the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Decadal trends in observed surface solar radiation and their causes in Brazil in the first two decades of the 21st century
Abstract. Numerous studies have investigated the long term variability of surface solar radiation (SSR) around the world. However, the large disparity in the availability of observational data between developed and least developed/developing countries leads to an underrepresentation of studies on SSR changes in the latter. This is especially true for South America, where few observational studies have investigated the SSR decadal trends, and usually only at a local or regional scale. In this study we use data from 34 stations distributed throughout all the regions of Brazil to present the decadal SSR decadal trends in the first two decades of the 21st century and investigate their associated causes. The stations were grouped into 8 composites according to their proximity. Our results show that in the North and Northeast Brazil a strong dimming occurred, with significant contributions from increasing atmospheric absorption, most likely due to anthropogenic emissions, and increasing cloud cover. In the Southeast and Midwest regions of Brazil near-zero trends resulted from competing effects of clear-sky processes and strong negative trends in cloud cover. In the South part of the Amazon and in Southern Brazil a statistically insignificant brightening was observed, with significant contribution from decreasing biomass burning emissions in the former and competing minor contributions in the latter. These results can contribute to deepen the knowledge and understanding of SSR decadal trends and their causes in South America, reducing the underrepresentation of this continent when compared to regions like Europe.
- Preprint
(2672 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-509', Anonymous Referee #1, 24 Mar 2024
This study aims at investigating the trends in solar radiation over Brazil and their causes over a period of about 20 years using data from 34 stations distributed over the country. A wealth of information from different independent sources is employed to explain the observed trends under all-sky and clear-sky conditions. The discussion of the results is based on a combination of qualitative and quantitative estimates of changes in different factors. Despite some issues that are addressed in my detailed comments below, the paper is well written and its findings are useful for understanding the recent variability of surface solar radiation over Brazil. The size of the country with different climate regimes and diversity of activities offer a good opportunity to investigate how different factors can influence solar radiation. I find this paper’s contents suitable for ACP. However, it its current form I cannot recommend acceptance for publication.
General comments
Two decades of data cannot justify the word “Decadal” in the title and generally in the manuscript. It can be simply changed to “Trends”. For some figures even less than 20 years of data are shown and discussed. The fact that trends are given per decade does not justify the term decadal in a broader sense.
The regional composites include, in most cases, a station installed in a big city where the irradiance levels should be expected to be lower compared to the rest of the stations. This means that the multi-station average of irradiance is biased low by the city station. This has not been discussed in the context of effects arising from factors dominant in the city stations. It would be interesting to see the variability of each yearly value in the trend-plots by adding either standard deviation bars, or high-low ranges, or even the individual data points from each station in smaller symbols of different color.
I am concerned about the use of the second method to identify days under clear-sky conditions. The cloud cover data are available in only two instances in the day (12 and 18 UTC) which challenges their representativeness for the entire day. I doubt that the method is trustful. The trends shown in Table 1 derived by the two methods are rather inconsistent and it seems that the second method is applicable only in 3 composites and only in 2 cases shows consistent results with the other method. Therefore I don’t see the value of using this synoptic cloud cover method.
Specific comments
18: The term “clear-sky processes” is used repeatedly in the paper without discussing somewhere which these processes are.
37-57: For the studies cited in the Introduction, it would be helpful for the reader to include the period each study has investigated.
70: It would be helpful for the reader to point to Table A1 also in this place.
84: It would be useful information to state what percentage of the hourly, monthly or yearly data were missing, not necessarily for each station but, for example, for each region, or at least on average for all stations.
119-121: Why such a small criterion (only 2 days per month) was used in constructing monthly mean data of AAOD? Two days per month can hardly be considered representative for the monthly value of AAOD. How much the dataset would be affected if a more representative criterion (e.g. 5-10 days in a month) would have been adopted?
133: Although the effect would be rather small, has the conversion from m2 to grid taken into account the variation of the area of the grids due to earth’s curvature?
251: Please state what are the confidence levels that are used in the analysis (e.g. 95%).
264: The linear trend lines could also be shown on these plots, as well as in Figures 3 and 4.
269-279: Please make this caption clearer. The text is too long and difficult to follow. Some information could be moved or already are in the main text (e.g., the clear-sky methods, the reference to Table A1. What are the numbers in square brackets under Synop Cloud cover?
296: Make clear that this statement refers to SSR “based on synop cloud cover” only.
296-302: According to Table A1, in the Southeast region not all stations provided synoptic cloud data. Why this region is treated differently from the Belem and Manaus regions and the respective synop-derived clear-sky time series is included in the plots? Moreover, for this region the agreement of the two methods worse both in terms of trends and the variability of the time series.
309-310: In fact, in only 2 stations the clear-sky trends with the two methods are similar. This is another indication that the synop method is questionable and does not provide consistent results in 6 of the 8 regions.
324: How the cloud cover data from different stations were combined to make the regional averages shown in Figure 4? Did you use simply the arithmetic mean? Generally I would be reluctant to averaging cloud cover data as yearly means, especially if these are used to compare trends in irradiance. Could effects on irradiance are non-linear and the variability of their averages (even as daily means) should not be directly comparable.
328: As cloud observations show large spatial variability, I would suggest to show in Figure 4 also the data from the individual stations (with smaller symbols of different color). You could also consider overlaying the data of Figure 2 to allow a direct comparison of the variability and the trends, keeping in mind the previous comment.
331: Comparing the trends of Figures 2 and 4 is based on the assumption that the cloud cover observations in each region are representative for all stations in the region, even for those without synop observations. Please discuss this briefly in the text. Note that this assumption has not been applied in the data of Figure 3.
340-341: Please elaborate a little more on the statement about the relevance of the CCRE on the derived SSR trends.
386: The term (0-1) in the y-axes titles is misleading since the plots show anomalies. It is better to remove it. Drawing the trend lines would help also in this figure.
412: Since the OMI AAOD is used, I wonder why the AOD from OMI was not used also in Figure 5a instead of the CAMS reanalysis, in order to maintain consistency between the two data products.
414: Figure 7 shows the AAOD in relative units while Figure 5a in absolute units, so direct comparison is very difficult. I suggest using absolute values in the trends.
Technical
The verb “fit” is used repeatedly (manly in the discussion sections) to indicate consistency or compliance between results or findings. It is better to be replaced with expressions like “comply with” or “consistent to“.
25: Replace “object” with “subject”
82: Replace “went” to “were”
193: Replace “occurred” with “are”
210: “both locations” which are those? Belem and Manaus?
265: I suggest rephrasing as follows: “In each composite, anomalies are in reference to the mean of the entire period (shown in table 1).”
281: Remove “composites” in parentheses, as already exists before the parentheses.
390: Replace “FABas” with “Fabs”
402-403: “magnification of the effects” is a bit misleading; better replace it with “magnification of trends”
426: Replace “verified” with “investigated” or “studied”
474: Replace “until where the” with “over which a” and move this explanation in line 470 where the term “decorrelation length” is first mentioned.
475: Delete duplicate “time”
494: Replace :optimize” with “increase”
539: Replace “visible” with “active” or “occurring”
551: Replace “is especially remarkable” with “occurs primarily”
568: Replace “as the” with “at the”
Citation: https://doi.org/10.5194/egusphere-2024-509-RC1 -
RC2: 'Comment on egusphere-2024-509', Anonymous Referee #2, 03 Apr 2024
Review of the paper “Decadal trends in observed surface solar radiation and their causes in Brazil in the first two decades of the 21st century” by Correa et al. The authors present a study where they have investigated the long-term variability of surface solar radiation in Brazil in the first two decades of the 21st century. The study deserves to be published considering the novelty of the treated region even if it analyses a short period and it uses a small dataset.
Below the authors can find a list of suggestions that must be addressed before publication especially related to the adopted methodology.
Major comments:
- Lines 29-33: the authors should cite an higher number of papers demonstrating to know the literature and supporting in this way the subsequent sentence “ However, many regions of the world…”.
- Line 42: “machine learning methods…” another study on this topic can be cited (https://doi.org/10.5194/essd-15-4519-2023). It applies machine learning methods using both ground-based observations and reanalysis dataset.
- Section 2: how have you checked the homogeneity of the SSR series?
- Lines 70-84: it is necessary to add more information about the steps to move from the hourly values to the annual values in order to prove the robustness of the applied method. If any you can also add a reference of another paper where this method has been applied.
- Figure 1: it would be interesting to see the orography of the region instead of the green colour to notice if the stations are located in plane or mountain areas. This could also help to understand how much the selected station are representative of the composite they belong.
- Line 251: considering the not so long period covered by the SSR data, why did you not calculate the SSR trends with Sen-Theil? Have you verified that the trends calculated with the two methods are similar?
- Line 258: have you considered the series as absolute series or have you calculated the anomaly series? You should specify this point.
- Caption figure 2: in the text there is not explained that the series are treated as anomalies. It should be specified. Moreover, which is the reference period? How have you treated the missing values?
- Line 290: even if the differences are evident between different regions it should be considered that they cover different periods and all of them cover less than 30 years so more caution is necessary when the regions are compared.
Minor comments:
- Line 70: “is collected and controlled by the Instituto…” add “by”
- Line 164: there is a reference for this method?
- Line 181: change “absorved” with “absorbed”
- Table 1: columns 7-8 there are too many digits
Citation: https://doi.org/10.5194/egusphere-2024-509-RC2 -
RC3: 'Comment on egusphere-2024-509', Anonymous Referee #3, 06 Apr 2024
Summary and general comments
The manuscript “Decadal trends in observed surface solar radiation and their causes in Brazil in the first two decades of the 21st century” explores a set of data to describe and explain trends in surface solar radiation (SSR) for different regions of Brazil, including few major cities, during the last 15 to 20 years. Compared to other regions of the world, indeed South America lack more studies on this topic. The short time series used limits the needed long-term analysis for the region, but the article brings to the discussion important aspects that may are under controls of the contemporary trends in SSR. Trends in aerosols particles loading and properties along cloud cover are the main factors used to explain trends in SSR. Surprisingly, the authors did no show any timeseries of AOD as they did for cloud cover. AOD trend is important, but it does not show important steps in AOD evolution over Brazil, which I consider to be important to the interannual variability analysis.
The authors use anthropogenic emissions in way that seems to exclude biomass burning emissions or to classify it in a different emission category, smoke emission in Brazil is mostly driven by human activity. At many parts of the manuscript, urban and industrial emissions would fit better than anthropogenic emissions.
The article is well organized, but the author may resort to more table to summarize their dataset and details, since there are many sources of data with a variety of features. The role of each product must be clear from the methods topic.
Before recommendation to be published, I would suggest the authors to think through their approach more carefully and analyze in more detail some aspects that I highlight in the specific comments below.
Specific comments
Line18: Clarify what do you mean by “clear-sky processes”
Line18: South of Amazon is still part of the North region; why was it treat separately? How does it compare with other parts of Amazon within north region?
Line26: I guess evidence that SSR is not constant are older than that. You may adjust the sentence to emphasize the role of the cited studies pioneering the studies that try to understand the trends in SSR over time, not that evidence that it is not constant.
Line32: “Wild et al. 202” Check the year.
Line54: It would be interesting to mention what these studies say about dimming and brightening. These studies pointed out to which direction, or that was not the case?
Line59-61: To which point this limited time series would tackle the existent gap. Be more precise.
Line70: “controlled by the instituto” instead of “controlled the instituto”
Line71: Instituto Nacional de Meteorologia instead of “Instituto Brasileiro de Meteorologia”
Line113-114: How does CAMS AOD product performs over South America? Any literature review on this?
Line119-121: Here, make clear here that this is just for AAOD from OMI.
Line123 “long term mean” it is important to clarify the time period that your long-term mean refers to, not just here but above also, when you describe the same procedure for SSR.
Line125-128: Need to clarify the role of CERES data in the study.
Line129 -137: All this need to have their use justified. What's the point of use Edgar if it does not include biomass burning emission, inform the reader the reason to include EDGAR, since it does not consider biomass burning emission.
Line143-149: Would this threshold remain the same as time goes by? High aerosol loading events you said that would not be a problem for long term trends, but within the decades analyzed there was a strong shift in the aerosol loading in the atmosphere, mainly in the center-west of Brazil and Amazonia.
Line155-157: How does this fit with the first method threshold?
Line 171-174: You are working with products with different resolutions, at certain point will be important to describe how did you manage to sample this product around each station. Did you take the pixels that contain the coordinates of the station, or you did an average over a specific area? It would be helpful if you organize these products in a table with their description.
Line 175: I guess it will make it easier by separate the first term as absorption produced by the system (atms + surface) and the second term just the absorption produced by the Surface. A question, did you describe the source for TOA SW?
Line187: In your figure one you could take a climate basis classification to support objectively this statement. My point, you must support in clearer way this coverage of different climate characteristics.
Line195-197: It would be helpful to include Reboita et al. climates domain to contextualize the positions of your sites.
Line205: You are mentioning that precipitation is tied to local and mesoscale, but soon you bring ITCZ as the most important large-scale elements to explain precipitation seasonality in the region, which is true. Suggest you adjust the sentence, so local, regional, and large-scale role can be equally acknowledged.
Line 212-213: It is true that Manaus and Belem are not strongly influenced by the most important biomass burning region in the Amazon, which is the southern portion of the rainforest ecosystem, but Manaus and Belem are also affected by smoke, mainly from the biomass burning season in the northeast portion of Amazon, and AOD are really significant at periods (See this: https://amt.copernicus.org/articles/12/921/2019/)
Line221: Try to display these regions in figure 1, it will make it easier for the reader.
Line235: It depends on the season, during summer local convection associated with sea breeze plays a major role in cloud diurnal cycle.
Line238: smoke aerosols are also anthropogenic, at least those from biomass burning in most of Brazil, so I'll recommend you replace anthropogenic here to Urban-Industrial emissions.
Line241; Frontal system instead of “fronts” and extra-tropical cyclones instead of “subtropical cyclones”
Line243: Again, replace anthropogenic to urban-industrial emissions.
Line261: “Figure 2 shows the all-sky SSR anomalies time series” instead of “Figure 2 shows the all-sky SSR time series.”
Line264 (Figure 2) Why not include the standard deviations for each case. This would justify one plot for each site, otherwise I would use just one figure to plot all the sites along with different colors.
Line318-319: Clear sky cases correspond to which fraction of all cases? Clear sky cases are expected to be associated with particular meteorological scenarios; how can these aspects affect your conclusion here?
Line321: At this stage, without any analysis on aerosols, clouds etc, this sentence sounds strange.
Line342-34: “…major cloud contribution…” you have to keep in mind that it is only about cloud cover, your clouds dataset does not allow you to infer change in other aspect of clouds. (High, low, middle clouds) Have a look on this article for Sao Paulo, https://rmets.onlinelibrary.wiley.com/doi/abs/10.1002/joc.6203
Line351: Why CWV is included in the analysis, I'm not sure that this was clarified previously, but it is important to explain the purpose of including CWV in this analysis.
Line353 (Figure 5) Colorbar dimension is missing (all figure with colorbar dimension is missing not just this). Also, it would be important to include in these maps an indication of area were the trends are statistically significant.
Line 360: Do you mean southern hemisphere winter? Not summer.
Line363: Surprisingly the impact of the reduction of smoke from south amazon on downwind region does not appear.
Line395-397: Negative trend for clear sky seems consistent with smoke loading trend, but which would be the explanation for positive trend under all sky conditions? It would be interesting to see the seasonal distribution(frequency) of you days for both clear-sky and all-sky conditions.
Line 405: cloudiest instead of cloudier
Line421- 424: Does this result find echoes in the literature? How about OMI uncertainty in region with lower aerosol loading? How about the significance of this trends? Smoke from biomass burning is an important source of regional absorption. I would expect a reduction in regional smoke to produce a reduction in regional AAOD.
Line446-449: Let's say that this may explain AAOD increase in the southeast of Brazil and along the region close to the coast, but how about the central and north of Amazonia, where is found the larges positive trend in AAOD (see figure 7). That's why needed to focus on areas with trends statistically significant. Again, biomass burning is also anthropogenic emissions, and it has important fraction of black carbon, south of Amazonia lower smoke would contribute to less black carbon in the atmosphere.
Line 464-465: It also would be important to bring some notion about the performance of satellite and models products that were use. Since they would indicate if there were known bias that can help to explain some aspects of your results.
Line 495-497: This was already said few sentences above.
Line502: How about the cities mentioned above, are they not influenced by mesoscale and synoptic scales?
Line 517-529: If you could think in a diagram (with good visual perspective) that summarize this it would good.
Line 574 (Figure 9)- there is a missing plot (column 1, row 2)
Line 655-657: But you also have to take in to account the fact that cloud cover trend was not evaluate for this region.
Line 700-701: This discussion is generic about this; the author's need to search for more evidence in the literature to support this. Actually, I found that the author barely explores regional literature on their discussion, I mean studies that try to analyze trends in aerosols in South America.
Line 711-712: This is hard to say, there are many open aspects, cloud cover for south amazon was not evaluated, which can play an important role on these results. So, I would recommend the author to be more cautious here.
Line714-716: That's true, however, the limited time frame of the time series is still a challenge for SSR trend evaluation as has been done for north hemisphere regions.
Citation: https://doi.org/10.5194/egusphere-2024-509-RC3
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
166 | 37 | 15 | 218 | 8 | 8 |
- HTML: 166
- PDF: 37
- XML: 15
- Total: 218
- BibTeX: 8
- EndNote: 8
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1