Quantitative Methods for Stochastic High Frequency Spatio-Temporal and Non-Linear Analysis: Assessing Health Effects of Exposure to Extreme Ambient Temperature.
Liss, Alexander.
2015
-
Abstract: Extreme weather events, such as heat waves and cold spells, cause
substantial excess mortality and morbidity in the vulnerable elderly population, and cost
billions of dollars. The accurate and reliable assessment of adverse effects of extreme
weather events on human health is crucial for environmental scientists, economists, and
public health officials to ensure proper protection of... read morevulnerable populations and efficient
allocation of scarce resources. However, the methodology for the analysis of large national
databases is yet to be developed. The overarching objective of this dissertation is to
examine the effect of extreme weather on the elderly population of the Conterminous US
(ConUS) with respect to seasonality in temperature in different climatic regions by
utilizing heterogeneous high frequency and spatio-temporal resolution data. To achieve
these goals the author: 1) incorporated dissimilar stochastic high frequency big data
streams and distinct data types into the integrated data base for use in analytical and
decision support frameworks; 2) created an automated climate regionalization system based
on remote sensing and machine learning to define climate regions for the Conterminous US;
3) systematically surveyed the current state of the art and identified existing gaps in the
scientific knowledge; 4) assessed the dose-response relationship of exposure to temperature
extremes on human health in relatively homogeneous climate regions using different
statistical models, such as parametric and non-parametric, contemporaneous and
asynchronous, applied to the same data; 5) assessed seasonal peak timing and
synchronization delay of the exposure and the disease within the framework of
contemporaneous high frequency harmonic time series analysis and modification of the effect
by the regional climate; 6) modeled using hyperbolic functional form non-linear properties
of the effect of exposure to extreme temperature on human health. The proposed climate
regionalization method algorithmically forms eight climatically homogeneous regions for
Conterminous US from satellite Remote Sensing inputs. The relative risk of hospitalizations
due to extreme ambient temperature varied across climatic regions. Difference in regional
hospitalization rates suggests presence of an adaptation effect to a prevailing climate. In
various climatic regions the hospitalizations peaked earlier than the peak of exposure.
This suggests disproportionally high impact of extreme weather events, such as cold spells
or heat waves when they occur early in the season. These findings provide an insight into
the use of high frequency disjoint data sets for the assessment of the magnitude, timing,
synchronization and non-linear properties of adverse health consequences due to exposure to
extreme weather events to the elderly in defined climatic regions. These findings assist in
the creation of decision support frameworks targeting preventions and adaptation strategies
such as improving infrastructure, providing energy assistance, education and early warning
notifications for the vulnerable population. This dissertation offers a number of
methodological innovations for the assessment of the high frequency spatio-temporal and
non-linear impacts of extreme weather events on human health. These innovations help to
ensure an improved protection of the elderly population, aid policy makers in the
development of efficient disaster prevention strategies, and facilitate more efficient
allocation of scarce resources.
Thesis (Ph.D.)--Tufts University, 2015.
Submitted to the Dept. of Civil Engineering.
Advisor: Elena Naumova.
Committee: Magaly Koch, Gilbert Metcalf, and Kurt Pennell.
Keywords: Environmental health, Statistics, and Remote sensing.read less - ID:
- xk81jz04q
- Component ID:
- tufts:21477
- To Cite:
- DCA Citation Guide EndNote
