Rooftop Temperature Bias Study
50 of the NWS's 11,700 Coop stations are located on the roofs of
Rooftop exposures has been known for years to create measurement
biases as compared to the traditional ground-based measurement sites.
For example, in 1942, one of the world's leading climatologist,
Helmet Landsburg, wrote in his textbook "Physical Climatology,"
"conclusions on climate derived form records of roof stations
may be no means representative of those at the ground level."
Since surface temperature measurements are intended to represent
conditions where people live, work, commute, and grow their food
there is interest in quantifying the biases associated with rooftop
measurements. Although some site specific studies have been performed,
no large scale attemps have ever been conducted to measure differences.
In the interest of adherence to exposure standards and represnetativeness
of surface observations, the NWS commissioned a study of rooftop
temperature biases, beginning in 1999 and scheduled for completion
in 2000. The results will be published and used to determine future
NWS policy of its own rooftop stations.
The proposal, attached below, is funded by NOAA's Environmental Data
and Information Services.
Proposal Title: A Comparison of Rooftop and Standard Ground-Based Temperature
continuity of surface air temperature measurements are critical
for many NOAA activities including short term weather forecasting
and warning, climate monitoring and the prediction and assessment
of decadal to centennial climate change. Historically, most air
temperature observations have been taken over a grassy surface at
a height of approximately five feet above the ground. Non-standard
instrument siting, such as rooftop installation, introduces uncertainty
in both the accuracy and representativeness of temperature data
within the NWS. The continued expansion and availability of real-time
weather data from sources outside of NOAA which often use rooftop
installations, and the growing opportunities within NOAA to include
such data in forecasting, verification, research and climate applications,
makes it imperative to understand what effects rooftop siting has
on data quality, continuity and representativeness.
will study biases and uncertainties in temperature records caused
by rooftop instrument locations. Both National Weather Service stations
and non-NOAA sources will be investigated. Ground level to rooftop
temperature differences will be studied to show what, if any, differences
occur and whether these differences are significant, systematic,
predictable and a function of current weather conditions or if differences
are highly variable and inconsistent. Work will be performed at
the National Climatic Data Center to better document where and when
NWS weather stations have been situated on rooftops.
will take place during FY 1999. Most data analysis, assimilation
and presentation of results will follow in FY 2000.
To quantify biases
that rooftop temperature measurements exhibit under different synoptic
conditions and instrument exposures and to provide data users
with the effects these biases are having on the representativeness
of temperature data
temperature observations have historically provided the vast bulk
of temperature information for forecasting, climate change research,
and public information and business application.
exposure height for sensors is 5 feet above a ground surface representative
of the surrounding environment. The intent of this standard is to
provide representative measurements of air temperature where people
live, work, play, and grow their food. Standards for accurate and
representative weather observations and climatic data collection
encourage consistent and relatively uniform instrument siting and
ground-level exposure. However, there are numerous locations around
the country where instruments are exposed on rooftops instead of
on the ground. Rooftop sitings are usually the result of lack of
a nearby ground-based site or security considerations
The NWS currently
operates numerous rooftop stations nationwide. In the private sector,
rooftop station exposures are growing rapidly nationwide. Several
thousand such sites now are operating. In many cases, these stations
include real-time data accessability and widespread public dissemination
by the media.
observations with observations from standard ground-level weather
stations can pose significant issues for weather forecasting and
verification, weather and climate analysis, and important climate
applications such as energy demand planning and forecasting by large
Do rooftop temperatures accurately represent the air temperatures
that people experience at ground level? Does the representativeness
vary depending on time of year, exposure of instruments on the roof,
the type of roof (metal, stone, color, etc.), height of roof, latitude,
elevation above sea level, and surrounding climate and vegetation?
work has been done to document the magnitude oSeptember 24, 2013ware of the potential impacts associated
with rooftop observations. Rooftop weather stations (both NOAA and
non-NOAA) continue to be accepted as official sources of weather
observations. Rooftop temperature data published by NOAA can be
included in climatic data sets without obvious flags that alert
data users to this fact.
is an effort to quantify the bias that results from taking temperature
measurements on rooftops versus the standard ground-based exposure
for all four seasons and under different synoptic situations. This
goal will be accomplished by making a large number of temperature
comparisons between rooftop observations and nearby ground-based
will be evaluated as a function of varying weather conditions such
as cloud cover, wind speed, sun angle (time of year). To the extent
possible, the effects of building height, roof characteristics (area,
color), location of the sensor on the roof (east corner, center,
etc.) and surrounding objects and land surface characteristics will
also be assessed.
The project is
further defined by five goals:
i) To establish
the relationship of temperature observations from rooftop to ground
ii) To document in NCDC metadata files the dates and time periods
of NWS rooftop observations.
iii) To identify meteorological conditions associated with larger
temperature effects of the rooftop locations.
iv) To assess the impact of rooftop locations on defining the
climate (normals) of a location, and assess the potential impact
of rooftop observations on long-term climate research.
v) To present study results at national meetings to inform a wide
is a collaborative effort of several groups with a strong interest
in understanding continuity, biases and data quality issues associated
with rooftop temperature observations. Participants include Colorado
State University, the NWS and NESDIS (NCDC) in NOAA, and a media
participant (Bob Ryan, NBC, local television affiliate in Washington,
The study will
be conducted by collecting data from a large number of both rooftop
and ground based stations for one year and then performing analyses
of the data in the second year.
The results of
this study will be used to evaluate to what extent rooftop temperature
data are useful in weather and climate applications. If differences
are small and/or relatively predictable and consistent, relationships
to ground level measurements could be developed and applied. However,
if differences are large and unpredictable, efforts should be made
to terminate rooftop measurements.
The project includes
two types of data comparisons:
of rooftop observation to non-rooftop observations
ii) comparison of rooftop observation to field standard observations.
In the first
type, the rooftop and non-rooftop observations will be from instruments
that are in regular service as a part of a designated network. The
first network includes sites operated by the NWS and will include
ASOS and the Cooperative Observer Network (Coop). Observations will
include daily maximum and minimum temperatures.
In the second
type of comparison one field standard R.M. Young (RMY) aspirated
temperature system will be placed beside the rooftop instrument
and a second and third RMY will be placed at the surface near the
building with the rooftop instrument. One RMY will be used to make
an absolute temperature comparison with the rooftop instrument.
The other two RMYs will be used to make observations at the surface
nearby. One of the surface RMY systems will be placed on the North
side of the building and one at a location to be chosen. The two
at the surface will allow a definition of horizontal differences
in surface temperature and the "North" temperature. The
temperature on the North side of the building is of interest since
many Coop and first order observations in the historic climate records
came from such a location.
results from the NWS networks are of great importance to the NWS
as we are currently evaluating the future of NWS rooftop observations.
Comparisons with other networks are equally important since their
observations are being made available publicly to a wide audience.
C. Research Plan
is described in this section is ordered by the five specific goals
in section 1.
I) To establish
the relationship of temperature observations from rooftop to ground
The first step
is to identify the rooftop sites to be used in the study. At least
50 sites will be identified with one half in the NWS and one half
outside the NWS. Historical data will be acquired for as many sites
as possible. Several of the NWS sites have data for a few decades.
Data will be collected from all sites during the first year of the
study. In addition hourly data will be collected from the nearest
ASOS location to each site.
Three types of
analysis will be used in the data comparisons. The first is to form
the accumulated sum of the difference in temperature between the
rooftop instrument and the comparison instrument. The analytic model
to be used is to assume a linear relationship between them.
In a graphical
form of the accumulated temperature difference as a function of
time, the slope of the line is the bias between the two observations.
If the assumed relationship is correct, then the slope is steady
and easily determined. If the relationship is not correct, then
the slope varies with time and a variable relationship is identified.
The second analysis
is to correlate the two temperatures and calculate the fraction
of the variance which is explained by the correlation.
The third analysis
is to prepare frequency distribution of the temperature difference
and analyze them. A relationship with a stable slop will be characterized
as a nearly gaussian distribution with small tails. If the distribution
is skewed toward one side or has a limited number of large deviations
for the mean, then we plan to try to identify the weather conditions
which lead to large differences.
ii) To document
in NCDC metadata files the dates and time periods of NWS rooftop
The NCDC participation
in the project is primarily to improve and digitize files to allow
the project to verify the actual times and locations of rooftop
instruments in the NWS for Coop sites and for first order sites.
iii) To identify
meteorological conditions associated with larger temperature effects
of the rooftop locations.
The ASOS hourly
observations will be used to define weather conditions. The expectation
is that large differences are due to weather conditions or local
characteristics of the roof. Primary weather conditions are wind
speed and solar radiation. High wind speed and overcast skies will
lead to small temperature differences which are due to instrument
differences. Low wind speed and clear skies will lead to larger
temperature differences. These will be investigated individually
on days of occurrence to understand the causes.
iv) To assess
the impact of rooftop locations on defining the climate (normals)
of a location, and assess the potential impact of rooftop observations
on long-term climate research.
The results of
(iii) above are the important source for this assessment. Stable
biases will just produce a bias in the normal and long term climate
records. Variable relationships and highly skewed relationships
will have a very different effect on climate normals.
v) To present
study results at national meetings to inform a wide audience.
be presented to several meetings but two are known. The American
Meteorological Society has an annual meeting in January of each
year and the American Association of State Climatologists has an
annual meeting each August.
FY 1999 Work
To identify rooftop stations.
#2/Feb 99: Identify ground level station suitable for comparison
and obtain comparable data.
#3/Jun99: Conduct field study with RMY instruments.
A list of known NWS and non-NOAA rooftop weather stations.
#2/Sep 99: Data set collected.
Analyze co-located data from roof and surface.
#2/Jun 00: Analyze rooftop vs. Non-rooftop data.
#3/Jun 00: Present results to scientific community.