Skip to main content

Full text of "NASA Technical Reports Server (NTRS) 20150022492: Relationships Between Excessive Heat and Daily Mortality over the Coterminous United States"

See other formats


Relationships Between Excessive Heat and Daily 
Mortality over the Coterminous United States 


Bill Crosson 

Mohammad Al-Hamdan 

Science and Technology Institute, Universities Space Research Association 
National Space Science and Technology Center, Huntsville, AL 

Maury Estes, Sue Estes 
Earth System Science Center 
University of Alabama in Huntsville 

Dale Quattrochi 
NASA Earth Science Office 
Marshall Space Flight Center, Huntsville, AL 


Sixth Conference on Environment and Health 
American Meteorological Society 95 th Annual Meeting 

Phoenix, Arizona 
8 January 2015 



Objective and Methods 


> Objective: Examine the relationships between heat-related mortality and 
excessive heat at the daily, county scale 

> Science Questions: 

• Can heat-related deaths be clearly tied to excessive heat events? 

• What time lags are critical for predicting heat-related deaths? 

• Which of the many common heat metrics correlates best with heat-related 
deaths? 

> Methods: 

• Use meteorological reanalysis data to develop measures of extreme heat on a 
1/8° (~12 km) grid over the coterminous U.S. 

• Aggregate heat measures to county, daily level for coterminous U.S. 

• Use heat data to identify 'Extreme Heat Events' (EHEs) based on nine definitions 

• Examine the relationship between EHEs and heat-related mortality using CDC's 
National Center for Health Statistics 'Multiple Causes of Death 1999-2010' data 



Environmental Data: 

North American Land Data Assimilation System (NLDAS) 


> North American Land Data Assimilation System (NLDAS) 
combines meteorological data from models with observations 
from ground stations and satellites for the conterminous U.S. 

■ l/8 th degree (about 12 km) spatial resolution 

■ Available hourly for 1979-current 

■ Available from http://www.emc.ncep.noaa.gov/mmb/nldas/ 

> Hourly air temperature and humidity data have been used to 
create grid-level daily maximum and minimum temperatures 
and other heat-related variables such as Heat Index. 




■■■■■□□□□□□□□□□□□Ill 


Metrics of Excessive Heat 

1. Daily Maximum Air Temperature 


> Daily maximum air temperature, the highest temperature recorded at an 
observation site between midnight and midnight local standard time, is a 
traditional measure of heat, and one with which everyone is familiar. 

We used NLDAS hourly data to calculate daily maximum air temperature. 


T(°F) 


45-60 

60-65 

65-70 

70-75 

75-80 

80-84 

84-88 

88-91 

91-93 

93-95 

95-97 

97-99 

99-101 

101 -103 

103-105 

105-107 

107-109 

109-111 

111 -113 

113-115 



August 13, 2007 




Metrics of Excessive Heat 

2. Heat Index (HI) 

Heat Index or 'apparent temperature' combines air temperature and relative 
humidity (Steadman et al., J. Appl. Met. 1979). As humidity increases, the efficiency 
with which our bodies lose heat through evaporation of sweat is reduced. 
Commonly used as a heat stress indicator, HI is the basis for NWS heat warnings: 

f 90° F Extreme Caution 
HI > “1 105° F Danger 

L 130° F Extreme Danger 



HI is only calculated when 
air temperature > 80° F. 




Metrics of Excessive Heat 

3. Net Daily Heat Stress (NDHS) 


Net Daily Heat Stress is a new heat variable that gives an integrated measure of heat stress 
(and relief) over the course of a day, defined as: 

NDHS = 2(HI I -Hl h0t )-2(T C00| -T i ) 

i Y 1 J— ' 

heat stress heat relief 

> Sums are over the hours in a day and include only positive terms. Units are degree-hours. 

> The 'heat stress' term is only calculated when Hlj > Hl hot , where Hl hot is a threshold above 
which HI is considered a stressor, set to 90° F. 

> The 'heat relief term is only computed when Tj < T COO |, a temperature below which relief 
from heat occurs, set to 75° F. 

> If heat relief is greater than heat stress, NDHS is set to 0. 

Advantage over Temperature and Heat Index: 

NDHS may be a more appropriate heat stress measure than HI, an instantaneous variable not 
well-suited to capture the health impacts of multi-day heat events. 


6 




Metrics of Excessive Heat 

3. Net Daily Heat Stress (NDHS) 





Q) 



NDHS = Max(0, Red area - blue area) 





■■■■□□□□□□□□□□[ 


Metrics of Excessive Heat 

3. Net Daily Heat Stress (NDHS) 


NDHS 

(degree-hours) 


<50 
50-100 
100-125 
125-150 
150-175 
175-200 
200 - 225 
225 - 250 
250 - 275 
275 - 300 
300 - 325 
325 - 350 
350 - 375 
375 - 400 
400 - 425 
425 - 450 
450 - 475 
475 - 505 






Definitions of Extreme Heat Events 


We applied 9 definitions of Extreme Heat Events (EHE) for the 'warm season' of 
May-September. 

For all definitions, the variable must exceed the threshold on two consecutive days. 
Percentiles are determined for each county based on 1981-2010 NLDAS data for 
the warm season. 



Variable(s) 

Percentile 

Absolute threshold 

1 

Daily maximum temperature 

95 


2 

Daily maximum temperature 

95 ■ 

90° F 

3 

Daily maximum temperature 

99 


4 

Daily maximum and minimum 
temperatures 

95 


5 

Daily maximum and minimum 
temperatures 

95 

90° F for max. temperature 

6 

Daily maximum Heat Index 

95 


7 

Daily maximum Heat Index 

99 


8 

Net Daily Heat Stress 

95 


9 

Net Daily Heat Stress 

99 







Mortality Data 


Source: CDC's National Center for Health Statistics "Multiple Causes of 
Death 1999-2010', which includes: 

State and county of decedent's residence and place of death; 

Date of death; 

Underlying cause of death; 

Contributing factors of death; 

and many other demographic variables. 

In this study, a "heat-related death' is one for which heat was an 
underlying or contributing cause. 




Evaluation Statistics 


We applied several statistics for 9 regions and for the entire US to evaluate how well 
each of the EHE definitions indicates heat-related mortality. 

Using a 'forecast-outcome' analogy, an 'EHE exposure' is one person exposed to 
extreme heat for one day - think of this as a forecast of a heat-related death. (We 
calculate all statistics on a per 1 million exposures basis.) 

An 'EHE death' is one heat-related death occurring during an EHE. 

A 'non-EHE death' is one heat-related death occurring outside of an EHE. 

1. Equitable Threat Score (ETS): Takes into account correct forecasts, 'false 
alarms' (EHE exposures but no deaths) and missed events (non-EHE deaths) and is 
designed to avoid a tendency toward poor scores for rare events. 

2. Ratio of heat-related deaths on EHE days to heat-related deaths on non-EHE days 
ETS and the Ratio are calculated based on total heat-related deaths and exposures 
over all warm seasons, 1999-2010. 

3. Linear correlation coefficient between daily heat-related deaths and number of 
persons exposed to EHE 

Correlations are calculated based on daily heat-related deaths and exposures. 



Evaluation Statistics - Equitable Threat Score (ETS) 

Results for Entire US 


Results are shown for best temporal lag (usually 1 day). 

Best results: NDHS 99 th percentile and maximum Heat Index 99 th 
percentile EHE definitions. 

Worst results: Maximum temperature 95 th percentile EHE definition. 


0.07 

.063 .064 

0.06 h 

.042 

■ ■■■■■ ■ 


0.03 


0.02 



o.o: 1| ||p ;|| ||| III || ||| ||| || 

Max & min Max & min Max HI 95% Max HI 99% Max temp Max temp Max temp NDHS 95% NDHS 99% 
temp 95% temp 95% & 95%&90F 99% 95% 

90 F 



Evaluation Statistics - Ratio of Death Rates 

Results for Entire US 


Results are shown for best temporal lag (usually 1 day). 

Best results: Maximum Heat Index 99 th percentile and NDHS 99 th 
percentile EHE definitions. 

Worst results: Maximum temperature 95 th percentile EHE definition. 


14.3 14.2 



Max & min Max & min Max HI 95% Max HI 99% Max temp Max temp Max temp NDHS 95% NDHS 99% 
temp 95% temp 95% & 95%&90F 99% 95% 

90F 



Evaluation Statistics - Linear Correlation Coefficient 

Results for Entire US 


Results are shown for best temporal lag (usually 1 day). 

Best results: Maximum temperature 99 th percentile EHE definition. 

Worst results: Maximum temperature 95 th percentile EHE definition. 



Max & min Max & min Max HI 95% Max HI 99% Max temp Max temp Max temp NDHS95% NDHS99% 
temp 95% temp 95% & 95% & 90F 99% 95% 

90F 



Evaluation Statistics - Equitable Threat Score 


Best results were obtained using a lag of 1 day unless otherwise noted. 


.066 

Max. HI 99% 



Best metrics: 

NDHS: 5 regions 
Max. HI: 3 regions 
Max. Temp: 1 region 





Evaluation Statistics - Ratio of Death Rates 


Best results were obtained using a lag of 1 day unless otherwise noted. 






Evaluation Statistics - Linear Correlation Coefficient 


Best results were obtained using a lag of 1 day unless otherwise noted. 



.733 

NDHS 99% 


and 

Max. HI 99% 


.560 

NDHS 95% 


.552 

Max. & min. 
temp. 95%/90F 


.682 

Max. Temp 
95%/90F 


.743 

NDHS 99% 


.784 

Max. HI 99% 


.708 

NDHS 99% 


.623 

NDHS 99% 


Best metrics: 

NDHS: 6 regions 
Max. HI: 2 regions 
Max. Temp: 1 region 
Max & Min Temp: 1 region 


kV 








Summary and Conclusions 


> We used NLDAS meteorological data to identify all Extreme Heat Events, based 
on 9 definitions, during 1999-2010 for all US counties. 

> We used CDC's 'Multiple Causes of Death, 1999-2010' dataset to compute daily, 
county-level mortality counts for which heat was the underlying or a contributing 
cause. 

> We used three statistics to evaluate how well the 9 EHE definitions correlate with 
heat-related mortality within 9 US regions. 

> For the US, and for most regions, the Net Daily Heat Stress metric performed 
best, followed closely by daily maximum Heat Index. 

> For each evaluation statistic and each heat metric, EHE definitions using 99 th 
percentile performed better than definitions based on 95 th percentile. 

> A temporal lag of 1 day produced the best indicator of mortality in most cases. 




Acknowledgments 


♦ NASA Applied Sciences 

♦ CDC Environmental Public Health Tracking Program 

♦ Sigrid Economou and Mark Puckett of CDC for supplying 
the mortality dataset