GitLab or CMEMS data pages, which are also account-locked but provide information on acquiring access. If there exists a different page with the relevant contact information for HIRLAM consortia, consider adding it instead.
hirlam.org is also mentioned twice in the main text, but it is accompanied by a citation, so this URL does not require a change in the main text.
14. I do not see the Appendix section's title even though two Appendix figures are present. It could be a glitch in the template.
Citation: https://doi.org/10.5194/gmd-2023-63-RC1 -
RC2: 'Comment on gmd-2023-63', Anonymous Referee #2, 16 Aug 2023
The authors presented a comparison between the Fitch wind farm parameterization and the EWP within two models: WRF and HARMONIE. The results of both models are compared to some airborne measurements and to SAR. A control experiment for both parameterizations with no wind farms is used as a reference case. The authors provided their setup and post-processing files publicly, which is a credited effort and can be helpful to other researchers to try regenerate the same results.
In general, I was not convinced that this manuscript can be a standalone publication, and I recommend a major revision. My main comments are listed below, and hopefully they be of help to the authors.
- The objective of the paper is to compare the performance of Fitch and EWP in the model HARMONIE. The major concern here is:
how is this different from previously published studies that compared Fitch and EWP in WRF?
For example, Pryor et al (2020, https://doi.org/10.1175/JAMC-D-19-0235.1) did a similar comparison as well as some of the authors of this manuscript (Larsén and Fischereit, 2021, https://doi.org/10.5194/gmd-14-3141-2021).
Using a different model (WRF vs HARMONIE) is irrelevant in this context. Both Fitch and EWP depend on the incoming flow (e.g., wind speed, wind direction, turbulence, etc) whether this flow is resolved by WRF or by HARMONIE. It is true that both models (WRF and HARMONIE) will resolve different flow fields and will transport the wake of the wind farms differently, but this is a comparison between WRF and HARMONIE, which is not the purpose of this study. As far as Fitch and EWP are concerned, the results and the conclusions presented here are not new to what is already known in the literature using WRF simulations.
It is stated that this manuscript is a part of a series of manuscripts using HARMONIE to model all the wind turbines in Europe. I agree with the authors that the first order of business is to make sure that Fitch and EWP are correctly implemented by comparing them to field measurements and to other models (in this case WRF). However, I suggest that this manuscript be summarized and included as an implementation-check-up section in upcoming manuscripts. - As a follow up to the last point, the introduction section misses any previously published differences between Fitch and EWP. It will be helpful for the reader that a study focusing on a comparison between Fitch and EWP discusses what the literature has to say about this issue.
- Can you please elaborate more on the annotations of the variables in equations 2—4? For instance, what does a subscript (rh) mean. What is an overbar? Even though these equations are explained in more detail in the literature (e.g. Volker et al 2015), for closure it is best to explain the meaning of the used variables.
- In line 114: “The later initialization causes the simulations follow more closely the boundary conditions”:
Can you please elaborate more what this means? A later initialization means a shorter spin-up period. If this means that after 12 hours of initialization, the model starts to deviate significantly from field measurements, does this raise a concern about the model setup? Can you please explain in case I understood this incorrectly? - Line 167: “Previous studies found that simulation results are not very sensitive to variation in fr between 1.5 and 1.7”:
Which “simulation results” are not sensitive? Wind speed, power production, TKE, ..? This study (https://doi.org/10.1175/MWR-D-23-0006.1) showed different conclusions. Can you please provide a proof that the range 1.5 –1.7 does not cause a significant difference at least within the context of the current comparison? - Line 203: “while WRF slightly overestimates the wind”
Can you please elaborate why? This does not seem to be the case in Larsén and Fischereit (2021, Fig. 1) for the same day (14 October 2017) and for only a 10-min time difference (17:10 there against 17:20 here). Can you please explain this discrepancy? - Figures 8 and 11 are difficult to read, especially the wind direction row. Can you please make it clearer by making the sub-figures bigger? Maybe split this into multiple figures?
- Line 217: “The wind direction, WD, is slightly off in HARMONIE, especially in the earlier transects up to 15:50”
Any suggestions or explanations why this happened? - In Fig. 9, wakes are much shorter in HARMONIE (top two sub-figures) compared to WRF (bottom two). Also, the direction of the wake in HARMONIE is a bit off compared to WRF and to the wind direction indicated in Fig. 8. Can you please explain why?
- Figure 10, the differences between HARMONIE and WRF are large. Can this compromise the comparison between Fitch and EWP for the two models?
Citation: https://doi.org/10.5194/gmd-2023-63-RC2 - The objective of the paper is to compare the performance of Fitch and EWP in the model HARMONIE. The major concern here is:
Jana Fischereit et al.
Jana Fischereit et al.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
297 | 77 | 10 | 384 | 19 | 5 | 3 |
- HTML: 297
- PDF: 77
- XML: 10
- Total: 384
- Supplement: 19
- BibTeX: 5
- EndNote: 3
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1