Article Text
Abstract
Background/aim Epidemiological studies show that changes in air temperature affect mortality and morbidity. So far these changes have been observed mainly on a temporal scale while spatial changes have rarely been considered. We used land-use regression (LUR) models to predict the spatial variability of temperature and relative humidity and to detect urban heat islands in the Augsburg region, Germany.
Methods We measured temperature and relative humidity at more than 80 sites between December 2012 and October 2017 in the city centre of Augsburg and the two surrounding counties. Seasonal averages were calculated for monitors with at least 14 days of measurements. Discontinuous measurements were adjusted using data from a reference station. We compiled a large set of predictors, which will be offered as potential explanatory variables to separately model the spatial variation of temperature and relative humidity. As geographic predictors we will offer traffic and land use variables, altitude, population, building density and sky view factor. As remote sensing predictors we will offer albedo, normalised difference vegetation index and. K-fold cross-validation will be used to validate our models.
Results For each season, we could include five rounds of measurements ranging from 29 to 73 available monitors. The seasonal averages of the monitors ranged between 13°C and 23.2°C for the summers and between −2.4°C and 7.1°C for the winters. For the springs they ranged between 5.2°C and 14°C, while between 2°C and 19.3°C during autumns. The final LUR models are intended to reflect the spatial distribution of temperature and relative humidity in the study area for the different seasons and will be presented at the conference. Air temperature values are expected to increase within the city centre where building density and traffic rates are higher. Instead, values of relative humidity are supposed to decrease with scarce presence of water bodies and vegetation.
Conclusion We are developing spatial models to predict seasonal mean temperature and relative humidity for a typical city of Southern Germany. On the basis of these models, we aim to investigate potential health effects in subsequent epidemiological analyses as we will apply the final LUR models to the residential addresses of our KORA (Cooperative Health Research in the Augsburg Region) participants.