Thursday, May 28, 2015

New paper finds a large warming bias in Northern Hemisphere temperatures from 'non-valid' station data

A new paper published in the Journal of Atmospheric and Solar-Terrestrial Physics finds that the quality of Northern Hemisphere temperature data has significantly & monotonically decreased since the year 1969, and that the continued use of 'non-valid' weather stations in calculating Northern Hemisphere average temperatures has created a 'positive bias' and "overestimation of temperatures after including non-valid stations." 

The paper appears to affirm a number of criticisms of skeptics that station losses, fabricated/infilled data, and positively-biased 'adjustments' to temperature data have created a positive skew to the data and overestimation of warming during the 20th and 21st centuries. 

Graphs from the paper below show that use of both valid and 'non-valid' station data results in a mean annual Northern Hemisphere temperature over 1C warmer at the end of the record in 2013 as compared to use of 'valid' weather station data exclusively. 

In addition, the paper shows that use of the sharply decreasing number of stations with valid data produces a huge spike in Northern Hemisphere temperatures around ~2004, which is in sharp contrast to much more comprehensive satellite data showing a 'pause' or even cooling over the same period, further calling into question the quality of even the 'valid' land-based stations (urban heat island effects perhaps?).


"The number of valid weather stations is monotonically decreasing after 1969" is shown by the dashed line, and has resulted in an "overestimation of temperature after including non-valid stations" shown by the solid line, especially a spike in temperature in the early 21st century that is not found in satellite temperature records. 
Using temperature data from "valid" stations only, and a base period of 1961-1990, the warmest temperatures were in the first half of the 20th century.
Using a base period of 1800-2013 (including 'non-valid' stations) shows a temperature spike beginning in the early 21st century, but this is not found in the much more accurate and spatially comprehensive satellite records. 
"The computed average by using all stations [including invalid stations, dashed line] is always greater than from using only valid [stations, solid line at bottom of chart]. Percentage of valid stations has steadily declined since 1969 [shown in grey shaded area]. 

Highlights

Introduce the concept of a valid station and use for computations.
Define indices for data quality and seasonal bias and use for data evaluation.
Compute averages for mean and five point summary plus standard deviations.
Indicate a monotonically decreasing data quality after the year 1969.
Observe an overestimation of temperature after including non-valid stations.

Abstract

Starting from a set of 6190 meteorological stations we are choosing 6130 of them and only for Northern Hemisphere we are computing average values for absolute annual MeanMinimumQ1, MedianQ3,Maximum temperature plus their standard deviations for years 1800–2013, while we use 4887 stations and 389 467 rows of complete yearly data. The data quality and the seasonal bias indices are defined and used in order to evaluate our dataset. After the year 1969 the data quality is monotonically decreasing while the seasonal bias is positive in most of the cases. An Extreme Value Distribution estimation is performed for minimum and maximum values, giving some upper bounds for both of them and indicating a big magnitude for temperature changes. Finally suggestions for improving the quality of meteorological data are presented.

No comments:

Post a Comment