To estimate evapotranspiration (ET) from space the accurate retrieval of Surface Temperature and several land surface parameters (LSP) is of vital importance. This however had lead in the past to a large caveat: namely, the consistency between the different input parameters. When consistency between input parameters is not considered, this ignores the fact that that without goodness inputs no decent output can provided. When LSP- dynamics contradict each other, the output of the model cannot be representative of reality. At present however, the LSPs used in large scale ET estimations originate from different single-sensor retrieval-approaches and even from different satellite sensors.

In response, the Earth Observation Land Data Assimilation System (EOLDAS) was developed. EOLDAS uses a multi-sensor approach to couple different satellite observations/types to radiative transfer models (RTM), consistently. It is therefore capable of synergistically estimating a variety of LSPs. Considering that ET is most sensitive to the temperatures of the land surface (components), the goal of this research is to expand EOLDAS to the thermal domain. This research not only focusses on estimating LST, but also on retrieving (soil/vegetation, Sunlit/shaded) component temperatures, to facilitate dual/quad-source ET models.

To achieve this, The Soil Canopy Observations of Photosynthesis and Energy (SCOPE) model was integrated into EOLDAS. SCOPE couples key-parameters to key-processes, such as photosynthesis, ET and optical/thermal RT. In this research SCOPE was also coupled to MODTRAN RTM, in order estimate BOA component temperatures directly from TOA observations. This paper presents the main modelling steps of integrating these complex models into an operational platform. In addition it highlights the actual retrieval using different satellite observations, such as MODIS and Sentinel-3, and meteorological variables from the ERA-Interim.

Area B - Board 185

The Land Surface Temperature (LST) varies rapidly in space and time. The LST temporal variations are driven by the earth’s diurnal and annual cycles and are conveyed differently from place to place based on differences in geographic location, topography, solar geometry, meteorology, and the surface’s ability to conduct, convect and store heat. Accurate knowledge of the LST diurnal temperature cycle (DTC) and annual temperature cycle (ATC) can help us understand the thermal behavior of the earth’s surface and its physical properties and also to identify and correct land surface model deficiencies. In recent years, several models have been proposed so as to approximate the LST DTC and ATC. The primary driver of the LST DTC is the rotation of the earth around its axis, which results in the diurnal cycle; while that of the LST ATC is the seasonal variation of the incoming solar radiation due to the earth’s axial obliquity. A major advantage of DTC and ATC models is that they can be reduced to a set of key parameters (e.g. the LST mean, diurnal or annual amplitude, peak time and phase shift) that are gap-free and provide a generalized characterization of the cloud-free LST temporal dynamics that is more informative that individual LST measurements.

In this work (Sismanidis et al., 2018) we use the sine-based ATC model of Bechtel (2015) and a five-year time series of diurnal (half-hourly) geostationary LST from the Spinning Enhanced Visible and Infrared Imager (SEVIRI), so as to map the annual and diurnal spatiotemporal dynamics of mainland Europe (10°W, 32°N to 30°E, 50°N). The LST data are from the Land Surface Analysis Satellite Applications Facility (Land-SAF) and to process them we divided them into 48 groups according to their acquisition time (i.e., 00:00, 00:30, 01:00, ..., 23:30 UTC). Next we stacked the data of each group according to their acquisition day (d ∈ [1, 2, ..., 365]) and fitted each group’s stack with the sine-based ATC model of Bechtel (2015). The fitting was done in a pixel wise manner and for each pixel the half-hourly Annual Cycle Parameters (ACP), i.e. the mean annual surface temperature (MAST), the yearly amplitude of surface temperature (YAST), and the phase shift relative to the spring equinox (Theta), were retrieved. The MAST and YAST are self-explanatory, while Theta can be seen as a measure of the heat uptake latency of the Earth’s surface. The last step of our processing was to compensate the longitudinal differences of the MAST, YAST and Theta, so as to make the derived ACP time-consistent. This was done by converting the UTC time of each pixel to local solar time and then interpolating (linearly) the half-hourly MAST, YAST and Theta values in the temporal domain.

The derived MAST, YAST and Theta maps provide a complete representation of how mainland Europe responds to the heating of the sun and the LST decay at night and reveal how Europe’s biogeographic regions (e.g. Atlantic, Continental, Mediterranean and Pannonian) differ in that respect. Our results further suggest that the diurnal MAST, YAST and Theta exhibit distinct temporal profiles that are influenced by the pixel’s topography, climate and surface’s characteristics and that the derived ACP can provide an observation-based spatially consistent background for studying and characterizing the surface’s thermal behavior in space and time.


References:
Bechtel, B. (2015), A New Global Climatology of Annual Land Surface Temperature, Remote Sens., 7(3), 2850–2870, doi:10.3390/rs70302850.
Sismanidis, P., B. Bechtel, I. Keramitsoglou, and C. T. Kiranoudis (2018), Mapping the Spatiotemporal Dynamics of Europe’s Land Surface Temperatures, IEEE Geosci. Remote Sens. Lett., 15(2), 202–206, doi:10.1109/LGRS.2017.2779829.

Area B - Board 186

Over the past decades, soil moisture observation through various satellite mission has become widely available in both active and passive sensors onboard. Global and regional scale monitoring of soil moisture from remotely sensed data have a great importance in terms of both climatic as well as environmental perspective. The spaceborne sensors use the microwave emission from the earth’s surface as it is mostly influenced by the temperature and water content in the soil. However, the products of the satellite-derived soil moisture have a spatial resolution on the order of tens of km which is often too coarse for the local scale monitoring and application. Therefore, the goal of this study is to assess the soil moisture evaluation through disaggregation of Advanced Microwave Scanning Radiometer 2 (AMSR2) over an area identified for the Critical Zone Observatory (CZO) in parts of the Ganga basin. The AMSR2 provides the soil moisture in different bands such as C band (6.9 GHz/7.3 GHz) and X band (10.7 GHz) with a spatial resolution of 25 km retrieved through Land Parameter Retrieval Model (LPRM) algorithm. Here, the downscaling of AMSR2 pixel is based on an algorithm called ‘triangle method’ which relates the soil moisture with MODIS derived Normalized Difference Vegetation Index (NDVI) and Land Surface Temperature (LST). The downscaling was done for the C1 (6.9 GHz) and X band (10.7 GHz) with a temporal interval from September 2017 to December 2017. The AMSR2 soil moisture was downscaled to 1 km since the LST and NDVI used here has 1 km spatial resolution. The results were compared to the in-situ data collected over the CZO for the same period. It shows very small variations over the region due to much less variability of NDVI and LST. Our results also show great sensitivity towards the precipitation observed from the automatic weather station deployed in the field. Both the C1 and X band derived soil moisture values do not show much spatial variation throughout the area. The RMSE between the observed and downscaled value was calculated and it lies within 0.07 to 0.16 for both the AMSR2 bands. However, the C1 band shows a better correlation towards the field irrigated with the canal water than tube well. Although the result of the downscaling shows some insights into the spatial variation of soil moisture throughout the CZO, we are working to incorporate the additional dependable parameters for a better accuracy.

Area B - Board 187

Anthropogenic climate change is leading to more frequent and extreme weather events such as heat waves. This can be further exacerbated by the Urban Heat Island (UHI) effect, and with rapid urbanization found in countries like China, an increasing proportion of the population may be exposed to extreme heat-stress events without access to the resources they need. There are two routes to action this problem: heat response plans which target vulnerable populations at the time of events with local aid in the community; and environmental design strategies which direct long term urban construction to reduce heat absorption and retention. Both of these strategies require an understanding of the distribution of persistent ‘hot spots’ within urbanised areas.

The use of satellite retrieved Land Surface Temperature (LST) to analyse the UHI effect is investigated, including methods to determine the areas most affected by extreme heat. This is put into the context of potential climate services, and how they could benefit residents in large and developing cities. Four case study cities in China are investigated: Xi’an, Shanghai, Beijing and Guangzhou. LSTs from NASA's MODIS instrument on board the Aqua satellite are used, due to the twice daily, 1km coverage of this instrument, and an overpass time close to the hottest part of the day. ESA's CCI Land Cover product for years 2003 - 2014 is also used. Methods for spatial and temporal analysis are investigated, with initial spatial analysis showing that LST correlates well with land use as expected, and that there is a condensed hot spot in the centre of the cities which is more pronounced at night. City regions are further analysed for pixel anomalies against the urban background, and to identify any changes in LST over time. Finally in-situ data from land stations including daily minimum and maximum near surface air temperature are incorporated into the analysis for comparison and to identify if these data combined can form the building blocks of an urban heat climate service.

Area B - Board 188

Land Surface Temperature estimation is important task for city development planning. Providing analysis of thermal maps of city it is possible to define places with temperature anomalies and define urban heat islands (UHI) , that have decreasing effect on air quality in this areas and low comfortability for citizens. Analysis of land surface temperature maps within land cover maps [1] gives the possibility to identify possible UHI’s and provide changes in the city’s development planning to increase cityzens life quality. Thermal anomalies and thermal islands are disasters of the city which need to be detected, his is one of the tasks for realization of Smart City in the project Horizon2020 Era-Planet SMURBS. Also land surface temperature is one of the essential variables that can be used for SDG’s indicators calculation in GEOEssential project.
In this work, we provide analysis for Rivne city using classification map of land cover and the average surface temperature during the summer of 2018. Classification map is created using Sentinel-1 and Sentinel-2 10 meters satellite images [3] with deep learning neural network classification approach [4,5] developed in Space Research Institute of NANU and SAU]. Land Surface temperature map build on Landsat-8 30 meters 10-th thermal band images. We also carried out a segmentation of the territory of the city by type of buildings, infrastructure, development density, density of green spaces, dividing them with sufficiently large areas without artificial objects or main transport solutions, thus we created 600 separate zones. This segmentation give us possibility to analyze problems in city planning and find solution to improve the situation in these areas. For the beginning we calculated zonal statistics for development density and tree density in this areas. Analysis of this statistics indicate high livelihood of buildings, low percents of coverage with healthy trees and as well as indicates the areas in which there are almost no trees, this primarily concerns the problem of tree felling and uncontrolled development of the city, which leads to a significant excess of surface temperature in comparison with the whole city, that is, the emergence of UHI’s. All areas with more than 80% building density and less than 10% of tree cover have very high summer mean land surface temperature, that can be more than 30 degrees for morning time. This selection criterion identifies most of the zones with the highest surface temperature.
This study made it possible to determine the living areas in which it is necessary to make landscaping in the first place because the average surface temperature there is too high, which creates a discomfort for the inhabitants. Also, there we found zones of industrial character, which have a negative impact on local residential buildings. Discovered parks and alleys are the best for recreation and areas where greenery is in poor condition, which is why they do not fulfill the duties of urban trees.

References
1. Kussul, N., Lavreniuk, M., Sumilo, L., Kolotii, A., Rakoid, O., Yailymov, B., ... & Vasiliev, V. (2018, July). Assessment of Sustainable Development Goals Achieving with Use of NEXUS Approach in the Framework of GEOEssential ERA-PLANET Project. In XVIII International Conference on Data Science and Intelligent Analysis of Information (pp. 146-155). Springer, Cham.
2. Shumilo L. , Kolotii A., Lavreniuk M. , Yailymov B. (2018 July). Use of land cover maps as indicators for achieving sustainable development goals. In IGARSS, Valencia, Spain. – 2018. – P. 830-833
3. Kussul, N., Shelestov, A., Lavreniuk, M., Butko, I., & Skakun, S. (2016, July). Deep learning approach for large scale land cover mapping based on remote sensing data fusion. In Geoscience and Remote Sensing Symposium (IGARSS), 2016 IEEE International (pp. 198-201). IEEE.
4. Kussul, N., Lavreniuk, M., Skakun, S., & Shelestov, A. (2017). Deep learning classification of land cover and crop types using remote sensing data. IEEE Geoscience and Remote Sensing Letters, 14(5), 778-782.
5. Lavreniuk, M., Shelestov, A., Kolotii, A., Vasiliev, V., Shumilo, L., & Kussul, N. (2018, April). Automated System for Crop Mapping in Amazon Web Services based on Sentinel Data. In EGU General Assembly Conference Abstracts (Vol. 20, p. 16629).

Area B - Board 189

Urban surface temperature (LST) is used as a key indicator for various urban climate issues. Moderate Resolution Imaging Spectroradiometer (MODIS) with 1 km resolution LST provides four times LSTs in a day. In addition, MYD/MOD21 LST products using the ASTER Temperature Emissivity Separation (TES) technique are provided, distinguished from the existing MYD/MOD11 LST products, which used the split-window algorithm. MODIS LSTs are widely used, but have limitations in not able to be obtained below cloud areas. Especially in humid regions, the cloud area is very wide and the contaminated LSTs are vast. In this study, MODIS LSTs were estimated by using air temperatures observed from dense weather stations in Seoul, Korea, and the built model was applied to cloudy areas to estimate urban LSTs under cloudy condition. In-situ hourly air temperatures (0-23h) observed from 25 weather stations of Korea Meteorological Administration (KMA) in Seoul, solar radiation, aspect, slope, elevation, Local climate zone, mean land surface temperature of study period and the ratio of impervious area were used as input variables. Random forest machine learning was used to estimate the MODIS daytime Aqua LSTs (MYD11A1 and MYD21A1, respectively) in clear sky condition. The developed models were applied to the cloudy regions, and we validated the LST estimation result using observed Land surface temperature from weather station of KMA which was not included in modeling. The validation results showed Root Mean Square Errors (RMSE) of 2.4 and 4.4◦C and correlation coefficients of 0.60 and 0.53 for estimating the clear-sky LSTs for MYD21A1 and MYD11A1 respectively, and RMSE of 6.1 and 3.9◦C and correlation coefficients of 0.48 and 0.56 for estimating cloudy-sky LSTs for MYD21A1 and MYD11A1 respectively.

Area B - Board 190

The land surface emissivity (LSE) is an intrinsic property of natural materials that is deemed particularly relevant to estimate the surface energy budget. The energy budget is one of the key processes in climatic, hydrological, ecological, and biogeochemical models. Besides, a full exploitation of the thermal infrared imagery to notably estimate the land surface temperature (LST) relies on the knowledge of LSE. Therefore, an accurate retrieval of LSE from remotely sensed observations remain a primary objective of the scientific community.
Herein, the proposed work aims at improving the methodology for processing high resolution thermal infrared satellite data of Landsat like sensors; i.e.: only one or two channels in thermal wavelength [10-12.5μm], nadir direction of acquisition, high resolution with simultaneous visible and near infrared (VNIR) eventually medium infrared (MIR) data at compatible resolution. For such, we are revisiting in the frame of this study a classic method based on a statistical relationship between the vegetation indices derived from (VNIR) bands and the LSE.
In order to account for vegetation canopies, we simulated a range of emissivity vs. vegetation indices thanks to the SAIL and the SAIL-thermal radiative transfer models. Indeed previous studies considered either spectrum from soil and leaf samples obtained from spectral libraries without accounting for the canopy processes (i.e. the cavity effect). Other previous studies considered in-situ measurements of emissivity, but in any cases only very few measurements (only specific vegetation cases) were available.
In order to simulate emissivity for vegetation stand, we used leaf spectra and soil spectrum based on the new ECOSTRESS spectral library and the USGS / ONERA database. The two databases encompass a set of leaf and soil reflectance in a spectrum ranging from visible to thermal infrared, i.e. [0.3µm to 15μm]. The other parameters required as input parameters for SAIL and SAIL-thermal radiative transfer model were chosen in order to represent a wide range of vegetation canopies, soils and geometric configurations. In final, we obtain a synthetic and database of coherent canopy reflectance and emissivity covering solar and thermal infrared (TIR) spectra.
This yields a background for investigating the possible relationships with a minimum degree of confidence between VNIR bands (Landsat bands as starting point) and TIR emissivity.
In parallel to this effort, we analyze the measurements acquired during the hyper-spectral airborne campaign AHSPECT that took place in 2015 on a summer day and a fall day in the South-West of France. The knowledge of land cover types for the areas fly over by the aircraft with a typical metric resolution allows us to extract the associated spectrum per each land unit. Owing to a large extraction over manifold land cover types, the robustness of the relations theoretically found between VNIR and TIR data are verified and even consolidated.
As a result we concluded on the possibility to estimate accurately (with associated error) the LSE at high resolution with associated VNIR remotely sensed data.

Area B - Board 191

Number of different methods has been developed and successfully used for Land Surface Temperature (LST) estimation; among them, some require algorithms that exploit the different atmospheric absorption in adjacent thermal infrared (TIR) channels. The nowadays accuracy of these two-channel methods, validated for current satellite sensors with different central wavelength and width of the TIR channels, is typically below 1.2K.
In this work, we aim to present the approach followed to select the optimal TIR bands for the LST inversion problem and to provide Split Window (SW) coefficients that can be used to retrieve LST from the Thermal Airborne Spectrographic Imager (TASI-600) airborne data. TASI is a commercial hyperspectral infrared sensor produced by ITRES, Canada, with 32 TIR bands covering the range between 8 and 11.5 µm with a spectral resolution of 109.5 nm. TASI spectral bands have been numerically analysed in order to select the optimal bands for the algorithm design. The outcomes of this analysis can be exploited for the optimal band selection for new satellite TIR hyperspectral sensors.
The SW algorithm according to the Jiménez-Muñoz and José A. Sobrino 2008 formulation with seven coefficients was calibrated and validated using a datasets of profiles applied to configure MODTRAN to simulate Top of the Atmosphere (TOA) brightness temperatures. These datasets were selected from a training database compiled by Borbas et al. (SeeBor database) which consists of 15,704 clear-sky profiles of temperature, Total Column Water Vapour (TCWV) and ozone; and other ancillary data such as spectral emissivity, elevation, site latitude and longitude, skin temperature, surface pressure and International Geosphere-Biosphere Programme (IGBP) classification. Simulations have been performed on a subset of the SeeBor dataset resized on Europe.
The algorithm coefficients have been retrieved by multiple regression analysis and their statistical significance was evaluated by simultaneous hypothesis using a Student’s t test and a P-value lower than 0.05. Among the TASI 32 spectral bands the optimal regression was obtained by using TASI band 19 (10.034 µm) and band 28 (11.051 µm). For the overall land cover types and for the optimal bands selection we obtained an R2 of 0.93 and a RMSE of 0.635K. The algorithm, when applied to the single IGBP classes, showed a good performance for the majority of IGBP classes, except for barren profiles. This low performance for the Barren class is anyhow physically consistent; since barren emissivities have larger spectral variations in the window region between 10 and 12 µm due to the variable abundances of vegetation and soil units within the IGBP Barren class.
To date, the SW algorithm validation is foreseen by using real TASI data acquired in the Basilicata region in Italy on the urban and industrial district.

Area B - Board 192

This study proposes an integrated use of multi-satellite data for the calibration of a distributed hydrological model to improve water resources assessment in the Yangtze River basin. In particular, a new methodology is used to constrain an internal model variable of the distributed hydrological models at basin scale by using satellite data of land surface temperature.
The model algorithm, of the fully distributed hydrological model Flash–flood Event–based Spatially–distributed rainfall–runoff Transformation - Energy Water Balance model (FEST-EWB), solves the system of energy and mass balances in terms of a representative equilibrium temperature (RET) that is the land surface temperature that closes the energy balance equation and so governs the fluxes of energy and mass over the basin domain. This equilibrium surface temperature, which is a critical model state variable, is comparable to LST as retrieved from operational remote sensing data (MODIS) which is used for the calibration of soil and vegetation parameters. The effects on river discharge of the Three Gorges dam, the largest hydropower project in the world is evaluated. Observed flow duration curves, for the four available river cross sections, will be evaluated and compared to FEST-EWB simulated flow duration curves as well as observed as simulated hydrographs. The dynamic of the two large lakes (Poyang and Dongting), which change considerably their area during the seasons, will also be analyzed using MERIS and ASAR data to infer surface area changes and LEGOS altimetry data (Topex/Poseidon) for water level.
FEST-EWB is run in for the whole Yangtze River basin at spatial resolution of 0.009° and temporal resolution of 3 hours. Results are provided in terms of hourly evapotranspiration, soil moisture and land surface temperature maps and flood hydrographs at different river cross sections for the period between 2003 to 2006.

Area B - Board 193

The objective of this study is the evaluation of the potential of two Land Surface Temperature downscaling algorithms with respect to high resolution LST from LANDSAT 7 ETM+ and resampled MODIS LST (MOD11A1).
Four different LST sources have been compared over the Heihe river basin, an endorheic basin in China, characterized by a wide variety of ecosystems, from desert oases to irrigated croplands and wooded mountain ridges.
The first two LST sources are measurements provided by the ETM+ instruments (aboard satellite Landsat-7) at 30m spatial resolution every 16 days, and MODIS (aboard Terra at the lower resolution of 1000m. The other two sources are products of downscaling algorithms. The STARFM (Spatial and Temporal Adaptive Reflectance Fusion Model) algorithm merges ETM+ and MODIS images to obtain LST data with the spatial resolution of the former and the temporal frequency of the latter, involving neighbouring pixels in the process. On the other hand, the DisTrad (Disaggregation of radiometric Temperature) algorithm, establishes a link between LST and NDVI, in order to revert to the former anytime vegetation data is the only data available.
Globally, resampled MODIS and DisTrad perform better, not quite reaching the accuracy of ETM+ but all the same yielding an accurate approximation. On the other hand, STARFM struggles with the variety of land cover types, offering an acceptable performance in the desertic area, which is the most uniform of all. Categorizing the pixels according to their land cover type, it is found that vegetated areas, especially croplands, are the most difficult to interpret for the LST sources, with low performance in the early summer during the peak of the maize season. Furthermore, grouping pixels by their lighting condition (whether they are in light or in shadow) does not offer major results: data quality for shadowed pixels is quite worse than for lighted pixels, but the number of the formers is so low that the impact on the overall result is close to negligible.
Overall, MODIS and DisTrad are the best candidates to compensate for the low temporal frequency of ETM+ without losing too much accuracy. Furthermore, DisTrad application requires more input data and computing effort than re-sampling MODIS.

Area B - Board 194

Thermal infrared (TIR) remote sensing provides direct signature of land surface temperatures (LST) and surface energy balance (SEB) fluxes which are of prime interest for ecosystem water use and stress mapping. However, in landscape ecological studies the use of TIR remote sensing data is limited due to the unavailability of high spatial resolution TIR remote sensing data and a limited understanding of these data and associated difficulties in processing them into useful information.
In this paper we introduce the Luxembourg Airborne Observatory System (LAOS) that has been developed to facilitate at the landscape scale (100-1000 km2) flexible TIR image acquisition and accurate retrieval of LST and parameterisation independent estimation of evapotranspiration (ET). LAOS consists of a thermal hyperspectral camera (Telops Hyper-Cam LW) and an airborne observation platform operated from a third party aircraft (twin-engine Cessna T303). The data pre-processing scheme includes generation of radiance images from the raw FT-data and orthophoto mosaic generation. The science data flow comprises LST retrieval through dedicated temperature-emissivity separation scheme and ET estimation through a calibration-free SEB model (STIC: Surface Temperature Initiated Closure).
Until now, LAOS has been tested over grassland (Latisana site, Italy) and forest (Attert catchment site, Luxembourg). Examples of TIR-image based LST and ET retrieval results in comparison to collected ground data are discussed for these two sites. At the Attert site, multiple airborne TIR images were acquired over 22 forest plots during a summer day, converted to LST and ET using the STIC model in conjunction with meteorological data, and compared to in situ measurements of transpiration (sap velocity). Sap velocity and ET both had peak around 14:00, but declined at different times in the afternoon. Results showed that ET at 9 hrs were same for all three geology types present in the study area, but at 14-19 hrs ET was larger over sandstones as compared to the schist, suggesting a (at least partial) geologic control of eco-hydrological processes in the area.
At the Latisana site, multiple airborne TIR images were acquired over grassland plots, converted to LST and ET (same methods as above), and evaluated against EC tower and chamber measurements of infrared radiometric temperature (TR) and ET. Image LST were also compared to in situ water temperatures of the pool with deviations of less 0.5°K. Image LST over grassland were generally larger compared to the tower and chamber TR, but showed similar trends over the day. Chamber ET and image ET agreed relatively well in the morning with increasing deviations towards noon, where chambers might overestimate ET at hot temperatures.
These examples demonstrate that airborne land surface temperature and evapotranspiration can be mapped at very high resolution and multiple times during a day which allows unprecedented analysis of ecological processes across landscapes.

Area B - Board 195

Infrared surface spectral emissivity is a fundamental property of terrestrial materials, important in a host of environmental applications ranging from mineral mapping to land cover change determination, and in particular the remote sensing of land surface temperature (LST). Surface spectral emissivity has also been used to inform the chemical and physical composition determination of terrestrial and planetary materials. Multiple laboratories have now developed facilities for determining the surface spectral emissivity of samples from reflected infrared radiation measurements (making use of Kirchhoff’s Law), with the results often made publicly available such as in the ECOSTRESS (Ecosystem Thermal Radiometer Experiment on Space Station) spectral library. However, until now there has been relatively little work conducted in evaluating the uncertainties in these types of measurements and the consistency of measurements made in different laboratories. Here we intercompare surface spectral emissivity measurements in a ‘round-robin’ exercise using identical samples at 7 different international laboratories. During this exercise, 4cm-1 spectral resolution emissivity spectra were determined for three samples: distilled water, which has a uniformly high spectral emissivity, and two artificial samples constructed from gold and aluminium sheets laminated in polyethylene (PE), which show respectively both Lambertian and specular characteristics along with widely varying emissivity features across the 2.5 – 14 μm region. Here we present first results from the round-robin intercomparison exercise and discuss the implications for the uncertainty of surface spectral emissivity values determined in the laboratory and for the satellite LST products and LST field validation studies that rely on them.

Area B - Board 196

Understanding of land surface spectral and broadband emissivity is essential to the robustness of satellite LST products. This includes laboratory and field emissivity measurements, which can both support the use of in-situ radiometer-based LSTs used for satellite product validation, and which can act as a point of comparison to the emissivities assumed within the satellite LST products themselves. Laboratory measurements maybe the most accurate but can have limited ability to capture specific field conditions and their spatio-temporal variability. Field techniques have therefore been developed in an attempt to capture these variabilities, and these include the one-lid and two-lid box methods using broadband radiometers (Rubio et al., 1997; 2003), and the use of far higher spectral resolution portable infrared spectrometers to make measurements of surface emitted and reflected downwelling thermal emission from which surface spectral emissivity can be retrieved. Here we compare these approaches for a series of natural and man-made materials, providing independent evaluation using laboratory measurements of surface spectral emissivity made with a Bruker Vertex 70 Fourier Transform InfraRed (FTIR) Spectrometer and infragold integrating sphere. Outcomes have been used to inform the satellite LST products evaluations now using data from the recently deployed East African LST measurement site on the Kapiti Plains, Kenya.

Area B - Board 197

The Copernicus Global Land Service provides a wide range of satellite and earth observation derived products focused on monitoring of the terrestrial environment. For these products to be of use for environmental decision-making they must be validated using independent data sources to ensure they are of high quality and consistency.

Validation of satellite-derived products has traditionally been through intensive field campaigns, which enable the spatial variability of the variable of interest to be well-characterised over a particular site but are limited in time and space. In recent decades, numerous ground-based monitoring sites have been established as part of several wider ground-based observation networks. The spatial variability is not as well-characterised for these sites, but they provide a better sampling of data and provide useful information in terms of temporal consistency of products.

The Ground-Based Observations for validation of Copernicus Global Land Products (GBOV) project aims to develop and distribute robust in-situ datasets from these ground-based monitoring sites for a systematic and quantitative validation of all seven Copernicus Global Land Products (CGLPs). In GBOV, observations from identified monitoring sites are processed into various point-scale reference measurements and then upscaled in order to be more representative of the equivalent CGLP pixel, which will cover a much wider spatial area.

Here we present the methods used to upscale in situ Land Surface Temperature (LST) data for validation of the Copernicus Global Land Service LST product. We describe the derivation of in situ LST data from radiometric observations at several ground-based monitoring sites. We then outline the upscaling method used to estimate LSTs which are equivalent to a CGLP pixel from the derived point-scale in situ LSTs. We also present preliminary results comparing the validation of the Copernicus Global Land Service LST product with upscaled LST data instead of point in situ LST data.

Area B - Board 198

Land surface temperature (LST) is the temperature of the Earth’s surface, which can be obtained with satellite remote sensing. LST is a particularly useful quantity for various applications within climate research. Therefore, validation of satellite data sets against in situ data is important for ensuring the quality of satellite derived LST, which is usually based on investigating the absolute differences between the two data sets.
In the work presented here, LST data sets from several polar-orbiting and geostationary satellites, namely AATSR, GOES, MODIS and SEVIRI, are compared over two sets of in situ stations, run by Karlsruhe Institute of Technology (KIT) and by the National Oceanic and Atmospheric Administration (NOAA) (Surface Radiation Budget Network - SURFRAD stations). All satellite data sets were produced in a harmonized format developed in the framework of ESA’s GlobTemperature (GT) project. All validations follow a standardized approach, which ensures that the obtained results are directly comparable to each other.
The stations are located in Europe, Africa and USA, and cover different land surface types, climates, biomes and topographies. Three years of satellite LST data are validated over all stations and, thanks to longer data availability, for ten years over the SURFRAD sites.
The results of the validations show that the differences between satellite and in situ LST are strongly influenced by spatial homogeneity, land cover class and orography at the stations. For the years 2010 – 12, overall differences between satellite and in situ LST are often within 2 K, with generally better results during night. During day, when shadows are present and the influence of heterogeneous land covers is higher, the observed differences increase. AATSR and MODIS LST tend to exceed the corresponding in situ LST values. SEVIRI and GOES LST tend to be lower than the in situ LST, an effect that is stronger for GOES LST.
As LST differences varied considerably between individual stations, time series of them were also investigated per station. A seasonal cycle was observed for most data sets, but at some stations the LST differences vary considerably with time. This temporal variation was strongest for a station located on a mountain edge in a heterogeneous landscape and for a station experiencing strong seasonal cycles including rainy seasons. At two stations, the seasonal cycle was so strong that the daytime data was excluded from further analysis. At another station located in a valley, a considerable influence from topography was found. Heterogeneous land covers impacted spatial representativeness and affected the analyses at several stations negatively. At two stations, directional effects from bushes or trees were clearly visible in the analyses.
The results show that it is important to investigate the differences between the area-integrated satellite measurements and the in situ point measurement for each site separately in order to be able to interpret - and possibly account for - the many factors influencing the results.
In situ validation of LST is a well-established and reliable method to gain information on the quality of satellite LST data set and the large data base of satellite LST provided by the GT project makes it feasible to carry out standardised validations of several data sets in a comparable way. The validations presented here are providing the baseline for further validation work and a larger data base construction carried out within ESA’s Climate Change Initiative (CCI) LST project.

Area B - Board 199

Understanding the changing and complex urban thermal environment is key to addressing the health and sustainability of the cities in which more than half of the world’s population live. The monitoring and assessment of the thermal environment requires spatial resolution that so far has precluded air temperatures from being a viable parameter in most cities. Land surface temperatures (LSTs) offer the ability through satellite remote sensing to investigate the urban environment in a robust and consistent manner. Additionally land surface emissivity (LSE) is required to enable accurate LST estimation and characterise broad-scale thermal infra-red properties of materials.
In this work, the first Optimal Estimation (OE) retrieval of simultaneous LST and LSE data optimised to be robust for highly complex surfaces is presented. It uses the thermal channels of the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) instrument with a spatial resolution of 90 m. The retrieval was first tested in a simulated environment, trained against the ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) emissivity spectral library and tested against the Spectral Library of impervious Urban Materials (SLUM) dataset. In these simulations the uncertainties retrieved are better than 1 K in LST and 0.015-0.017 for LSE.
This work also marks the first usage of an inverse method with ASTER data. Although this method was primarily conceived to provide a robust retrieval for heterogeneous urban regions, it has generic applicability to all land surface types and as such an assessment of its accuracy and uncertainty was conducted at key validation sites. Validation of the LST and LSE was undertaken through two studies. Firstly through inter-comparison with the JPL Temperature-Emissivity Separation (TES) method at a validation site in Algodones, USA, and secondly through assessment with an in-situ validation site in Gogabeb, Namib Desert. The inter comparison results agreed well between the TES and OE methods, particularly in ASTER Channel 12 and with very low retrieval radiance residuals across the channels. The OE retrieval also performs well against the in-situ validation data from Gogabeb, which assessed the retrievals performance in both LST and LSE. In both of these assessments the surfaces understudy have large emissivity spectral inter channel variations, a feature common in the urban environment and therefore a useful test for a method aimed at urban applications. These results show the retrieval of robust and scientifically meaningful LST and LSE data for a heterogeneous environment from ASTER.

Area B - Board 200

Land surface temperature (LST) is of fundamental importance for terrestrial ecosystem and land surface processes, such as, vegetation health, Earth's surface energy balance, and global/regional hydrological cycle. In 2016, LST has been recognized by Global Climate Observing System (GCOS) as an Essential Climate Variable (ECV) due to its importance to the global climate system. Meanwhile, GCOS also documented its desired requirements for LST to be used as an ECV for climate studies, including, spatial/temporal resolutions, data uncertainty and data stability. Currently, global and regional LSTs are measured by various remote sensing instruments, such as, MODIS onboard NASA Terra/Aqua missions, VIIRS onboard NOAA S-NPP/NOAA-20 missions, SLSTR onboard Copernicus Sentinel-3 series, and various agencies' geostationary platforms. Because of the value of long-term LST data, significant efforts are underway to develop satellite based climate data records (CDR) for LST, such as, ESA's Climate Change Initiative (CCI) and the GlobTemperature project. As a prototype, GlobTemperature has combined LST estimated from ATSR and AATSR data to produce a 17-year (A)ATSR LST CDR dataset. While it is crucial to produce long term LST records, it is also equally important to assess the quality and stability of the existing satellite based LST records to ensure these data can meet the requirements set by the GCOS. In this study, we use triple-allocation methods to assess the uncertainty of two satellite based LST data (i.e., (A)ATSR LST CDR and MODIS LST), CMIP5 simulations and three different land surface reanalysis LST data (i.e., MERRA-2, ERA-Intrim, NCEP). In addition, we also assess the stability of these long term LST data records based on time series analysis. These analysis provide a preliminary analysis of existing long term LST datasets. These preliminary assessment results can help guide future development of LST CDRs and provide valuable information for future LST applications for climate studies.

Area B - Board 201

Sentinel-3A was launched on 16 February 2016 and was the first twin of the Sentinel-3 mission. Sentinel-3B was recently launched, 25 April 2018, but data is still not available. Both satellites carry onboard Sea and Land Surface Temperature Radiometer (SLSTR) sensor which performs Thermal InfraRed (TIR) measurements with its two bands (S8 and S9) situated on 10.9 and 12.0 µm and with a spatial resolution of 1 Km. Thermal bands were designed to provide a noise equivalent difference temperature (NeΔT) below 50 mK.
Until the availability of Sentinel-3B data, the main aim of this work is to contribute to the quality data only of the Sentinel-3A SLSTR sensor with calibration of thermal bands and validation of Land Surface Temperature (LST) data. Images between years 2017 and 2018 have been analyzed for cal/val through in-situ measurements obtained with fixed stations situated over Spain. Currently, three sites have been identified and equiped with thermal radiometers: the agricultural area of Barrax (39.05N, 2.1W), the marshland area in the National Park of Doñana (36.99N, 6.44W), and the semi-arid area of the National Park of Cabo de Gata (36.83N, 2.25W). These sites have different characteristics in terms of variety of crops, biomes and soil type. Precision of in-situ measurements has been computed for each station (e.g. surface heteorogeneity, emissivity uncertainties, sensor calibrations...) showing LST uncertainties of 0.5-0.8 K.
LST from SLSTR data was derived using Split-Window (SW) algorithm which requieres emissivities input. We adopted the Normalized Difference Vegetation Index (NDVI) Thresholds Method to estimate the emissivity instead of the biome class method proposed in the LST level-2 product provided by ESA in order to compare differences between both inputs. But, previously to application of LST algorithm, a Vicarious Calibration (VC) was performed to test the nominal sensor calibration throught Radiative Transfere Equation.
Preliminary VC result shows a near zero bias in both bands with a precision below 1.0 K in terms of brightness temperature at 300 K. SW algorithm provides Root Mean Square Errors (RMSE) around 2.0 K when compared to in situ values of LST. Comparison between LST level-2 product and SW algorithm shows minimal emissivity extraction influence with RMSE differences below 0.5 K.

Area B - Board 202

The International Space Station (ISS) is the greatest scientific project humankind has ever carried out and, simultaneously, the largest artificial object on the Earth’s orbit. Germany participates in research aboard the ISS with a variety of experiments, ranging from research on the atmosphere to space medicine, and ornithology. While several multispectral and radar sensors are aboard the ISS carrying out American, Japanese, or Russian experiments, the European remote sensing community has seen the ISS orbit so far rather less suitable for scientific instruments performing Earth observation tasks. Unlike satellites on polar orbit, the ISS has an inclined equatorial orbit, sun-asynchronous, and with variable nadir. At second glance, the dynamic orbit of the ISS combines some unique features: First of all, it allows Earth observation at different day and night-times. Secondly, it is approx. 400 km closer to the Earth’s surface than most satellite orbits. With “DESIS” (German abbreviation: DLR Earth Sensing Imaging Spectrometer) the first German sensor was recently attached at the ISS hyperspectrally sensing the Earth surface from the Multiple User System for Earth Sensing (MUSES). The talk will present the framework of the upcoming second German remote sensing experiment on the ISS: “DIEGO – Dynamic Infrared Earth Observation on the ISS Orbit“. The sensor will collect high-resolution, multi-spectral data in visible, near infrared (VNIR), and mid-wave infrared (MWIR), as well as long-wave infrared (LWIR) of the Earth’s surface. DIEGO will close the existing gap in high-resolution thermal infrared Earth observation. The resolution within the LWIR spectra will exceed 60 m. The sensor will provide eleven spectral bands with six in LWIR, two in MWIR, and three in VNIR. Hence, DIEGO will provide the possibility of gaining higher accuracies when calculating surface temperature from space by using the radiative transfer or Planck’s law. Better certainties are necessary to deduce essential climate variables like fire disturbance via the fire radiative power, but also measurements for cloud, land, and sea surface temperatures. The DIEGO mission data will be made available to the scientific and research community via web GIS and data hubs. Its sensor characteristics makes it a suitable successor of the ECOSTRESS mission prolonging thermal infrared remote sensing from the ISS in the next decade. Furthermore, one of the main goals of DIEGO is to spread and deepen knowledge about key technologies of space flight and Earth observation in school lessons and break new ground in digital education and experimental learning for knowledge transfer.

Area B - Board 203

Land use, vegetation and other surface characteristics significantly influence Land Surface Temperature (LST) and the diurnal temperature cycle (DTC). However, LST is also influenced by the current and previous weather situation, e.g. atmospheric transparency (cloud, aerosol, and water vapor) and soil moisture availability. Therefore, instantaneous LST values are of limited use for analyses of annual and inter-annual variations and short term synoptic effects need to be separated from more permanent surface characteristics. Since October 2017 the Land Surface Analysis Satellite Application Facility (LSA SAF) operationally generates the Derived Land Surface Temperature (DLST) product, which represents 10-day thermal characteristics of the land surface and provides a generalized characterization of LST temporal dynamics. DLST consist of two sub-products: maximum and median temperature composites and so-called Thermal Surface Parameters (TSP), the parameters of a DTC model fitted to the composites. While maximum temperature composites approximate the hottest (driest) day in a composite-interval, median composites approximate the most frequently encountered (i.e. typical) day. Together composites and TSP provide 10-day syntheses of LST and spatially smooth fields of maximum and minimum temperature.

The only input required by the DLST algorithm is LSA SAFs operational LST product for MSG/SEVIRI (MLST); both DLST sub-products have the same spatial resolution (3 km) but considerably fewer missing pixels and are spatially more continuous. However, TSP require about 10 times less storage space than MLST composites and have direct physical meaning, e.g. minimum temperature (i.e. at sunrise), temperature amplitude and time of maximum temperature. Therefore, TSP are particularly well suited for studies of thermal surface characteristics, e.g. Surface Urban Heat Islands (SUHI), and are readily combined with land surface parameters varying on similar time scales, e.g. vegetation indices. The presentation briefly introduces the LSA SAF DLST (LSA-003) product and demonstrates some practical applications; it concludes by highlighting a recently published SUHI study that applied an adapted version of the algorithm to 1 km MODIS LST.

Area B - Board 204