Climate Change Indicators in the United States Technical Documentation Growing Degree Days

Growing Degree Days

Identification

1. Indicator Description

This indicator examines changes in annual growing degree days from 1948 to 2020 at 305 locations
across the contiguous 48 states. Growing degree days are calculated using surface air temperature data
from meteorological stations overseen by the National Oceanic and Atmospheric Administration
(NOAA). Studies have documented a relationship between growing degree days and pollen season
length and start date for grass, oak, and birch pollen: as growing degree days increase, grass pollen
season lengthens and oak and birch pollen seasons begin earlier (Zhang et al., 2015; Lo et al., 2019).
Because allergies are a major public health concern, observed changes in growing degree days, which
serve as a proxy for changes in pollen season length and start date, provide insight into ways in which
climate change may affect human well-being. More broadly, growing degree days also affect plant
growth, agricultural production, and the spread and impact of plant diseases and pests.

2. Revision History

April 2021: Indicator published.

Data Sources

3. Data Sources

Data for this indicator come from NOAA's Global Historical Climatology Network (GHCN) Daily database.
This integrated database of land surface stations across the globe provides daily climatological data from
numerous sources. Available data include maximum and minimum surface air temperatures from the
climate monitoring stations that constitute the U.S. Climate Reference Network. Data availability varies
by station; this analysis only used stations that provide minimum and maximum daily temperatures.

4. Data Availability

The GHCN-Daily data employed in the analysis are available for download at:

https://doi.org/10.7289/V5D21VHZ. The data employed in the current analysis were downloaded in
March 2021 (GHCN-Daily version 3.28); as the GHCN-Daily data set is continuously updated, data
downloaded on future dates could differ from the data shown by this indicator.

Individual weather station data are maintained by NOAA's National Centers for Environmental
Information (NCEI), and the data are distributed on various computer media (e.g., anonymous FTP sites),
with no confidentiality issues limiting accessibility. Specifically, the data for this indicator can be
obtained online via FTP at: ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/dailv. Appropriate metadata and
"readme" files are also available at this link.

Technical Documentation: Growing Degree Days

-------
Methodology

5. Data Collection

Systematic collection of weather data in the United States began in the 1800s. Since then, observations
have been recorded from 23,000 stations. At any given time, observations are recorded from
approximately 8,000 stations. Some of these stations are automated stations operated by NOAA's
National Weather Service. The remainder are Cooperative Observer Program (COOP) stations operated
by other organizations using trained observers and equipment and procedures prescribed by NOAA. For
an inventory of U.S. weather stations and information about data collection methods, see:
www.ncdc.noaa.gov/data-access/land-based-station-data. the technical reports and peer-reviewed
papers cited therein, and the National Weather Service technical manuals at: www.weather.gov/coop.
Sampling procedures are also described in Kunkel et al. (2005) and in the full metadata for the COOP
data set, available at: www.weather.gov/coop. Variables that are relevant to this indicator include
observations of daily maximum and minimum temperatures.

The GHCN-Daily database includes the most complete collection of U.S. daily climate summaries
available (NOAA, 2021a). Its U.S. collection includes a dozen separate data sets archived by NCEI. NCEI
explains the variety of databases that feed into the GHCN for U.S.-based stations in online metadata and
at: www.ncdc.noaa.gov/data-access/land-based-station-data/land-based-datasets/global-historical-
climatology-network-monthlv-version-4. The GHCN-Daily database contains the earliest observations
available for the United States, as well as the latest measurements available from the climate
monitoring stations that make up the U.S. Climate Reference Network.

The currently active U.S. stations in GHCN-Daily update data through real-time data feeds. There is
continual reprocessing of the data, and all data are subject to change; however, changes to data values
for U.S. stations are rare beyond 60 days from the end of a given month.

6. Indicator Derivation

This analysis is based on an approach published by Dr. Yong Zhang of the Environmental and
Occupational Health Sciences Institute at Rutgers University. His publication (Zhang et al., 2015)
describes the relationship between growing degree days and pollen upon which this indicator is based.
Specifically, the authors found that as growing degree days increase, grass pollen season lengthens and
oak and birch pollen seasons begin earlier. Drawing on this relationship, this indicator measures changes
in growing degree days as a proxy for changes in grass pollen season length and oak and birch pollen
season start date. Earlier start dates have also been correlated with longer season length (Lo et al.,
2019; Anderegg et al., 2021). Grass, birch, and oak are not the only pollen types collected at NAB
stations, but they are collected at a majority of monitoring stations, which is useful context for this
national-scale indicator. Data for ragweed, mugwort, and other plant species studied by Zhang et al.
(2015) are also measured, but data availability is sparser.

Growing degree days are calculated using daily temperature data from 305 NOAA monitoring sites in the
contiguous 48 states for the period 1948-2020. These stations were selected based on the following
criteria for data availability:

• 95 percent of the years from 1948 to 2020 must have one day per month with available data for
both minimum and maximum temperature. Years with months without any data were removed

Technical Documentation: Growing Degree Days

-------
from the analysis. This eliminates stations that were not operational during part of the period of
record or had long periods of incomplete data.

• Each station must have no more than 30 consecutive days of missing data. This capped the
number of consecutive days that EPA would interpolate over.

• Each station must have 95 percent completeness overall (i.e., data for 95 percent of all days
during the period of record). This ensured that there were no stations with excessive instances
of missing data.

EPA selected 1948 as a start date because it enabled inclusion of most stations from the U.S. Historical
Climatology Network, which is a key contributing database to the GHCN-Daily. In addition, pre-1948
weather data have limitations as documented in Kunkel et al. (2005). The year 1948 is an established
starting point used by other EPA indicators that draw data from GHCN-Daily.

The calculation of growing degree days relies on a widely used averaging method, described in
publications such as McMaster and Wilhelm (1997). After downloading the daily temperature data for
each of the weather stations employed in the analysis, EPA averaged the maximum and minimum daily
temperatures for each day and subtracted a base temperature of 50°F. A negative value is assumed to
be zero (i.e., there is no such thing as a negative growing degree day). For stations that had missing days
of data, EPA used linear interpolation to estimate the number of growing degree days on those days,
based on actual growing degree days calculated for the surrounding dates with available data. As noted
in the criteria above, this approach was limited to periods of no more than 30 days, and in practice,
most interpolation was conducted over much shorter periods.

For each year at each station, EPA aggregated daily degrees above the base temperature to calculate an
annual growing degree day total. EPA then used the series of annual growing degree day totals to
calculate a long-term trend for each station, using Sen's slope regression. To provide more context for
understanding the magnitude of the observed changes, Figure 1 of this indicator reports trends as
percentage increases or decreases, computed from the value for the last year of the regression line
relative to the value of the first year of the regression line.

EPA selected 50°F as a baseline temperature for this analysis. Different plant species naturally have
different temperature requirements, but for a broad indicator like this one, where multiple species are
of interest, it is most useful to set a single baseline. In the absence of using an observation-based model
using several inputs, one can use a defined base threshold temperature representative of many places
and plants. Two main temperatures—32°F and 50°F—are often used and cited (e.g., by the USA National
Phenology Network). EPA chose 50°F to better represent accumulated heat relative to pollen types
examined by Zhang et al. (2015) (www.usanpn.org/data/agdd maps). Crimmins and Crimmins (2019)
provide further discussion of how 50°F represents a point of accumulated heat that often aligns with
flowering and related activities that are further into the year than simply emergence or initiation of
greening.

7. Quality Assurance and Quality Control

The GHCN-Daily data are subject to a strict quality assurance and quality control process, described at
https://www.ncdc.noaa.gov/ghcn-dailv-methods (NOAA, 2021b). During each reprocessing cycle, the
data are checked for formatting inconsistencies such as impossible months or days and invalid
characters in data fields. Next, a sequence of fully automated quality assurance procedures identifies
daily values that violate one of the quality tests. These tests identify a variety of data problems,

Technical Documentation: Growing Degree Days

-------
including the excessive duplication of data records; exceedance of physical, absolute, and climatological
limits; excessive temporal persistence; excessively large gaps in the distributions of values; internal
inconsistencies among elements; and inconsistencies with observations at neighboring stations. Data
that fail a given quality control check (0.3 percent of all values) are marked with flags, depending on the
type of error identified. GHCN-Daily does not contain adjustments for biases resulting from historical
changes in instrumentation and observing practices.

Analysis

8. Comparability Over Time and Space

Growing degree days have been calculated using the same methods for all locations and throughout the
period of record. The analysis was limited to weather stations that did not move during the period of
record. NOAA follows strict protocols to ensure consistent data collection instrumentation over time
and across the country.

9. Data Limitations

Factors that may impact the confidence, application, or conclusions drawn from this indicator are as
follows:

1. This indicator presents information on changes in growing degree days as a proxy for changes in
pollen season length and start date for grass, oak, and birch based on work published by Zhang
et al. (2015). However, the indicator does not analyze pollen data, and therefore should be
viewed as a rough approximation for changes in these pollen season characteristics.

Importantly, the length of the pollen season does not necessarily scale linearly with growing
degree days, and the relationships demonstrated in the literature cannot be guaranteed to be
exactly the same in other locations or under different conditions. That is why this indicator is
presented as a screening-level proxy for pollen season.

2. The growing degree day measure reflects cumulative conditions that support plant
development, but as a broad indicator, it does not consider plant species-specific temperature
thresholds, does not incorporate upper temperature limits into the calculation (as some more
sophisticated analyses do), and does not capture potentially important effects in the sequencing
of weather conditions for plant development. There are factors other than growing degree days
that also affect pollen season duration and start date; some of those factors reflect phenological
cycles of plant activity, and some are unrelated to climate such as local plant composition,
geographic location (latitudinal position), and proximity to urban areas (Lo et al., 2019). As a
result, the link between growing degree days and pollen season timing is not precise.

3. EPA is aware of other analyses that have restricted the calculation of growing degree days to a
defined "pollen season" window. However, because this is a broad indicator designed to be
relevant to a variety of plant species, which may differ in the timing of their "pollen seasons,"
EPA has elected to calculate growing degrees across the entire calendar year for this high-level
summary indicator.

Technical Documentation: Growing Degree Days

-------
4. A change in total growing degree days does not necessarily change the intensity of the pollen
season—though it may have a relation for some plant species.

10. Sources of Uncertainty

Uncertainty has not been calculated directly for this indicator. However, because growing degree days
are based solely on temperature measurements, and because NOAA weather stations measure
temperature with precise, well calibrated instruments and protocols in place to minimize error, any
uncertainty in growing degree days would be expected to be minimal.

Section 12 discusses the level of statistical confidence in station-specific long-term rates of change
calculated via linear regression.

11. Sources of Variability

Inter-annual temperature variability results from normal year-to-year variation in weather patterns,
multi-year climate cycles such as the El Nino-Southern Oscillation and Pacific Decadal Oscillation, and
other factors. Temperature patterns also vary spatially. This indicator provides information on changes
in growing degree days using location-specific trends, as shown in Figure 1.

12. Statistical/Trend Analysis

As noted above, Figure 1 of this indicator uses Sen's slope regression to assess the slope of any long-
term trend present at each station. This type of regression is useful for a screening-level analysis such as
the one presented here. Of the station-specific trends shown in Figure 1, 170 (55.7 percent of stations)
are significant to a 95 percent level (Mann-Kendall p-value < 0.05). Higher-magnitude increases in
growing degree days—such as those that tend to be prevalent in the western United States—are
generally more statistically significant than lower-magnitude trends.

EPA examined these trends further using the Durbin-Watson test for serial correlation (autocorrelation)
of the regression residuals. Of the 170 stations that were significant to a 95 percent level (p < 0.05)
according to the Mann-Kendall test, 106 showed autocorrelation (p-value of the Durbin-Watson test <
0.1, indicating that the test resulted in an extreme value [indicating autocorrelation] and there is a low
probability that such an extreme value could have been observed in a non-autocorrelated data set [the
null hypothesis]). A block bootstrap (using four blocks) on the Mann-Kendall tau was applied to those
106 sites that had both significant autocorrelation and significant trends. A Mann-Kendall bootstrap
block length of four was chosen using the formula nA(l/3), where n is the number of years in the record.
The Mann-Kendall test indicated a significant trend in only 23 of the 106 sites after applying the block
bootstrap. Of the 23 sites with a significant trend, in all but seven cases the trend was increasing. Thus,
when autocorrelation and bootstrapping results are considered, a total of 87 stations (64 + 23) (29
percent) had statistically significant trends.

For reference, Figure TD-1 shows the data and Sen's slope trend of annual growing degree days for five
stations representing the minimum, maximum, median, 25th percentile, and 75th percentile slopes out of
the entire distribution of regression slopes. EPA has included this figure to give a sense of the shape of
the data for a representative sample of sites. Visually, Figure TD-1 suggests that a linear regression may
be at least a reasonable first-order characterization of the data.

Technical Documentation: Growing Degree Days

-------
Figure TD-1. Annual Growing Degree Days for Five Sample Sites, 1948-2020

to
>-
03
"D
0)
0)

L_

ao
~o

GD

c

o

L-

Median slope

5,COO
4,000
3,000

v = 1.6x -1419.9























~~^























1340 1950 1960 1970 1950 1990 2000 2010 2020

25th percentile slope

5,000
4,000
3,000

y = -0.02x + 1927.5







































1940 1950 1960 1970 1930 1990 2000 2010 2020

75th percentile slope

5,000
4,000
3,000
2,000
1,000

y = 3.05x-4731.0











































_































1940 1950 I960 1970 1930 1990 2000 2010 2020

Minimum slope

4,000













V = -3

6*+11219.6











































1940 1950 I960 1970 19S0 1990 2000 2010 2020

Maximum slope

5,000
4,000
3,000
2,000
1,000

y= 14k - 262B7.4













































	



















194Q 1950 1960 1970 19B0 1990 2000 2010 2020

Data source: NOAA. 2021. Global Historical
Climatology Network Daily: Data access.
Accessed March 2021.
www.ncdc.noaa.aov/ahcnd-data-access.

Year

No attempt has been made to aggregate the results of this indicator into overall national or regional
trends, as doing so would require consideration of uneven station density and the influence of
topography in areas not directly represented by a weather station.

References

Anderegg, W., J. Abatzoglou, L. Anderegg, L. Bielory, P. Kinney, and L. Ziska. 2021. Anthropogenic climate
change is worsening North American pollen seasons. P. Natl. Acad. Sci. USA 118(7):e2013284118.
www.pnas.org/content/118/7/e2013284118.

Crimmins, M.A., and T. Crimmins. 2019. Does an early spring indicate an early summer? Relationships
between intraseasonal growing degree day thresholds. J. Geophys. Res.-Biogeo. 124(8):2628-2641.

Technical Documentation: Growing Degree Days

6

-------
Kunkel, K.E., D.R. Easterling, K. Hubbard, K. Redmond, K. Andsager, M.C. Kruk, and M.L. Spinar. 2005.
Quality control of pre-1948 Cooperative Observer Network data. J. Atmos. Ocean. Tech. 22:1691-1705.

Lo, F., C. Bitz, D. Battisti, and J. Hess. 2019. Pollen calendars and maps of allergenic pollen in North
America. Aerobiologia 35:613-633. https://link.springer.com/article/10.1007/slQ453-019-09601-2.

McMaster, G.S., and W.W. Wilhelm. 1997. Growing degree-days: One equation, two interpretations.
Agr. Forest Meteorol. 87(4):291-300.

NOAA (National Oceanic and Atmospheric Administration). 2021a. Global Historical Climate Network
Daily: Description. Accessed March 2021. www.ncdc.noaa.gov/ghcn-dailv-description.

NOAA (National Oceanic and Atmospheric Administration). 2021b. Global Historical Climate Network
Daily: Methods. Accessed March 2021. www.ncdc.noaa.gov/ghcn-dailv-methods.

Zhang, Y., L. Bielory, T. Cai, Z. Mi, and P. Georgopoulos. 2015. Predicting onset and duration of airborne
allergenic pollen season in the United States. Atmos. Environ. 103:297-306.

Technical Documentation: Growing Degree Days

7

-------