1
Geophysical Research Letters
Supporting Information for
Changes in the frequency of observed temperature extremes largely driven by a
distribution shift
Ronak N. Patel*, David B. Bonan*,
Tapio Schneider
*
*
California Institute of
Technology, Pasadena, CA, USA
Contents of this file
Figures S1 to S
5
Additional Supporting Information (Files uploaded separately)
Captions for
Movies
S1
to S
2
Introduction
This
supporting
information includes
additional figures
and videos
that were not suited
to being in the main text. The methods used to generate these figures can be found in
the main text.
Information about the
used
ECMWF ERA5 reanalysis data can be found at
Hersbach
et al. (2023)
.
2
Figure S1
.
Observed extreme warm days and associated most slowly changing
components from 1955 to 2021. The percent of days each year from 1955 to 2021 above
the historical (a) 90th and (b) 99th percentiles for each of the IPCC AR6 land regions
(solid lines). The mos
t slowly varying component of this change (Methods) is plotted with
a purple dashed line. The historical period is defined as 1961
-
1990. Every 20 years
starting in 1960 is marked with a vertical grey line.
3
Figure S2
.
The
(a)
standard deviation
and (b) skewness for the
daily maximum
temperature distribution for the period 1961
-
1990
in the
Berkeley Earth Surface
Temperatures
dataset.
4
Figure S
3
.
The slowest low
-
frequency components and corresponding low
-
frequency
patterns in surface
temperature extremes
for
ERA5
. The first (a) low
-
frequency
component (LFC) and (c) corresponding low
-
frequency pattern (LFP) of the percent of
days in a year where the daytime maximum temperature exceeds the historical (1961
-
1990) 90th percentile threshold. (b
-
d) Same as (a
-
c) but for the 99th percentile. Note
different color bar limits.
Ocean data points are masked in grey.
5
Figure S
4
.
The square of the Pearson correlation coefficient between the expected
number of days above the historical (a) 90th and (b) 99th percentiles based on a shift of
the historical temperature distribution compared to the observed number of days for
1955 to 202
1
in
ERA5
. The isoline of R
2
= 50% is marked in green. (c
-
d) Same as (a
-
b),
but for the root mean squared error. (e
-
f) Same as (a
-
b) but for the bias of only having a
distribution shift. (g) Linear least squares regression slope of the standard deviation of
the annual temperature dis
tribution from 1955 to 2021. (h) Same as (g) but normalized
relative to the historical (1961
--
1990) standard deviation.
Ocean data points are
masked
in grey
.
6
Figure S
5
.
Similar to Figure 4, but for shifting the distribution by the
annual median
warming relative to the historical baseline (1961
-
1990).
The square of the Pearson
correlation coefficient between the expected number of days above the historical (a)
90th and (b) 99th percentiles based on a shift of the historical temperature distribution
compared to the observed number of days for 1955 to 202
1
in
the
Berkeley Earth Surface
Temperatures
dataset
. The isoline of R
2
= 50% is marked in green. (c
-
d) Same as (a
-
b),
but for the root mean squared error. (e
-
f) Same as (a
-
b) but for the bias of only having a
distribution shift.
Video
S1
.
Change in the percent of days above the historical 90
th
percentile relative to
the 1961
-
1990 baseline for each year from 1950 to 2021
in the Berkeley Earth Surface
Temperatures dataset
.
Video
S
2
.
Change in the percent of days above the historical 99
th
percentile relative to
the 1961
-
1990 baseline for each year from 1950 to 2021
in the Berkeley Earth Surface
Temperatures dataset.