A New Approach for Satellite-Based Probabilistic Solar Forecasting with Cloud Motion Vectors

Round 1

Reviewer 1 Report

I reviewed this MS with great interest, probabilistic forecasting is in my opinion an underdeveloped field and these kind of studies are well needed. In this particular work, the authors provide an enhanced probabilistic forecast not based solely on a statistical approach or a time series analysis but also add deterministic information from satellite derived CMV.

I don't have any major concerns about this paper and I believe it is worth publishing in Energies, but there are just some minor comments/suggestions that I would like to give to the authors.

1. Parameters used for the gaussian error in the Monte-Carlo method. One would think these parameters seem to have a high impact on the results and performance of the different probabilistic methodologies analysed but according to P10L344 initial guesses are used for both locations. I would like to read a bit more in the paper about how these initial values were guessed in order to understand why were they chosen, especially the radius of the monitoring perimeter.
2. Please rephrase P3L109 that reads "as a stationnarization pre-processing..." in some other way that explains it in other words since it seems a bit confusing in its present form, at least for me.
3. P5L207-210 mentions different CDF depending on the elevation of the sun and the level of Kc but I could not find any other mentions to these later in the paper. What ranges are used for the sun elevation? based on what? and for kc?
4. P6L218-219 you mention that GHI is computed from kc and the clear sky GHI, could you give additional explanations about which clear sky model are you using for this? Even if it was included in the soda data it would be good to give a reference about this clear sky model.
5. P9L290-293 are a bit confusing to me, this does not seem to be the same deterministic CMV approach mentioned previously in the performance assessment section, right? if this is a different method and not properly a performance assessment I believe it should be moved maybe to section 2.
6. Please explain in further detail the threshold for considering high or low variability days in table 3. Maybe in L353 it could be added especifically "Low variability days were considered those with a mean kc variation below 0.075..." or something similar.

 

Some spellcheck errors/typos:

P3L98 should it be "values" instead of "valued"?

P8L284-285 "devation" to  "deviation"

P8L271 "we will also..." to "we will also use..."

P8L285 "with respect..." to "with respect to..."

P9L317 "availability", do you mean "reliability"?

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The proposed study has a core and good approach topic for existing research, CMV forecasting. As there was no study for probabilistic CMV forecasting, this study can make significant insights to these studies. However, I have some questions about methodology, and it is expected that reinforcement will be required.

1. The proposed study used a new methodology including identification of cadidate pixels to estimate probabilistic forecasting. However, I wonder this new methodology shows different results compared to simple Monte-carlo simulation using bias distribution. This study used CH-PeEn as reference, but I think monte-carlo using CMV should be used as reference because the new methodology was proposed.

2. Previous CMV studies showed the influence of smoothing. Cited CMV extraction paper[42] used smoothing to CMV, but smoothing to resulting GHI or clearness index images should be considered. As this smoothing can reduce spatial uncertainty, significant performance increase can be identified. I wonder this study considered it. The study below can be a reference.

Kühnert, J.; Lorenz, E.; Heinemann, D. Satellite-Based Irradiance and Power Forecasting for the German Energy Market; Academic
Press: Cambridge, MA, USA, 2013; ISBN 9780123971777
Oh, M.; Kim, C.K.; Kim, B.; Yun, C.; Kang, Y.-H.; Kim, H.-G. Spatiotemporal Optimization for Short-Term Solar Forecasting Based on Satellite Imagery. Energies 2021, 14, 2216. https://doi.org/10.3390/en14082216

The following are additional small questions.

It can be a great help if overall flowchart is added in the manuscript.
It seems like the spatial area is tens of km, but the radius of the monitoring perimeter is 1km in Table2. I wonder if it's a typo or if there's some other reason. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The manuscript has been well modified and appears to be sufficient for publication.