Total Pageviews

Friday, January 18, 2019

Sampling Errors in the Instrumental Surface Temperature Record

To calculate 
the surface 
average 
temperature 
of our planet, 
the Earth's surface
is divided into 
5 degree longitude by 
5 degree latitude grids.
( this divides the globe's surface 
into 2592 grid cells )

Unfortunately, a majority 
of the surface grids 
have no thermometers,
or are missing some data 
needed to compile 
the global average.

For a majority of grids.
missing data are "infilled"
( wild guessed ) 
by government bureaucrats
with science degrees.

There is no way 
to double check
(verify) 
their guesses, 
or to prove 
them wrong 
(falsify them.)

But ... the global
surface average,
with all that infilling, 
can be compared 
with two other
measurement 
methodoloiges,
weather satellites 
and weather balloons.

Weather satellites and balloons 
show LESS warming than the 
surface data.

They also correlate well
with each other.

That means surface 
temperature data
are outliers, 
so you'd expect
them to be rejected.

But not in 
climate 
junk "science",
where the 
weather satellites
and weather balloons 
are ignored
simply because 
they show 
less warming 
than surface data !

A real scientist 
would consider 
surface temperature 
data quality, with all 
of that infilling,
and the lack of 
correlation with 
satellite and balloon
data, to be low quality ! 

But never mind that, 
for now !

Surface temperature 
data quality are low
for other reasons too
 -- even if we deliberately 
ignore all the infilling 
and lack of correlation 
with other methodologies ! 




The surface 
temperature record 
is built on 
two measurements, 
taken daily at each 
monitoring station, 
specifically the 
maximum 
temperature (Tmax),
and the 
minimum 
temperature (Tmin). 

These daily readings 
are then averaged
to calculate a daily 
mean (average) temperature:  
Tmean = ( Tmax + Tmin ) / 2. 

The surface instrumental 
temperature record 
is the basis for all 
the ( fake ) alarm 
over climate change. 

Therefore, the accuracy 
of the surface temperature 
record is important. 

Are two daily measurements 
of temperature sufficient 
to accurately determine 
the average temperature
of ouyr planet's surface, 
and temperature trends? 

The right answer is "no".



It is important to understand 
that air temperature is a signal.

The process of measuring 
a signal is known as sampling. 

A rapidly changing signal 
must be sampled 
more frequently 
than a slow 
rate of change signal.

Exactly how frequently 
a signal must be sampled, 
was answered in 1928 ,
by Harry Nyquist, 
a Swedish-born 
American Electronics 
Engineer. 

A few years later, 
with the help 
of Claude Shannon, 
the Nyquist-Shannon 
Sampling Theorem 
was published. 

According to the 
Nyquist-Shannon 
Sampling Theorem, 
we must 
sample the signal 
at a rate at least 
two times 
the highest frequency 
component of the signal.

If we sample signals 
less often than
the Nyquist Rate, 
our samples 
will contain error. 




The Nyquist 
Sampling Theorem 
is essential science for 
digital audio, digital video, 
industrial process control, 
medical instrumentation, 
flight control systems, 
digital communications, 
etc.

Climate "science", and the 
surface instrumental 
temperature record, 
completely ignore Nyquist.

The historical temperature data 
use only two samples / day.

NOAA, in their USCRN 
( US Climate Reference Network ) 
has determined that it is 
necessary to sample 
at 4,320-samples/day 
to eliminate Nyquist errors. 

4,320-samples/day 
equates to 1-sample 
every 20 seconds. 

This rate can only be achieved 
by automated sampling 
with electronic instruments. 

But most of the instrumental
surface temperature record 
is comprised of 
historical readings 
of mercury max/min 
thermometers, 
taken long before 
automation was an option. 

Today, despite the 
availability of automation,
the instrumental record 
still uses Tmax and Tmin 
        ( effectively 2- samples/day ) 
instead of a 
Nyquist-compliant 
sampling rate. 

The reason for this 
is to maintain 
compatibility 
with the older 
historical record. 




The maximum 
and minimum 
temperatures happen 
whenever they happen,
not at exactly 
the same time 
of each day. 

The fact that 
Tmax and Tmin 
happen at 
irregular times 
during the day 
causes its own kind 
of sampling error. 

The instrumental
surface temperature 
record uses 
the worst method 
of sampling possible 
– resulting in 
maximum error.

Calculating 
mean temperature 
with 2-samples/day 
rarely yields 
the correct mean
calculated with 
five-minute samples.

The trend 
sampling error 
may be of a
similar magnitude 
to the claimed 
temperature trends!




It is certain that every 
mean temperature, 
and derived trend 
in the record ,
contains 
significant error, 
if it was calculated 
with two temperature
samples/day,
violating the 
Nyquist-Shannon 
Sampling Theorem,
known for over 80 years 
and essential science 
to every field 
of technology 
that involves 
signal processing. 

This calls 
into question 
the scientific value 
of the instrumental 
surface temperature 
record and the 
practice of using 
Tmax and Tmin 
to calculate 
mean values 
and long-term 
temperature trends.