EUSTACE (EU Surface Temperatures for All Corners of Earth) is a Horizon2020 project that has produced a spatially complete, near-surface air temperature (NSAT) analysis for the globe for every day since 1850. The analysis is based on both satellite and in situ surface temperature observations over land, sea, ice and lakes, which are combined using state-of-the-art statistical methods. The use of satellite data enables the EUSTACE analysis to offer improved estimates of NSAT in regions that are poorly observed in situ, compared with existing in-situ based analyses. This presentation provides a brief overview of the EUSTACE project, and illustrates how satellite infrared land surface temperature (LST) data – developed within the framework of the European Space Agency (ESA) Data User Element (DUE) GlobTemperature project – have been used in EUSTACE. Satellite infrared LSTs represent the temperature of the Earth’s ‘skin’, which can differ from the corresponding NSAT by several degrees or more, particularly during the hottest part of the day. A global empirical regression-based approach is used to transform the satellite observations to NSAT. Three models are used, which utilise different combinations of satellite LST observations (day time only, night time only, and day + night time) in order to maximise spatial coverage, as infrared LST observations are clear-sky only. The regression models are trained on daily in situ observations of NSAT and use other predictors known to influence the LST-NSAT relationship, such as vegetation and snow cover. An assessment of the satellite NSAT estimates using station data not used in the model training indicates a global median satellite-station difference of within ±0.1 K and a root-mean-square difference (RMSD) of between 2.8 and 4.9 K, depending on the model used. The satellite NSAT estimates are provided with uncertainties, which are propagated through the model and provided to users as random, locally-correlated and systematic uncertainty components.

One of the main objectives of the Sentinel-3 (S3) mission is to measure sea- and land-surface temperature with high-end accuracy and reliability in support of environmental and climate monitoring in an operational context. Calibration and validation are thus key criteria for operationalization within the framework of the Sentinel-3 Mission Performance Centre (S3MPC).
Land surface temperature (LST) has a long heritage of satellite observations which have facilitated our understanding of land surface and climate change processes, such as desertification, urbanization, deforestation and land/atmosphere coupling. These observations have been acquired from a variety of satellite instruments on platforms in both low-earth orbit and in geostationary orbit. Retrieval accuracy can be a challenge though; surface emissivities can be highly variable owing to the heterogeneity of the land, and atmospheric effects caused by the presence of aerosols and by water vapour absorption can give a bias to the underlying LST. As such, a rigorous validation is critical in order to assess the quality of the data and the associated uncertainties.
Validation of the level-2 SL_2_LST product from S3A, which became freely available on an operational basis from 5th July 2017 builds on an established validation protocol for satellite-based LST. This set of guidelines provides a standardized framework for structuring LST validation activities. The protocol introduces a four-pronged approach which can be summarised thus: i) in situ validation where ground-based observations are available; ii) radiance-based validation over sites that are homogeneous in emissivity; iii) intercomparison with retrievals from other satellite sensors; iv) time-series analysis to identify artefacts on an interannual time-scale. This multi-dimensional approach is a necessary requirement for assessing the performance of the LST algorithm for the Sea and Land Surface Temperature Radiometer (SLSTR) which is designed around biome-based coefficients, thus emphasizing the importance of non-traditional forms of validation such as radiance-based techniques. Here we present examples of the ongoing routine application of the protocol to operational Sentinel-3 LST data.

The impacts of anthropogenic and climatic changes on the water and energy fluxes between land and atmosphere are diverse and complex. Land surface temperature (LST) influences the partition of energy between ground and vegetation and determines the surface air temperature. It is thus dependent on the soil moisture, vegetation, and albedo conditions. As such, it can act can act as global indicator of the status and trends in evapotranspiration regimes which are often linked to land use and cover changes, whether these are gradual or sudden. Building on previous studies (Muro et al 2018), we use the MODIS LST archive (2000–2017) and the Greenbrown time series analysis R package to study the trends of the LST in the African continent. The LST time series data is divided into a seasonal component, noise, and a monotonic trend. Results show that large parts of Sub-Saharan Africa, such as Uganda, Tanzania and South Africa are experiencing increases in LST, significant at 95% confidence interval. In most cases these are product of the expansion of agriculture on wetlands, deforestation, or land degradation. Other countries such as South Sudan or Senegal show decreases in LST. Restoration of degraded grasslands, reforestation or expansion of agriculture on arid lands are some of the underlying causes such decreases in LST. Despite that LST often displays quick daily variations, we provide evidence of how solid statistical analysis of long records of time series analysis can be successfully used to monitor and study the effect of land cover changes in evapotranspiration regimes at continental scale.

High spatial resolution land surface temperature (LST), evapotranspiration and water stress essential for field-scale irrigation water management, flash drought warning, and yield forecast in arid and semi-arid ecosystems. The availability of high spatial resolution LST from low Earth observing satellites like Landsat-8 helps in determining vegetation water use and stress at parcel scale. However, coarse temporal resolution (16-day revisit) and loss of information due to overcast sky conditions during satellite overpass limits its use for routine monitoring and mapping. Good quality daily LST is available from recently launched Sentinel-3 (S-3), although the relatively coarser spatial resolution (1 km) limits its application for water footprinting at the field scale. However, Sentinel-2 (S-2) with 13 spectral channels in the visible/near infrared (VNIR) and short wave infrared (SWIR) spectral range has high spatial and temporal resolution (20 m; 3 days repeativity), and the information contained in the SWIR bands tend to have immense potential for mapping water use and stress.
This study demonstrates an integration of SWIR-based water stressed index (SIWSI) into a physically-based surface energy balance (SEB) model for simultaneous estimation of LST and vegetation water use. SEB model called Surface Temperature Initiated Closure (STIC) was modified to ingest SWIR information, wherein estimation of LST was done in an inverse-iterative mode (iLST) alongwith simultaneous estimation of evapotranspiration (E), evaporation (EE), transpiration (ET), and water stress. Ecosystem water use was estimated by accumulating E and its components.
The new method (named STICiLST) was tested and evaluated across an aridity gradient in the North Australian Tropical Transect (NATT) using time-series of 8-day MODIS Terra visible and LST products in conjunction with eddy covariance measurements from 2010 to 2016. An intercomparison of iLST with MODIS Terra LST in five sites in the EC grid revealed an error of 3 – 6 K, with a high systematic error in the humid tropical ecosystem (Howard Springs) which progressively declined in the arid ecosystem (Alice Springs). Although the root mean square error (RMSE) in instantaneous E was 25 – 40% of the observed mean, the percent RMSE was relatively less when E was accumulated annually (20% RMSE). In general, the humid ecosystem revealed a tendency of high systematic errors in E (35%) as compared to the arid and semi-arid ecosystems (25 – 30%). This initial evaluation points towards dedicated exploitation of SWIR data for helping to constrain elements of the surface energy balance and ubiquitous field scale water status observations needed in both ecology and agriculture from current S-2, future sensors like TRISHNA (planned Indo-French mission), Sentinel-8, as well as reconstruction of LST at high spatial resolution from Indian RS sensors (e.g., LISS-3, AWiFS) and MODIS LST at 500 m spatial resolution.

Confidence in satellite-derived LST products and the downstream information (such as evapotranspiration estimates) derived from them comes in part from their evaluation against independent ground-based data. Such independent data are ideally provided by high performance radiometers and associated equipment installed semi-permanently at a well characterised LST validation site, where long-term multi-seasonal LST records can be built up for comparison to the satellite LST archives. To date, there are a limited number of such sites and they do not yet cover the full range of Earth’s major biomes. Many LST validation sites have also been located in regions having relatively constant emissivity, rather than in biomes experiencing a pronounced wet and dry season that produces significant intra-annual variations in vegetation cover. For the purpose of evaluating satellite LST products in more seasonally varying vegetated regions we have set up a new and recently completed long-term, LST validation site in Kenya. This work has been conducted as part of the UK Space Agency funded Pest Risk Information Service (PRISE) International Partnership Project, and the validation site is a collaboration between King’s College London/National Centre of Earth Observation (NCEO) and the International Livestock Research Institute (ILRI). Additional instrumentation has been provided by NASA’s Jet Propulsion Laboratory (JPL-NASA).

The new long-term LST validation site is the first in East Africa and is in a 50 km2 region of relatively homogeneous savannah. It is located at the ILRI research ranch on the Kapiti planes, south-east of Nairobi. The vegetation in the area grows in accordance with Kenya’s bimodal rainfalls, consisting of short-rains (October/November) and long-rains (March/April to May/June), and is also subject to wildlife and controlled grazing by cattle. Four towers host a series of eight high performance longwave infrared radiometers, along with associated equipment such as cloud and vegetation monitoring RGB cameras and a single CO2/CH4/H2O eddy-covariance system. The radiometer towers have been sited to ensure the data are useful for geostationary as well as polar-orbiting LST evaluation and are spread over an area equivalent to around three Meteosat pixels. With Meteosat-8 now repositioned over East Africa and operating contemporaneously with Meteosat-11 still located over West Africa, the true impact of satellite view angle on LST retrievals can, for example, now be inferred via satellite comparisons to the in situ LST data. Similarly, comparisons between polar-orbiting and geostationary satellite LSTs in this savannah region can be conducted supported by the additional knowledge coming from the LST validation site data record. Along with details of the LST validation site setup and examples of the data collected, here we will present early comparisons between these types of satellite-derived LST products and the LST validation site information.

The sites of high-temperature geothermal energy resources are commonly marked by thermal manifestation area where fumaroles, steaming grounds, hydrothermally altered grounds, hot springs, volcanic gas vents, craters, and mud pools occur. Surface thermal signatures may be used to define the evolution of development of shallow structures present on both volcanic districts and geothermal areas and possibly related to tectonic activity along active faults. These areas must be commonly subjected to geophysical monitoring where the measurements of shallow and surface temperatures may play an important role. The Italian Tyrrhenian margin is characterized by several high and medium enthalpy geothermal systems and in Tuscany several geothermal systems with a large-scale steam-dominated geothermal anomaly are presents. For example The Larderello–Travale geothermal system, known as Larderello geothermal field, is the site of a large-scale steam-dominated geothermal anomaly. At this area superheated steam is present at depths over 3.5 km and with temperatures exceeding 350 °C, whereas the deep reservoir of the Mt Amiata geothermal fields is in a two-phase (liquid + vapour mixture) state with temperatures of 300–350 °C. The different geological assets and thermal behavior of these areas correspond to surface thermal anomalies that may be very different both for the maximum temperatures and for the size and distribution on the ground. Remote sensing represents an effective and expedite tool to acquire infrared images when there are large extents of areas affected by thermal anomalies. However medium range (drone) to long range (satellite) thermal images are affected by errors and underestimates due both to the distance and to atmospheric absorption so they require to be calibrated with short range field measurements in order to accurately correct for these effects. In this context, systematic global cataloging of thermal anomalies as measured from high spatial resolution spaceborne sensors and acquired with thermal camera installed on drones are still in their infancy.
Our goal is to create and analyze a catalog of data by examining the magnitude, frequency and distribution of Tuscany geothermal districts thermal signatures at the highest available spatial resolution ranging from the hundreds of meters per pixel of satellite thermal images to the tenths/hundreds of centimeters per pixel of the thermal images taken by the drones.
The maps obtained by satellite data acquiring thermal imagery at suitable spatial resolution and drone measurements have been orthorectified and geocoded in order to be superimposed on digital terrain models or on the orthophotogrammetric mosaic obtained after processing photos acquired by drone. This allows, for example, to follow the evolution of thermal anomalies, that may represent a modification of the current state of the geothermal field and a possible hazard for both the population and for industrial assets
Thermal infrared measurements can also be used for environmental surveys for the detection of sites with abusive waste landfill or the relief of areas subject to hydrogeological disaster. The accurate temperature measurements in land targets recorded with a variety of instruments during regular field campaigns can also offer a valid methodology for the calibration and validation of satellite data.
The drone is a FlyBit produced by Flytop (now Ingegneria Dei Sistemi, IDS). It is equipped with a thermal camera FLIR VUE PRO-R or a visible camera Sony Alpha 6000. The VUE PRO-R uses an uncooled microbolometer with a resolution of 640x512 pixel capable of a precision of ±5°C (or 5% of the measure). Field campaigns were performed using and handheld thermal camera FLIR SC640 having a resolution of 640x480 pixel, ±30mK of thermal resolution and a precision of ±2°C (or 2% of the measure) and a type K thermocouple. Satellite data are free and available from USGS web site.
Here we show the results obtained during two field campaigns and in particular the one held at the end of September 2018 during which the simultaneously acquisition by Landsat8 satellite and thermal data of FlyBit have been analysed. By removing the atmosphere contribute, from Landsat8 data we have produced a surface temperature map that will be compared with the ground field measurements and surface temperature map measured by VUE PRO-R on FlyBit.