Characterization of the spatial variability of bidirectional reflectance distribution function parameters over a heterogeneous rice paddy landscape

Abstract:

Surface reflectance can vary according to the solar and viewing geometry, and land surface properties. This variation can be characterized by the bidirectional reflectance distribution function (BRDF). To remove the angular effects, satellite reflectance images can be adjusted to nadir view with fixed solar zenith geometry using estimated BRDF parameters. However, a widely used BRDF parameter product (i.e. MCD43A1) is only available at relatively coarse spatial resolution. Therefore, the spatial resolution of MCD43A1 might be insufficient for adjusting higher resolution satellite reflectance products (e.g. Landsat 8 and Sentinel 2), especially over heterogeneous landscapes. As the spatial variability of BRDF parameters within a MCD43A1 pixel scale was not characterized in higher spatial resolution pixel scale, BRDF effects in fine resolution images were not fully evaluated over a heterogeneous landscape. From in situ the measurements and inverse mode of BRDF model, we estimated BRDF parameters in satellite pixel scales from 3 m to 500 m, which corresponds to the spatial resolutions of different satellite surface reflectance products (e.g. PlanetScope, Landsat 8, Sentinel 2, and MCD43A4). For this, we conducted multi-angular reflectance measurements using a hyperspectral imager on an unmanned aerial vehicle (UAV). In particular, we covered view zenith angles (VZA) of 0, ±15, ±30, ±45, and ±60 degrees for the same locations. We considered the field of view of the hyperspectral sensor and UAV flight heights at each VZA to cover one MCD43A1 pixel. The results reveal the uncertainty of BRDF parameters with spatial variation that might be averaged out at large scales. The proposed approach could be extended to a variety of land cover types in future studies to evaluate the variation of BRDF parameters with scales.