Cumulative Route Solar Radiation for Pedestrian Access to Urban Green Space: A Physics-Based Parametric Study with Surrogate Modeling

Abstract

Pedestrian accessibility to urban green space depends not only on walking distance but also on the environmental conditions experienced along the route, where solar exposure can raise the perceived cost of walking and influence route choice.Yet route-level solar exposure is rarely expressed in a form that designers can manipulate during early-stage planning, when street geometry and tree canopy are still being decided. This thesis develops Cumulative Route Solar Radiation (CRSR), a design-responsive accessibility metric defined as the spatially averaged short-wave solar radiation received at pedestrian height along a walking route. CRSR is computed through a physics-based Radiance simulation workflow in Ladybug Tools for a representative street segment in Rainier Valley, south Seattle. A 3×2×2 full-factorial experiment varies street aspect ratio, tree spacing, and canopy transmittance, and an artificial neural network (ANN) surrogate trained on a 5,000-sample Latin Hypercube dataset extends estimation across the parameter space. Across the twelve scenarios, reductions in route-averaged CRSR relative to the Baseline range from 3.8% to 37.3%. Street aspect ratio produces the strongest effect, followed by tree spacing and canopy transmittance, with combined interventions yielding a less-than-additive response. The surrogate reproduces simulated CRSR with a holdout coefficient of determination of 0.917 and a mean absolute percentage error of 6.60%, and benchmarking against linear, polynomial, and tree-based models confirms the non-linear structure of the CRSR response and situates the ANN surrogate within a broader set of candidate approximation methods. CRSR thus offers a reproducible, route-integrated radiation metric that complements existing walking-impedance and perceived-distance frameworks for climate-responsive accessibility analysis. Because the surrogate is trained on a single reference geometry and validated against simulation rather than field measurement, multi-segment surrogate results are exploratory and require further calibration.