AN INNOVATIVE SAMPLING METHOD FOR PREDICTINGBUILDING DAMAGE AFTER EARTHQUAKES

Building damage in seismic regions exhibits spatial correlation, where neighboring buildingsamples share similar information. Treating them as independent observations can lead to insufficientdata. To address this, we propose a two-stage sampling approach that selects well-dispersed, smallsizedsamples with substantial information. We use the Halton iterative partitioning method to choosesample locations, and individual buildings are randomly selected for each location. These selectedsample buildings are then used in a spatial regression model to predict damage ratios. Our approachoutperforms other methods in terms of spatial balance and prediction error, as demonstrated using bothsimulated and real datasets from the ۲۰۱۷ Sarpol-e Zahab earthquake in Iran.

کلیدواژه ها:

Spatial Sampling ، Spatial Regression Model ، Spatial Correlation ، Seismic Damage