New approaches in multi-angular proximal sensing of vegetation: Accounting for spatial heterogeneity and diffuse radiation in directional reflectance distribution models

•AMSPEC-MED automated multi-angular hyperspectral system set in tree-grass ecosystem•Ecosystem 3-D and FOV ray casting models simulate heterogeneous sensor's observations.•Diffuse radiation included in Reflectance Distribution Functions (RDF) improves model.•The model achieves the unmixing of tree and grass hemispherical-directional RDFs.•Comparison to MODIS BRDF product and hand held spectral data cast r2 ∈ [0.74, 0.88].

The development of tower-mounted automated multi-angular hyperspectral systems has brought new opportunities and challenges for the characterization of the Bidirectional Reflectance Distribution Function (BRDF) on a continuous basis. This study describes the deployment of one of these systems in a Mediterranean savanna ecosystem (AMSPEC-MED), and proposes new approaches for modeling of directional effects. In this study, a Hemispherical-Directional Reflectance Distribution Function (HDRDF) was introduced in order to quantify the effect of diffuse radiation on the estimation of BRDF. The HDRDFs of the two covers of the ecosystem - trees and grasses - were un-mixed using a 3-Dimensional (3-D) model of the observed scene. Up-scaling HDRDF estimates to MODIS BRDF product (r2 ∈ [0.74, 0.86]) and down-scaling to hand held spectral measurements (r2 = 0.88) achieved a reasonable accuracy (RMSE ∈ [1.81, 3.14]). Despite the uncertainties in the estimation of diffuse irradiance and the 3-D representation of the scene, HDRDF un-mixing demonstrates the potential of automated multi-angular proximal sensing to study vegetation properties in heterogeneous ecosystems and the correction of directional effects of different sources.