>Figure 1 provides a comparison of GISTemp smoothing radius data for January, 1978 through April, 2008: the blue curve with 250km radius smoothing, the red with 1200km.
Figure 1: GISTEMP Smoothing Radius Comparison – 250km (Blue) vs 1200km (Red) – January, 1978 to April 2008 with Linear Trends
The 1200km radius curve is based on the standard data distributed through the GISS website. Source here:
The data for the 250km radius curve was created using the GISS map page, the only way I could find to acquire the data. (Do not attempt this at home if you have something worthwhile to do.)
Data Sources: Land: GISS analysis & Ocean: Hadl/Rey 2
Map Type: Anomalies
Mean Period: Each month
Time Interval: Begin: Each year & End: Same year
Base Period: Begin: 1951 & End: 1980
Smoothing Radius: 250km
Projection Type: Regular
In the upper right-hand corner of each map is the monthly anomaly data. I input those monthly values into a spreadsheet by hand. (Again, a tedious task.) Sample Anomaly Map (Mercator and Polar Projections) and Zonal Means Plot outputs follow as Figures 2a, 2b, and 2c.
Figure 2a: GISS Global Temperature Anomaly Map – April 2008 – 250km Radius Smoothing – Mercator Projection
That thick red line across the bottom across the bottom of Figure 2a is not a lower border. It is the anomaly data at the South Pole stretched by the Mercator projection. Refer to Figure 2b.
Figure 2b: GISS Global Temperature Anomaly Map – April 2008 – 250km Radius Smoothing – Polar Projections
Figure 2c: GISS Global Temperature Zonal Means Plot – April 2008 – 250km Radius Smoothing
Don’t let the 3 deg C anomaly at the South Pole concern you. The temperature forecast this week for the Amundsen-Scott South Pole Station is in the range of -40 to -60 deg C.
Back to the comparison of radius smoothing: Figure 3 illustrates the same data as Figure 1, but it’s been smoothed with a 7-month filter. Eliminating the month-to-month noise helps illustrate the amplification inherent in the 1200km smoothing.
Figure 3: GISTEMP Smoothing Radius Comparison – 250km (Blue) vs 1200km (Red) – January, 1978 to April 2008 with Linear Trends – Smoothed With 7-Month Filter
Subtracting the 250km radius data from the 1200km provides another view of the ever increasing difference between the two data sets. Refer to Figure 4.
Figure 4: GISTEMP Smoothing Radius Comparison – 1200km Radius MINUS 250km Radius – January, 1978 to April 2008 with Polynomial Trend – Smoothed With 7-Month Filter
Based on the polynomial curve, it appears the divergence of the 1200km radius data began about 1989-1990. This is near the time that the number of reporting climate stations began to decrease, reducing the land area coverage. Refer to the gray areas in Figures 2a and 2b. Does this indicate the amplification by the 1200km radius smoothing is more a function of the reduced land area coverage than of polar amplification? Or did the high frequency of El Ninos from 1990 to 2007 cause Polar amplification and excessive high-latitude warming in the Northern Hemisphere that hasn’t been suppressed yet? Or both?
COMPARISON WITH HADLEY CENTRE DATA
Figure 5 provides a comparison of the GISTemp data (250km Radius Smoothing) with the Hadley Centre global temperature anomaly data, HADCrut3GL. They look like twins.
Figure 5: Global Temperature Anomaly Data Comparison – GISTemp With 250km Radius Smoothing versus HADCrut3GL – January 1978 to April 2008
The difference between HADCrut3GL and GISTemp with the reduced 250km smoothing, Figure 6, appears to be declining, but if the step change synched with the 97/98 El Nino was removed, the difference between the two would be relatively flat.
Figure 6: Global Temperature Anomaly Data Comparison – GISTemp With 250km Radius Smoothing MINUS HADCrut3GL – January 1978 to April 2008
The step change in Figure 6 is consistent with the step change that results when Smith and Reynolds SST data (ERSST.v2) is subtracted from Hadley Centre data (HADSST2GL). Refer to Figure 7. The “Hadl/Rey 2” data employed by GISS must be comprised of Hadley Centre data prior to a given transition date, Smith and Reynolds after, where the transition date is prior to 1997.
Figure 7: Step Change Evident in HADSST2GL Minus ERSST.v2 – January 1950 to November 2007