Weather indices for heat-health watch and warning systems

Abstract

BACKGROUND AND AIM: Extreme heat events have important impacts on human life and health, and have been shown to increase daily mortality and hospitalization. The heat waves of recent years led many countries to establish their heat health watch and warning systems (HHWWS) to prevent the effects of heat. The existing HHWWS usually contain weather indices and their corresponding thresholds. However, the weather indices, which are defined as linear combinations of lagged weather variables, are often unrelated to the variable of interest (over-mortality or over-hospitalization in case of HHWWS). Thus, in order to reduce the subjectivity, it is of interest to construct weather indices which are linked to the variable of interest. METHODS: Indeed, according to the definition of the weather indices, deriving indices can be seen as a dimension reduction problem. Considering the relationship with variable of interest, the most common example is the supervised dimension reduction methods, in our case, it is more about supervised principal component analysis (SPCA) based methods. RESULTS:Therefore, the main objective of this study is to determine weather indices by using SPCA based methods. The comparison results show that the obtained weather indices have better performance than the classic indices in warning system. CONCLUSIONS:These weather indices linked with health variables are then expected to be useful to improve current indices. KEYWORDS: weather indices, HHWWS, heat wave, mortality