Midlatitude North America

We limit the study area for North America to the latitudinal band between the Tropic of Cancer and 53° N; north of this band, population and station density becomes sparse, a fact manifested in CRU2 as well as GHCN V2 data gaps

Figure 5.4. Same as Figure 5.1, but for North America; (a) North American average temperature anomaly relative to the base period 1950-99 from CRU2 observations and the CGCM historical and future scenarios. (b) H&CSI displayed as percentage of North American grid points with summer temperatures above the 90th or below the 10th percentiles of their local summer 1950-99 climatology. HSI (CSI) is displayed in positive (negative) values. By definition, during the base period (delineated by broken orange lines), the mean values are 10% of the area experiencing unusually hot or cold summers. For color version, see plate section.

Figure 5.4. Same as Figure 5.1, but for North America; (a) North American average temperature anomaly relative to the base period 1950-99 from CRU2 observations and the CGCM historical and future scenarios. (b) H&CSI displayed as percentage of North American grid points with summer temperatures above the 90th or below the 10th percentiles of their local summer 1950-99 climatology. HSI (CSI) is displayed in positive (negative) values. By definition, during the base period (delineated by broken orange lines), the mean values are 10% of the area experiencing unusually hot or cold summers. For color version, see plate section.

over Canada. North American average temperature and H&CSIs derived from observations and the CGCM are displayed in Figure 5.4.

The first three decades of the twentieth century were the coldest on record, with the summers of 1903, 1907, and 1915 experiencing widespread cold anomalies affecting at least one-half of North America, a scale not reached again until 1992 - the largest summer temperature extreme of the second half of the century. We note that these, as well as smaller-scale recent cold outbreaks (e.g., 1976, 1982, 1993, and 2004), were associated with locally wet summers (result not shown). On the other hand, the 1930s appear to have experienced the warmest and most consistently warm conditions on record, at least until very recently; 7 of 11 (1930-40) summers were more extensively warm (none cool) than expected, and all 11 were warmer than average (see Figure 5.6a for the spatial pattern). The summers of 1934, 1936, and 1937 experienced by far the most severely extensive hot spells of the century (more than 30% of North America was under hot spell conditions). These summers were also exceptionally dry (result not shown for the individual summers; see Figure 5.6b for the 11-year average precipitation anomaly). No summers during the 1930s saw cold outbreaks that even remotely approached their expected spatial extent of 10%.

The 1940s were generally and mildly cool. Temperatures in the 1950s were unusually variable, but the warm summertime temperature anomalies associated with the 1950s drought,5 although large, were not as spatially extensive, temporally persistent, or exclusive as in the 1930s; they were balanced out by comparably large cold anomalies, a fact reflected in near-zero average temperature anomalies. The 1960s and 1970s saw less variability, with predominantly cool conditions. Temperatures in North America became more variable in the 1980s and 1990s while warming into the start of the twenty-first century. In the midst of general warmth, a cold spell covered 50% of midlatitude North America in 1992, the first time such widespread cold was observed since 1907. The cold summer of 1992 was followed by the mostly cool 1993, both perhaps related to the Mount Pinatubo volcanic eruption6 (Robock, 2000). But even in the cool summer of 1993 (CSI = 23%), the expected area was hot (HSI = 11%) and the decade that followed (1994-2003) was akin to the 1930s, with 9 out of 10 summers exceeding the mean extent of heat and 3 summers (1998,2002, and 2003) with hot spell conditions covering more than 30% of the continent for the first time since the 1930s. The last observed summer (2004) was both anomalously warm (HSI = 22%) and cold (CSI = 27%). The recent midwestern heat waves of 1995 and 1999 each resulted in HSI values of under 30%.

The questions as to what global climate conditions are conducive to intense summer heat outbreaks over North America and whether these conditions are reproduced by the model can be addressed by considering maps of correlation coefficients between HSI and local Ta2m over the globe (Fig. 5.5). In Reanalysis and the CGCM, spatially extensive hot spells over North America

5 The bulk of the 1950s warmth was coincident with drought, which was mostly concentrated in the fall and winter seasons.

6 Both these summers, 1992 and 1993, were also anomalously wet over the regions corresponding to the largest observed cooling: Great Plains in 1992, and the northern plains and northwestern United States in 1993 (result not shown). The summer of 1993 was considerably wetter than that of 1992.

Mid Latitude Cyclone Over The Usa

Figure 5.5. Correlation coefficient between North American HSI and local JJA Ta2m over the globe in (a) Reanalysis, 1948-2003, and (b) the CGCM commit run, 2000-99. For color version, see plate section.

Figure 5.5. Correlation coefficient between North American HSI and local JJA Ta2m over the globe in (a) Reanalysis, 1948-2003, and (b) the CGCM commit run, 2000-99. For color version, see plate section.

tend to organize preferentially around the midwestern United States,7 and they tend to be associated with cool temperatures in the northeast midlatitude and tropical eastern Pacific, warm temperatures in the subtropical western Pacific, and a general tendency toward warm temperatures in the tropical and northwestern Atlantic. Other patterns apparent in Reanalysis are characterized by

7 Principal components analysis (PCA) can be used to find coherent patterns of variability that optimally explain the variance of a spatially correlated field; e.g., summer temperature. Principal components analysis applied to CRU2 (as well as to Reanalysis and CGCM) data results in a similar Midwest/Great Plains pattern of summer heat variability (results not shown). Observed H&CSIs are correlated with this main PC mode (correlation = 0.9). This leading mode explains 43% of summer temperature variance (PC2 explains 18%). In CRU2, large peaks (dips) in HSI (CSI) tend to historically line up with those of PC1, indicating that hot (cold) summers tend to have spatially consistent temperature anomalies. The 1930s heat closely followed this pattern. Most other anomalous summers throughout the record did too. Recently, however, the spatial pattern of warming exhibited a rather different structure (see below).

mostly weak positive correlations in patches around the globe, reflecting the general recent global warming. Except for the propensity towards positive correlations dictated by the observed warming trend, the main patterns are reproduced by the model's ''commit'' climate run, which is stationary by design. The association between HSI and the preferred heat wave location appears to be weakened in Reanalysis, probably because the recent warming over North America did not manifest itself in the preferred heat wave region of the midwestern and north-central United States (see below). Other differences may be due in part to the larger noise level in the shorter reanalyzed climatic sample. As an aside, we have also examined multidecadal changes in global correlation patterns. Modeled correlations computed for consecutive 50-year periods by using the model's historical as well as B1 and A2 runs (figures not shown) plainly resemble the main patterns of Figure 5.5. They also indicate a consistent change in the characteristic pattern of the HSI amounting to a progressive migration of North American heat outbreaks toward the north of the study region. In summary, North American hot summers tend to have a preferred spatial signature that is typically expressed in conjunction with specific sea surface temperature (SST) patterns, and the model appears able to realistically reproduce the observed dynamic structure.

The marine temperature patterns associated with extensive North American summer heat spells are analogous to the anomalous sea surface temperature ''forcings'' for the 1930s Dust Bowl drought as outlined by Schubert et al. (2004). It is also well known that large-scale summer heat can be caused or exacerbated by atmospheric drought through soil moisture deficit (see Alfaro et al., 2006). This was the case in the hot and dry 1930s, when precipitation, especially in summer, was scarce8 (see Figure 5.6a, b). Although more extreme and persistent than other hot summers on record, the spatial structure of the 1930s heat followed the typical pattern of hot summers over North America (result of PCA not shown). However, the most recent period of summer heat observed over North America, although on the scale approaching that of the 1930s, was not associated either with prolonged large-scale drought or with an anomalously cold tropical Pacific.

In fact, the most recent warm period, 1994-2004, was very much unlike the 1930s (Figure 5.6c). The central and eastern United States were mostly wet over this period9 (Figure 5.6d). Despite this recent wetness, the central and eastern United States have not cooled as would be naturally expected. Such expectation

8 Drought reconstructions show that the 1930s Dust Bowl drought was the most severe and widespread such event to strike the United States since 1700 (Cook et al., 1999).

9 In fact, the entire country has been generally wet over the last quarter century.

-120 -100 -80
-120 -100

-120

-100

-120

-100

-120

-100

-120

-100

Figure 5.6. Summer temperature (a, c) and precipitation (b, d) anomalies for two 11-year periods: 1930-1940 (a, b) and the most recent period of record, 1994-2004 (c, d). Notice the reversed color scale on the precipitation plots. Canadian precipitation records suffer from missing data at the end of the record, accounting for the visible discontinuity along the US-Canada border in (d). For color version, see plate section.

is supported by Figure 5.7a, which shows observed local correlations between summer temperatures and precipitation. Meanwhile, much of the recent warming over midlatitude North America has been due to warmer summers all around this central and eastern US wet spot (Figure 5.6c,d). Recent summer warmth over California and the rest of the mountainous West is probably related to the strong warming observed in the West during winter and spring, which resulted in changes in the snow-to-rain ratio (Knowles et al, 2006) as well as the spring's earlier arrival and related changes in surface hydrology (Cayan et al., 2001), drying the soil in summer without appreciable changes in precipitation amounts. These same changes in summer temperatures and hydrology have also resulted in increased wildfire activity over the western mountains (Westerling et al., 2006). The western summer warming (Figure 5.6c) is reflected in increasing values of the North American HSI, but this pattern is generally atypical of extensive North American summer heat.

Strong and extensive Canadian warming, apparent in Figure 5.6c, did not strongly affect the earlier results because only far southern Canadian data, for

Figure 5.7. Correlation coefficient between local June-August average temperature and precipitation in (a) observations (GHCN V2 and CRU2 evaluated over the period of record, 1900-2004, or slightly shorter based on data availability for the individual grids) and (b) the model (the commit run, 100 years). For color version, see plate section.

Figure 5.7. Correlation coefficient between local June-August average temperature and precipitation in (a) observations (GHCN V2 and CRU2 evaluated over the period of record, 1900-2004, or slightly shorter based on data availability for the individual grids) and (b) the model (the commit run, 100 years). For color version, see plate section.

reasons of data quality given above, were included in the North American index calculation. Had more extensive Canadian data been included, the recent warming would have appeared larger. This observation is consistent with the model result projecting heat waves to spread progressively towards the north (not shown).

Was this article helpful?

0 0

Post a comment