Radiative transfer modelling is an essential part of the Earth Observation measurement data processing and control chain. Radiative transfer simulation tools are notably used for Level-1 data calibration and Level-2 geophysical product validation. As mentioned in the joint presentation “Comparison of three 1D radiative transfer models over Libya-4 against well-calibrated radiometers”, verifying calibration with an accuracy below 3% can currently not be done reliably with common 1D radiative transfer models. Achieving high accuracy levels requires a 3D radiative transfer model to relax the “smooth, flat, homogeneous Earth assumption” implicitly made in 1D models and account for the structural effects of complex geometry (e.g. cloud, terrain or vegetation structure).

The radiative transfer simulation effort for Earth Observation is scattered among subcommunities. Although lands, atmosphere, ocean colour and cryosphere subcommunities share most of their radiative transfer physics, no 3D radiative transfer model currently integrates all the physics in a way such that they could be easily combined to account for subtle coupling effects with high accuracy. An assessment of radiative transfer model user requirements, done in the framework of the Metrology for Earth Observation and Climate-3 (MetEOC-3) project, highlighted similarities and differences between Earth Observation subcommunities [1]. Contributions from the planetology community also showed that interest in a common simulation platform goes beyond Earth Observation.

Taking successful cross-community software development efforts as an example, the Eradiate radiative transfer model will leverage modern software design approaches to provide a common radiative transfer simulation platform for Earth Observation applications based on the aforementioned requirements. A high-level API, written in Python, will provide end-users with simple tools to quickly setup and solve radiative transfer problems using community-specific concepts and scene setups. A low-level community-agnostic API, written in C++ and Python, will allow developers to build specific radiative transfer simulation tools or extend the high-level API. The core library will be written in modern C++ and use a modular architecture inspired by computer graphics concepts to isolate software parts, thus improving maintainability and testability. This multi-level design aims at smoothing the learning curve by allowing end-users to keep using the concepts they are familiar with, while providing common concepts for developers to couple physics from currently compartmentalised subcommunities. It also provides great flexibility, by making it simple to add, remove, modify, activate or disable model components.

In addition, Eradiate [2] aims at providing highly accurate and efficient radiative transfer equation integration methods and radiative property computation algorithms. Great care will be taken in the selection and implementation of algorithms and numerical methods. Interfaces to standard radiative property databases will be provided, and software design will make it possible to write new interfaces. Extensive documentation and comprehensive training material for each API level will ease learning and setting up the simulation platform. Unit and regression testing will be run using continuous integration to ensure software quality. Finally, the software will be open and free for any contributor to modify. This should pave the way to the establishment of a developer and user community, benefiting from scientific advances made in the entire Earth Observation community.

In this presentation, we will review the current progress and plans for the development of the Eradiate radiative transfer model.

References:
[1] Leroy, V., and Govaerts, Y. M. (2018). Radiative Transfer Model User Requirements Analysis in Support to Calibration/Validation Activities (issue 1). Rayference. URL: http://www.eradiate.eu/docs/eradiate-user_requirements_analysis-1-20180705.pdf
[2] Eradiate Website: http://www.eradiate.eu/

Leaf Area Index (LAI) is a parameter that describes vegetation flux and structure, and can influence the biophysical and biogeochemical functioning processes that underpin terrestrial ecosystems. LAI derived from Earth Observation (EO) can provide data at the spatial and temporal scales required for climate observations, change detection and modelling of vegetation functioning processes. The quality of EO products are paramount for their use in these applications, where detailed traceable and reliable uncertainty information is increasingly becoming required alongside data provision.

This contribution aims to evaluate the quality of LAI products using a model-based approach capable of benchmarking an LAI processing chain, as well as determining conformity to accuracy requirements and contributions to uncertainty required to determine uncertainty traceability. Widely implemented LAI validation protocols are simulated using a 3D Monte Carlo Ray Tracing Radiative Transfer Model (RTM) in a variety of realistic canopies and under a set of scenarios that tests the major contributions towards an LAI uncertainty budget. In particular, Digital Hemispherical Photography (DHP) is focused on in since it is one of the most widely used in-situ methods available.

Four main contributions towards uncertainty are tested, which are required if full uncertainty traceability is to be achieved: 1) the algorithms used to compute LAI (LAI from 57.5°, and Miller’s algorithm), as well as algorithms used to correct Plant Area Index (PAI) to LAI and clumping, 2) the definition of LAI, 3) the field measurement protocol with respect to sampling design and camera placement, 4) sites-specific factors such as canopy heterogeneity, LAI and the ecosystem type. Uncertainty budgets are developed with respect to uncertainty contributions, and conformity tested against quality requirements given a specific conformity testing method. Given the current lack of detailed and traceable uncertainty information, as well as standard protocols for assessing quality, the 3D modelling approach based on radiative transfer modelling is demonstrated to be a useful tool to EO and validation communities to better quantify and understand the quality of their data.

Understanding the Global Climate Change and the way it will affect the Earth and preserving biodiversity are two major challenges of the 21st century.
Forests cover 30% of continental surfaces and are a major contributor to the carbon cycle. Sequestration of carbon in forest biomass appears as an important mechanism – in conjunction with reduced emissions - for mitigating climate change. Adequate management of forest resources could strengthen the action of forests as a carbon sink. Forests also play a key role for biodiversity conservation and energy supply. Sustainable management of forest ecosystems is thus critical for the future of mankind.
However, the lack of consistent information on forest structure and biomass and on their dynamics at global scale hampers the development of ecological models and of earth-vegetation-atmosphere interaction global models that are essential to improve knowledge on forest ecosystem functioning and to address the challenge of forest sustainable management.
By its ability to measure the vertical structure of land cover, airborne Lidar technology is highly qualified for forest applications. NASA’s ICESat1 mission (2003-2009) paved the way for Earth observation from space with Lidar systems. ICESat2, launched in September 2018, although primarily designed for ice monitoring will also provide vegetation height measurements worldwide. Vegetation profiles with high vertical resolution should also become soon available with GEDI (NASA) and MOLI (JAXA) missions that will be installed on the ISS at the end of 2018 and in 2019 or 2020, respectively. Although data coverage will be limited from 52°S to 52°N, both these missions will contribute to ecosystem structure observation from space in conjunction with radar missions, especially the BIOMASS ESA’s mission. Combined with Sentinel 2 imagery, which provides spectral information with high spatial and temporal resolutions, they will also enable to evaluate the complementarity between information on 3D structure, composition and phenology for a comprehensive monitoring of forest ecosystem conditions.

LEAF (Lidar For Earth and Forests) mission has been under study at CNES since 2008 and could contribute in the future to the international effort to sustain global carbon and biodiversity monitoring systems.
The system under study combines a NIR full-waveform Lidar profiler and a very high resolution multispectral imager. CNES has supported scientific projects in order to refine instrument specifications and design a system that will meet user requirements. Due to the high cost associated with the development of an airborne prototype, priority has been given to simulation approaches.

The aim of this contribution is to present LEAF mission concept and to focus on the recent “LEAF ExpeVal” experiment, designed to consolidate and validate the simulation approaches developed to model a realistic Lidar signal on various forest types, including complex tropical forests.
These approaches rely on DART, a radiative transfer model developed at CESBIO. A Lidar module was introduced into DART in the early 2010s and has undergone constant improvement since then. However, simulating a realistic signal also requires to work on model inputs and to develop methods to build detailed forest scenes.
The objective of the LEAF-Expeval experiment was twofold: first, to validate the simulations by comparing simulated and experimental Lidar waveforms on a tropical forest; second, to develop a new approach to simulate spaceborne Lidar data from small footprint airborne Lidar data (ALS data).
Achieving the first objective is a key step to refine LEAF system specifications in order to ensure that the system will provide accurate height measurements and vegetation profiles on tropical forests, which are the most constraining environment for a space Lidar. Field and airborne data were collected in Paracou experimental forest in French Guiana. The processing and analysis steps include (1) the construction of forest scenes from TLS (Terrestrial Lidar Scanner) and reflectance measurements on a dense forest plot, (2) the estimation of the calibration coefficient of the airborne system, coefficient that is used to convert digital count into physical units, (3) the simulation of airborne Lidar data on the forest scene using DART and (4) the comparison of simulated and real waveforms.
As ALS data are now widely available, achieving the second objective would enable the simulation and evaluation of LEAF data on a variety of forest ecosystems. This part of the study relies on Lidar simulations realized on temperate broadleaved forest scenes created using Allostand (Amap) and Genesis software. Small footprint lidar point clouds were simulated using DART and were in turn used to simulate large footprint lidar waveforms. The resulting waveforms were compared to reference data, i.e. large footprint waveforms obtained directly with DART on the same forest scenes. First results are encouraging and showed that large footprint waveforms obtained from the processing of small footprint data were very similar to reference waveforms, even with a change of waveband between lidar point clouds and large footprint waveform.

Clouds are of main importance for climate and meteorological questions. For simplicity and computational reasons, clouds are often considered as homogeneous plane parallel layers, which usually is far from the reality. In the solar or in the infrared range, the effects of finite dimensions and cloud heterogeneities need to be assessed because they can have large impacts on cloud optical property retrieval.
3DMCPOL is a forward Monte-Carlo model able to compute radiative transfer in a three-dimensional atmosphere, accounting for polarization using Stokes formalism. Initially developed for solar radiation (Cornet et al., 2010), it has been then extended to thermal infrared radiation (Fauchez et al., 2014). The radiative transfer is computed using the local estimate method to optimize the time to compute reflectances/radiances. The surface can be either Lambertian or heterogeneous with bidirectional function for ocean or snow. The model participated and was improved during the International Polarized Radiative Transfer (IPRT) on 3D cloud cases (Emde et al., 2018).
Several cloud remote sensing applications using 3DCMPOL will be presented. We will show how the use of simulated total and polarized reflectances in the framework of POLDER3/PARASOL or of the multi-angular airborne radiometer Observing System Including PolaRisation in the Solar Infrared Spectrum (OSIRIS) allow to assess errors on optical thickness retrieval, cloud microphysical parameters and aerosol above cloud retrievals. We will also present studies led in the context of MODIS/AQUA in the solar and infrared range to estimate biases according to the sensor scale. Finally, we will present recent developments concerning an integrated simulator using high resolution cloud physical modelling and 3DMPOL to simulate measurements of future satellite missions as for example for the Multi-Viewing Multi-Channel Multi-Polarization Imaging (3MI) mission on-board the EUMETSAT Polar System - Second Generation.

This work presents results from the development and application of accurate radiative transfer modeling (RTM) for the assessment and minimization of uncertainties in typical optical observations of coastal waters. Notably, these waters are characterized by rather complex bio-optical properties due to concurrent radiance contributions from non-covarying in-water optically active components, sea-bottom and nearby land. As such, the remote sensing of coastal waters still represents an open challenge to the creation of fit-for-purpose Climate Data Records.
The different radiance contributions to a satellite sensor observing coastal waters are simulated making use of the RTMs included in the ARTEMIS-OC software package, namely the numerical code based on the Finite Element method for the simulation of ocean color satellite data (FEMRAD-OC) and the Novel Adjacency Perturbation Simulator for Coastal Areas (NAUSICAA) full three-dimensional (3D) backward MC code.
The plane-parallel FEMRAD-OC code is used to simulate radiance contributions at the sensor, which can be reasonably assumed as horizontally invariant. The 3D NAUSICAA code is instead used to simulate the adjacency radiance Ladj, defined as the difference in the radiance at the sensor between the case accounting for the non-uniformity of the underlying reflecting surface and the case assuming a uniform surface.
All simulations are performed for a stratified atmosphere (i.e., by modeling the vertical distribution of gas molecules and aerosols) bounded by a Fresnel reflecting surface, fully accounting for multiple scattering, sea surface roughness, slanted illumination and observation conditions. NAUSICAA simulations additionally include a non-uniform reflecting surface, with detailed description of the coastal morphology.
In order to reduce computing time for MC simulations, the adjacency radiance is modeled as:

Ladj={ρl⋅ κl −Rrs⋅ κw}⋅ C−W (1)

where the albedo of the land ρl and the remote sensing reflectance Rrs are both assumed isotropic and spatially homogeneous; the functions C and W designate the cumulative radiance contribution at the sensor originating from the land assumed as an ideal Lambertian reflector and as a Fresnel wind-roughened sea surface, respectively; and parameters κl and κw are scaling factors to adjust the radiance reflected by an ideal Lambertian reflector to that reflected by an actual land surface and to that leaving the water volume, respectively. The term W is often called the Fresnel mask.
The proposed modeling allows decoupling land and water optical properties from the atmospheric scattering, while accurately describing the anisotropy of the sea surface. Once the functions C and W are computed for given geometric and atmospheric inputs, Eq. (1) allows a straightforward evaluation of the AE for a wide variety of land and water spectral signatures. The latter can be extrapolated from satellite-derived or in situ measured data.
Statistical uncertainties on MC computations were set to not exceed the average radiometric resolution of OC sensors. MC systematic uncertainties were evaluated to be always lower than the statistical uncertainty of NAUSICAA simulations with a 99.7% level of confidence.
The methodology is currently applied to analyze the impact of AE in OC calibration/validation activities: key results are illustrated.

Radiative transfer models play a critical role for Earth Observation level-1 data calibration activities and level-2 geophysical product validation. The accuracy of these models is driven by both (inter)calibration requirements and generation of highly realistic synthetic data sets for retrieval algorithm benchmarking and evaluation. In the framework of the Metrology for Earth Observation and Climate-3 (MetEOC-3) project, the metrology community has assessed current user needs in terms of radiative transfer modelling for the support of these CalVal activities. Needs are notably driven by the unprecedented radiometric accuracy of Sentinel data, whose accuracy can be around 1% (e.g. 1% to 3% for the S2/MSI instrument). We assess here whether it is possible to reach such an accuracy level with current 1D radiative transfer models. Specific model features are reviewed in detail, such as molecular absorption due to the presence of water vapour (whose abundance in the lower part of atmosphere might make it interfere with aerosol scattering).
Three different surface-atmosphere coupled 1D Radiative Transfer Models are compared over Libya-4, a very spatially uniform and temporally stable CEOS Pseudo Invariant Calibration Site made of oriented sand dunes. These models are 6SV, LibRadTran and RTMOM. They rely on different numerical methods to solve the radiative transfer equation. The successive order of scattering method is used by the 6SV model. Within LibRadTran, different radiative transfer equation solvers can be selected. Monte Carlo method has been used for this study. Finally, RTMOM relies on the Matrix Operator Method. These models also use different approaches to account for the coupling between the surface and the atmosphere and to compute the gaseous transmittance. In particular, the estimation of this latter quantity is based on different version of the HITRAN database according to the model. The comparison therefore focuses on the ability of these models to (i) reliably simulate molecular absorption in the presence of aerosols, (ii) account for surface reflectance anisotropy and radiative coupling with the atmosphere and (iii) simulate satellite observations for any viewing and illumination conditions. The three models are operated in such a way that they use exactly the same surface reflectance model and aerosol single-scattering properties.
The comparison has been first performed against a reference Line-By-Line Radiative Transfer Model. The purpose of this comparison is to highlight how molecular absorption is handled by the different models, assuming a standard mid-latitude summer atmosphere. The models have next been compared against observation timeseries acquired over Libya-4 by a series of well-calibrated instruments such as MODIS, MERIS, MSI, OLI, … Results show that in average, the three models agree within a ±2% range with respect to observations but differences sometimes exceed 2% between the models to a great extent for large illumination zenith angles or in spectral regions subject to important molecular absorption, i.e. when the total column gaseous transmittance is typically smaller than 0.95. These models are therefore not suited to verify calibration accuracy better than 3% over Libya-4. A new generation of Radiative Transfer Model capable of meeting this type of requirements is thus needed. The specific requirements of this new generation of models are addressed in a joint presentation of this session, entitled “Eradiate: A New-Generation 3D Radiative Transfer Model For Earth Observation Applications”.