Adjusting HadCRUT for choice of SST data set and UAH polar data

[skierinvermont]

There has been much discussion on the blogosphere as to the divergence of HadCRUT and GISS 1998-2010 and especially 2002-2010.

In response to this, a while ago I developed global temperature indexes using GISS 60S60N and HadCRUT 60S60N data and UAH polar data to show that when adding in UAH polar data to HadCRUT, it largely reconciles the two. As others pointed out however, some differences remain, especially 2002-2010 where GISS 60S60N warmed faster than HadCRUT60S60N. After quickly demonstrating the result of polar substitution with UAH data as I have previously shown, I will demonstrate that much of the remaining difference between GISS and HadCRUT is due to the different SST data sources they use. GISS uses the Reynolds satellite SST analysis (OISSTv2) and HadCRUT uses HadSST2.

1980-2010:

GISS6060+UAH polar is not significantly different from the original GISS, indicating GISS's polar extrapolations are accurate enough in the long-run. Short term differences exist such as the ones below, but those short-term differences have evened out in the long-run.

1997-2010 trends:

GISS: .143C/decade

GISS6060+UAH polar: .109C/decade

HadCRUT6060+UAH polar: .080C/decade

HadCRUT: .038C/decade

Thus, for 1997-2010 inclusion of UAH polar data brings HadCRUT 40% closer to GISS and GISS 32% closer to HadCRUT. The remaining difference of .029C/decade is 28% of the original difference. Thus 72% of the difference between HadCRUT and GISS 1997-2010 is due to the poles. GISS has warmed too quickly, while HadCRUT has warmed too little, assuming UAH is correct. However, 28% of the difference remains due to 60S60N.

2002-2010 trends:

GISS: .037C/decade

GISS6060+UAH polar: -.016C/decade

HadCRUT6060+UAH polar: -.063C/decade

HadCRUT: -.064C/decade

Substitution of UAH for the poles brings GISS down 52% to HadCRUT and doesn't increase HadCRUT at all. 52% of the difference is explained by the polar extrapolations, but 48% remains due to differences 60S60N.

Thus I went searching for the remaining differences in the 60S60N region.

GISS uses OISSTv2 SST data. HadCRUT uses HadSST2. Using the KNMI climate explorer I found the monthly anomalies for each source in the latitude band 60S-60N, imported into excel, and computed the following trends. The GISS SST data source has a more positive trend for both periods.

1997-2010:

OISSTv2: .010C/decade

HadSST2: .002C/decade

difference: .008C/decade

2002-2010:

OISSTv2: -.070C/decade

HadSST2: -.030C/decade

difference: .04C/decade

Assuming that the ocean between 60S and 60N composes 60% of global surface area I found substituting the GISS SST data source into HadCRUT would raise the HadCRUT trend an additional:

1997-2010: .005C/decade

2002-2010: .024C/decade

Thus explaining an additional percentage of the difference between the two:

1997-2010: 5%

2002-2010: 24%

Thus, differences in the two SST data sources for HadCRUT and GISS explain 24% of the divergence 2002-2010. 52% is due to polar extrapolation. Very little difference between the two remains for either the 1997-2010 period or the 2002-2010 period.

It is astonishing to me that after the countless hours certain blogs have spent analyzing the differences between HadCRUT and GISS, not once, has any of them brought up the fact that the GISS SST data source (which is satellite derived and commonly known as the Reynolds SST analysis) has warmed faster than the HadCRUT SST source (HadSST2).

[skierinvermont]

bump

I would think this would get some posts considering it concerns much discussion that has taken place here and elsewhere. SSTs constitute 70% of the data used in GISS and HadCRUT and yet almost no attention is paid to the fact that the GISS SST source has warmed faster 2002-2010 than the HadCRUT source, thus explaining much of the much discussed divergence.

Perhaps some would prefer to believe the myth of the magical warm bias of extrapolation than face the fact that satellite SST data is responsible for the divergence?

[skierinvermont]

Perhaps some would prefer to believe the myth of the magical warm bias of extrapolation than face the fact that satellite SST data is responsible for the divergence?

The response thus far certainly is giving that impression.