Modelling understory light availability in a heterogeneous landscape using drone-derived structural parameters and a 3D radiative transfer model

The light environment within vegetated landscapes is a key driver of microclimate, creating varied habitats over small spatial extents and controls the distribution of understory plant species. Modelling spatial variations of light at these scales requires finely resolved (< 1 m) information on topography and canopy properties. We demonstrate an approach to modelling spatial distributions and temporal progression of understory photosynthetically active radiation (PAR) utilising a three dimensional radiative transfer model (discrete anisotropic radiative transfer model: DART) where the scene is parameterised by drone-based data.

The study site, located in west Cornwall, UK, includes a small mixed woodland as well as isolated free-standing trees. Data were acquired from March to August 2019. Vegetation height and distribution were derived from point clouds generated from drone image data using structure-from-motion (SfM) photogrammetry. These data were supplemented by multi-temporal multispectral imagery (Parrot Sequoia camera) which were used to generate an empirical model by relating a vegetation index to plant area index derived from hemispherical photography taken over the same time period. Simulations of the 3D radiative budget were performed for the PAR wavelength interval (400 – 700 nm) using DART.

Besides maps of instantaneous above and below canopy irradiance, we provide models of daily light integrals (DLI) which are assessed against field validation measurements with PAR quantum sensors. We find relatively good agreement for simulated PAR in the woodland. The impact of simplifying assumptions regarding leaf angular distributions and optical properties are discussed. Finally, further opportunities which fine-grained drone data can provide in a radiative transfer context are highlighted.