patrickbdevaney/mia9 · Datasets at Hugging Face

text stringlengths 0 6.44k
than the MIA in both sunny and shaded locations, with a
standard deviation of 78F in sunny and 88F (4.48C) in shaded
locations. iButtons show that the 1088F threshold (shown by
the horizontal line in Fig. 3c) is exceeded at some locations
even for low values of the MIA heat index. At MIA heat index values greater than 1008F (37.88C) (which is close to the
95% value for a heat index over the entire period), almost
all iButton sites exceeded the 1088F threshold. The 1088F
threshold criteria have only been met 14 times over the last
70 years based on data at MIA (Fig. 4a). Since we know that
iButton data often exceed this threshold, we can model the
number instances of where this threshold would have been exceeded. We use the regression lines shown in Fig. 3c to calculate a “modeled” iButton average heat index (see methods,
section 2). Figure 4 also shows that the number of days in excess of 1088F heat index occurs nearly 1=3 of the year at the
iButton sensors. There is a trend of days in excess of 1088F
heat index increasing 5.7 days decade21
.
4. Conclusions and discussion
This study uses hyperlocal data collected with citizens in
Miami–Dade County to characterize hyperlocal urban heat
exposure. We find that while minimum temperatures at sites
within the county are well-represented by measurements at
MIA, maximum temperature throughout the county were well
in excess of MIA measurements. The heat index computed
for the subset of iButtons that had humidity measurements exceed the threshold level (1088F) for heat advisories that is used
by the National Weather Service for nearly one-third of the
year, with a trend of approximately 6 more days per decade.
While we used only the threshold level to indicate exceedances, and not duration (2 h1), these data indicate that extreme
heat exposure is both a persistent and increasing issue in the
county. This trend is, of course, an oversimplification since it
assumes that the overall relationship between MIA and the
county remains constant with time. The county has significantly
urbanized over time with less greenspace and more impermeable surface (Miami–Dade County Open Data Hub 2021),
which would suggest an increasing influence of urbanization on
866 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 62
Unauthenticated \| Downloaded 02/29/24 07:14 AM UTC
heat exposure over time and in the localized areas of our observations. As the degree of urbanization intensifies with expansion toward the west, we would expect that differences between
the airport and the urban area would likely decrease.
We interpret differences between maximum temperatures in
the iButtons and MIA as representing meaningful differences
in temperatures recorded by the sensors. The NWS Automated
Surface Observing Systems (ASOS) instrument at the airport is
located at the southwest corner of the Miami airport (well outside the footprint of urban Miami), placed at 2 m above the
ground and is well ventilated with no buildings or structures in
the immediate area. The iButtons are requested to be placed
under cover, either in treed (with canopy) or open canopy
(sunny) locations and are generally placed at 0.5–2-m height. A
number of factors likely contribute to the overall higher maximum temperatures recorded by iButtons. Hyperlocal conditions
such as proximity to ground surface, greenness, and permeable/
impermeable surfaces have been shown to impact the maximum
temperatures in urban environments (Schwaab et al. 2021), and
likely contribute to the variation we observed (Fig. 3). These
site-to-site differences contain important information about
heat exposure that is relevant to the impacts that residents experience in their daily lives. We expect we did not see large differences among “sunny” and “shady” conditions in this analysis
simply because variation among more specific microenvironmental conditions requires further analyses. A forthcoming
study will evaluate the influence of site characteristics on siteto-site differences.
It is no surprise that there is generally higher potential heat
exposure in urban areas of Miami, particularly in the warm
season. The question is how hot is too hot? The choice of local or regional threshold to reveal extreme heat is generally
based on a statistical approach, such that conditions are only
rarely exceeded (Robinson 2001; Zhang et al. 2012; McElroy
et al. 2020). The NWS advisory level of 1088F used in Miami–
Dade County is a statewide metric based on heat-related emergency room admissions in select counties across the state. NWS
offices do have discretion about issuing advisories, and forecasters use available data and models to evaluate whether
an advisory should be issued (R. Molleda 2023, personal
communication). However, this is a subjective decision, and
there is a lack of available quantitative data to evaluate
both the spatial extent and duration of possible excess heat
exposure in parts of the forecast region. Our data can be
used to help inform those decisions. Our results show that
maximum heat index values are on average 118F higher than
the airport. While the error attributed to some low-cost sensors should also be considered [;58F (2.88C)], this can be
used as guidance for the overall (average) urban heat island
effect in the county and indicates that localized areas of the
county are at potentially near-dangerous levels throughout
at least the warm season of the year (Fig. 4b).
This finding also brings into question the concept of whether
heat “spells” or “waves” are even an appropriate framework to
assess human health impacts in this region. For example, simply
lowering the threshold to a heat index of 1068F (41.18C) based
on MIA would result in 56 heat advisories since 2000, and for
1048F (408C) this would be 188, as compared with 7 days (or
0.1%) for the 1088F threshold. The nonlinearity of the heat index makes the number of advisories extremely sensitive to the
FIG. 2. The average maximum (top plot of each panel) and minimum (bottom plot of each panel) recorded temperature for the airport
(black), iButton sensors in sunny locations (red), and iButton sensors in shaded locations (blue) for (a) September–October 2018, (b) June–
December 2019, and (c) June 2020–January 2021. iButton data are only plotted for those days on which data were recorded.
JULY 2023 CLEMENT ET AL. 867
Unauthenticated \| Downloaded 02/29/24 07:14 AM UTC
choice of a single threshold. This may have the effect of increasing the number of heat advisories to a level where this becomes
an ineffective communication or outreach tool, referred to by
NWS as “warning fatigue.” Alternatively, if we choose a locally
defined 95th percentile of MIA heat index as the threshold for
extreme heat, as is commonly done (Turek-Hankins et al. 2020),
This value is 100.88F (38.28C) for the entire period from 1950 to
2019, and 102.18F (38.98C) for the post-2000 period (which is
warmer overall). In this case we get 1382 advisories of the
whole period, and 409 since 2000. Again, these high numbers of threshold exceedances also become problematic for
looking at extreme heat as happening in “waves” with a return to “normal” offering relief. Further, these criteria need
to be reevaluated regularly as the climate warms and as local conditions that impact the urban heat island evolve
(Hess and Ebi 2016; Issa et al. 2021; McKinnon et al. 2021).
We therefore suggest, simply based on the heat data, that using a single threshold criterion for the entire state that includes
such a wide range of climate zones such as Florida is not appropriate for characterizing the danger of heat exposure. It is further critical to recognize that this approach may miss heat

than the MIA in both sunny and shaded locations, with a

standard deviation of 78F in sunny and 88F (4.48C) in shaded

locations. iButtons show that the 1088F threshold (shown by

the horizontal line in Fig. 3c) is exceeded at some locations

even for low values of the MIA heat index. At MIA heat index values greater than 1008F (37.88C) (which is close to the

95% value for a heat index over the entire period), almost

all iButton sites exceeded the 1088F threshold. The 1088F

threshold criteria have only been met 14 times over the last

70 years based on data at MIA (Fig. 4a). Since we know that

iButton data often exceed this threshold, we can model the

number instances of where this threshold would have been exceeded. We use the regression lines shown in Fig. 3c to calculate a “modeled” iButton average heat index (see methods,

section 2). Figure 4 also shows that the number of days in excess of 1088F heat index occurs nearly 1=3 of the year at the

iButton sensors. There is a trend of days in excess of 1088F

heat index increasing 5.7 days decade21

.

4. Conclusions and discussion

This study uses hyperlocal data collected with citizens in

Miami–Dade County to characterize hyperlocal urban heat

exposure. We find that while minimum temperatures at sites

within the county are well-represented by measurements at

MIA, maximum temperature throughout the county were well

in excess of MIA measurements. The heat index computed

for the subset of iButtons that had humidity measurements exceed the threshold level (1088F) for heat advisories that is used

by the National Weather Service for nearly one-third of the

year, with a trend of approximately 6 more days per decade.

While we used only the threshold level to indicate exceedances, and not duration (2 h1), these data indicate that extreme

heat exposure is both a persistent and increasing issue in the

county. This trend is, of course, an oversimplification since it

assumes that the overall relationship between MIA and the

county remains constant with time. The county has significantly

urbanized over time with less greenspace and more impermeable surface (Miami–Dade County Open Data Hub 2021),

which would suggest an increasing influence of urbanization on

866 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 62

Unauthenticated | Downloaded 02/29/24 07:14 AM UTC

heat exposure over time and in the localized areas of our observations. As the degree of urbanization intensifies with expansion toward the west, we would expect that differences between

the airport and the urban area would likely decrease.

We interpret differences between maximum temperatures in

the iButtons and MIA as representing meaningful differences

in temperatures recorded by the sensors. The NWS Automated

Surface Observing Systems (ASOS) instrument at the airport is

located at the southwest corner of the Miami airport (well outside the footprint of urban Miami), placed at 2 m above the

ground and is well ventilated with no buildings or structures in

the immediate area. The iButtons are requested to be placed

under cover, either in treed (with canopy) or open canopy

(sunny) locations and are generally placed at 0.5–2-m height. A

number of factors likely contribute to the overall higher maximum temperatures recorded by iButtons. Hyperlocal conditions

such as proximity to ground surface, greenness, and permeable/

impermeable surfaces have been shown to impact the maximum

temperatures in urban environments (Schwaab et al. 2021), and

likely contribute to the variation we observed (Fig. 3). These

site-to-site differences contain important information about

heat exposure that is relevant to the impacts that residents experience in their daily lives. We expect we did not see large differences among “sunny” and “shady” conditions in this analysis

simply because variation among more specific microenvironmental conditions requires further analyses. A forthcoming

study will evaluate the influence of site characteristics on siteto-site differences.

It is no surprise that there is generally higher potential heat

exposure in urban areas of Miami, particularly in the warm

season. The question is how hot is too hot? The choice of local or regional threshold to reveal extreme heat is generally

based on a statistical approach, such that conditions are only

rarely exceeded (Robinson 2001; Zhang et al. 2012; McElroy

et al. 2020). The NWS advisory level of 1088F used in Miami–

Dade County is a statewide metric based on heat-related emergency room admissions in select counties across the state. NWS

offices do have discretion about issuing advisories, and forecasters use available data and models to evaluate whether

an advisory should be issued (R. Molleda 2023, personal

communication). However, this is a subjective decision, and

there is a lack of available quantitative data to evaluate

both the spatial extent and duration of possible excess heat

exposure in parts of the forecast region. Our data can be

used to help inform those decisions. Our results show that

maximum heat index values are on average 118F higher than

the airport. While the error attributed to some low-cost sensors should also be considered [;58F (2.88C)], this can be

used as guidance for the overall (average) urban heat island

effect in the county and indicates that localized areas of the

county are at potentially near-dangerous levels throughout

at least the warm season of the year (Fig. 4b).

This finding also brings into question the concept of whether

heat “spells” or “waves” are even an appropriate framework to

assess human health impacts in this region. For example, simply

lowering the threshold to a heat index of 1068F (41.18C) based

on MIA would result in 56 heat advisories since 2000, and for

1048F (408C) this would be 188, as compared with 7 days (or

0.1%) for the 1088F threshold. The nonlinearity of the heat index makes the number of advisories extremely sensitive to the

FIG. 2. The average maximum (top plot of each panel) and minimum (bottom plot of each panel) recorded temperature for the airport

(black), iButton sensors in sunny locations (red), and iButton sensors in shaded locations (blue) for (a) September–October 2018, (b) June–

December 2019, and (c) June 2020–January 2021. iButton data are only plotted for those days on which data were recorded.

JULY 2023 CLEMENT ET AL. 867

Unauthenticated | Downloaded 02/29/24 07:14 AM UTC

choice of a single threshold. This may have the effect of increasing the number of heat advisories to a level where this becomes

an ineffective communication or outreach tool, referred to by

NWS as “warning fatigue.” Alternatively, if we choose a locally

defined 95th percentile of MIA heat index as the threshold for

extreme heat, as is commonly done (Turek-Hankins et al. 2020),

This value is 100.88F (38.28C) for the entire period from 1950 to

2019, and 102.18F (38.98C) for the post-2000 period (which is

warmer overall). In this case we get 1382 advisories of the

whole period, and 409 since 2000. Again, these high numbers of threshold exceedances also become problematic for

looking at extreme heat as happening in “waves” with a return to “normal” offering relief. Further, these criteria need

to be reevaluated regularly as the climate warms and as local conditions that impact the urban heat island evolve

(Hess and Ebi 2016; Issa et al. 2021; McKinnon et al. 2021).

We therefore suggest, simply based on the heat data, that using a single threshold criterion for the entire state that includes

such a wide range of climate zones such as Florida is not appropriate for characterizing the danger of heat exposure. It is further critical to recognize that this approach may miss heat