Tracing rays from leaves to sky

Abstract

Light-simulation tools—exemplified by Radiance—are widely used for quantitative daylight studies and are increasingly adopted in agrivoltaics (agri-PV) to handle complex geometry via ray tracing. Yet, beyond typical workflows three practical limitations persist: spectrally resolved skies are restricted to the visible band; soft-shadow (penumbra) rendering relies on runtime-intensive solar-disk sampling; and fast, integrated canopy models remain scarce. We present a Radiance-compatible Python framework that adds: (i) atmosphere-specific sun–sky generation across the solar spectrum; (ii) efficient, equal-area sampling of the solar disk; and (iii) a simple canopy reconstruction tailored to narrow-trained orchards. To improve spectral fidelity, resolution, and range, we couple SMARTS-derived spectra to a Perez-based sky, leveraging Radiance's multispectral rendering. We deterministically sample the sun's finite extent using a Fibonacci lattice, yielding stable penumbra without prohibitive runtimes. The canopy model parameterizes porosity and seasonal development at a daily rate. Canopy representation matters: opaque–static models, common in agri-PV simulations, systematically underestimate light levels and miss spatiotemporal patterns needed to diagnose suboptimal conditions. Comparatively, a porous–dynamic model led to ≈26% higher seasonal light levels, with gains attaining ≈100% early in the season and converging to ≈16% after foliage matured. While penumbra is limited under conventional PV modules, penumbra-capable renderings enable exploration of design pathways—narrower cell layouts (half-cell and beyond) with greater module–canopy separation—that smooth lighting extremes.