NRT Product Available since Nov 2020
Land Surface Temperature (LST) is the radiative skin temperature over land. LST plays an important role in the physics of land surface as it is involved in the processes of energy and water exchange with the atmosphere. LST is useful for the scientific community, namely for those dealing with meteorological and climate models. Accurate values of LST are also of special interest in a wide range of areas related to land surface processes, including meteorology, hydrology, agrometeorology, climatology and environmental studies. Land Surface Emissivity (EM), a crucial parameter for LST retrieval from space, is independently estimated as a function of (satellite derived) Fraction of Vegetation Cover (FVC) and land cover classification.
This Operational product is documented in the following documents:
Please see Product Peer-Review publications in References.
The use of LSA SAF products in publications is kindly requested to be duly acknowledged:
LST was provided by the EUMETSAT Satellite Application Facility on Land Surface Analysis (LSA SAF; Trigo et al., 2011)
Trigo, I. F., C. C. DaCamara, P. Viterbo, J.-L. Roujean, F. Olesen, C. Barroso, F. Camacho-de Coca, D. Carrer, S. C. Freitas, J. García-Haro, B. Geiger, F. Gellens-Meulenberghs, N. Ghilain, J. Meliá, L. Pessanha, N. Siljamo, and A. Arboleda, 2011: The Satellite Application Facility on Land Surface Analysis. Int. J. Remote Sens., 32, 2725-2744, doi: 10.1080/01431161003743199
The all-sky LST product provides skin temperature estimate for all land pixels within the SEVIRI disk, every 30 min, for both clear and cloudy conditions (Martins et al, 2019). It is a combination of two already operational algorithms: (i) the clear sky component is derived from MSG level 2 product, MSLT (LSA-001), while; (ii) the cloudy sky component is obtained from the energy balance algorithm currently in use for the estimation of MSG 30-minute evapotranspiration (MET-v2, LSA-311). The identification of cloudy pixels is based on the cloud mask generated by the Nowcasting and Very Short Range Forecasting Satellite Application Facility (NWC SAF) software.
The clear-sky component uses the estimates from MLST (LSA-001), which is derived from SEVIRI/MSG using a Generalized Split-Window (GSW) algorithm (Trigo et al, 2008).
For cloudy pixels, LST is derived from the surface energy balance model used to derive the actual evapotranspiration (MET; LSA-311), as well as the surface sensible (MH, LSA-304) and latent heat (MLE, LSA-305) turbulent fluxes. The model incorporates estimates of shortwave and longwave radiation fluxes, land surface albedo, soil moisture and vegetation, as well as near surface meteorological information obtained from ECMWF operational forecasts (Ghilain et al, 2019, 2020). The model assumes that each pixel is composed of different tiles, each representing a particular surface type, according to the ECOCLIMAP-II database (e.g, Faroux et al., 2013), namely: bare soil, snow, deciduous broadleaved trees, evergreen needle-leaved trees, evergreen broadleaved trees, crops, irrigated crops, grass, bogs and marshes, rocks, open water, urban. In practice, each SEVIRI pixel may include a maximum of four tiles, e.g., the dominant among the different vegetation types, one water tile, one snow tile, one bare ground (soil, rocks and urban). Pixel values are weighted values of its tiles. Values over inland waters are masked out in this product version.
The MLST-AS product output and accompanying quality flag, are stored on single files on HDF5 (Hierarchical Data Format, version 5) format, for the SEVIRI full disk, every 30 min. HDF5 is a machine independent standard for storing/sharing scientific data. In this format, each file contains also the necessary information for manipulating the data. For more information on this data format see https://www.hdfgroup.org/
Each component of the product has its own sources of uncertainty. The clear sky component has a well described framework to treat its uncertainty (Freitas et al., 2010), identifying the uncertainty sources associated to inputs and to the regression model, and their propagation to the final product; these include surface emissivity, atmospheric total column water vapour, the viewing geometry and to a lesser extent the sensor radiometric noise.
For cloudy pixels, the main sources of uncertainties are: a) the physical formalism of the algorithm itself, b) the errors associated with each input of the algorithm, in particular from the Downwelling Surface Shortwave Flux (MDSSF, LSA-201), Downwelling Surface Longwave Flux (MDSLF, LSA-204), Soil Moisture (SSM, internal product) and Albedo (MDAL; LSA-101) and c) the surface heterogeneity and land cover classification used in the algorithm. Some of the model parameters are also major sources of uncertainty, such as the roughness lengths, the minimum canopy resistance, and the phase and amplitude of the heat storage capacity. Misclassification of clouds also affect the quality of the cloudy part of the scheme, as some the inputs of the ET model critically depend on the cloud classification, namely the downwelling surface radiative fluxes. Nevertheless, DSSF and DSLF errors tend to partially compensate each other in these cases.
The comparisons to in situ stations show an overall accuracy of 0.0 K and a Root Mean Squared Difference (RMSD) of 2.9 K. For clear sky situations, the accuracy is -0.2 K and the RMSD is 2.8 K. For cloudy sky, accuracy is 0.2 K and the RMSD is 2.8 K. These show that the product performance is within the optimal requirements, and that cloudy sky estimates have similar quality to the clear-sky product. There are however some situations where optimal requirements are not met. There is a clear difference in the performance of MLST-AS between Bare/Crops/Grassland and forest stations, with the latter showing systematically worst errors. Regions with high aerosol optical depth also exhibit worse comparisons against independent all-sky LST datasets such as ERA5-Land.