An Assessment of the Weather Research and Forecasting Model for Solar Irradiance Forecasting under the Influence of Cold Fronts in a Desert in Northwestern Mexico

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Thanks for the authors' efforts in preparing this manuscript. 

The experiment design and results support your conclusions, BUT I have some comments you may consider revising or updating this manuscript.

1. For Table 1 (opt2), what is the "WRF single" microphysics scheme? Did you mean the "WRF simple ice"?

2. Please re-plot figures 3, 4, 5, 6, and 7. Those figures are tough to read, and the image quality is poor for publishing.

3. In section 2.5, how many data points were used to calculate those metrics? What is the number of N?

4. From Line 565-569, in this conclusion, did you check the forecast result? Was the tropical storm captured correctly in the model?

5. Following the above question, this manuscript only shows the results from CF01 and CF05; what are the results from the rest? Also, do the "overall" statistic results still support opt2 as the optimal configuration? 

6. This study picks several cold-front cases; why not include satellite data to compare? It should help explain how different schemes produce cloud coverage, which would help explain the solar irradiance forecast or estimation. 

 

Author Response

Comments 1: For Table 1 (opt2), what is the "WRF single" microphysics scheme? Did you mean the "WRF simple ice"?

Response 1: Thanks for the observation. The name "WRF single", refers to  WRF Single Moment 6 Class. Instead of "WRF single", the name WSM6 has been updated in table 1.

Comments 2: Please re-plot figures 3, 4, 5, 6, and 7. Those figures are tough to read, and the image quality is poor for publishing.

Response 2: We totally agree and all the images has been updated to improve the quality. This changes, also includes some particular observations requested by the other reviewer.

Comments 3: In section 2.5, how many data points were used to calculate those metrics? What is the number of N?

Response 3: The number of data was N=720 for each cold front event simulated. Derived from this observation I´ve updated  the line <330 >.

Comments 4: From Line 565-569, in this conclusion, did you check the forecast result? Was the tropical storm captured correctly in the model?

Response 4: Thank you for pointing this out. The forecast result was indeed checked. This was a key part of the analysis for the CF05 event. So I'd like to give a brief context of this tropical storm as reported by the NHC (reference [38] https://www.nhc.noaa.gov/data/tcr/EP202020_Odalys.pdf):

 ODALYS was a tropical storm that lived from November 3 to 5, during November 6 to 7 it remained as a remmmanent low, and finally it dissipated during November 8. The center of the system was located at the closest point to the southern tip of the Baja California Peninsula, 2,000 km west-southwest.

Our simulation in WRF, ran from November 4 to 9 (the first 24 hours of this period were excluded for the statistical analysis). Thus, we can confirm that our DOMAIN-01 (Figure 3) was able to capture the tropical storm during its final phase (November 5-8). Also, it is important to mentioned that this report points that while the remnant low phase, the storm was steered by the near surface winds around a strong ridge, before dissipate. This is one of the reasons we detected an anomalous transport of moisture from south latitudes to our location during the cold front displacement. Even days after dissipated, this above normal moisture was detected in our ground-based meteorological data.

Comments 5: Following the above question, this manuscript only shows the results from CF01 and CF05; what are the results from the rest? Also, do the "overall" statistic results still support opt2 as the optimal configuration? 

Response 5: I am grateful for your contribution to this discussion. As previously indicated in section 3 (line 358), the events CF01, CF02, CF03, and CF04 exhibit comparable statistical values. Following an exhaustive examination of the statistical and graphical data, it is our conclusion that CF01 represents a typical case, while also serving as an illustrative comparison with other moisture-related event, such as CF05. This assertion is corroborated by the data presented in Table 1. 
In the graphical analysis and statistical results of CF01 to CF04, the opt2 resulted in the most favorable prediction, or at least the second most favorable prediction with the smallest difference. As an example, when all configurations (opt0, opt1, opt2, and opt3) are considered and the average Pearson correlation value is calculated for each event (excluding CF05), the value ranges from 0.79 to 0.85 in Tamb and from 0.81 to 0.95 in GHI. Similarly, the mean value of all configurations for mean absolute error (MAE) ranges from 1.45 to 5.14 °C in Tamb and from 23.01 to 57.80 W m⁻² in GHI. 

Comments 6:This study picks several cold-front cases; why not include satellite data to compare? It should help explain how different schemes produce cloud coverage, which would help explain the solar irradiance forecast or estimation. 

Response 6: 

 We appreciate this comment, as it contributes significantly to the discussion of the methodology for comparing the results presented in the manuscript. However, it is important for us to clarify that, although the possibility of comparing satellite data to relate cloud variability during simulated frontal events was contemplated, the lack of in situ measured cloud cover data and the spatial and temporal resolution of the satellite data were the main impediments to do so satisfactorily. Nevertheless, some comparison tests of the satellite data from the NASA-Power platform (Satellite/Reanalysis/Model. Reference [24]) were performed; it was possible to detect a kind of systematic error in terms of the NASA POWER global horizontal irradiance and relative humidity data compared with the measurements obtained on surface by our weather station during the passage of the frontal events. It was observed that, under the influence of cold fronts, the temporal resolution of the NASA-Power information available to us did not allowed to clearly observe the temporal and spatial variability of a mesoscale phenomenon compared to the short-term forecast. In most of the observed short periods (30-120 minutes), an apparent overestimation of GHI satellite data versus surface measured data was detected, indicating that some local conditions or conditions related to the interaction of near-surface low levels circulation could influence the measurements during the passage of the mesoscale event. This overestimation “phenomenon” and the apparent measurement errors of satellite data in the comparison with surface measurements are also corroborated in the following papers we consult for longer periods of analysis and not specifically focused on mesoscale events:

• Rodrigues, G. C., & Braga, R. P. (2021). Evaluation of NASA POWER reanalysis products to estimate daily weather variables in a hot summer mediterranean climate. Agronomy, 11(6), 1207.
• Marzouk, O. A. (2021). Assessment of global warming in Al Buraimi, sultanate of Oman based on statistical analysis of NASA POWER data over 39 years, and testing the reliability of NASA POWER against meteorological measurements. Heliyon, 7(3).

Based on this finding, our manuscript chose to keep the NASA-Power database as a reference to compare the measured data in addition to those recorded by surface instruments.  Including satellite data in this type of comparison is a strong possibility for future work for us. However, because we did not have the capacity to do so for the reasons mentioned above, and because our work was focused on evaluating the solar parameterization and to a lesser extent the Microphysic parameterizations, we chose to focus on the analysis of the variability of humidity and temperature associated with the GHI variations during the event.

To emphasize this point, we have updated line <250> and added references  [26] and [27].

 

***All modifications to the manuscript are highlighted in the word file. 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Review of the manuscript atmosphere-3073776 “An assessment of the WRF model for solar irradiance forecasting under the influence of cold fronts in the northwestern desert of Mexico”  by J.E. López-Velázquez et al.

 

The authors have developed an advanced method for predicting short-term changes in global horizontal irradiance in Northwestern Mexico. Predicting weather changes is of great importance for many aspects of human life, including the use of solar energy. Thus, I recommend publication of this article after minor revision. I believe it could be improved if the authors took into account the following comments.

 

COMMENTS.

 

 

 

1)      Line 17. The authors should clarify that WRF means Weather Research and Forecasting.

2)      Line 18. Please explain that GHI means Global Horizontal Irradiance.

3)      Line 374.  “…values above 500 W.m^-2 for most of the options…”. Use a different notation for W×m^-2 here and below.

4)      Line 380. “The scatter plot for GHI shows correctly with 0 W.m^-2, all the predicted values in the

night period”. What does it mean “0 W×m^-2”?

5)      Figures 4 and 5. Please increase the font size of the Y-axis legend.

6)      Section 3.2, Taylor diagrams. It worth to describe Taylor diagrams in more detail. E.g. describe what values are plotted on the X-axis and Y-axis.  Describe that the semi-circle contours indicate the RMS values etc.

7)      Figures 6 and 7. Please show that one figure describes event CF01 and the other one – CF05.

8)      Tables 3 and 4 show that for all GHI predictions the MBIAS values ​​are negative, i.e. all forecasts underestimate the actual values. The question arises: is it possible to adjust the forecasting techniques in such a way as to remove this negative shift?

 

 

 

 

 

Author Response

Comments 1: Line 17. The authors should clarify that WRF means Weather Research and Forecasting.

Response 1: Thanks for pointing this out. This observation has been updated in line 17.

Comments 2: Line 18. Please explain that GHI means Global Horizontal Irradiance.

Response 2: same response for this comment, it has been updated in line 18.

Comments 3: Line 374.  “…values above 500 W.m^-2 for most of the options…”. Use a different notation for W×m^-2 here and below.

Response 3: We totally agree and have update this notation overall the manuscript.

Comments 4: Line 380. “The scatter plot for GHI shows correctly with 0 W.m^-2, all the predicted values in the night period”. What does it mean “0 W×m^-2”?

Response 4: We completely agree with your comment about this line and thank you for pointing it out. We committed and impresicion trying to refer that in Figure 4 and 5, most of the lower values for GHI are plotted in the night period (circles), which indicates coherent results, because during this night period the values of GHI ranges from 0- 30 W m^-2. 

Following this comment, we updated line <382> in the manuscript. 

Comments 5: Figures 4 and 5. Please increase the font size of the Y-axis legend.

Response 5: Thanks for this observation. This has been updated, also the image has more quality now. 

 

Comments 6: Section 3.2, Taylor diagrams. It worth to describe Taylor diagrams in more detail. E.g. describe what values are plotted on the X-axis and Y-axis.  Describe that the semi-circle contours indicate the RMS values etc.

Response 6: Thank you for pointing this out, we agree with your comment. To emphasize about this in the manuscript, we have updated line 342 to 347.

 

Comments 7: Figures 6 and 7. Please show that one figure describes event CF01 and the other one – CF05.

Response 7: We attended to this recommendation. Figures 6 and 7 are updated, both have better quality now and the name of the events is on it.

 

Comments 8: Tables 3 and 4 show that for all GHI predictions the MBIAS values ​​are negative, i.e. all forecasts underestimate the actual values. The question arises: is it possible to adjust the forecasting techniques in such a way as to remove this negative shift?

Response 8: We appreciate this comment, as it contributes significantly to the discussion of the methodology to improve the results presented in the manuscript. However, it is important for us to clarify that the short range forecast, which captures this mesoscale event (cold fronts associated with different air masses), has an inherent variability that cannot be easily corrected by the analysis of the MBIAS if we want to improve the hourly short range forecast. In addition, we can see in Tables 3 and 4 that Tamb and Hrel are not consistent in the underestimation, especially for the MBIAS in CF05.

In response to your comment: if we consider the MBIAS as a possible indicator to adjust the forecast results, it would be recommendable to implement it for a medium range forecast focusing on periods of five days or more, or under the permibisility of less accuracy in short term variability.  

 

***All modifications to the manuscript are highlighted in the word file. 

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thanks for the authors' effort in fixing the issues.

Here is only one more thing that you need to fix.

1. Please update your Figs 4 and 5, especially the words of the color legend. The words are pretty small. 

Author Response

 

Comment 1. Please update your Figs 4 and 5, especially the words of the color legend. The words are pretty small. 

Response 1:

Thanks for the observation. Figure 4 and 5 (line 399 & 424) have been updated with a larger font size in the legend.

Author Response File: Author Response.pdf