The SPART model: A soil-plant-atmosphere radiative transfer model for satellite measurements in the solar spectrum

Radiative transfer models (RTMs) of vegetation canopies can be applied for the retrieval of numerical values of vegetation properties from satellite data. For such retrieval, it is necessary first to apply atmospheric correction to translate the top-of-atmosphere (TOA) satellite data into top-of-canopy (TOC) values. This atmospheric correction typically assumes a Lambertian surface reflection, which introduces errors if the real surface is non-Lambertian. Furthermore, atmospheric correction requires atmospheric characterization as input, which is not always available. In this study, we present an RTM for soil-plant-atmosphere systems to model TOC and TOA reflectance as observed by sensors, and to retrieve vegetation properties directly from TOA reflectance skipping the atmosphere correction processes with the inversion mode of the RTM. The model uses three computationally efficient RTMs for soil (BSM), vegetation canopies (PROSAIL) and atmosphere (SMAC), respectively. The sub-models are coupled by using the four-stream theory and the adding method. The resulting 'Soil-Plant-Atmosphere Radiative Transfer model' (SPART) simulates directional TOA spectral observations, with all major effects included, such as sun-observer geometries and non-Lambertian reflectance of the land surface. A sensitivity anaylsis of the model shows that neglecting anisotropic reflection of the surface in coupling the surface with atmosphere causes considerable errors in TOA reflectance. The model was validated by comparing TOC and TOA reflectance simulations with those simulated with the atmosphere-included version of the DART RTM model. We show that the differences between DART and SPART are less than 7% for simulating TOC reflectance, and are less than 20% (less than 10% at most bands) for simulating TOA reflectance. The model performance in retrieving key vegetation and atmospheric properties was evaluted by using a synthetic dataset and a satellite dataset. The inversion mode allows estimating vegetation properties along with atmospheric properties and TOC reflectance with reasonable accuracy directly from TOA observations, and remarkable accuracy can be achieved if prior information is used in the model inversion. The model can be used to investigate the sensitivity of surface and atmospheric properties on TOC and TOA reflectance and for the simulation of synthetic data of existing and forthcoming satellite missions. More importantly, it facilitates a quantitative use of remote sensing data from satellites directly without the need for atmospheric correction.