the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Daily satellite-based sunshine duration estimates over Brazil: Validation and inter-comparison
Maria Lívia Lins Mattos Gava
Simone Marilene Sievert da Costa Coelho
Anthony Carlos Silva Porfírio
Abstract. The broad geographical coverage and high temporal and spatial resolution of geostationary satellite data provide an excellent opportunity to collect information on variables whose spatial distribution and temporal variability are not adequately represented by the in situ networks. This study focuses on assessing the effectiveness of two geostationary satellite-based sunshine duration (SDU) datasets over Brazil, given the relevance of SDU to various fields, such as agriculture and energy sectors, to ensure reliable SDU data over the country. The analyzed datasets are the operational products provided by the Satellite Application Facility on Climate Monitoring (CMSAF), that uses data achieved with the Meteorological Satellite (Meteosat) series, and by the Satellite and Meteorological Sensors Divison of the National Institute for Space Research (DISSM/INPE), that employs Geostationary Operational Environmental Satellite (GOES) data. The analyzed period ranges from September 2013 to December 2017. The mean bias error (MBE), mean absolute error (MAE), root mean squared error (RMSE), correlation coefficient (r) and scatterplots between satellite products and in situ daily SDU measurements provided by the National Institute of Meteorology (INMET) were used to access the products performance. They were calculated on a monthly basis and grouped into climate regions. The statistical parameters exhibited a uniform spatial distribution, indicating homogeneity within a given region. Except for the Tropical Northeast Oriental (TNO) region, there were no significant seasonal dependencies observed. The Mean Bias Error (MBE) values for both satellite products were generally low across most regions in Brazil, mainly between 0 and 1 hour. The correlation coefficient (r) results indicated a strong agreement between the estimated values and the observed data, with an overall r value exceeding 0.8. Nevertheless, there were notable discrepancies in specific areas. The CMSAF product showed a tendency to overestimate observations in the TNO region, with MBE consistently exceeding 1 hour for all months, while the DISSM product exhibited a negative gradient of MBE values in the west-east direction, in the northern portion of Brazil. The scatterplots for the TNO region revealed that the underestimation pattern observed in the DISSM product was influenced by the sky condition, with more accurate estimations observed under cloudy skies. Additional analysis suggested that the biases observed might be attributed to the misrepresentation of clear-sky reflectance. In the case of the CMSAF product, the overestimation tendency observed in the TNO region appeared to be a result of systematic underestimation of the Effective Cloud Albedo. The findings indicated that both satellite-based SDU products generally exhibited good agreement with the ground observations across Brazil, although their performance varied across different regions and seasons. The analyzed operational satellite products present a reliable source of data to several applications, being an asset due to its high spatial resolution and low time latency.
- Preprint
(5841 KB) - Metadata XML
- BibTeX
- EndNote
Maria Lívia Lins Mattos Gava et al.
Status: closed
-
RC1: 'Comment on egusphere-2023-1195', Anonymous Referee #2, 18 Aug 2023
The reviewed study assesses the overall quality of Sunshine Duration (SDU) estimates through geostationary satellite data over Brazil. This is done by comparing CMSAF’s product, composed by Meteosat measurements and DISSM/INPE’s product, obtained from GOES series’ data, against SDU estimates from in-situ measurements from meteorological stations. The analysis is based on usual statistical metrics, guided by a climate regions separation and leveraging data enhancements by the way the results are displayed.
General Comments:
- The data presented, introduces the overall behavior of SDU products on climate regions in South America. I acknowledge the contribution of this study to the progress in satellite estimates of SDU in the region. This kind of assessment is meaningful for a broad range of economic and infrastructure activities developed in the study region.
- The authors revisited the literature properly and brought to discussion the available measurements over the study region and their respective limitations. All the results were discussed accordingly. Additionally, the long description of the regions’ aspects was relevant for this discussion.
- Their results were presented in a clear, concise, and well-structured way. Some of the figures embedded a lot of relevant information from the way the data was presented. It was a smart choice to do so. Their interpretation was coherent and supported by the data presented. Substantial conclusions were reached in accordance with their analysis.
Specific Comments:
- Particularly in Figures 4, 6, 7 and 8, it was not completely clear to me why just the middle month of every season of the year were shown. Is there a specific reason for this choice? It might be better to clear that out in the text.
- Figures 6 and 7 are very illustrative of the products’ general behavior. In most of those Difference vs. Ground-truth plots there are clusters of counts close to the 0.0 line, centered around larger SDU, which means the products present better performance with longer periods of clear sky. However, on “g” plots (mostly 6g but 7g as well), the data seems more spread and less clustered. Does that suggest that the products account for cloudiness better in July for most of the regions of interest?
Technical Corrections:
The questions [2 - 4] after line 75 should be re-written for correctness. I understand that there would be no loss in meaning with those changes.
2: “Are there regions in which… ?”
3: “Are there seasonal variations… ?”
4: “Could deficiencies in the retrieval be traced back to their source?”
Line 173 (p.7) could be incorporated to the previous paragraph.
If the DNI (...) of a non-sunny slot (Kothe et al., 2017). The daily SDU in hours is then derived by Eq. (5)
Line 193 (p. 7). Suggestion for re-writing:
Presents high SDU values, on average, over the whole year. With highest values in the winter, along with the smallest variability, indicating predominance of clear sky days
→ On average, this region presents high SDU over the whole year, compared to the others. In the winter, the highest values are reached and the variability is low, due to the predominance of clear sky days.
Figura 4’s legend. Suggestion for avoiding confusion:
Figure 4. Spatial distribution of monthly MBE (h) between (...)
→ Figure 4. Spatial distribution of monthly MBE (hours) between (...)
Citation: https://doi.org/10.5194/egusphere-2023-1195-RC1 - AC1: 'Reply on RC1', Maria Lívia Lins Mattos Gava, 29 Aug 2023
-
RC2: 'Comment on egusphere-2023-1195', Anonymous Referee #1, 29 Aug 2023
This study evaluates two satellite-based data records of sunshine duration for the region of Brazil. The first product is a Meteosat-based climate data record by CM SAF, the other product is a GOES-based product by DISSM / INPE. Both datasets are compared against in-situ measurements from ground stations all over Brazil. The analyses is done for different climatic regions and seasons.
General comments
- Sunshine duration is an important parameter as there are long-time series of measurements and it is easy to communicate to the general public, as this parameter is well known and easy to understand. This study helps to estimate the quality of satellite-based sunshine duration and shows that satellite-based products can be a good complement to station-based data.
- The study is very well structured and written. It gives an excellent overview about the two different retrieval techniques used to derive sunshine duration from satellite. The applied approach is clearly explained and the results are easy to understand and well presented by meaningful figures. Results and figures support the conclusions in an adequate way.
Specific comments
- It might be good to include a paragraph describing some of the specific challenges, which have to be considered while comparing satellite-data to station-based data. For the Meteosat-based product Brazil is right at the outer edge of the observed region, which comes along with larger uncertainties in the data due to different effects (parallax effect, larger footprint, different atmospheric effects, …). Thus, the observing geometry is different to a GOES-based product. How representative are station-based measurements for a grid point of 4 km by 4 km (or even larger for the CM SAF product at the edge of the Meteosat disc)?
- CM SAF released SARAH version 3.0 in May 2023. Especially with some improvements in underlying auxiliary data, such as water vapour and surface albedo, it is expected to perform better in the region of Brazil.
Technical corrections
- line 206: ‘summer’ could be confusing for some readers, as we are on the southern hemisphere
- Fig. 8: I learned that it should be avoided, wherever possible, to use rainbow colour scales. Maybe you consider this at least for the next time.
Citation: https://doi.org/10.5194/egusphere-2023-1195-RC2 - AC2: 'Reply on RC2', Maria Lívia Lins Mattos Gava, 04 Sep 2023
Status: closed
-
RC1: 'Comment on egusphere-2023-1195', Anonymous Referee #2, 18 Aug 2023
The reviewed study assesses the overall quality of Sunshine Duration (SDU) estimates through geostationary satellite data over Brazil. This is done by comparing CMSAF’s product, composed by Meteosat measurements and DISSM/INPE’s product, obtained from GOES series’ data, against SDU estimates from in-situ measurements from meteorological stations. The analysis is based on usual statistical metrics, guided by a climate regions separation and leveraging data enhancements by the way the results are displayed.
General Comments:
- The data presented, introduces the overall behavior of SDU products on climate regions in South America. I acknowledge the contribution of this study to the progress in satellite estimates of SDU in the region. This kind of assessment is meaningful for a broad range of economic and infrastructure activities developed in the study region.
- The authors revisited the literature properly and brought to discussion the available measurements over the study region and their respective limitations. All the results were discussed accordingly. Additionally, the long description of the regions’ aspects was relevant for this discussion.
- Their results were presented in a clear, concise, and well-structured way. Some of the figures embedded a lot of relevant information from the way the data was presented. It was a smart choice to do so. Their interpretation was coherent and supported by the data presented. Substantial conclusions were reached in accordance with their analysis.
Specific Comments:
- Particularly in Figures 4, 6, 7 and 8, it was not completely clear to me why just the middle month of every season of the year were shown. Is there a specific reason for this choice? It might be better to clear that out in the text.
- Figures 6 and 7 are very illustrative of the products’ general behavior. In most of those Difference vs. Ground-truth plots there are clusters of counts close to the 0.0 line, centered around larger SDU, which means the products present better performance with longer periods of clear sky. However, on “g” plots (mostly 6g but 7g as well), the data seems more spread and less clustered. Does that suggest that the products account for cloudiness better in July for most of the regions of interest?
Technical Corrections:
The questions [2 - 4] after line 75 should be re-written for correctness. I understand that there would be no loss in meaning with those changes.
2: “Are there regions in which… ?”
3: “Are there seasonal variations… ?”
4: “Could deficiencies in the retrieval be traced back to their source?”
Line 173 (p.7) could be incorporated to the previous paragraph.
If the DNI (...) of a non-sunny slot (Kothe et al., 2017). The daily SDU in hours is then derived by Eq. (5)
Line 193 (p. 7). Suggestion for re-writing:
Presents high SDU values, on average, over the whole year. With highest values in the winter, along with the smallest variability, indicating predominance of clear sky days
→ On average, this region presents high SDU over the whole year, compared to the others. In the winter, the highest values are reached and the variability is low, due to the predominance of clear sky days.
Figura 4’s legend. Suggestion for avoiding confusion:
Figure 4. Spatial distribution of monthly MBE (h) between (...)
→ Figure 4. Spatial distribution of monthly MBE (hours) between (...)
Citation: https://doi.org/10.5194/egusphere-2023-1195-RC1 - AC1: 'Reply on RC1', Maria Lívia Lins Mattos Gava, 29 Aug 2023
-
RC2: 'Comment on egusphere-2023-1195', Anonymous Referee #1, 29 Aug 2023
This study evaluates two satellite-based data records of sunshine duration for the region of Brazil. The first product is a Meteosat-based climate data record by CM SAF, the other product is a GOES-based product by DISSM / INPE. Both datasets are compared against in-situ measurements from ground stations all over Brazil. The analyses is done for different climatic regions and seasons.
General comments
- Sunshine duration is an important parameter as there are long-time series of measurements and it is easy to communicate to the general public, as this parameter is well known and easy to understand. This study helps to estimate the quality of satellite-based sunshine duration and shows that satellite-based products can be a good complement to station-based data.
- The study is very well structured and written. It gives an excellent overview about the two different retrieval techniques used to derive sunshine duration from satellite. The applied approach is clearly explained and the results are easy to understand and well presented by meaningful figures. Results and figures support the conclusions in an adequate way.
Specific comments
- It might be good to include a paragraph describing some of the specific challenges, which have to be considered while comparing satellite-data to station-based data. For the Meteosat-based product Brazil is right at the outer edge of the observed region, which comes along with larger uncertainties in the data due to different effects (parallax effect, larger footprint, different atmospheric effects, …). Thus, the observing geometry is different to a GOES-based product. How representative are station-based measurements for a grid point of 4 km by 4 km (or even larger for the CM SAF product at the edge of the Meteosat disc)?
- CM SAF released SARAH version 3.0 in May 2023. Especially with some improvements in underlying auxiliary data, such as water vapour and surface albedo, it is expected to perform better in the region of Brazil.
Technical corrections
- line 206: ‘summer’ could be confusing for some readers, as we are on the southern hemisphere
- Fig. 8: I learned that it should be avoided, wherever possible, to use rainbow colour scales. Maybe you consider this at least for the next time.
Citation: https://doi.org/10.5194/egusphere-2023-1195-RC2 - AC2: 'Reply on RC2', Maria Lívia Lins Mattos Gava, 04 Sep 2023
Maria Lívia Lins Mattos Gava et al.
Data sets
Satellite-derived sunshine duration product - DISSM/INPE M. L. L. M. Gava, S. M. S. C. Coelho, and A. C. S. Porfírio https://doi.org/10.5281/zenodo.7958199
GOES 13 - visible reflectance M. L. L. M. Gava, S. M. S. C. Coelho, and A. C. S. Porfírio https://doi.org/10.5281/zenodo.7963354
Surface Radiation Data Set - Heliosat (SARAH) - Edition 2.1 U. Pfeifroth, S. Kothe, J. Trentmann, R. Hollmann, P. Fuchs, J. Kaiser, and M. Werscheck https://doi.org/10.5676/EUM_SAF_CM/SARAH/V002
Maria Lívia Lins Mattos Gava et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 158 | 43 | 14 | 215 | 6 | 5 |
- HTML: 158
- PDF: 43
- XML: 14
- Total: 215
- BibTeX: 6
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1