Land remote sensing capabilities in the optical domain have dramatically increased in the past decade, owing to the unprecedented growth of space-borne systems providing a wealth of measurements at enhanced spatial, temporal and spectral resolutions. Yet, critical questions remain as how to unlock the potential of such massive amounts of data, which are complementary in principle but inherently diverse in terms of products specifications, algorithm definition and validation approaches. Likewise, there is a recent increase in spatiotemporal coverage of in situ reference data, although inconsistencies in the used measurement practices and in the associated quality information still hinder their integrated use for satellite products validation. In order to address the above-mentioned challenges, the European Space Agency (ESA), in collaboration with other Space Agencies and international partners, is elaborating a strategy for establishing guidelines and common protocols for the calibration and validation (Cal/Val) of optical land imaging sensors [1]. Within this paper, this strategy will be illustrated and put into the context of current validation systems for land remote sensing. A reinforced focus on metrology is the basic principle underlying such a strategy, since metrology provides the terminology, the framework and the best practices, allowing to tie measurements acquired from a variety of sensors to internationally agreed upon standards. From this general concept, a set of requirements are derived on how the measurements should be acquired, analysed and quality reported to users using unified procedures. This includes the need for traceability, a fully characterised uncertainty budget and adherence to community-agreed measurement protocols. These requirements have led to the development of the Fiducial Reference Measurements (FRM) concept, which is promoted by the ESA as the recommended standard within the satellite validation community. The overarching goal is to enhance user confidence in satellite-based data and characterise inter-sensor inconsistencies, starting from at-sensor radiances and paving the way to achieving the interoperability of current and future land-imaging systems.
[1] Niro, F.; Goryl, P.; Dransfeld, S.; Boccia, V.; Gascon, F.; Adams, J.; Themann, B.; Scifoni, S.; Doxani, G. European Space Agency (ESA) Calibration/Validation Strategy for Optical Land-Imaging Satellites and Pathway towards Interoperability. Remote Sens. 2021, 13, 3003.
The global market of Small UAV is knowing a fast growth, in Europe is valued at $197 million in 2015 and is forecast to reach as much as $3 billion by 2025 [1]. The recent advancements in technologies and sensors make possible a wide range of UAV models in various sizes, weights, and payload carrying capabilities to support the broad range of commercial and military applications. Moreover, since 2021 the emanation of a unique and agreed regulations made by the European Union Aviation Safety Agency (EASA) are supporting a more rational development and deployment of drones in various sectors, like land mapping, infrastructure, agriculture, transport, entertainment, and security. Toward more specific applications, we proposes an innovative approach for in-situ validation of optical satellite products, using small UAV system (hexa-rotors) equipped with a multispectral camera Sentinel-2 likes. In particular, we focus on the surface reflectance and Bidirectional Reflectance Distribution Function (BRDF) [2] validation using a peculiar payload and programming an acquisition plan capable to provide enough data and automatic, repeatable surveys on different surfaces.
BRDF describes the level of surface anisotropy, the correction techniques is conceived to normalize the observations to a standard angular configuration, or alternatively, as an intrinsic signature of the considered surface that can be used to retrieve relevant bio-geophysical variables. The parametrization of the Bidirectional Reflectance Distribution Function (BRDF modeling) is based on the inversion of the model with observed data (multi-angular measurements made by different type of devices such as field instruments or satellites) in order to find the model’s parameters which best fit the measurements. The calibrated model is then applied to predict reflectance at any sun-view geometry conditions. Although reflectance at any observation angle can be obtained using a ground-based multi-angular instrument, the complicated measurement process is time-, labor- consuming, and only "point" measurements can be retrieved [3].
At present, there is a lack of extensive in-situ validation data to assess the quality of BRDF correction. The proposed experimental activity demonstrates that the combination of a multispectral camera Sentinel-2 likes and the UAV hexa-rotor flight potentials (as the capability of setting different view azimuth angles and view zenith angles), can improve the capability of BRDF modelling and validation. In order to pursuit our goals, we identified the following implementation steps: the Integration of multispectral camera on UAV platform; the system calibration; the flight planning; the dataset processing; the BRDF modeling.
The payload carried by the UAV is the multispectral camera MAIA S2 [4], which permits the simultaneous acquisition of high-resolution images at various wavelength intervals in VIS/NIR (Visible/Near-Infrared) electromagnetic spectrum regions. In particular, the multispectral camera is designed to acquire in the same wavelength intervals of the ESA Copernicus Sentinel-2 (S2) satellite constellation. The imaging sensors has 1.2 Mpixel resolution, high sensitivity and global shutter technology, allowing the simultaneous acquisition of images free from motion artefacts. Moreover, the acquisition system includes a further module, the Irradiance Light Sensor (ILS), which measures the level of the down-welling light in each band, allowing the correction for light changes during the survey, such as those caused by clouds. The use of the ILS is crucial for radiometric correction when passing from digital number (DB) to reflectance. ILS provides irradiance data at the exact time of shooting for each image and in each spectral band, substantially improving the accuracy of reflectance values in multi-temporal surveys.
The UAV system is a hexa-rotor with a maximum take-off weight up to 6 kg. It is equipped with a Plug & Play System, which allows the pilot to switch between payloads. Real-time mission management is possible as well Autonomous Waypoint navigation. The flight unit has integrated Global Positioning System (GPS) and Inertial Measurement Unit (IMU) sensors and auto-stabilization of flight in manual mode. The system is characterized by a maximum flight time up to 20 minutes. The Italian Civil Aviation Authority (ENAC) has given the Certification of Design attesting the compliance to Italian and European (EASA) laws. Remote controlled gimbal for orienting MAIA camera between 0° and 90° respect to surface normal direction is possible as well as controls of orientation on the horizontal plane. These two latter properties are very useful for a complete BRDF characterization.
In this work, we present the results obtained by testing our UAV system over two different land cover type: vegetation and asphalt. The flight plan designed for BRDF modeling survey, provides acquisition in 12 view azimuth angles (0 °, 30 °, 60 °, 90 °, 120 °, 150 °, 180 °, 210 °, 240 °, 270 °, 300 °, 330 °) and for each of those 7 different view zenith angles are set on the camera gimbal (0 °, 10 °, 20 °, 30 °, 40 °, 50 °, 60 °), for a total of 84 measurements for each survey. After the preprocessing stage, we evaluate the variation of the spectral signature varying the view azimuth and zenith angle, producing the polar plot of the reflectance for each band. The dataset have been used to compare two semi-empirical models: the kernel-driven RossThick-LiSparse (Ross-Li) linear model [5], which has been used for the processing of MODIS land surface measurements [6], and the Rahman-Pinty-Verstraete (RPV) non-linear model [7]. The set of parameters have been derived for both the models and then validated respect the UAV measurements for each band, showing an overall R-squared of about 0.8.
The development and test activities here described have been carried out as part of the ESA-funded Quality Assurance for Earth Observation (IDEAS-QA4EO) framework contract (WP-2175, UAV for BRDF characterization) [8] and will be further implemented during the second phase. The UAV system is considered to be part of the SRIX4VEG initiative that will be held in Barrax (ES) during summer 2022 [9].
[1]: de Miguel Molina, B., & Oña, M. S. (2018). The drone sector in Europe. In Ethics and civil drones (pp. 7-33). Springer, Cham.
[2]: Bartell, F. O., Dereniak, E. L., & Wolfe, W. L. (1981, March). The theory and measurement of bidirectional reflectance distribution function (BRDF) and bidirectional transmittance distribution function (BTDF). In Radiation scattering in optical systems (Vol. 257, pp. 154-160). SPIE.
[3]: Müller, C., Hosgood, B., & Andreoli, G. (1998). Sensitivity analysis and quality assessment of laboratory BRDF data. Remote Sensing of Environment, 64(2), 176-191.
[4]: Nocerino, E., Dubbini, M., Menna, F., Remondino, F., Gattelli, M., & Covi, D. (2017). GEOMETRIC CALIBRATION AND RADIOMETRIC CORRECTION OF THE MAIA MULTISPECTRAL CAMERA. International Archives of the Photogrammetry, Remote Sensing & Spatial Information Sciences, 42.
[5]: Wanner, W., Li, X., & Strahler, A. H. (1995). On the derivation of kernels for kernel‐driven models of bidirectional reflectance. Journal of Geophysical Research: Atmospheres, 100(D10), 21077-21089.
[6]: Lucht, W., Schaaf, C. B., & Strahler, A. H. (2000). An algorithm for the retrieval of albedo from space using semiempirical BRDF models. IEEE Transactions on Geoscience and Remote sensing, 38(2), 977-998.
[7]: Rahman, H., Pinty, B., & Verstraete, M. M. (1993). Coupled surface‐atmosphere reflectance (CSAR) model: 2. Semiempirical surface model usable with NOAA advanced very high resolution radiometer data. Journal of Geophysical Research: Atmospheres, 98(D11), 20791-20801.
[8]: Quality Assurance Framework for Earth Observation (QA4EO) Web Portal. Available online: https://www.qa4eo.org/ (accessed on 20 Nov 2021).
[9] Surface Reflectance Intercomparison Exercise (SRIX4VEG) Web Portal. Available online: https://frm4veg.org/srix4veg/ (accessed on 20 Nov 2021).
During the last few years, unmanned aerial vehicle (UAV, or drone) technology has evolved tremendously. At the same time, miniaturization of hyperspectral sensors has made it possible to use these sensors on board small drones, providing completely new possibilities to carry out high quality remote sensing measurements with high spatial and spectral resolution. Yet a new drone application is performing calibration and validation (Cal/Val) measurements for optical satellite sensors, such as Sentinel-2. ESA has created the concept of Fiducial Reference Measurements (FRM), which have been defined to be a subset of in situ measurements used for satellite Cal/Val with robust and demonstrable traceability to SI or community standards following documented procedures with a clear uncertainty budget. In the context of the EMPIR (The European Metrology Programme for Innovation and Research) -funded MetEOC-4 project (Metrology for Earth Observation and Climate) lead by National Physical Laboratory NPL, UK, we will present our results on performing multiple hyperspectral drone campaigns for Sentinel-2 FRM measurements over one hectare forest site located in southern Finland.
The overall goal of the MetEOC-4 is to develop calibration and validation standards and methods, covering pre- and post- launch of observation systems as well as complimentary in situ networks, for land, ocean, and atmosphere. This will extend the capabilities of the SI beyond the laboratories, into the ‘field’ and build on the framework needed to underpin a global climate observing system. The challenge in several current satellite Cal/Val test sites is that the object is bright and non-vegetated (e.g. sand). We need extension of test sites to biophysical surfaces requiring challenging spectral/spatial corrections. Additionally, these vegetated sites need to be monitored under different illumination and phenological conditions over multiple time steps.
Our research group at the Finnish Geospatial Research Institute FGI has been utilizing drones in various remote sensing tasks since 2008. We have been developing quantitative and SI-traceable image processing chains for hyperspectral drone imagery. The outcomes of our processing chain are geometrically accurate 3D point clouds, radiometrically corrected homogeneous reflectance image mosaics and bi-directional reflectance distribution function (BRDF) information of the object.
In this study, we have performed seven hyperspectral drone campaigns between June and October 2021 with solar zenith angles between 37° and 70°. The test site is one hectare forest site located in Kirkkonummi, Southern Finland. The forest consists of mainly full-grown coniferous trees (height about 20 m), Norway spruce (Picea abies) and pine (Pinus sylvestris), with some deciduous trees (birch (Betula pendula)). We used a DJI Matrice 600 Pro drone with Specim AFX10 hyperspectral camera and Gremzy T7 gimbal. AFX10 is a pushbroom sensor with 400 to 1000 nm spectral range, 224 spectral bands, 5.5 nm spectral resolution, 1024 spatial pixels, integrated GPS/IMU, internal SSD storage and 2.1 kg weight. Each flight was performed to match Sentinel-2 overpass time within +-15 minutes, with drone flight time about 15 minutes. Flights were performed from the 90 meters flying height resulting in a 6 cm ground sampling distance (GSD). Illumination conditions were monitored before, during and right after the flights by making aerosol optical thickness (AOT) measurements with Microtops-II sunphotometer. Four reference reflectance panels of 1 m2 size were installed next to the drone take off area to be used as radiometric calibration reference panels for the AFX10 data. Additionally, we performed one drone liDAR campaign with DJI L1 lidar over the test forest to generate accurate 3D point cloud, digital terrain and surface models of the area. The flying height of the liDAR campaign was about 90 meters above ground and 70 meters above treetops, producing approximately 500 points / m2.
In this presentation, we will introduce our FRM drone campaigns, data processing methodologies and first results. We will discuss the impact of key factors such as solar geometry, atmospheric conditions, and phenology to the Cal/Cal process. We will analyse the impact of the uncorrelated errors in the comparison of ground to satellite observations and different ground measurement methods to derive an estimation of the uncertainty improvements with more flights. This work will extend the application of these techniques beyond the current single site, single time step scenario which will improve its scientific relevance and uptake by the international community.
Over the past decades, many radiative transfer models have been developed and are widely used for e.g. vicarious calibration and look-up table generation for atmospheric correction. Many of these models ship atmospheric property databases. Sub-components of these models have been extensively tested in ideal conditions but so far, no initiative similar to RAMI has been undertaken to systematically compare models when they are used to simulate actual satellite observation. The accuracy of these models has not been clearly assessed in realistic usage conditions. As well as for RAMI-1, the primary goal of RAMI4ATM will be to document the variability between coupled surface-atmosphere models when run under well-controlled, but realistic, conditions.
This new phase is oriented toward the support of calibration and validation activities relying on the use of radiative transfer models for the simulation of satellite observations in the visible, near and shortwave infrared spectral regions. It is therefore primarily directed at users of models used for calibration and validation activities. A series a standard cases have been foreseen including:
• The following Surface types will be included : Lambertian, RPV, Ross-Li, homogeneous discrete canopy;
• A standard atmospheric profile will be considered with the possibility to rescale the water vapour concentration;
• Fine and coarse aerosol types with different low and high optical thickness;
• Simulations will be performed in the blue, green, red, NIR and SWIR spectral regions corresponding to Sentinel-2A MSI bands B02, B03, B04, B08A, B11 and B12.
The expected outcome of this exercise is as follows:
• to allow users to cross compare their simulations with respect to the other participating models over a variety of ideal atmospheric scenarios including gas absorption, Rayleigh and Mie scattering;
• to inform the user community on the performance of the participating available models and the differences in their results;
• to help developers improve their models;
• to progressively build community consensus on the best ways to simulate the radiative transfer below and above the Earth’s atmosphere.
First comparison results will be presented at LPS22.
The correction of the atmospheric effects on optical satellite images is essential for quantitative remote sensing applications. Open and free data access to Copernicus Sentinel-2 (EC/ESA) and Landsat 8 (NASA/USGS) missions increased significantly the scientific interest on atmospheric correction (AC) and several approaches have been introduced by involving different radiative transfer models, single or multitemporal images, various algorithms to estimate aerosol properties and water vapour content, constant or diverse aerosol models, various sources of ancillary data, etc. These methodologies are usually validated independently by developers and/or users based on a certain number of sites with available reference data and/or are compared with results of other AC processors.
In order to investigate all the AC aspects and issues in an integrated way, a benchmark exercise (Atmosperic Correction Inter-compariosn eXercise, ACIX) was initiated in 2016 in the frame of CEOS Working Group on Calibration & Validation (WGCV) with the aim to compare the state-of-the-art AC processors. ACIX is a voluntary and open-access initiative to which every AC processor’s developer is invited to participate. ACIX-I was an initial attempt to study the variability of AC performances over diverse atmospheric and land cover conditions using Landsat 8 and Sentinel-2A input data. It was highly appreciated by the participants and considered as a useful tool to discover not only the assets and flaws of the approaches, but also ways to improve them. Thus, a second implementation of the experiment was requested to inter-compare the enhanced versions of the participating processors, but also to be expanded by including additional AC processors. In this second implementation, ACIX was split in two categories: Land and Aqua, with focus on the processors performing over land and water correspondingly. In this presentation, attention is given only to the Land part of the exercise.
The sites for the inter-comparison analysis over land were defined by investigating the full catalogue of AERONET sites for available measurements within 30min (±15min) from the satellites’ overpass. Eventually a total of 123 and 110 AERONET sites, which were distributed globally and representing various land cover types, made the site list for Sentinel-2A, -2B and Landsat 8 acquisitions correspondingly. Based on these available AERONET measurements, Aerosol Optical Depth (AOD) and Water Vapour (WV) retrievals were validated with the help of various statistical metrics. Regarding Surface Reflectance (SR) validation, as there is not yet any global network of systematic ground-based measurements, alternative approaches had to be adopted. Therefore, simulated SR reference dataset was computed over all the test sites by using the 6SV full radiative transfer code, with the required aerosol and water vapour information to have been acquired from AERONET. Moreover, measurements from the calibration dedicated network RadCalNet over La Crau (France) and Gobabeb (Namimbia) were involved in the SR validation. The observations in this case were processed to the same sun and sensor geometry, as well as spectrally integrated to the corresponding sensor spectral bands of Sentinel-2 and Landsat 8. The analysis results varied depending on the AC product compared, the reference dataset and the metrics. In this presentation an overview of the analysis and results will be given and discussed.
The validation of the Sea and Land Surface Temperature Radiometer (SLSTR) Land Surface Temperature (LST) operational product is one of the main objectives of the ESA Copernicus Space Component Validation for Land Surface Temperature, Aerosol Optical Depth and Water Vapour Sentinel-3 Products (LAW) project.
Five new ground LST stations were deployed in the LAW project to validate SLSTR LST over previously unrepresented biomes, which were selected after a land cover gap analysis study. The LST LAW stations are located in: KIT Forest – Germany, classified as closed broadleaved deciduous forest according to the ATSR LST Land Cover classification, which is used in the SLSTR LST product; Svartberget – Sweden, classified as open needle leaved deciduous or evergreen forest; Hyytiälä – Finland, classified as closed to open mixed broadleaved and needle leaved forest; Robson Creek – Australia, classified as closed to open broadleaved evergreen and/or semi-deciduous forest; and Puéchabon – France, classified as sparse vegetation. The KIT Forest station was deployed in August 2020 since it is in the premises of KIT Campus North, however the other stations were deployed in October 2021. Thus, all five stations are operative since November 2021. Due to the earlier deployment of the KIT Forest station, it provided a valuable dataset to test the tools created for monitoring and evaluating the LAW insitu LST data.
The main instruments of all five LST stations are two narrowband Thermal Infrared Heitronics KT15.85 IIP radiometers, which measure the ground brightness temperature and the sky brightness temperature at a zenith angle of 53°, respectively. These radiometers are used to estimate the in-situ LST via the radiative transfer equation. The Heitronics KT15.85 radiometer covers the spectral range from 9.6 to 11.5 µm and has an associated uncertainty of ±0.3 K. All radiometers used in the LAW stations were calibrated at laboratory against a primary reference blackbody at the Physikalisch-Technische Bundesanstalt (PTB – Germany) and against the KIT LandCal P80P blackbody certified by PTB. Additionally, the stations are equipped with a HydroVUE10 air temperature and relative humidity sensor, which provide useful information for monitoring the radiometric data.
The in-situ LST uncertainty for each station was estimated via propagation of errors in the radiative transfer equation. Additionally, the uncertainty introduced by each uncertainty source was analysed. For the larger KIT Forest dataset, it was obtained a mean in-situ LST uncertainty of 0.32 ±0.03 K, with values up to ±0.4 K. The main contribution to the total in-situ LST uncertainty was introduced by the ground KT15.85 radiometer, with a mean uncertainty of ±0.3 K, although the emissivity also showed a significant contribution, with uncertainties up to ±0.29 K. The lowest uncertainty contribution was introduced by the sky radiometer measurements, with a mean value of ±0.035 K. Similar results were obtained for the available period of the other four LAW stations, which showed a mean (maximum) in-situ LST uncertainty of ±0.33 K (±0.44 K), ±0.38 K (±0.46 K), ±0.32 K (±0.43 K) and ±0.35 K (±0.44 K) at Hyytiälä, Puéchabon, Svartberget and Robson Creek stations, respectively. Thus, the mean in-situ LST uncertainty for all new LAW stations is lower than ±0.4 K. The in-situ LST uncertainty for each LAW station will be reprocessed for a larger dataset before the ESA Living Planet Symposium 2022.