Optimal Averaging of Seasonal Sea Surface Temperatures and Associated Confidence Intervals (1860–1989)

View More View Less
  • 1 Climate Analysis Center, National Meteorological Center, NWS, NOAA, Washington, D.C.
  • | 2 Coupled Model Project, National Meteorological Center, NWS, NOAA, Washington, D.C.
  • | 3 Climate Analysis Center, National Meteorological Center, NWS, NOAA, Washington, D.C.
© Get Permissions
Full access

Abstract

Optimal averaging (OA) is used to compute the area-average seasonal sea surface temperature (SST) for a variety of areas from 1860 to 1989. The OA gives statistically improved averages and the objective assignment of confidence intervals to these averages. The ability to assign confidence intervals is the main advantage of this method. Confidence intervals reflect how densely and uniformly an area is sampled during the averaging season. For the global average, the early part of the record (1860–1890) and the times of the two world wars have largest uncertainties. Analysis of OA-based uncertainty estimates shows that before 1930 sampling in the Southern Hemisphere was as good as it was in the Northern Hemisphere. From about 1930 to 1950, uncertainties decreased in both hemispheres, but the magnitude of the Northern Hemisphere uncertainties reduced more and remained smaller. After the early 1950s uncertainties were relatively constant in both hemispheres, indicating that sampling was relatively consistent over the period. During the two world wars, increased uncertainties reflected the sampling decreases over all the oceans, with the biggest decreases south of 40°S. The OA global SST anomalies are virtually identical to estimates of global SST anomalies computed using simpler methods, when the same data corrections are applied. When data are plentiful over an area there is no clear advantage of the OA over simpler methods. The major advantage of the OA over the simpler methods is the accompanying error estimates.

The OA analysis suggests that SST anomalies were not significantly different from 0 from 1860 to 1900. This result is heavily influenced by the choice of the data corrections applied before the 1950s. Global anomalies are also near zero from 1940 until the mid-1970s. The OA analysis suggests that negative anomalies dominated the period from the early 1900s through the 1930s although the uncertainties are quite large during and immediately following World War I. Finally, the OA analysis shows significant positive global SST anomalies beginning in the late 1970s. The SST anomalies in the Indian Ocean and Southern Ocean poleward of 20°S make the strongest contributions to the positive global anomalies observed since the late 1970s. In contrast to the more recent period, the SST anomalies in the period from the early 1900s through 1940 were dominated by the anomalies in the Northern Hemisphere poleward of 20°N.

Abstract

Optimal averaging (OA) is used to compute the area-average seasonal sea surface temperature (SST) for a variety of areas from 1860 to 1989. The OA gives statistically improved averages and the objective assignment of confidence intervals to these averages. The ability to assign confidence intervals is the main advantage of this method. Confidence intervals reflect how densely and uniformly an area is sampled during the averaging season. For the global average, the early part of the record (1860–1890) and the times of the two world wars have largest uncertainties. Analysis of OA-based uncertainty estimates shows that before 1930 sampling in the Southern Hemisphere was as good as it was in the Northern Hemisphere. From about 1930 to 1950, uncertainties decreased in both hemispheres, but the magnitude of the Northern Hemisphere uncertainties reduced more and remained smaller. After the early 1950s uncertainties were relatively constant in both hemispheres, indicating that sampling was relatively consistent over the period. During the two world wars, increased uncertainties reflected the sampling decreases over all the oceans, with the biggest decreases south of 40°S. The OA global SST anomalies are virtually identical to estimates of global SST anomalies computed using simpler methods, when the same data corrections are applied. When data are plentiful over an area there is no clear advantage of the OA over simpler methods. The major advantage of the OA over the simpler methods is the accompanying error estimates.

The OA analysis suggests that SST anomalies were not significantly different from 0 from 1860 to 1900. This result is heavily influenced by the choice of the data corrections applied before the 1950s. Global anomalies are also near zero from 1940 until the mid-1970s. The OA analysis suggests that negative anomalies dominated the period from the early 1900s through the 1930s although the uncertainties are quite large during and immediately following World War I. Finally, the OA analysis shows significant positive global SST anomalies beginning in the late 1970s. The SST anomalies in the Indian Ocean and Southern Ocean poleward of 20°S make the strongest contributions to the positive global anomalies observed since the late 1970s. In contrast to the more recent period, the SST anomalies in the period from the early 1900s through 1940 were dominated by the anomalies in the Northern Hemisphere poleward of 20°N.

Save