Contributors: D. Ghent (UoL), K. Veal (UoL), G.Kirches (BC)

Issued by: University of Leicester/ Darren Ghent

Date: 07/10/2025

Ref: C3S2_313e_BC_WP3-DR-LST-LST_CCI-v3.00-1995-2024_202506_ATBD

Official reference number service contract: 2024/C3S2_313e_BC/SC1

<style>
.special_indent ul > li > ul {
    padding-left: 0;
}

.special_indent ul > li > ul > li > ul {
    padding-left: 1.5em;
}
</style>

History of modifications

List of datasets covered by this document

Acronyms

General definitions

Biome: The term biome in LST science is used to define the land cover classification. For some LST retrieval schemes coefficients may be categorised by biome. The ULeic / LandCover_cci (LCCS) hybrid biome classification contains for example 43 land cover classes.

Brightness temperature (BT): The Brightness Temperature is the temperature of a black body that would have the same radiance as the radiance actually observed with the radiometer. Specifically it indicates the directional temperature obtained by equating the measured radiance with the integral over wavelength of the Planck’s Black Body function times the sensor response (Norman and Becker, 1995).

Brokered Climate Data Record: Climate Data Record processed by a partner agency and then integrated into C3S. 

Calibration: Calibration is the process of quantitatively defining the system response to known, controlled system inputs (Guillevic et al., 2018).

Error: Result of a measurement minus a true value of the measurand. Note that in practice a true value cannot be determined and therefore a conventional true value is used instead (JCGM, 2008).

Fractional Vegetation Cover (FVC): The ratio of the vertical projection area of vegetation on the ground to the total vegetation area. For satellite data In simplified terms this is the fraction of a specified area that is covered by green vegetation. For LST retrieval this parameter has often been used to infer emissivity.

Granule: A group of pixels within an orbit that correspond to x number of pixels across the orbit track and y number of pixels along the orbit track.

Land Surface emissivity (LSE): Emissivity describes a material’s ability to emit the thermal energy which it has absorbed (Guillevic et al., 2018).

Land surface temperature (LST): Land Surface Temperature (LST) is the aggregated radiative skin temperature derived from thermal radiation of all objects comprising the surface, measured in situ and estimated from satellite. It is a basic determinant of the terrestrial thermal behaviour, as it controls the effective radiating temperature of the Earth’s surface (Norman and Becker, 1995; Guillevic et al., 2018).

Reference standard: Measurement standard designated for the calibration of other measurements standards for quantities of a given kind in a given organization or at a given location (Guillevic et al., 2018).

Satellite azimuth angle: The length of the arc of the horizon (in degrees) intercepted between North and the direction of the satellite from the observation point measured clockwise from the reference direction. This is frequently referred to as the Viewing Azimuth Angle (VAA).

Satellite zenith angle: The angle between a straight line from a point on the earth's surface to the satellite and a line from the same point on the earth's surface that is perpendicular to the earth's surface at that point. This is frequently referred to as the Viewing Zenith Angle (VZA).

Skin temperature: The temperature of a layer of a medium of depth equal to the penetration depth of the electromagnetic radiation at the given wavelengths (Norman and Becker, 1995). Surface brightness and radiometric temperatures are the effective temperature that a radiometer would measure near the surface, including emissivity effects and reflected downwelling radiance.

Split-Window (SW): Refers to the use of adjacent infrared bands to correct for atmospheric effects based on differential absorption (Wan and Dozier, 1996).

Total column water vapour (TCWV): Amount of water (depth of vertical column of unit-crossectional area) which would be obtained if all the water vapour in a specified column of the atmosphere were condensed to liquid. The equivalent term precipitable water (PW) is also frequently used.

Uncertainty: A parameter associated with the result of a measurement, that characterizes the dispersion of the values that could reasonably be attributed to the measurand, that is the value of the particular quantity to be measured (JCGM, 2008).

Executive summary

Land Surface Temperature (LST) refers to the skin temperature of the Earth’s surface and serves as an approximation of the thermodynamic temperature based on radiance measurements at regional and global scales. LST is typically derived from remote sensing measurements in the thermal infrared (TIR) supporting the assessment of the Earth’s radiative energy budget. Recognised as an Essential Climate Variable (ECV) by the Global Climate Observing System (GCOS) LST is a critical parameter for characterising the Earth’s climate system.

Many climate applications and scientific challenges depend on LST knowledge such as climate modelling, data assimilation, numerical weather prediction, forecasting and reanalyses, drought monitoring and crop management, and land-atmosphere feedbacks. LST is a significant factor in understanding energy balance as it controls energy partition into latent and sensible heat and is an indicator of strong surface warming trends.

The Copernicus Climate Change Service (C3S) delivers a Climate Data Record (CDR) for LST for the time period 1995 – 2024 inclusive, which will be continued forward in time as an Interim Climate Data Record (ICDR). The C3S LST CDR provides the global-scale information for climate applications. An objective of the C3S product is to ease the use of LST data for global climate applications, trend analysis and model evaluation. The output Level-3S global monthly datafiles are provided in NetCDF CF-compliant format and include LST as the main variable, with accompanying variables on LST uncertainty and quality flags. Two datafiles are provided for each month, one for daytime LST and one for night-time LST. Data are produced and output at 0.01˚ spatial resolution and masked by land cover class to differentiate land from water.

The C3S LST is brokered from both the Climate Change Initiative Land Surface Temperature (LST_cci) project and the UK Earth Observation Climate Information Service (EOCIS). LST_cci provides the CDR covering the years 1995 – 2023, and EOCIS provides the Interim Climate Data Record (ICDR) which covers the full 12-months of 2024. The data record is a multi-sensor product which is called the Infrared Climate Data Record (IRCDR) with a version number of 3.00. All data across the multi-sensor CDR and ICDR are consistent in terms of intercalibration, calibration database, retrieval algorithm, uncertainty model, and cloud detection approach. The contribution of each of the individual sensors to the global monthly multi-sensor IRCDR can be summarised as:

• ERS-2 Along Track Scanning Radiometer (ATSR-2): 1995-06-01 to 2002-05-31
• Envisat Advanced Along Track Scanning Radiometer (AATSR): 2002-06-01 to 2012-03-31
• EOS Terra Moderate Resolution Imaging Spectroradiometer (MODIS): 2012-04-01 to 2018-11-30
• Sentinel-3B Sea and Land Surface Temperature Radiometer (SLSTR): 2018-12-01 to 2024-12-31.

This Algorithm Theoretical Basis Document (ATBD) describes the algorithms used to generate the C3S LST product version 3.00. Section 1 introduces the four missions from which the C3S LST product is derived. Section 2 catalogues both the input Level 1 data and the auxiliary data used in the generation of the LST product. Section 3 describes the algorithms used to generate the output data, including the retrieval algorithm, the uncertainty model and the cloud masking approach. A flowchart is included to illustrate all the main steps of the processing chain. Finally, section 4 briefly describes the output format and provides illustrations of the main variables of the output product. A separate Product User Guide and Specification (PUGS) document details the full format and specifications for the LST product. The Product Quality Assessment Report (PQAR) document provides information on the quality assurance of the LST product, including verification and validation details.

Missions and Instruments

The brokered IRCDR is a multi-sensor dataset encompassing the missions ATSR-2, AASTR, Terra-MODIS and SLSTR-B. The University of Leicester C3S processing chain on the UK JASMIN facility processes each of these datasets and combines them into the IRCDR. Here a brief summary of each sensor is provided. 

ATSR-2 and AATSR

The Along Track Scanning Radiometer (ATSR) series of instruments include ATSR-2 and AATSR (Advanced Along-Track Scanning Radiometer). These were launched on board European Space Agency (ESA) sun synchronous, polar orbiting satellites ERS-2 in April 1995, and Envisat (Environmental Satellite) in March 2002, respectively. The last of these instruments – AATSR – provided its final data on 8th April 2012. These ATSRs therefore provide approximately 17 years of data. Continuation of this sensor series occurred, albeit with a data gap of approximately 4 years, with the launch of the Sea and Land Surface Temperature Radiometer (SLSTR) sensors on board Sentinel-3 satellites (see Section 1.3).

All ATSR instruments used similar orbits and equator crossing times ensuring a high level of consistency. With a swath width of 512 km, AATSR was able to provide approximately 3-day global LST coverage with a repeat cycle of 35 days. The overpass of AATSR was 10:00 (local solar time) in its descending node and 22:00 (local solar time) in its ascending node. For ATSR-2 the overpass times were 10:30 and 22:30 in the descending and ascending nodes respectively. The orbit of the ATSRs was very stable in local crossing times and no notable orbital drifts occurred.

AATSR had good radiometric accuracy of less than 0.1 K in the mid-range of surface temperatures for both 11 and 12 μm brightness temperatures, which makes its data ideal for use in climate data records. It achieved excellent radiometric stability by using two blackbodies for onboard calibration, which were viewed on each scan cycle. Stirling Cycle coolers maintained the infrared detectors with low radiometric noise. All three ATSRs had similar specifications with near-infrared (NIR) / infrared (IR) channels at 1.6, 3.7, 11 and 12 μm. Both ATSR-2 and AATSR had three additional visible channels at 0.55, 0.66 and 0.87 μm for extending the application of ATSR data into the land domain. A distinguishing feature of the ATSRs was the dual-angle (DA) capability (nadir and forward at an angle of ~55° to nadir). However, only the nadir view has generally been utilised in LST retrievals, LST_cci included. The rationale on the use of the nadir view only is provided in Soria and Sobrino (2007), which assessed both SW and DA over topographically flat and homogeneous rice fields and found DA algorithms to be less accurate.

More details can be found at https://atsrsensors.org/.

MODIS

The MODIS (Moderate Resolution Imaging Spectroradiometer) instrument was launched on board the sun-synchronous, near-polar orbiting satellite Terra (EOS AM-1) on 18 December 1999. The instrument provides a pair of observations each day acquiring data in 36 spectral bands. Terra-MODIS acquires data at approximately 10:30am (local solar time) in its descending node and at approximately 10:30pm (local solar time) in its ascending node. The swath width of the instrument, 2330 km, enables the satellite to view almost the entire surface of the Earth every day. The spatial resolution of the thermal bands is 1 km, with both land surface temperature and land surface emissivity being core products from this instrument.

More details can be found at https://modis.gsfc.nasa.gov/about/.

SLSTR

The Sea and Land Surface Temperature Radiometer (SLSTR) on board the Sentinel satellites 3-A and 3-B is based on the principles of AATSR. The Copernicus Sentinel-3 constellation responds to the requirement for an operational and near-real-time monitoring of  LST over a period of 15 to 20 years. Sentinel-3A was launched ion 16^th February 2016, and Sentinel-3B was launched on 25^th April 2017.

SLSTR  is  designed  to  retrieve  global  sea-surface  temperatures  to  an accuracy of better than 0.3 K and global land surface temperature to an accuracy of less than 1 K. Like AATSR a dual view capability is  maintained with SLSTR - the nadir swath being 1420 km, and the backward view having a swath width of 750 km. This supports a maximum revisit time of 4 days in dual view and 1 day in single  view. There are  nine spectral channels including two additional bands optimised for fire monitoring and improved cloud detection. The spatial resolution of SLSTR is 500 m in the visible and shortwave infrared channels and 1 km in the thermal infrared channels. The baseline retrieval for the operational ESA SLSTR LST product consists of a nadir-only split-window algorithm with classes of coefficients for each land cover-diurnal  (day/night) combination.

More details can be found at https://sentiwiki.copernicus.eu/web/s3-slstr-instrument.

Input and auxiliary data

Level-1B input data

In this section the source input Level 1 data is presented for each of the four missions which constitute the IRCDR: ATSR-2, AASTR, Terra-MODIS and SLSTR-B.

ERS-2 ATSR-2

Envisat AATSR

Terra MODIS

Sentinel-3 SLSTR-B

Auxiliary data

In this section the auxiliary data used in the retrieval and cloud clearing algorithms for the IRCDR processing are presented. The ERA5 data are used both in the derivation of coefficients for the LST retrieval algorithm and the cloud detection algorithm. The CAMEL dataset is used as the source for the emissivity in the LST retrieval algorithm. The OSTIA dataset is used to distinguish between sea-ice and open water.

ERA5

CAMEL emissivity

OSTIA sea-ice coverage

Algorithms

LST Retrieval Algorithm

The most common and simplest deterministic approach for retrieving LST is the the split-window (SW) method. This can be used without having to input external atmospheric profiles or having to explicitly solve the radiative transfer equation. The basis for this method revolves around the use of two bands typically in the 10 – 12 μm atmospheric window region, whereby atmospheric effects are compensated for by using the differential absorption characteristics between those bands. This approach has been extensively used to compute sea surface temperature (Merchant et al., 1999) where the emissivity of water is well known. Over land the assumption is that emissivities in the SW bands being used are stable and well known and can be assigned values based on relationships with vegetation indices or from a land-cover classification map (Hulley et al., 2019).

The Generalised Split Window (GSW) algorithm is one formulation of the SW approach, which is used for each sensor in the multi-sensor IRCDR. The generalised split window algorithm is a view-angle dependent split-window algorithm proposed for LST retrieval by Wan and Dozier (1996). It is based around channels in the 11 and 12 µm regions. The performance of the generalized split-window LST algorithm requires knowledge about the band emissivities for real land surfaces. In the LST_cci GSW method, emissivity information is used explicitly rather than incorporating this information implicitly through land cover coefficients. In the operational MODIS implementation, band averaged emissivities for each of the two channels centred around 11 and 12 µm are separately calculated:

\varepsilon_{i} = \dfrac{\int_{\lambda_1}^{\lambda_2}\Psi(\lambda)\varepsilon(\lambda)B(\lambda, T_s)d\lambda}{\int_{\lambda_1}^{\lambda_2}\Psi(\lambda)B(\lambda, T_s)d\lambda} \quad (eq. 1)

where λ₁ and λ₂ are the upper and lower bounds of the channel λ, ε(λ) is the channel emissivity, Ψ(λ) is instrument response function, and B(λ,T_s) is the radiance emitted by a blackbody at surface temperature T_s. The emissivity per channel is assigned on a pixel basis according to land cover class. In cases of mixed pixels the emissivity is calculated based upon the proportion of the pixel assigned to each land cover class. To determine LST, both the mean emissivity of the two channels (11 and 12 µm) and the difference in emissivity between them are required. The mean emissivity of the two thermal channels (ε_mean) is derived as:

\varepsilon_{mean} = 0.5(\varepsilon_{11} + \varepsilon_{12}) \quad (eq. 2)

The difference in emissivity between the two thermal channels (Δε) is calculated as:

\Delta\varepsilon = \varepsilon_{11} - \varepsilon_{12} \quad (eq. 3)

LST is now calculated as:

T_s = C + \left(A_1 + A_2\dfrac{1 - \varepsilon_{mean}}{\varepsilon_{mean}} + A_3\dfrac{\Delta\varepsilon}{\varepsilon^2_{mean}}\right)\dfrac{T_1 + T_2}{2} + \left(B_1 + B_2\dfrac{1 - \varepsilon_{mean}}{\varepsilon_{mean}} + B_3\dfrac{\Delta\varepsilon}{\varepsilon^2_{mean}}\right)\dfrac{T_1 - T_2}{2} \quad (eq. 4)

where C, A₁, A₂, A₃, B₁, B₂ and B₃ are retrieval coefficients. T₁ and T₂ are the 11 and 12 µm brightness temperatures.  The coefficients for GSW are dependent on satellite viewing angle and water vapour. Error analysis (Ghent at el., 2019) shows that viewing angle and atmospheric column water vapour must be considered in the retrieval to achieve highest accuracy over the wide atmospheric and surface conditions. The bands for water vapour have a width of 15 kg⋅m^-2 so that the first water vapour band spans from 0 to 15 kg⋅m^-2. The bands for satellite zenith angle have a width of 5°. The retrieval coefficients are linearly interpolated between viewing angle and water vapour bands to minimise step changes.

Radiative Transfer Modelling

The LST retrieval algorithm (Equation 4) requires pre-computed retrieval coefficients for the LST processing chain. These coefficients are derived with linear regression on data simulated using radiative transfer modelling. An advantage of this approach is independence from in situ measurements, which is an important consideration for the objective of developing a Climate Data Record. A critical consideration for the implementation is for fast processing of sufficient numbers of profiles to adequately characterize the entire range of potential atmospheric states representative of each biome class (Ghent et al., 2017).

Radiative Transfer for TOVS (RTTOV) is a fast Radiative Transfer Model (RTM) (Saunders, 2001). It is an efficient radiative transfer forward model for the visible, infra-red and microwave wavelengths. In contrast to models using a line-by-line methodology (Dudhia, 2017), RTTOV conceptualizes the simulation in terms of channel radiances. It therefore requires both an Instrument Spectral Response Function (ISRF) and a pre-calculated set of coefficients relating the channel to sensitivities to various atmospheric parameters. These coefficients parameterize the gas contributions to transmittances associated with the profile. This approach results in significantly increased computational speed for RTTOV compared to the line-by-line methodology (Matricardi et al., 2009), so that a number of profiles sufficient to characterise the range of atmospheric states representative of each land cover class can be achieved (Ghent et al., 2017).

RTTOV Version 12.3 is used for the brokered dataset from LST_cci. Retrieval coefficients are derived using forward modelling. Specifically, regressions between the skin temperature and the Top-Of-Atmosphere (TOA) radiances are used to populate a Calibration Database.

Calibration Database for Determining Retrieval Coefficients

Globally robust, traceable retrieval coefficients for the GSW algorithm are generated to adequately characterise a wide range of potential atmospheric states representative of each land cover. Simulated brightness temperatures and LSTs are derived from RTTOV given inputs of vertical atmospheric profiles, surface and near-surface conditions, surface emissivities, and the spectral response function of the sensor of interest.

A calibration database consisting of a range of LST, TOA Radiance, Atmospheric profiles and varied emissivity is required to both train and test the algorithm. This database must be representative and portable so as to provide consistency for different groups developing algorithms. The LST_cci Calibration Database (CCICDB) is designed to be the main environment in which any thermal IR coefficient-based algorithm can be trained and tested.

The generation of the simulated data set is based on modelled atmospheric data and ancillary sources for emissivity, fractional vegetation and biome information. All of these inputs must be used with a common forward model to result in useful and robust data at both the surface LST level and for the observed TOA brightness temperatures.

The calibration dataset is constructed from the following data sources:

• ECMWF Re-analysis 5 (ERA5) atmospheric profile data (C3S, 2017)
• Combined ASTER and MODIS Emissivity for Land (CAMEL) emissivity data (v2) (Borbas et al., 2018)
• RTTOV Forward Model (v12.3) (Saunders, 2001)
• ULeic / LandCover_cci (LCCS) hybrid biome map (Ghent et al., 2024)

In order to construct a robust and representative dataset a rigorous procedure is developed. The flow of input data and the significant processing steps are outlined in Figure 1.

Figure 1: Flow chart for the generation of the simulated true Top-Of-Atmosphere Brightness Temperature values. The input data include the ERA5 profiles and the CAMEL emissivities. These are configured to produce the Simulation Truth dataset (SIM-TRUTH) and run through the RTTOV radiative transfer model to produce the simulated channel TOA BTs (SIM-OBS) which form the LST_cci Calibration Database (CCICDB).

The process starts with the ERA5 profile data. The initial parameter considered is the atmosphere, primarily focussing on the water vapour as the most significant influence on the retrieval of LST. The profiles are drawn from primarily cloud free data, where the cloud fraction field in ERA5 is less than 10%. The ERA5 data is gridded at 0.25° and contains both full water vapour profiles as well as the Total Column Water Vapour (TCWV) values. Averaging the full profile can result in an overly smoothed product, which might not be representative of real atmospheres. To avoid this the ERA5 data is gridded to cells of 2.5x2.5° and the mean TCWV values calculated. The mean total column water vapour is then compared to all the TCWV values in the 2.5x2.5° grid cell. This grid cell size is significantly finer than the resolution used in the Round Robin assessment (Perry et al., 2020), and as such results in a higher number of points spatially. A further selection criterion involves only 10% of the points above 60°N and below 60°S are selected to avoid polar samples having an unrepresentative influence on the dataset.

The profile that corresponds to the closet matching TCWV value is selected as the representative profile for the grid cell. This methodology ensures that not only the profile is physically realistic, but also that it represents the most likely conditions for the majority of the grid cell, which is essential when comparing the profile to other parameters such as biome and emissivity which are sourced from different datasets. Any variation in the profiles within the grid cell is explored in the sensitivity on water vapour. The sensitivity is explored by comparing the variation of water vapour for each grid cell of the selected dataset with the standard deviations of the climatology of water vapour. All other information such as LST, elevation, and other gas profiles are also selected using the index of this profile.

The second step of processing involves emissivity from CAMEL. The emissivity is obtained at the locations selected in the ERA5 data and averaged from 0.05° to the 0.25° spatial resolution to match the ERA5 data used in the simulation. This does reduce the impact of the most extreme emissivity values, but results in values more representative of those expected at the 1 km resolution of the IRCDR. All of the data is collated and the biome added. All parameters are input to the RTTOV v12 forward model along with the sensor Instrument Spectral Response Function (ISRF). RTTOV then simulates the BTs for each sensor for the given simulated truth.

The Calibration Database requires two datasets (“Training” and “Testing”) both representative for the perturbations of parameters. The multi-year data is sub-divided (Perry et al., 2020) so that odd years are assigned to the “Training” partition of the Calibration Database and even years to the “Testing” partition. The “Training” set are the profiles and surface conditions used to develop the algorithm coefficients and train the retrievals. The “Testing” set is used to assess the performance and to derive robust uncertainty estimates for the retrievals. For both subsets an intra-annual temporal sampling method is based on three daily samples per month on the 5th, 15th and 25th day.

Cloud Masking

A consistent cloud detection method is applied to all sensors contributing to the multi-sensor IRCDR. This algorithm (named UOL_3) is a semi-Bayesian cloud masking approach using the probability of clear-sky conditions which has been developed at University of Leicester (Ghent et al., 2017). A pixel-level cloud mask is derived using a combination of simulated brightness temperatures and observational climatology. The approach is equally valid for both day and night-time retrievals as this method is independent of visible wavelength information.

This cloud masking algorithm uses atmospheric profile data to predict clear-sky conditions for the coincident space and time of a given satellite sensor observation. Coincident clear-sky brightness temperatures are derived by bilinear interpolation between surrounding ECMWF ERA5 profile locations and a temporal interpolation between the hourly analysis fields. On a spatial plane these modelled profile data correspond to the tie-point grid of the respective instrument and orbit granules.

For instance, in the case of Terra-MODIS an observational climatology for the full years 2001 - 2021 is acquired for each 5x5° grid cell for each of the land covers and diurnal conditions (day/night). This is stratified by the 43 biomes of the ULeic / LandCover_cci (LCCS) hybrid biome map (Ghent et al., 2024) described in Section 3.4.2. The mean and standard deviations for clear-sky conditions are stored in a Look up table (LUT). Using RTTOV, expected clear-sky brightness temperatures / brightness temperature differences are simulated for these profile data. To calculate the clear-sky probability at each pixel location a probability density function (PDF) assuming a normal distribution is constructed. The simulated mean brightness temperatures for the corresponding granule and the standard deviation of the brightness temperature from the observational climatology define the PDF as shown in Figure 2 (Bulgin et al., 2014). The observational climatology corresponds to 5x5° grid cells for the given month, land cover and diurnal state. A per-pixel cloud mask is generated from comparing the pixel brightness temperatures/brightness temperature differences with the pixel probability density functions. Pixels are identified as cloudy if the combined probabilities are less than a set of confidence thresholds. The thresholds themselves are simply for converting the probabilities into a binary mask. For daytime observations, the cloud flag is set if either the observed 12 µm brightness temperature or 11 - 12 µm brightness temperature difference fall outside of the 95% confidence levels of the corresponding simulated PDFs. For night-time observations, the 12 µm brightness temperature and the 11 - 3.7 µm differences are used. For granules where insufficient profile data are available to simulate the expected brightness temperatures then the operational cloud flags are used instead. This is also the case where incompatibilities between the atmospheric and surface states result in an RTTOV error (which is a rare occurrence).

Figure 2: For each granule of an orbit (left), the expected 12 µm brightness temperature is simulated from coincident profiles. The PDF of observed 12 µm brightness temperatures for each land cover-diurnal condition, given the space and time position, is also determined (top-right in green). This PDF is then re-centred around the expected BT mean for the granule from the simulations, and this re-centred PDF represents the expected clear-sky conditions (bottom-right in green) (from Bulgin et al. 2014). 

Auxiliary Data

Emissivity

The Combined ASTER and MODIS Emissivity for Land (CAMEL) database is a global monthly mean emissivity dataset. It assimilates both ASTER Global Emissivity Database retrieved values and University of Wisconsin-Madison MODIS Infra-red Emissivity dataset values. The CAMEL dataset contains 13 emissivity values at different wavelengths from 3.6 to 14.3 µm at a resolution of 0.05° (Borbas et al., 2018). The dataset is created from satellite observations, and thus is appropriate for use in satellite retrievals to be representative of realistic materials observed from space at the scale of a satellite pixel.

CAMEL emissivity data is used in the GSW algorithm to calculate LST. Figure 3 illustrates the average global emissivity from CAMEL for the 10.8 and 12 μm channels of MODIS, which shows non-climatic temporal trends in both the CAMEL V2 and V3 datasets. This is due to a trend in temperature being artificially introduced into the emissivity when the algorithm was moved from ASTER to MODIS Temperature-Emissivity Separation (TES). In LST_cci, all split-window algorithms use a climatology of CAMEL V2 to remove these non-climatic temporal trends. Emissivity from CAMEL is also employed in the Calibration Database for determining retrieval coefficients for all SW approximation algorithms (Section 3.2.1).

Figure 3: Global stability assessment of the CAMEL V2 (left (a)) and V3 (right (b)) emissivity datasets. For V2, years after 2016 are taken from a climatology of emissivity derived from 2004 - 2016.

Land Cover

Land cover information for LST_cci products is provided by the ULeic / LandCover_cci (LCCS) hybrid biome map (Ghent et al., 2024). The baseline for this map comes from Land Cover CCI (LC_cci), who produced maps mainly from the MERIS FR time series, the MERIS RR dataset and SPOT Vegetation (SPOT-VGT) (Kirches et al., 2019). Land cover maps in Land Cover CCI are derived using a classification model based on the GlobCover unsupervised classification chain. In order to obtain land cover maps both globally consistent and regionally tuned, the GlobCover processing chain includes machine learning classification steps and a multi-year strategy.

For LST_cci, the aim is consistency across the CCI projects, while also retaining the most appropriate system for thermal infrared data. The LC_cci classification has been modified to sub-divide the bare soil classes into distinct sub-classes based on soil taxonomy. The rationale here is that emissivity variability is highest for different types of bare soil and for biome-based algorithms distinguishing between different bare soil types is crucial for improving robustness of the retrieval algorithms. The LCCS hybrid product merges the LC_cci classification with bare soil differentiation of six additional classes. This maintains consistency across the CCI programme in the use of a single land cover dataset, while optimising its implementation in LST_cci.

The LC_cci dataset is gridded on a global grid with a spatial resolution of 1/360°. Since this is higher spatial resolution than the majority of the LST_cci sensors, the data was re-binned to 1/120° to better match the IRCDR dataset. The re-gridding was performed using a 2-dimensional histogram approach, first sub dividing the LC_cci data into the 1/120° grid cells, and then finding the median land classification index for the LC_cci pixels in each 1/120° cell. Bare soil values in the LC_cci dataset were identified and the corresponding pixel found. These new values, coded as 200 – 207, have been incorporated into the LCCS Hybrid scheme (Table 1).

Table 1: ULeic / LandCover_cci (LCCS) hybrid biome definition.

Uncertainty Characterisation

Generally, for each LST pixel, three components of uncertainty are provided, representing the uncertainty from effects whose errors have distinct correlation properties (Ghent et al., 2019):

• uncorrelated (no correlation of error component between cells);
• locally systematic (correlation of error component between “nearby” pixels);
• [large-scale] systematic (correlation of error component between “distant” pixels).

Locally correlated errors are modelled via spatio-temporal correlation length scales that determine how an observation influences the analysis in the vicinity of its time-space location. Systematic errors represent a bias between different sources of data, the magnitude of which is conditioned by the uncertainty attributed to systematic effects.

This approach is both a necessary minimum, since locally systematic effects are significant, and preclude use of a simple random/systematic model and an approximation, in that there are several effects that have a systematic aspect, and all of these are required to be partitioned into either the locally systematic or systematic component. This three-component model can be applied to all satellite processing levels (L1, L2, L3, and L4).

Uncorrelated uncertainty

The uncorrelated (sometimes called the random) component of L1 channel uncertainty can be denoted as u_ran(y_c). The effect of this combined across all channels needs to be propagated through the retrieval to give a contribution to the estimate of uncertainty from uncorrelated effects u_ran(x) in the retrieved surface temperature. The assumption is that the radiance noise is Gaussian and small enough to consider that the law of propagation of uncertainty is adequate for this propagation, which means:

u_{ran,y}(x) = \sqrt{\sum_{c=1}^{n}\left(\dfrac{\partial R}{\partial y_c}u_{ran}(y_c)\right)^2} \quad (eq. 5)

where u_ran,y(x) is the uncorrelated uncertainty due to effects from radiometric noise, R is the measurement equation, y_c is the observation in channel c and u_ran(y_c) is the uncorrelated component of the Level-1 channel uncertainty (which is the expected radiometric noise for channel c). 

Emissivity is an auxiliary input to all estimates of thermodynamic temperature from BTs, whether explicit or implicit. For LST, there is a potentially significant uncorrelated error component caused by the pixel-to-pixel variations in emissivity not captured in emissivity auxiliary information because it is related to variability on the ground that is not captured in emissivity atlases/models. The associated uncertainty can be estimated as:

u_{ran,\varepsilon}(x) = \sqrt{\sum_{c=1}^{n}\left(\dfrac{\partial R}{\partial \varepsilon_c}u_{ran}(\varepsilon_c)\right)^2} \quad (eq. 6)

where u_ran,ε(x) is the uncorrelated uncertainty due to effects from surface emissivity, ε_c is the emissivity in channel c and u_ran(ε_c) is the uncorrelated component of the input emissivity. Therefore, some estimate of the uncertainty in emissivity per channel is required. In practice, this is estimated as the magnitude of pixel-to-pixel scale emissivity variability within areas that, based on same land cover classes being treated as having a common emissivity. Emissivity errors are estimated per land class based on both existing literature and validation studies (Ghent et al., 2019).

The total uncorrelated component is the acquired by adding the individual components in quadrature.

Locally correlated uncertainty

Atmospheric fields are correlated on timescales greater than 1 day and length scales greater than 100 km, and it is assumed that errors in estimates of these fields from NWP are correlated on the same scales. For coefficient based retrieval methods the retrieval ambiguity is a contributor of residuals in the fit. For radiative-transfer based retrieval coefficients, simulated-retrieved and simulation-input surface temperatures can be compared. The standard deviation of this input and output difference is an estimate of the magnitude of this locally correlated form of uncertainty. The calculation of the uncertainty can be done on stratified data to parameterise the variations in magnitude of this form of uncertainty. For each range of satellite viewing angle and water vapour (being the primary sources of variability), the uncertainty is estimated as:

u_{loc,fit}(x) = \sqrt{Var(\hat{x} - x_{in})} \quad (eq. 7)

where u_{loc_fit}(x) is the uncertainty due to coefficient fitting, and x is the double difference between the simulated surface temperature minus the retrieved surface temperature and simulated surface temperature minus the numerical weather prediction surface temperature.

In the GSW algorithm, the model coefficients are calibrated for bands of satellite zenith angle and bands of total column water vapour. As such, an estimate of the state TCWV is required for the LST retrieval, which is used to select which coefficients are applied in the model. To characterize this effect, it is necessary to propagate the uncertainty in TCWV through the LST model. A commonly used measure of uncertainty in Numerical Weather Prediction Models is the ensemble spread that is generally available together with the variable best estimate. The ensemble systems consist of a set of model runs with perturbed initial conditions; some systems also include perturbations to the model physics, more than one model within the ensemble or different physical parametrization schemes (WMO, 2012). Processing of ensemble data can be quite demanding and, therefore, a climatology of the TCWV spread has been used to approximate the instantaneous spread of TCWV with dependence on actual TCWV, latitude and month.

LST retrieval assumes an emissivity which may be driven by auxiliary land classification information or observed vegetation indices. Across a particular land cover class area, there may be a mean difference between the assumed and true mean emissivity. This is thus a locally correlated effect on the scales of emissivity variability. The form of the propagation to L2 uncertainty is estimated as:

u_{loc,\varepsilon}(x) = \sqrt{\sum_{c=1}^{n}\left(\dfrac{\partial R}{\partial \varepsilon_c}u_{loc}(\varepsilon_c)\right)^2} \quad (eq. 8)

where u_loc,ε(x) is the locally correlated uncertainty due to effects from surface emissivity, ε_c is the emissivity in channel c and u_loc(ε_c) is the locally correlated component of the input emissivity. This locally correlated component is based on pixels for the same land cover having the same error characteristics. This does not capture sub-pixel variability for any given pixel within a land cover, which is captured above in the uncorrelated component, and for which high resolution emissivity data are used to quantify the error properties. The correlation length scale is dependent on the source of the uncertainty. As a result, atmospheric and surface related uncertainties are considered and provided separately and propagated as either correlated or uncorrelated uncertainties as appropriate for a given product. The total locally correlated component is then derived by adding the individual components in quadrature.

Systematic uncertainty

This includes components such as the uncertainty in the radiative transfer model. It is assumed here that known corrections have been applied by data producers, either at L1 or in the retrieval process to L2. After these corrections have been applied then the remainder can be described as an uncertainty in the bias of the satellite observed surface temperatures relative to other data sources of temperature. Knowledge of the satellite engineering specifications and/or validation performance may allow a reasoned estimate of the likely magnitude of residual biases, which are not known and therefore have not been included in the corrections.

Since the different components are independent of each other they are combined in quadrature for a total uncertainty per pixel in the product.

Harmonisation for Climate Data Records

Cross-calibration of BTs using Infrared Atmospheric Sounding Interferometer (IASI) spectra

In the hypothetical situation where two sensors are observing the same field of view, with the same spectral band, at the same time and from the same view angle, one would expect the BTs to be the same. If this is not the case, that would imply intercalibration differences between the two sensors.

Harmonisation of L1 data across all sensors for the LST_cci IRCDR entails adjustment of the BTs to a reference sensor. The Global Space-based Inter-Calibration System (GSICS) have used the Infrared Atmospheric Sounding Interferometer (IASI) as the reference sensor in a calibration of SEVIRI radiances for example. IASI is a Fourier transform spectrometer and provides infrared spectra with high resolution (0.5 cm^-1 after apodisation, L1C spectra) between 645 cm^-1 and 2760 cm^-1 (3.6 μm to 15.5 μm). The IASI spectra are multiplied by the spectral response function of the instrument being harmonised. These IASI radiances are then compared to matched radiances from the non-reference instrument, the resulting differences analysed and a bias and uncertainty in the bias calculated. The radiances from the non-reference instrument are then aligned with IASI.

IASI is an across-track scanning system with scan range of ±48° 20´, symmetrically with respect to the nadir direction. A nominal scan line covers 30 scan positions towards the Earth and two calibration views. One calibration view is into deep space, the other is observing the internal black body. The scan starts on the left side with respect to the flight direction of the spacecraft.

The elementary (or effective) field of view (EFOV) is the useful field of view at each scan position. Each EFOV consists of a 2 × 2 matrix of so-called instantaneous fields of view (IFOV). Each IFOV at nadir is circular with a diameter of 12 km at nadir. At the edge of swath the IFOV is elliptical with size 39 km across track and 20 km along track.

In LST_cci access to radiances for all instruments is not available so the inter-calibration is performed with BTs. The steps required are:

1. Collocate IASI pixels with instrument pixels
   1. Use L1c IASI swath data
   2. Use instrument data on Level-3 Uncollated (L3U) grid
2. Multiply IASI spectra with instrument spectral response function to produce instrument equivalent IASI BTs
3. Analyse differences – use results of analysis to calibrate instrument BTs and propagate uncertainty

The intercalibration corrections are applied to Terra-MODIS and are carried out to the BTs prior to the LST retrievals. No intercalibration corrections are applied to the ATSRs or SLSTRs. Assessment of such intercalibrations to the ATSRs and SLSTRs found any correction would be negligible with uncertainties greater than the correction itself. Thus any correction would be counter-productive.

Multi-sensor Time Correction

Differences in overpass time between sensors of a series results in step changes in LST. This occurs for example between ATSR-2, overpass time 10:30 and 22:30, and AATSR with overpass time of 10:00 and 22:00. Since there is a data gap between the end of AATSR mission and start of SLSTR mission, LST from Terra-MODIS can be used to fill the gap. For the multi-sensor IRCDR spanning the missions ATSR-2, AASTR, Terra-MODIS and SLSTR-B a time correction is applied by estimating the BTs that would be observed at the selected nominal time for the whole IRCDR.

RTTOV version 12.3 was used to simulate the BTs. Simulations are done on a UTC hourly grid, at tie points (centre pixel of a group of 5x5 pixels on the L1 swath). For each observed pixel on the swath, the simulations are interpolated in space and time to the observation time and to the nominal time at each observation location. The simulations are based on ERA5 profiles as input to derive the typical morning heating rate (evening cooling rate) in terms of BTs, which are then used to interpolate in time.

A climatological standard deviation of simulated - observed BT differences was used to estimate the uncertainty on the BT correction. After cloud clearing at the observation time, the differences are binned by latitude, land cover class, satellite zenith angle, and calendar month. Standard deviations of the binned differences are calculated. The uncertainty assigned is the standard deviation of the bin the observed BT falls in.

As the difference of two simulations is used to calculate the correction, the uncertainty on the BT correction (σ²) is:

\sigma^2 = \sigma^2_{t_{obs}} + \sigma^2_{t_{nom}} - 2\rho\sigma_{t_{obs}}\sigma_{t_{nom}} \quad (eq. 9)

where σ²_tobs is the uncertainty on the simulation at observation time and σ²_tnom is the uncertainty on the simulation at the nominated time and ρ is the coefficient of correlation between the simulations at the two times. The uncertainty on the simulation at observation time, σ²_tobs, is estimated from the differences between simulations and observations. The uncertainty on the simulation at the nominated time, σ²_tnom, could be assumed equal to σ²_tobs. If errors in the simulation at observation time and nominated time are correlated (ρ) then the uncertainty in the difference should reduce by the covariance.

For the IRCDR the data of two of the sensors are corrected with the other two uncorrected. Nominal equatorial overpass times of 10:00 and 22:00 are selected, which correspond to the overpass times of AATSR and SLSTR-B and so these two datasets are not corrected. For ATSR-2 and Terra-MODIS a time correction of 30 minutes is applied to every pixel to bring these sensors to the same nominal equatorial overpass times.

Multi-sensor Infrared Climate Data Record (IRCDR)

To fill the gap between the end of the Envisat mission and the Sentinel-3B mission, requires an instrument of equivalent spatial resolution with a Local Equatorial Crossing Time (LECT) close in time to AATSR and SLSTR and of sufficient high quality. Terra-MODIS was chosen as it meet these minimum requirements. In addition, since Terra-MODIS has been operational from late in the ATSR-2 mission, through AATSR, and to-date through SLSTR, cross-calibration can be performed. Terra-MODIS also has a common LECT (10:30 and 22:30) with ATSR-2 so knowledge of the temporal correction between ATSR-2 and AATSR is also applicable for Terra-MODIS and AATSR. Note, AATSR and SLSTR have the same LECTs (10:00 and 22:00).

Sentinel-3B is used rather than Sentinel-3A since it was found that the temporal stability of Sentinel-3A does not meet GCOS climate requirements (GCOS-245, 2022). This is due to a change from Collection 3 to Collection 4 of the input L1 data (Good et al., 2025).

Five steps have been taken in the development of the ATSR-SLSTR CDR:

• Step 1: gap bridging
• Step 2: time difference correction
• Step 3: apply consistent Level-2 algorithms
• Step 4: restrict satellite viewing angle
• Step 5: selection of dataset during overlap periods

The approaches for each of these steps are summarised here, and illustrated in Figure 4:

Step 1:

• Intercalibration work for each input sensor vs Metop-IASI, through Quasi Simultaneous Nadir Overpasses (QSNOs) over Oceans and at high latitudes (Greenland, Antarctica, and over sea-ice) and SNOs where possible elsewhere.
• All sensors contributing to the IRCDR have been assessed for intercalibration corrections with respect to the reference sensor IASI (see Section 3.6.1), with corrections applied where non-negligible.

Step 2:

• The effect of the temporal mismatch between instruments on the LST time series is characterised and the resulting correction tuned.
• BTs for each pixel are simulated at the actual and nominated overpass times using RTTOV version 12.3 and interpolated in space and time to the observation time and to the nominal time at each observation location.
• Corrections of 30 minutes are made for ATSR-2 and Terra-MODIS to the overpass time of AATSR and SLSTR-B, so that the much of the time series is uncorrected (i.e. at its native overpass time).

Step 3:

• A common algorithm is applied across all sensors, in this case the GSW algorithm as described in Section 3.1.
• The uncertainty components described in Section 3.5 are implemented across all sensors consistently.
• Cloud masking is consistent, with the probabilistic cloud mask applied for each of the sensors.

Step 4:

• The satellite viewing angles of the sensors is restricted to that of the lowest common denominator – the ATSRs to avoid discontinuities in the time-series.

Step 5:

• The switch over between sensors in the daily data record occur at crossover between months. This is to preserve consistency when propagated to monthly products. Therefore the IRCDR comprises of the following sensors over the respective time windows:
  • ATSR-2: 1995-06-01 to 2002-05-31
  • AATSR: 2002-06-01 to 2012-03-31
  • Terra-MODIS: 2012-04-01 to 2018-11-30
  • SLSTR-B: 2018-12-01 to 2024-12-31

Figure 4: Flowchart of the full IRCDR processing chain from input Level 1 data through to the final Level 3 output product.

Output data

Level-3S Global, Monthly IRCDR Product

The structure of the file names is described in detail in Product User Guide and Specifications (PUGS) of the C3S LST Products (C3S, 2025) Here a summary of the most useful information is provided.

Each output follows the LST_cci file name structure below, with an additional "cds" suffix at the end:

ESACCI-LST-<Processing Level>-LST-<Product String>[…]<Indicative Date>[<Indicative Time>]-fv<Version Number>cds.nc

The key elements of the LST_cci filenames are described in Table 2:

Table 2: Description of key elements in the filename convention.

An example of an IRCDR filename following this format is:

ESACCI-LST-L3S-LST-IRCDR_-0.01deg_1DAILY_DAY-20100101000000-fv2.00cds.nc

 The format of the data files is NetCDF-4. Within the files are data (known as variables) and metadata (known as attributes). A summary of the key variables within the IRCDR product files is provided in Table 3.

Table 3: Summary of the variables in the IRCDR NetCDF output files.

Figures 5 and 6 illustrate the main output data layers of the Level-3S Global Monthly IRCDR Product, the LST and uncertainty components respectively.

Figure 5: Monthly land surface temperature in June 2024 for (a) daytime and (b) nighttime and in December 2024 for (c) daytime and (d) nighttime.

Figure 6: Uncertainty on the monthly daytime LST for June 2024: (a) total uncertainty, (b) uncertainty due to uncorrelated errors, (c) uncertainty due to errors correlated on atmospheric scales, (d) uncertainty due to errors correlated on surface scales.

References

Borbas, E. E., G. Hulley, M. Feltz, R. Knuteson and S. Hook (2018). "The Combined ASTER MODIS Emissivity over Land (CAMEL) Part 1: Methodology and High Spectral Resolution Application." Remote Sensing 10(4): 643. https://doi.org/10.3390/rs10040643

Bulgin, C. E., Sembhi, H., Ghent, D., Remedios, J. J. and Merchant, C. (2014). "Cloud clearing techniques over land for land surface temperature retrieval from the Advanced Along Track Scanning Radiometer". International Journal of Remote Sensing, 35 (10). pp. 3594-3615.

Copernicus Climate Change Service (C3S) (2017). "ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate". Copernicus Climate Change Service Climate Data Store (CDS), 5th August 2019. https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels?tab=overview

Copernicus Climate Change Service (C3S) (2025). "Product User Guide and Specification (PUGS) v1.0 - CDR and ICDR Land Surface Temperature v3.00". C3S2_313e_BC_WP3-DR-LST-LST_CCI-v3.00-1995-2024_202506_PUGS_v1.0 (not yet published)

Coppo, P., B Ricciarelli, F., Brandani, J., Delderfield, M., Ferlet, C., Mutlow, G., Munro, T., Nightingale, D., Smith, S., Bianchi, P., et al. (2010). "SLSTR: A high accuracy dual scan temperature radiometer for sea and land surface monitoring from space". J. Mod. Opt. 57, 1815–1830

Donlon, C.J., Martin, M., Stark, J., Roberts-Jones, J., Fiedler, E., and Wimmer, W., (2012). The Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) system. Remote Sensing of the Environment. doi: 10.1016/j.rse.2010.10.017 2011.

Dudhia, A. (2017). "The Reference Forward Model (RFM)". Journal of Quantitative Spectroscopy and Radiative Transfer, Volume 186, Pages 243-253, ISSN 0022-4073, https://doi.org/10.1016/j.jqsrt.2016.06.018.

Edwards, T., R. Browning, J. Delderfield, D. J. Lee, and K. A. Lidiard, (1990). "The Along Track Scanning Radiometer - Measurement of sea-surface temperature from ERS-1". The Journal of The British Interplanetary Society, 43, 160-180

Feltz, M.; Borbas, E.; Knuteson, R.; Hulley, G.; Hook, S. (2018). "The Combined ASTER MODIS Emissivity over Land (CAMEL) Part 2: Uncertainty and Validation". Remote Sens. 10, 664. https://doi.org/10.3390/rs10050664

GCOS-245. (2022). "The Global Observing System for Climate: ECVs Requirements (GCOS-245)". https://library.wmo.int/records/item/58111-the-2022-gcos-ecvs-requirements-gcos-245.

Ghent, D., Corlett, G., Goettsche, F., & Remedios, J. (2017). "Global land surface temperature from the Along-Track Scanning Radiometers". Journal of Geophysical Research – Atmospheres, 122, 12167-12193

Ghent, D., Veal, K., Trent, T., Dodd, E., Sembhi, H., and Remedios, J. (2019). "A New Approach to Defining Uncertainties for MODIS Land Surface Temperature". Remote Sensing, 11, 1021. doi: 10.3390/rs11091021

Ghent, D. J., Anand, J. S., Veal, K., Remedios, J. J. (2024). "The operational and climate land surface temperature products from the Sea and Land Surface Temperature Radiometers on Sentinel-3A and 3B". Remote Sensing, 16, 3403.

Good, E., Blannin, B., Waring, A., Veal, K., and Ghent, D., (2025). "Sidebar 2.1: Super extreme land surface temperature hotspots", [in “State of the Climate in 2024”], Bull. Amer. Meteor. Soc., 106 (8), S37–S39, https://doi.org/10.1175/BAMS-D-25-0102.1

Good, S.; Fiedler, E.; Mao, C.; Martin, M.J.; Maycock, A.; Reid, R.; Roberts-Jones, J.; Searle, T.; Waters, J.; While, J.; Worsfold, M. (2020). The Current Configuration of the OSTIA System for Operational Production of Foundation Sea Surface Temperature and Ice Concentration Analyses. Remote Sens., 12, 720. doi: 10.3390/rs12040720

Guillevic, P.; Göttsche, F.; Nickeson, J.; Hulley, G.; Ghent, D.; Yu, Y.; Trigo, I.; Hook, S.; Sobrino, J.; Remedios, J.; Román, M.; Camacho, F. (2018). "Land Surface Temperature Product Validation Best Practice Protocol", Version 1.1 https://lpvs.gsfc.nasa.gov/PDF/CEOS_LST_PROTOCOL_Feb2018_v1.1.0_light.pdf

Hulley, G.C., Ghent, D., Göttsche, F.M., Guillevic, P.C., Mildrexler, D.J., & Coll, C. (2019). "Land Surface Temperature". In G.C. Hulley, & D. Ghent (Eds.), Taking the Temperature of the Earth (pp. 57-127): Elsevier

Hunt, S.E.; Mittaz JP, D.; Smith, D.; Polehampton, E.; Yemelyanova, R.;Woolliams, E.R.; Donlon, C. (2020). "Comparison of the Sentinel-3A and B SLSTR Tandem Phase Data Using Metrological Principles". Remote Sens. 12, 2893

Joint Commitee for Guides in Metrology (JCGM), (2008). "Evaluation of Measurement Data - Guide to the Expression of Uncertainty in Measurement".

Kirches, G., Brockman, C., Boettcher , M., Peters, M., Bontemps, S., Lamarche, C., Schlerf , M., and Santoro, M. (2016), Land Cover CCI PRODUCT USER GUIDE VERSION 2, http://data.ceda.ac.uk/neodc/esacci/land_cover/docs/ESACCI-LC-PUG-v2.5.pdf

Llewellyn-Jones, D., et al. (2001), "AATSR: Global-change and surface temperature measurements from Envisat", ESA Bulletin-European Space Agency (105), 11-21.

Matricardi, M. (2009). "Technical Note: An assessment of the accuracy of the RTTOV fast radiative transfer model using IASI data", Atmos. Chem. Phys., 9, 6899-6913, doi:10.5194/acp-9-6899-2009.

Merchant, C.J., Harris, A.R., Murray, M.J., Zavody, A.M. (1999). "Toward the elimination of bias in satellite retrievals of sea surface temperature 1. Theory, modeling and interalgorithm comparison". J. Geophys. Res. Oceans 104, 23565–23578.

MODIS Characterization Support Team (MCST), (2017). "MODIS 1km Calibrated Radiances Product". NASA MODIS Adaptive Processing System, Goddard Space Flight Center, USA: http://dx.doi.org/10.5067/MODIS/MOD021KM.061

Norman, J.M. and F. Becker. (1995). "Terminology in thermal infrared remote sensing of natural surfaces". Agricultural and Forest Meteorology, 77: p. 153-166.

Perry, M., Ghent, D., Jimenez, C., Dodd, E., Ermida, S., Trigo, I. F., and Veal, K. (2020). "Multi-Sensor thermal infrared and microwave land surface temperature algorithm intercomparison". Remote Sensing, 12(24), 4164

Saunders, R. (2001). "RTTOV-7 - Science and Validation Report", EUMETSAT.

Savtchenko, A., Ouzounov, D., Ahmad, S., Acker, J., Leptoukh, G., Koziana, J. and Nickless, D. (2004). "Terra and Aqua MODIS products available from NASA GES DAAC". Advances in Space Research, 34(4), pp.710-714

Smith, D. L., et al. (2001). "Calibration of the AATSR instrument", Advances in Space Research, 28(1), 31-39

Smith, D.; Hunt, S.E.; Etxaluze, M.; Peters, D.; Nightingale, T.; Mittaz, J.; Woolliams, E.R.; Polehampton, E. (2021). "Traceability of the Sentinel-3 SLSTR Level-1 Infrared Radiometric Processing". Remote Sens., 13, 374.

Soria, G. and J.A. Sobrino. (2007). "ENVISAT/AATSR derived land surface temperature over a heterogeneous region". Remote Sensing of Environment, 111(4): p. 409-422.

Wan, Z. and Dozier, J. (1996). "A generalised spilt-window algorithm for retrieving Land-Surface Temperature from space", IEEE Trans. GeoSci. Remote Sens.

WMO (2012). "Guidelines on Ensemble Prediction Systems and Forecasting". WMO-No.1091

Xiong, X.; Angal, A.; Chang, T.; Chiang, K.; Lei, N.; Li, Y.; Sun, J.; Twedt, K.; Wu, A. (2020). "MODIS and VIIRS Calibration and Characterization in Support of Producing Long-Term High-Quality Data Products". Remote Sens., 12, 3167. https://doi.org/10.3390/rs12193167

Related articles