[LINK TO HOME PAGE]

THIS POST IS A COMPARISON OF GLOBAL MEAN TEMPERATURES FROM FOUR SOURCES, THREE RECONSTRUCTIONS FROM THE INSTRUMENTAL RECORD AND ONE DIRECT MEASUREMENT OF LOWER TROPOSPHERE TEMPERATURE FROM SATELLITE BASED MICROWAVE RADIOMETRY. THE IMPLIED UNCERTAINTY IS COMPARED WITH THE UNCERTAINTIES IN HADCRUT4 DESCRIBED BY COLIN MORICE. 

FIGURE 1: TEMPERATURE DATA 1979-2018 FOR EACH CALENDAR MONTH

FIGURE 2: DETRENDED CORRELATION AMONG TEMPERATURE ANOMALIES

FIGURE 3: DECADAL WARMING TRENDS 1979-2018 FOR EACH CALENDAR MONTH

FIGURE 4: CORRELATION AMONG DECADAL WARMING RATES

FIGURE 5: ANIMATION OF MORICE UNCERTAINTY IN HADCRUT4: 1979-2018

FIGURE 6:  CORRELATION ANALYSIS OF HADCRUT4 WITH MORICE UNCERTAINTIES

FIGURE 7: CORRELATION TABLE 

1. Uncertainties in global mean temperature estimation and and implications for trend uncertainty are studied by comparing four different temperature datasets, three temperature reconstructions from the instrumental record and one set of satellite microwave radiometry. They are identified with 3-letter acronyms as HAD (Hadley Center HadCRUT4), GIS (Goddard Institute of Space Studies), BRK (Berkeley Earth Climate Research), and UAH (University of Alabama Satellite data).  The common time span 1979-2018 is used for all four data sources restricted by data availability for UAH. The temperatures are delivered as deseasonalized temperature anomalies that should have no seasonal cycle remaining in the data. The analysis is carried out for each calendar month separately. In related posts at this site it is shown that trend behavior of temperature varies significantly among the calendar months [LINK] . The uncertainties found in this analysis are compared with uncertainties in the HadCRUT4 temperature reconstructions reported by Colin Morice  (
2. Morice, Colin P., et al. “Quantifying uncertainties in global and regional temperature change using an ensemble of observational estimates: The HadCRUT4 data set.” Journal of Geophysical Research: Atmospheres 117.D8 (2012).
3. Figure 1 presents the four temperature series BRK, GIS, HAD, and UAH in a GIF format that cycles through the twelve calendar months. The animation appears to show differences among the four estimations of global mean temperature with some variance of these differences among the calendar months.
4. Figure 2 is a graphical summary of the visual information in the GIF animation in Figure 1. The left panel is a graphical presentation of detrended correlations among the three global mean temperature reconstructions BRK, GIS, and HAD. The red line in left panel shows a near perfect correlation between BRK and GIS and the black and blue lines display the somewhat lower correlations of HAD against the two nearly identical datasets GIS and BRK. These correlations appear to differ among the calendar months. The right panel of Figure 2 displays a similar comparison of the three surface temperature reconstruction series (BRK GIS HAD) with the UAH satellite based microwave sounding measurements for the lower troposphere.
5. Figure 3 and Figure 4 present the corresponding analysis of these four datasets (BRK GIS HAD UAH) for decadal trends computed as OLS linear trend values within a moving ten-year window that moves through the time series one year at a time. Figure 3 presents these trends for all four data sources graphically in a GIF animation that cycles through the calendar months and Figure 4 displays their correlations in a graphical format.
6. Correlations among the four decadal trend series shown in Figure 3 are summarized in Figure 4. The left frame of Figure 4 presents correlations among the three surface reconstructions UAH, GIS, and BRK. The right frame shows the correlation of the three surface reconstructions with UAH satellite data.
7. An extension of this uncertainty analysis is presented in Figure 5 and Figure 6 with the Morice uncertainties in the HadCRUT4 global mean temperature data. In 2012 the Hadley Centre completed their work on estimating the uncertainty in temperature in their reconstruction and published the results online [LINK] . The full text of Colin’s paper is available online [LINK ]. More detail on Colin Morice’s work is presented in a related post [LINK] . The Morice uncertainties are presented graphically as animation in the two brief video presentations in Figure 5. The top frame compares temperatures and the bottom frame compares their decadal trends. The data for these graphics are derived from the Morice variances in a Monte Carlo simulation that creates four different data series possible under the uncertainty described by the variance. The variance used in the Monte Carlo simulation is derived from the 95% confidence intervals reported by Colin Morice. Figure 6 is a graphical display of the correlations among the four Monte Carlo series for both temperature and decadal trends. The correlations displayed graphically in Figure 2, Figure 4, and Figure 6 are tabulated in Figure 7. The tabulation shows as follows:
8. Figure 7 Item#1: A near perfect correlation is found between BRK and GIS in both the temperature and decadal trend time series. Item#2: The comparison of nearly identical series BRK and GIS with HAD reveals lower correlations of both BRK and GIS against HAD in both the temperature and decadal trend time series with significant seasonal differences. The correlations are low in summer and high in winter for both temperature and decadal trends. Item#3: Comparison of the three surface temperature reconstruction (BRK GUS HAD) with satellite data for lower troposphere temperature (UAH). Here the summer correlations for both temperature and decadal trends are lower than in Item#1 and Item#2 although the winter correlations are strong. Item#4: Correlations among the four different Monte Carlo simulations of HAD with the Morice uncertainties show weaker correlations than the comparison of different measurement methods.
9. CONCLUSION: Differences among global mean temperature sources and whether surface reconstructions or satellite lower troposphere temperature measurements are within surface reconstruction uncertainty reported for the HAD by Morice. We therefore conclude that no significant difference is found among these four datasets that can be ascribed to measurement methods. An oddity of the findings of this study is the extreme correspondence between the BRK (Berkeley Earth) and GIS (Goddard Institute of Space Sciences) reconstructions.

[LINK TO HOME PAGE]