함수형 공간 자료를 위한 추정 및 예측모형과 응용

Abstract

Particulate matter consisting of the chemical compounds harmful to human body and floating in the air as the invisible dust has been affecting on the air pollution or the various disease by inhaling into the human body. According to the Korea Meteorological Administration’s open portal, the particulate matter is measured at 28 sites and is widely used in forecasting and warning systems related to the particulate matter. In this paper, we will introduce an approach that can estimate the fixed effect and predict the future value by incorporating the spatial approach with the functional approach, assuming that the nature of the measurement for the particulate matter is a functional nature. While the existing studies have focused on the longitudinal data approaches, these approaches did not consider the characteristics of the spatial data that can have dependency between the sites and of the irregularity that the sites are not uniformly distributed across the spatial domain. To address this issue, we consider extending the existing approach to the functional data approach and applying the spatial approach such as kriging. Specifically, we consider the weighted mean being insensitive to extreme observations to account for the fixed effect from the real data obtained from 28 sites for 2021 year. To do this, we estimate the covariance process and compute the optimal weight through the empirical variogram analysis. Also, we apply a kriging method that can predict the value associated with the particulate matter at any given observational site.