Last modified on May 30, 2025 17:05

Contributors: T.Usedly (DWD)

Issued by: Deutscher Wetterdienst / Tim Usedly

Date: 01/08/2024

Ref: C3S2_D312a_Lot1.2.3.7_202408_PQAR_ECV_SRB_SLSTR_v1.2

Official reference number service contract: 2021/C3S2_312a_Lot1_DWD/SC1

History of modifications

List of datasets covered by this document

Related documents

Acronyms

List of tables

List of figures

General definitions

Table 1: Summary of variables and definitions

Table 2: Definition of processing levels

Table 3: Definition of various technical terms used in the document

Table 4: Definition of statistical measures used in the document

Scope of the document

This document provides a description of the product validation results for the Sea and Land Surface Temperature Radiometer (SLSTR) v4.0 based Interim Climate Data Record (ICDR) of the Essential Climate Variable (ECV) Surface Radiation Budget (SRB).

The dataset produced by RAL Space and Brockmann Consult (BC) under the Copernicus Climate Change Service (C3S) programme ranges from 01/2017 – 12/2023 and provides an Interim Climate Data Record (ICDR) to the brokered Thematic Climate Data Record (TCDR) from European Space Agency Cloud Climate Change Initiative (ESA’s Cloud_cci).

The TCDR is a brokered product based on processing of the (Advanced) Along-Track Scanning Radiometer ((A)TSR) onboard ERS-2 and Envisat by RAL Space for the ESA Cloud_cci programme and ranges from 06/1995 – 04/2012. Detailed validation methodology and results are presented in the Cloud_cci Product Validation and Intercomparison Report [D1].

The ICDR is derived with a five-year gap from SLSTR onboard the Sentinel-3A and -3B satellites spanning from 01/2017 – 12/2023

Executive Summary

The Sea and Land Surface Temperature Radiometer onboard Sentinel-3A has provided data since January 2017. The launch of Sentinel-3B in October 2018 makes it possible to deliver not only individual data from both satellites but also a merged Sentinel-3A/3B product. The merged version (until 12/2023) is validated against measurements from ground stations from the Baseline Surface Radiation Network (BSRN). Depending on the temporal availability and specific variable up to 37 stations were used to provide the best possible global coverage. In addition to the merged SLSTR version, a second version on a different grid (equal area in addition to equal angle) is provided for the period from 07/2022 to 12/2023 and also validated against the same reference dataset as the equal angle version of SLSTR.

Validation to these SLSTR derived products is described in the following chapters of this document: Chapter 1 provides a summary of the product validation methodology while chapter 2 presents the validation results. A detailed validation methodology can be found in the Product Quality Assurance Document (PQAD) [D2]. Chapters 3 and 4 discuss possible application specific assessments and compliances with user requirements respectively.

Overall the SLSTR data meets the breakthrough/target GCOS requirement for the horizontal and temporal resolution. However, all variables do not meet the threshold GCOS requirements in terms of accuracy (Table 1-1); the values of the absolute bias are 13.05 W/m² (13.60 W/m²) for Surface Incoming Shortwave radiation (SIS) for equal angle grid (equal area grid), 13.36 W/m² (15.73 W/m²) for Surface Outgoing Longwave Radiation and do meet the threshold requirement (10 W/m²). Biases for Surface Reflected Shortwave Radiation (SRS) are 14.65 W/m² (14.00 W/m²) and Surface Downwelling Longwave Radiation (SDL) also do not meet the requirement due to a bias of 20.56 W/m² (18.40 W/m²). Comparison with the most continental and representative stations meet the threshold requirement by GCOS, outliers are mainly due the stations at high altitudes, high latitudes or Islands.

1. Product validation methodology

Detailed information about the validation methodology can be found in the corresponding PQAD [D2], section 3. The validation process is separated into three parts: Data preparation (section 1.1), validation (section 1.2) and evaluation (1.3).

1.1 Data preparation

SLSTR data is provided at a regular latitude-longitude grid with 0.5°x0.5° spatial resolution and monthly means. The validation is based on the comparison with monthly means of available ground station measurements of the BSRN (Ohmura, 1998). A nearest neighbor technique is used for a comparison, using the ground stations coordinates to extract the nearest grid point of the gridded dataset to calculate the bias and further statistical measures.

1.2 Validation

Validation is done by calculating each station’s mean bias and absolute bias as well as a correlation of all available station-months. In addition, Standard deviation and Fraction of months outside the validation target values are calculated.

1.3 Evaluation

The previously calculated absolute bias is used as evaluation against the requirements defined by the Global Climate Observing System (GCOS) in The 2022 GCOS ECVs Requirements (GCOS 245) [D3]. They are summarized in table 1-1.Comparison with CERES satellite data

Table 1-1: Summary of requirements for Surface Radiation Parameters: SIS, SDL and SOL based on GCOS [D3]

2. Validation results

Sections 2.1 – 2.4 show the validation results for the four variables: (i) Surface Incoming Shortwave Radiation (SIS), (ii) Surface Reflected Shortwave Radiation (SRS), (iii) Surface Downwelling Longwave Radiation (SDL) and (iv) Surface Outgoing Longwave Radiation (SOL).

2.1 Surface Incoming Shortwave Radiation

Figure 2-1: Plot (a, left) shows bias (black dots) and absolute bias (white triangle) for the comparison of SLSTR SIS and BSRN ground stations. Green shaded area marks the 10 W/m² threshold requirement by GCOS. Plot (b, right) shows the number of available months per station

Figure 2-1 shows bias and absolute bias for the SLSTR SIS compared to measurements from 37 stations. Most of the absolute biases are within the ± 10W/m² area with an overall small positive bias (1.82 W/m²). Four stations show a significant positive bias: Boulder (1689 m, east end of the Rocky Mountains), Darwin Met Office (North coast of Australia), Lanyu Island (Island close to Taiwan) and Reunion Island (Island east of Madagascar). A negative bias is seen in Izana (2373 m, central Spain). What all of these stations have in common is that they are either located at high altitude/topographic terrain or small islands/coastal areas and are questionable in terms of representativeness. In addition, Boulder and Darwin Met Office only rely on 19 and 17 months, respectively. Figure 2-2 indicates that almost all stations with absolute biases >10 W/m² are at high altitudes, islands/coasts or at high latitude (see Antarctica).

Figure 2-2: Temporal average of SIS from SLSTR and stations from the BSRN. Green dots indicate surface stations where the SLSTR SIS data meets the target accuracy (absolute bias less than 10W/m²), black triangles correspond to surface stations, where the SLSTR data does not meet the target accuracy

The correlation of all available 1706 station-months shows good agreement with a correlation of 0.97 and a small positive bias (red line slightly below the blue line) (figure 2-3).

Figure 2-3: Correlation between SLSTR SIS and BSRN ground stations. Blue line defines a perfect correlation (45°), the red line marks the slope

2.1.1 Surface Incoming Shortwave Radiation (equal area grid)

Figure 2-4: Plot (a, left) shows the bias (black dots) and absolute bias (white triangle) for the comparison of SLSTR SIS (equal area grid) and BSRN ground stations. Green shaded area marks the 10 W/m² threshold requirement by GCOS. Plot (b, right) shows the number of available months per station

The reduced validation period (07/2022 – 12/2023) for the SLSTR SIS version on the equal area grid shows generally the same pattern (Figures 2-4 to 2-6). Stations in Izana and Reunion Island are the biggest outliers, while most of the stations´ biases are slightly positive but within the 10 W/m² area. The overall bias is higher compared to the SLSTR SIS on the equal angle grid (4.81 W/m²) but this might be related to a reduced number of months (312) and a different selection of stations. In addition to the previously mentioned outliers, Ny Alesund, another outlier, is located at Spitzbergen, Norway at a very high latitude of 78.93 °N.

Figure 2-5: Temporal average of SIS (equal area grid) from SLSTR and stations from the BSRN. Green dots indicate surface stations where the SLSTR SIS data meets the target accuracy (absolute bias > 10W/m²), black triangles correspond to surface stations, where the SLSTR data does not meet the target accuracy

Figure 2-6: Correlation between SLSTR SIS (equal area grid) and BSRN ground stations. Blue line defines a perfect correlation (45°), the red line marks the slope

2.2 Surface Reflected Shortwave Radiation

Figure 2-7: (a) Bias (black dots) and absolute bias (white triangle) for the comparison of SLSTR SRS and BSRN ground stations. Green shaded area marks the 10 W/m² threshold requirement by GCOS. (b) Number of available months per station

Figures 2-7 to 2-9 show the bias and absolute bias for the SLSTR SRS data. For the outgoing radiations (SRS and SOL) a reduced number of stations is available for the period from 10/2018 on. The majority of stations show a negative bias (-5.22 W/m² on average). Outliers are again noticed for Izana (negative bias) and Boulder (positive bias). Stations south of 60°S latitude show generally a negative bias (Concordia Station, Georg von Neumayer Station and Syowa – all located on the Antarctica continent) and absolute biases are higher than 10 W/m². Other outliers are seen for the high-latitude stations Barrows and Ny-Alesund as well as the high altitude station of Izana. Stations with higher values tend to have a negative bias (see Figure 2-9), while satellite data with generally low values show good agreement with BSRN data.

Figure 2-8: Temporal average of SRS from SLSTR and stations from the BSRN. Green dots indicate surface stations where the SLSTR SRS data meets the target accuracy (absolute bias less than 10W/m²), black triangles correspond to surface stations, where the SLSTR data does not meet the target accuracy

Figure 2-9: Correlation between SLSTR SRS and BSRN ground stations. Blue line defines a perfect correlation (45°), the red line marks the slope

 2.2.1 Surface Reflected Shortwave Radiation (equal area grid)

   Figure 2-10: (a) Bias (black dots) and absolute bias (white triangle) for the comparison of SLSTR SRS (equal area grid) and BSRN ground stations. Green shaded area marks the 10 W/m² threshold requirement by GCOS. (b) Number of available months per station

Comparison of SLSTR SRS data on the equal area grid shows the same pattern with an overall negative bias (-8.99 W/m²). Stations with absolute biases outside of 10 W/m² are generally the same as those mentioned previously while European continental stations (Budapest-Lorinc, Hungary and Payerne, Switzerland) meet the target accuracy requirements for SLSTR (see Figure 2-11). Correlation shows a slight negative bias with a correlation of 0.93.

  Figure 2-11: Temporal average of SRS (equal area grid) from SLSTR and stations from the BSRN. Green dots indicate surface stations where the SLSTR SRS data meets the target accuracy (absolute bias less than 10W/m²), black triangles correspond to surface stations, where the SLSTR data does not meet the target accuracy

    Figure 2-12: Correlation between SLSTR SRS (equal area grid) and BSRN ground stations. Blue line defines a perfect correlation (45°), the red line marks the slope

2.3 Surface Downwelling Longwave Radiation

 Figure 2-13: Plot (a, left) shows the bias (black dots) and absolute bias (white triangle) for the comparison of SLSTR SDL and BSRN ground stations. Green shaded area marks the 10 W/m² threshold requirement by GCOS. Plot (b, right) shows the number of available months per station

 Compared with the previous variables, in the case of SDL most of the stations´ biases are positive (31/37) with more significant outliers for the high-altitude stations of Izana (2373 m), Sonnblick (3109 m) and Yushan (3858 m). Those three stations are clearly visible in figure 2-15 in a pronounced cloud of data points situated separately below the 1-1 line. The majority of the 1701 months shows a slight positive bias for values from 200 W/m² upwards, while generally lower values (<200 W/m²) are underestimated by SLSTR (majority of points belongs to Concordia Station, Antarctica. 26/37 stations have higher absolute bias than 10 W/m² (figure 2-14).

 Figure 2-14: Temporal average of SDL from SLSTR and stations from the BSRN. Green dots indicate surface stations where the SLSTR SDL data meets the target accuracy (absolute bias > 10W/m²), black triangles correspond to surface stations, where the SLSTR data does not meet the target accuracy

 Figure 2-15: Correlation between SLSTR SDL and BSRN ground stations. Blue line defines a perfect correlation (45°), the red line marks the slope. The pronounced cloud of data below the 1-1 line are the outliers for the high-altitude stations of Izana (2373 m), Sonnblick (3109 m) and Yushan (3858 m).

 2.3.1 Surface Downwelling Longwave Radiation (equal area grid)

 Figure 2-16: (a) Bias (black dots) and absolute bias (white triangle) for the comparison of SLSTR SDL (equal area grid) and BSRN ground stations. Green shaded area marks the 10 W/m² threshold requirement by GCOS. (b) Number of available months per station

Figures 2-16 to 2-18 show the same patterns for the SLSTR SDL equal area grid version with reduced months/stations availability. 312 station-months (compared to 1701) result in similar correlation coefficients (0.89 and 0.88 for equal area grid). However, it is worth mentioning that the selection of stations differs for the two comparisons and the absence of the Concordia Station removes parts of the negative bias.

 Figure 2-17: Temporal average of SDL (equal area grid) from SLSTR and stations from the BSRN. Green dots indicate surface stations where the SLSTR SDL data meets the target accuracy (absolute bias > 10W/m²), black triangles correspond to surface stations, where the SLSTR data does not meet the target accuracy

Figure 2-18: Correlation between SLSTR SDL (equal area grid) and BSRN ground stations. Blue line defines a perfect correlation (45°), the red line marks the slope. The pronounced cloud of data below the 1-1 line are the outliers for the high-altitude stations of Izana (2373 m), Sonnblick (3109 m)

 2.4 Surface Outgoing Longwave Radiation

Figure 2-19: (a) Bias (black dots) and absolute bias (white triangle) for the comparison of SLSTR SOL and BSRN ground stations. Green shaded area marks the 10 W/m² threshold requirement by GCOS. (b) Number of available months per station

The SLSTR SOL product exhibits the smallest bias (0.08 W/m²) compared to the other variables. This minimal bias arises because there is no overall trend, as observed with variables like SDL; instead, there is a balanced distribution (see figure 2-19). Correlation between SLSTR and BSRN is 0.97 while most of the months/stations tend to have a small negative bias which is compensated by measurements from Georg von Neumeyer and Izana Stations with positive biases (see figure 2-21).

 Figure 2-20: Temporal average of SOL from SLSTR and stations from the BSRN. Green dots indicate surface stations where the SLSTR SOL data meets the target accuracy (absolute bias > 10W/m²), black triangles correspond to surface stations, where the SLSTR data does not meet the target accuracy

Figure 2-21: Correlation between SLSTR SOL and BSRN ground stations. Blue line defines a perfect correlation (45°), the red line marks the slope

 2.4.1 Surface Outgoing Longwave Radiation (equal area grid)

Figure 2-22: (a) Bias (black dots) and absolute bias (white triangle) for the comparison of SLSTR SOL (equal area grid) and BSRN ground stations. Green shaded area marks the 10 W/m² threshold requirement by GCOS. (b) Number of available months per station

Validation for the SLSTR SOL product on the equal area grid for the period from 07/2022 – 12/2023 is limited to eight available stations (and five for the whole time period) and should therefore be treated with caution. Overall bias is -2.63 W/m² and absolute bias 15.73 W/m² which is mainly due to the Station at Izana, which has a larger impact due to the small number of stations (Figures 2-22 to 2-24).

 Figure 2-23: Temporal average of SOL (equal area grid) from SLSTR and stations from the BSRN. Green dots indicate surface stations where the SLSTR SOL data meets the target accuracy (absolute bias > 10W/m²), black triangles correspond to surface stations, where the SLSTR data does not meet the target accuracy

Figure 2-24: Correlation between SLSTR SOL (equal area grid) and BSRN ground stations. Blue line defines a perfect correlation (45°), the red line marks the slope

3. Application(s) specific assessments

This section is not applicable. There are no additional application specific assessments known since the dataset has just been published.

4.Compliance with user requirements

The GCOS requirements [D3] for the ECV Surface Radiation Budget are used to evaluate the compliance for different users needs. Tables 4-1 and 4-2 show the requirements as well as the results.

GCOS defines three requirements depending on user’s needs:

• Goal (G): The strictest requirement, indicating no further improvements necessary

• Breakthrough (B): Intermediate level between threshold and goal. Breakthrough indicates that it is recommended for certain climate monitoring activities

• Threshold (T): Minimum requirement

The SLSTR ICDR meets the breakthrough/target requirement (closely) for the horizontal/temporal resolution, respectively. However, all variables do not meet the requirements in terms of accuracy. Most of the continental stations are within the threshold requirement of 10 W/m², outliers stand out due to high altitudes, locations on islands or high latitude (e.g. Antarctica, Svalbard, Alaska)

It is worth mentioning that the GCOS requirements, defined by the World Meteorological Organisation (WMO), are not focused on satellite-based data records but also on climate models. Satellite-based data records, especially historical observing systems, are often not able to achieve the requirements.

Table 4-1: Results of evaluation against GCOS requirements for SLSTR SIS

Table 4-2: Results of evaluation against GCOS requirements for SLSTR SRS

Table 4-3: Results of evaluation against GCOS requirements for SLSTR SDL

Table 4-4: Results of evaluation against GCOS requirements for SLSTR SOL

References

Ohmura, A., et al. (1998), Baseline Surface Radiation Network (BSRN/WCRP): New precision radiometry for climate research, Bulletin of the American Meteorological Society, 79(10), 2115-2136. DOI:https://doi.org/10.1175/1520-0477(1998)079<2115:BSRNBW>2.0.CO;2

Related articles

• Page:
  SRB CLARA Product Family: Product Quality Assessment Report (PQAR)
• Page:
  COBRA: Product User Guide and Specification (PUGS)
• Page:
  ERB TSI: Product User Guide and Specification (PUGS)
• Page:
  CP: Climate Change Initiative (CCI)
• Page:
  SRB CCI-ICDR: System Quality Assurance Document (SQAD)