Satellite-based time-series of sea-surface temperature since 1981 for climate applications

The diurnal cycle in SST is typically in the peak-to-peak range of 0.1 K to 0.5 K.

“The diurnal cycle in sea surface temperature has been empirically characterised from sub-daily drifting buoy variability8 and by remote sensing9, and is typically in the peak-to-peak range of 0.1 K to 0.5 K.”

• cites [8] Morak‐Bozzo, S., Merchant, C. J., Kent, E. C., Berry, D. I. & Carella, G. Climatological diurnal variability in sea surface temperature characterized from drifting buoy data. Geosci. Data J. 3, 20–28 (2016). sourcePartly supported

  The Abstract says the authors used sub-daily drifting-buoy records to characterize the SST diurnal cycle, supporting that part of the citing sentence. The Results (§1.2) discuss how its amplitude varies and give a 0.1 K start-to-end difference for one seasonal example, but our copy does not state that the typical peak-to-peak range is 0.1–0.5 K. (LLM, conf=0.82)

  Experimental — accuracy not yet measured
• cites [9] Zhang, H. et al. Comparison of SST diurnal variation models over the Tropical Warm Pool region. J. of Geophys. Res. Oceans 123, 3467–3488 (2018). sourcePartly supported

  Section 4.3 (DV Cycles; page unavailable) reports an average amplitude of 0.4–0.6 K in the original satellite data, which overlaps the claimed range. But the paper considers sensitivity-scaled data more realistic and gives their average amplitude as 0.8–0.9 K, so it does not support 0.1–0.5 K as the typical peak-to-peak range. The Introduction’s 0.1–0.5 K figure describes the cool-skin effect, not the diurnal cycle. (LLM, conf=0.85)

  Experimental — accuracy not yet measured

Bayesian methods of cloud screening were used in the dataset production.

“Bayesian methods of cloud screening of imagery11,12”

• cites [11] Merchant, C. J., Harris, A. R., Maturi, E. & MacCallum, S. Probabilistic physically-based cloud screening of satellite infra-red imagery for operational sea surface temperature retrieval. Quarterly Journal of the Royal Meteorological Society 131(611), 2735–2755 (2005). sourceSupported
• cites [12] Bulgin, C. E., Mittaz, J. P. D., Embury, O., Eastwood, S. & Merchant, C. J. Bayesian cloud detection for 37 Years of Advanced Very High Resolution Radiometer (AVHRR) Global Area Coverage (GAC) data. Remote Sens. 10(1), 97 (2018). sourceSupported

• AssumptionThe authors state this as an assumption: In the cloud detection calculation, P(x|c) is assumed equal to P(x).

In situ SST measurements and upper layers of ocean models typically reflect SST at depths between ~10 cm and ~10 m.

“In situ SST measurements2 and the upper layers of ocean models typically reflect SST at depths between ~10 cm and ~10 m.”

• cites [2] Kent, E. C. et al. A Call for New Approaches to Quantifying Biases in Observations of Sea Surface Temperature. Bull. Amer. Meteor. Soc. 98, 1601–1616 (2017). sourcePartly supported

  The section “What Is SST and How Is It Measured?” says drifting buoys measure at nominal depths of 10–20 cm and moored buoys at about 1 m, supporting the in situ portion of the claim. It does not establish ~10 m as the typical lower boundary; it says measurements down to about 20 m have historically been treated as SST. The cited passage also does not establish the stated depth range for ocean-model upper layers. (LLM, conf=0.83)

  Experimental — accuracy not yet measured

Three major volcanic eruptions caused two periods of elevated stratospheric sulfate aerosol (1982-84, 1991-93) with impacts on infra-red brightness temperatures significant for SST retrieval (e.g., >0.03 K).

“Three major volcanic eruptions (El Chichon22; Pinatubo and Hudson23) caused two periods of elevated stratospheric sulfate aerosol (1982–84, 1991–93) with impacts on infra-red brightness temperatures that are significant for SST retrieval24 (e.g., >0.03 K).”

• cites [22] Matson, M. The 1982 El Chichón Volcano eruptions – A satellite perspective. J. Volcanol. Geotherm. Res. 23(1–2), 1–10 (1984). sourceNot checked — no full text of the source
• cites [23] Lambert, A. et al. Measurements of the evolution of the Mt. Pinatubo aerosol cloud by ISAMS. Geophys. Res. Lett. 20(12), 1287–1290 (1993). sourceNot checked — no full text of the source
• cites [24] Merchant, C. J., Harris, A. R., Murray, M. J. & Zavody, A. M. Toward the elimination of bias in satellite retrievals of skin sea surface temperature 1. Theory, modeling and inter-algorithm comparison. Journal of Geophysical Research 104(C10), 23565–23578 (1999). sourceCould not verify

  The retrieved passages contain bibliographic details and the paper’s title, but no substantive passage addressing volcanic sulfate aerosols or their effect on infrared brightness temperatures. Searches for volcanic aerosols, stratospheric aerosols, brightness-temperature bias, and 0.03 K found no supporting passage in our copy. (LLM, conf=0.74)

  Experimental — accuracy not yet measured

Total SST uncertainty and uncertainty components were estimated context-specifically.

“context-specific estimation of total SST uncertainty and uncertainty components16,17”

• cites [16] Bulgin, C. E., Embury, O., Corlett, G. & Merchant, C. J. Independent uncertainty estimates for coefficient based sea surface temperature retrieval from the Along-Track Scanning Radiometer instruments. Remote Sens. Env. 178, 213–222 (2016). sourceSupported
• cites [17] Bulgin, C. E., Embury, O. & Merchant, C. J. Sampling uncertainty in gridded sea surface temperature products and Advanced Very High Resolution Radiometer (AVHRR) Global Area Coverage (GAC) data. Remote Sens. Env. 178, 287–294 (2016). sourcePartly supported

  The Introduction (p. 4) says this paper models sampling uncertainty, one component of the uncertainty budget, while a companion paper addresses context-specific uncertainties from other error sources. Section 4.2 (p. 16) describes estimating sampling uncertainty from clear-sky coverage and SST variability. This source therefore supports context-specific estimation of a component, but not the attributed estimation of total SST uncertainty and its other components. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

Harmonisation of sensor calibrations was used in the dataset production.

“harmonisation of sensor calibrations13”

• cites [13] Giering, R. et al. A novel framework to harmonise satellite data series for climate applications. Remote Sens. 11(9), 1002 (2019). sourceSupported

Under low-wind, strong-sun conditions, the diurnal cycle in SST can be ~5 K.

“Under low-wind, strong-sun conditions, it can be ~5 K10.”

• cites [10] Gentemann, C. L., Minnett, P. J., Le Borgne, P. & Merchant, C. J. Multi-satellite measurements of large diurnal warming events. Geophysical Research Letters 35(22), L22602 (2008). sourceSupported

Physical modelling of time-of-day adjustments of retrieved SSTs was used to minimise the aliasing of daily SST cycles into long-term trends.

“physical modelling of time-of-day adjustments of retrieved SSTs to minimise the aliasing of daily SST cycles into long-term trends15”

• cites [15] Embury, O., Merchant, C. J. & Corlett, G. K. A reprocessing for climate of sea surface temperature from the Along-Track Scanning Radiometers: initial validation, accounting for skin and diurnal variability. Remote Sens. Env. 116, 62–78 (2012). sourceSupported

Estimates of standard uncertainty are provided for every SST at all product levels.

“Estimates of standard uncertainty4 (which may be considered as the standard deviation of the estimated error distribution) are provided for every SST39 at all product levels.”

• cites [39] Merchant, C. J. et al. Uncertainty information in climate data records from Earth observation. Earth System Science Data 9, 511–527 (2017). sourcePartly supported

  The SST example in the Conclusion says standard uncertainty is given at pixel level, and §5.1 defines it as the standard deviation of the estimated error distribution. But §9 describes per-datum uncertainty estimates at all product levels as an aspiration that is challenging and not fully solved. The source therefore does not support the claim that estimates are provided for every SST at all levels. (LLM, conf=0.86)

  Experimental — accuracy not yet measured

There is a diurnal cycle in SST that has been empirically characterised using satellite observations and drifting buoys.

“There is a diurnal cycle in SST that has been empirically characterised using satellite observations9 and drifting buoys8.”

• cites [9] Zhang, H. et al. Comparison of SST diurnal variation models over the Tropical Warm Pool region. J. of Geophys. Res. Oceans 123, 3467–3488 (2018). sourceSupported
• cites [8] Morak‐Bozzo, S., Merchant, C. J., Kent, E. C., Berry, D. I. & Carella, G. Climatological diurnal variability in sea surface temperature characterized from drifting buoy data. Geosci. Data J. 3, 20–28 (2016). sourceSupported

The physical origin of the correlated errors is understood to arise from the imperfectly accounted-for influence of the atmospheric state on the estimated SST.

“The decorrelation length scales are not yet quantified in detail, but the physical origin of the correlated errors is understood to arise from the imperfectly accounted-for influence of the atmospheric state on the estimated SST41.”

• cites [41] Merchant, C. J., Horrocks, L. A., Eyre, J. R. & O’Carroll, A. G. Retrievals of sea surface temperature from infrared imagery: origin and form of systematic errors. Quarterly J. Roy. Meteorol. Soc. 132, 1205–1223 (2006). sourceSupported

The quality indicator is influenced by the confidence we have that the SST uncertainty estimate for a given SST is valid.

“The quality indicator is influenced by the confidence we have that the SST uncertainty estimate for a given SST is valid39.”

• cites [39] Merchant, C. J. et al. Uncertainty information in climate data records from Earth observation. Earth System Science Data 9, 511–527 (2017). sourceSupported

The uncertainty from sub-sampling is parameterised in terms of the fraction of the cell observed and the variability in SST in the observed part of the cell.

“The uncertainty is parameterised effectively in terms of the fraction of the cell observed and the variability in SST in the observed part of the cell17.”

• cites [17] Bulgin, C. E., Embury, O. & Merchant, C. J. Sampling uncertainty in gridded sea surface temperature products and Advanced Very High Resolution Radiometer (AVHRR) Global Area Coverage (GAC) data. Remote Sens. Env. 178, 287–294 (2016). sourceSupported

Errors in satellite-derived SSTs do not all fall neatly into those arising from random and systematic effects, since errors introduced in the retrieval are locally correlated between pixels.

“Errors in satellite-derived SSTs do not all fall neatly into those arising from random and systematic effects, since errors introduced in the retrieval are locally correlated between pixels40.”

• cites [40] Mittaz, J. P. D., Merchant, C. J. & Woolliams, E. Applying Principles of Metrology to Historical Earth Observations from Satellites. Metrologia 56(3), 032002 (2019). sourceSupported

The SST at 10.30 or 22.30 local mean solar time is a good approximation to the daily mean.

“The SST at 10.30 or 22.30 local mean solar time is a good approximation to the daily mean8.”

• cites [8] Morak‐Bozzo, S., Merchant, C. J., Kent, E. C., Berry, D. I. & Carella, G. Climatological diurnal variability in sea surface temperature characterized from drifting buoy data. Geosci. Data J. 3, 20–28 (2016). sourcePartly supported

  Section 1.2 (Results; page unavailable) reports that the arithmetic mean of SST anomalies at 10:30 a.m. and 10:30 p.m. closely agrees with the daily mean. It does not establish that the SST at either time individually is a good approximation to the daily mean. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

Evaluated total uncertainty is less than 0.8 K for almost all SSTs.

“The evaluated total uncertainty is less than 0.8 K for almost all SSTs”

The mean density of good-quality observations is 13 km−2 yr−1.

“The mean density of good-quality observations is 13 km−2 yr−1.”

• Also stated in: § Background & Summary · plain

  “The number of such SST estimates obtained is 13 km−2 yr−1 on average.”

The dataset is maximally independent from in situ SSTs over the latter two decades of the record.

“despite maximal independence from in situ SSTs over the latter two decades of the record.”

• Also stated in: § Background & Summary · plain

  “A high degree of independence is achieved for the period 1995 to the end of the record, exploiting ATSR-2 and AATSR.”

The L4 analysis represents a multi-satellite estimate of daily mean SST at 20 cm depth.

“The L4 analysis represents a multi-satellite estimate of daily mean SST at 20 cm depth.”

The spatial area represented by pixel SST estimates is between 1 km2 and 45 km2.

“The spatial area represented by pixel SST estimates is between 1 km2 and 45 km2.”

Multi-annual global observational stability relative to drifting buoy SSTs is within 0.003 K yr−1 of zero, with 95% confidence interval −0.0026 to 0.0004 K yr−1.

“Multi-annual observational stability relative to drifting buoy measurements is within 0.003 K yr−1 of zero with high confidence, despite maximal independence from in situ SSTs over the latter two decades of the record.”

• Also stated in: § Background & Summary · plain

  “The multi-annual global observational stability for the time series, relative to drifting buoy SSTs is, with 95% confidence, in the range −0.0026 to 0.0004 K yr−1.”

Adjustments are provided to convert instantaneous skin SST to a depth of 20 cm, nominally corresponding to drifter and historic bucket temperature measurements.

“Here, adjustments are provided to convert the instantaneous skin SST to a depth of 20 cm, nominally corresponding to drifter and historic bucket temperature measurements.”

A climate data record of global sea surface temperature (SST) spanning 1981-2016 has been developed from 4 × 1012 satellite measurements of thermal infra-red radiance.

“A climate data record of global sea surface temperature (SST) spanning 1981–2016 has been developed from 4 × 1012 satellite measurements of thermal infra-red radiance.”

• Also stated in: § Background & Summary · plain

  “This paper presents a climate data record (CDR) of global SST spanning 1981–2016 derived from 4 × 1012 satellite measurements of thermal infra-red (TIR) radiance.”

The TIR measurements were collected by 11 AVHRRs and three ATSRs.

“The TIR measurements were collected by two series of sensors on Earth-orbiting satellites: 11 Advanced Very High Resolution Radiometers (AVHRRs) and three Along-Track Scanning Radiometers (ATSRs).”

The spatial footprint of TIR observations used for good-quality SST retrieval is between 1 km × 1 km and 18 km × 2.5 km.

“The spatial footprint of the TIR observations used for good-quality SST retrieval is between 1 km × 1 km (the resolution of the ATSR imagery at nadir view) and 18 km × 2.5 km (the maximum footprint of AVHRR global area coverage (GAC) pixels used).”

Independence from in situ measurements is lost to a significant degree for the period 1981 to 1995 because in situ SSTs were used as a calibration reference on large scales.

“Prior to that period, it has been necessary in this v2.1 CDR to use in situ SSTs as a “calibration” reference on large scales, so that to a significant degree, independence from in situ measurements is lost when considering the period 1981 to 1995.”

Satellite observations are obtained at local times of day that change through the record.

“The satellite observations are obtained at local times of day that change through the record (Fig. 3 and Table 1).”

For each skin SST estimate in the L2P products, a total uncertainty estimate is provided, which is the standard uncertainty from all sources of error combined.

“For each skin SST estimate in the L2P products, a total uncertainty estimate is provided, which is the standard uncertainty from all sources of error combined.”

The products include an adjustment that, when added to the skin SST, gives an estimate of SST at a depth of 20 cm.

“For this reason, the products include an adjustment which, when added to the skin SST, gives an estimate of the SST at a depth of 20 cm.”

The feature resolution is coarser than the grid cell size.

“Note that the feature resolution is thus coarser than the grid cell size.”

The desert dust index is only available to check for the ATSR series sensors, since it relies on having dual-view observations.

“The desert dust index is only available to check for the ATSR series sensors, since it relies on having dual-view observations.”

An adjustment is calculated for time-of-day effects.

“For this reason, an adjustment is also calculated for time-of-day effects.”

From 1991 onwards, there has always been a mid-morning satellite observing at close to 10.30 or 22.30 local mean solar time needing minimal adjustment.

“Moreover, from 1991 onwards, there has always been a mid-morning satellite observing at close to this local time needing minimal adjustment.”

The SST of a 0.05° cell is often calculated from pixels that do not fill the cell because of cloud cover.

“The SST of a 0.05° cell is often calculated from pixels that do not fill the cell, because of cloud cover”

ESA SST CCI ATSR L2P v2.1 spans 1991 to 2012.

“spanning 1991 to 2012”

ESA SST CCI AVHRR L2P v2.1 spans 1981 to 2016.

“spanning 1981 to 2016”

The uncertainty estimates for SST are valid under clear-sky, low-aerosol conditions.

“the uncertainty estimates for SST are valid under clear-sky, low-aerosol conditions”

The temporal adjustment is an estimate of the change in SST between the observation time and the nearest of 10.30 or 22.30 local mean solar time.

“For these reasons the temporal adjustment is an estimate of the change in SST between the observation time and the nearest of 10.30 or 22.30 local mean solar time.”

The most significant quality factors are undetected cloud and coarse-mode aerosol (primarily desert dust).

“The most significant quality factors are undetected cloud and coarse-mode aerosol (primarily desert dust):”

For aggregated data on spatio-temporal scales coarser than the CCI SSTs, the errors contributing to the total uncertainty cannot be assumed to be fully independent between SST values.

“Where applications involve use of the aggregated data on spatio-temporal scales coarser than the CCI SSTs, the errors contributing to the total uncertainty cannot be assumed to be fully independent between SST values.”

0.001% to 0.5% of records, depending on the sensor, show flag or mask inconsistencies.

“0.001% to 0.5% of records (depending on the sensor) show flag or mask inconsistencies”

There were detrimental increases in rainfall in the South China Sea and western Pacific Ocean, related to warmer SSTs in these regions.

“there were detrimental increases in rainfall (from increased convergence) in the South China Sea and western Pacific Ocean, related to warmer SSTs in these regions.”

The CCI analysis SSTs tend to be warmer than the comparator in the Southern Hemisphere.

“The CCI analysis SSTs tend to be warmer than the comparator in the Southern Hemisphere”

The results of the automated inspection of the dataset show a high degree of technical compliance.

“The results of the automated inspection of the dataset show a high degree of technical compliance.”

Unscreened and unadjusted-for desert dust events cause intermittent negative biases of magnitude 1 K in CCI AVHRR SSTs in the north east tropical Atlantic, Red Sea and Gulf of Arabia.

“Unscreened and unadjusted-for desert dust events cause intermittent negative biases of magnitude 1 K in CCI AVHRR SSTs in the north east tropical Atlantic, Red Sea and Gulf of Arabia”

Regional variability in the SST CCI AVHRR results in latitudes affected by dust aerosols manifests as a cool bias in the SST CCI analysis results in the zone from 0° to 20° N.

“Regional variability in the SST CCI AVHRR results in latitudes affected by dust aerosols manifest as a cool bias in the SST CCI analysis results in the zone from 0° to 20° N.”

Spot checking shows that files containing only fill value data originate from periods of outgassing or satellite commissioning-phase acquisitions.

“spot checking shows that these files originate from periods of outgassing or are satellite commissioning-phase acquisitions.”

The ATSR-series SST observations show excellent stability, especially during the ATSR-2 and AATSR periods.

“The excellent stability of the ATSR series observations, especially for the ATSR-2 and AATSR periods, is emphasised in panel b.”

The periods May 1982, October to December 1982, early August 1983 and late September 1983 show biases not accounted for within the stated uncertainties that introduce artefacts in the global-mean CCI analysis SST appearing to be in the range of 0.1 to 0.5 K.

“The following periods show biases not accounted for within the stated uncertainties that introduce artefacts in the global-mean CCI analysis SST appearing to be in the range of 0.1 to 0.5 K: May 1982, October to December 1982, early August 1983 and late September 1983.”

In atmosphere-only simulations forced with CCI analysis SSTs, the differences from compatible simulations forced with HadISST.2.2.0.0 were generally smaller than differences arising from changes in model resolution.

“The differences were generally smaller than differences arising from changes in model resolution.”

In the monsoon regions, colder Arabian Sea SSTs in the L4 analysis reduced the moisture flux over the western Ghats mountains.

“In the monsoon regions, colder Arabian Sea SSTs in the L4 analysis reduced the moisture flux over the western Ghats mountains”

Around 0.005% of the SST files contain only fill value data.

“Around 0.005% of the SST files contain only fill value data”

Some periods of degraded calibration of the AVHRR sensors cause temporary observational instabilities in the CDR.

“Some periods of degraded calibration of these sensors cause temporary observational instabilities in the CDR.”

The sensitivity of ATSR-series sensors to desert dust events is much less than that of AVHRR.

“the sensitivity of ATSR-series sensors to these events is much less.”

Flag or mask inconsistencies only appear at quality level 1 ("bad_data").

“these inconsistencies only appear at quality level 1 (“bad_data”)”

Warmer SST around the Maritime Continent and cooler SST around the equatorial Atlantic and eastern-boundary upwelling regions were associated with reduced bias in cloud regimes.

“Warmer SST around the Maritime Continent and as well as cooler SST around the equatorial Atlantic and upwelling regions of at the eastern boundaries of ocean basins were associated with reduced bias in cloud regimes”

CCI AVHRR SSTs are generally better during the period of overlap with ATSRs from the 1990s onwards.

“CCI AVHRR SSTs are also generally better during the period of overlap with ATSRs from the 1990s onwards.”

For the 12 μm channel, more confidence is placed in ATSR-2 than AATSR.

“therefore for that channel we have more confidence in ATSR-2.”

Different overpass times differentially sample SST, generating non-climatic signals if not adjusted for.

“Different overpass times differentially sample SST, generating non-climatic signals if not adjusted for.”

Single-view coefficient-based retrievals are associated with geographical biases arising from this information content deficit.

“Particularly for day-lit scenes, where only the 11 and 12 µm channels are used, single-view coefficient-based retrievals are associated with geographical biases arising from this information content deficit38.”

• cites [38] Merchant, C. J., Harris, A. R., Roquet, H. & Le Borgne, P. Retrieval characteristics of non-linear sea surface temperature from the Advanced Very High Resolution Radiometer. Geophysical Research Letters 36(17), L17604 (2009). sourceNot checked — no full text of the source

SSTs derived from IR radiances are sensitive to variation in temperature of the skin layer of the ocean.

“SSTs derived from IR radiances are sensitive to the variation in temperature of the skin layer of the ocean5.”

• cites [5] Minnett, P. J., Smith, M. & Ward, B. Measurements of the oceanic thermal skin effect. Deep-Sea Research Part II: Topical Studies in Oceanography. 58(6), 861–868 (2011). sourceNot checked — no full text of the source

AVHRRs are single-view sensors, which reduces the information content available for determining SST compared to the dual-view ATSRs.

“Unlike the ATSR-series, the AVHRRs are single-view sensors, which reduces the information content available for determining SST compared to the dual-view ATSRs.”

• cites [38] Merchant, C. J., Harris, A. R., Roquet, H. & Le Borgne, P. Retrieval characteristics of non-linear sea surface temperature from the Advanced Very High Resolution Radiometer. Geophysical Research Letters 36(17), L17604 (2009). sourceNot checked — no full text of the source

A high level of independence from in situ SST observations is preserved, despite use of prior information for the inverse.

“Thus, a high level of independence from in situ SST observations is preserved, despite use of prior information for the above inverse.”

Cloud detection is applied to the satellite imagery to minimise cloud biases in SSTs.

“Cloud detection is applied to the satellite imagery to minimise cloud biases in SSTs.”

• AssumptionThe authors state this as an assumption: In the cloud detection calculation, P(x|c) is assumed equal to P(x).

Taken together, the uncertainty and stability statistics suggest the dataset gives a detailed representation of SST variability on a range of space and time scales relevant to climate applications.

“Taken together, these statistics suggest that the dataset gives a detailed representation of SST variability on a range of space and time scales of relevance to climate applications.”

The influence of the prior SST on the retrieved SST is controlled to be <5% for SSTs given a quality indication of 4 or 5.

“we choose to minimise the influence of the prior SST on the retrieved state by adopting an inflated uncertainty in the prior SST that is sufficiently small to provide useful regularisation of the inverse and sufficiently large that the influence of the prior on the retrieved SST is controlled to be <5% for the SSTs given a quality indication of 4 or 5 (see below).”

Adding the depth adjustment makes the resulting SST nominally comparable to what would be measured by a drifting buoy at the satellite observation time.

“By adding this adjustment, the resulting SST is nominally comparable to what would be measured by a drifting buoy at the satellite observation time.”

There is no correlation of the sub-sampling effect between cells, so it contributes to the uncorrelated component of uncertainty in the L3U SST products.

“There is no correlation of this effect between cells, so this contributes to the uncorrelated component of uncertainty in the L3U SST products.”

The uncertainty from the time and depth adjustments is calculated and included in the total uncertainty provided for the daily mean depth SST estimate.

“The uncertainty from this adjustment is also calculated and included in the total uncertainty provided for the daily mean depth SST estimate.”

Sub-sampling is another source of uncertainty in the gridded SST.

“and therefore the sub-sampling is another source of uncertainty.”

• AssumptionRests on an assumption: Users typically treat the gridded SST as a value representative of the cell as a whole.

SSTs with relatively high uncertainty can still be flagged as good quality, provided there is nothing to indicate that the assumptions made in estimating the uncertainty are compromised.

“SSTs with relatively high uncertainty can still therefore be flagged as good quality, provided there is nothing to indicate that the assumptions made in estimating the uncertainty are compromised.”

The quality levels of pixels with large chi-square are downgraded.

“The quality levels of pixels with large chi-square are therefore downgraded.”

The parameterisation of B affects the location and degree of smoothing of SST features that inevitably occurs when in-filling the gappy observations of SST of a given day.

“A key aspect of the variational assimilation scheme is the parameterisation of B, which affects the location and degree of smoothing of SST features that inevitably occurs when in-filling the gappy observations of SST of a given day.”

Aliasing of the diurnal cycle with varying times of observation will produce spurious inter-annual trends if not adjusted for.

“Aliasing of this diurnal cycle with varying times of observation will produce spurious inter-annual trends if not adjusted for.”

Errors that are independent (uncorrelated) between observations arise from the instrumental noise in the satellite observations of brightness temperature.

“Errors that are independent (uncorrelated) between observations arise from the instrumental noise in the satellite observations of brightness temperature.”

The quality levels of SSTs with low sensitivity are downgraded.

“For this reason, the quality levels of SSTs with low sensitivity are downgraded.”

B influences the feature resolution of the analysis.

“B influences the feature resolution of the analysis.”

All products have been validated against in situ measurements of SST according to the Product Validation Plan.

“All products have been validated against in situ measurements of SST according to the Product Validation Plan57.”

• cites [57] SST-CCI Product Validation Plan (PVP) SST_CCI-PVP-UOL-001, Issue 2, 4 February 2014, available at, http://www.esa-sst-cci.org.Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The flag or mask inconsistencies do not affect the recommended uses of the dataset.

“and so do not affect the recommended uses of the dataset”

Because only AVHRRs are available during the first decade of the CDR, the CCI analysis SST anomalies show an exaggerated positive trend in dust-affected regions through the time series, thought to be about 0.01 to 0.02 K yr−1.

“Since only AVHRRs are available during the first decade of the CDR, whereas ATSRs were available from 1991 to 2012, the CCI analysis SST anomalies show an exaggerated positive trend in these regions through the time series, thought to be about 0.01 to 0.02 K yr−1.”

Some of the larger CCI-versus-in-situ SST discrepancies during the 1980s reflect the relative sparsity and quality of the in situ network at that time, as well as artefacts in the CCI SSTs.

“Note that some of larger discrepancies during the 1980s reflect the relative sparsity and quality of the in situ network at that time, as well as artefacts in the CCI SSTs discussed further in the usage notes that follow.”