Summer 2008 in North America - post-mortem analysis
Temperature
In the summer (JJA) of 2008, surface air temperature (SAT) averaged over the contiguous United States was 72.86°F, which was 1.01°F lower than during the summer of 2007, but 0.56°F higher than the 1971-2000 average (Fig. 1). Both the direction and magnitude of the temperature change were anticipated with a relatively high degree of confidence (see our forecast issued in March 2008 and its update issued in April 2008).
The spatial distribution of SAT anomalies during the summer (JJA) of 2008 is presented in Fig. 2. A center of anomalously warm temperatures is located in eastern Canada. Average summer temperatures in that region were up to 4°C above the 1971-2000 average. As for the contiguous United States, positive temperature anomalies were observed in the West, reaching 1.5°C in California and Nevada. Colder than normal summer temperatures were observed in the Great Lakes region and in the area from the Central Plains to New Mexico.
During the winter of 2008, the dominant storm track was the one with the jet stream steering frequent storms from the Pacific Northwest toward Utah and Colorado and then northeast, toward the Great Lakes (see the post-mortem analysis). This pattern continued through the spring and early summer bringing anomalously cold air to the northern United States, from the Pacific Northwest to the Midwest. As the warm season progressed, the storm track was slowly moving northward, pushed by subtropical high pressure belt. By July 2008, an anomalously high pressure center firmly settled over the western U.S., resulting in a heat wave. Denver, CO, for example, set a record of 24 consecutive days with 90°F-plus temperature, obliterating the previous record of 18 set in 1874 and tied in 1901.
The distribution of SAT anomalies during the central summer month of July (Fig. 3) was the closest to our forecast map. In August, heavy monsoonal rains brought some cooling to the Central Plains. While temperatures in most western states were again above normal in August, temperatures across much of the eastern half of the U.S. were below normal. In particular, a persistent low 500-hPa height anomaly (not shown) and colder than average temperatures were observed over the Great Lakes region throughout most of the summer (Fig. 2).
Temperature anomalies over North America are often associated with sea-surface temperatures (SST) in the North Pacific, particularly in those years when the ENSO signal is weak. The Pacific Decadal Oscillation (PDO) represents a primary pattern of SST anomaly distribution in the North Pacific. In summer 2008, the PDO index dropped precipitously to the level not seen since 1955 (Fig. 4).
Unlike the winter season, however, the summer PDO pattern exerts surprisingly little influence on North American temperature. Correlation coefficients between the summer PDO index and North American SATs usually do not exceed |0.4|, with some exceptions. Rules 52 and 264, which describe relationships between SST patterns in the North Pacific and SAT patterns over North America, worked well this summer.
The North Atlantic Oscillation (NAO) index was also sharply lower this summer. The index dropped in April and remained negative through August (the last month available). A negative NAO index is often associated with positive temperature anomalies in the Southeastern U.S. (rule 419). This summer, however, temperatures in the Southeast were close to average. Apparently other factors, such as SST anomaly patterns in the North Pacific and precipitation pattern over North America (rule 294) offset the NAO influence.
In assessing the accuracy of seasonal climate forecasts it is important to emphasize the difference between the NCEP/NCAR Reanalysis and data from individual stations. For example, according to the data from GISS, the mean summer (JJA) SAT anomaly in 2008 at Dallas/Fort Worth Airport (WBAN 03927, 32°9'N / 97°0'W) was 2.0°C (base period: 1971-2000). After the homogeneity adjustment by GISS, this anomaly was reduced to 1.30°C. Interestingly, the amount of adjustment was steadily increasing from -0.2°C in 1985 to +0.7°C in 2004 and then remained at this level in more recent years. But even after the adjustment, the SAT anomaly of 1.3°C is hard to find anywhere in the vicinity of Dallas (Fig. 2), where SAT anomalies according to the Reanalysis data are less than 0.5°C.
Even more striking examples can be found in the Great Lakes region, where temperatures this summer were markedly below the 1971-2000 average (Fig. 2). Thus, according to the Reanalysis data, the SAT anomaly at the grid point 45°N, 80°W (closest to Toronto, Canada) was -0.99 std (standard deviations). However, according to the data available at the Chicago Mercantile Exchange (and provided there by EarthSat), the summer SAT anomaly for Toronto in 2008 was +0.72 std, that is, the difference between the two datasets was 1.71 std! It is not clear what adjustments (if any) were used by EarthSat. As shown in Fig. 6, the EarthSat data shows a strong positive trend, whereas the Reanalysis data exhibits a weak negative trend. When the trends are removed, the correlation coefficient between the two time series is 0.95.
Our forecasts tend to correspond better to the Reanalysis data. When translating our forecast maps into weather derivatives forecast for individual stations, an adjustment for the difference between the Reanalysis and station data may be applied. But this adjustment is not always easy. For example, the difference between the station (GISS) data for Houston, TX, and Reanalysis data for the nearby grid randomly varies between -1.5 to +1.5 std, and the correlation between the two time series is only 0.6. Various problems with station data are often discussed at ClimateAudit and Climate Science blogs.