Engineering-seismological planning of observation points for Vs30 mapping via Poisson-disk sampling under access constraints

We present an engineering-seismological workflow for planning observation points in Vs30 mapping under access and safety constraints. The design is based on Poisson-disk sampling (PDS), which enforces a minimum inter-point distance to avoid clustering and introduces a practical service radius rrr that allows safe local relocation around restricted zones. On a controlled 100×100 synthetic testbed with four litho-zones, we compare random, regular grid, and PDS layouts at sparse designs (N = 25, 15, 7). Interpolation uses ordinary kriging and machine-learning models (SVR, XGBoost, k-NN) trained with Residual (RF) Bootstrap augmentation. PDS delivers more uniform coverage, fewer interpolation artifacts, and consistently better accuracy/robustness (MAE/RMSE/correlation/R²), with the strongest gains at N ≤ 15. The results support PDS as a default survey-planning tool for engineering microzonation and site-class assignment, improving field logistics while preserving uniform coverage around constrained infrastructure.

Перевод пока недоступен