AN ANALYSIS OF WARM SEASON NORTHERN HEMISPHERE
UPPER AIR OBSERVATIONS FROM 1957 TO 1995

Glenn D. Carrin

WSO Shreveport, LA (formerly of WSFO Albuquerque)

It has been noted that tropospheric temperatures and near-surface humidity over the tropical western Pacific increased from 1974 to 1988 (Gutzler, 1992). Interestingly, the tropospheric warming was characterized by decreasing lapse rates. Gutzler assumed that the stabilizing effect of these decreasing lapse rates would be offset by the destabilizing influence of the increasing near-surface humidity over the warm tropical waters. However, Gutzler concluded his study by identifying its obvious limitations of the non-global spatial coverage, and the short period of record which was limited to radiosonde observations taken since the last major change in the method employed to measure and report humidity by U.S.-operated stations.

Several publications have dealt with the utility of radiosonde climate archives as well as the changes that have taken place in instrumental accuracy and biases during the existence of the radiosonde program (cited by Gutzler). Even so, an analysis of radiosonde observations taken over a period of several decades from sites across the northern hemisphere may provide useful albeit generalized information on atmospheric trends. In this study, temperature and thickness data from across the northern hemisphere are analyzed in an attempt to estimate the tropospheric trends of the past four decades.

1. Data

Upper air data from over 90 stations across North America were obtained from the "Radiosonde Data of North America 1946-1995" CD-ROMs (1996). The data consisted of 00Z observations taken from June 1 through August 31 for the years 1957 through 1995. The available years of 1946-1956 were excluded for two reasons. First, data were not available from many of the sites used in this study during those years. Second, these late-day observations were usually time-tagged between 02Z and 04Z, presumably enough of a time discrepancy to prevent them from reliably correlating with the 00Z data being studied. Moreover, the quality of the 00Z data varied, being divided into the four categories explained below.

The majority of sites fell into the first category. Each of these sites recorded a nearly complete set of data for the time period under review. A high percentage of data from each summer and for each year was available, with no more than a few flights missing in any given summer.

About two dozen sites comprised the second category. One to five years of summer data were missing from each site. The missing years were typically scattered throughout the data set, and posed few problems to the trend calculations because a five year running mean smoothed out the missing summers.

The third category contained a few sites in which some years had less than fifty observations during a given summer. This primarily occurred with the Caribbean and Mexican sites, and the years affected were typically from the mid-1980s onward. Attempts to overcome these deficiencies were more challenging, so a three-tiered approach was employed. To illustrate, assume that stations A, B, and C each recorded only forty upper air observations during a particular summer. The observations from station A were evenly distributed among the three months, therefore, the observations were retained unchanged. At station B, twenty observations were recorded in June, ten in July, and ten in August. Because of this, ten of the June observations were randomly deleted so that the remaining observations would be evenly distributed among the three summer months. This edited set of data would be retained for use in the study. However, the number of observations at station C for June, July, and August were distributed as thirty, eight, and two respectively. In this case, the observations were considered to be too unevenly distributed to edit, and the entire data set for that summer at station C was eliminated.

The fourth category contains eight sites which were derived by combining upper air observations from separate but nearby sites. For example, the east Texas (ETX) data set was derived from observations recorded at Shreveport, Louisiana and Longview, Texas.  The derived sites are listed at the bottom of Table One.

By a conservative estimate, over one quarter of a million upper air observations were included in this study. A FORTRAN program was used to calculate each location’s average 500 MB temperature (T5), 700 MB temperature (T7), and 700-500 MB thickness (k) for each summer. This reduced the quarter-million observations to approximately 10,300 annual summer parameters.

Using Quattro Pro, five-year running means of T5, T7, and k were calculated for each site in order to smooth the data. Next, a linear fit calculation was applied to the five year running mean plot. To determine estimates of the temperature and thickness trends at each location, the change in temperature at both levels (DT5, DT7) and the change in thickness (Dk) were calculated by subtracting the beginning value of each best fit line from the ending value.

Table One lists the final trend estimates, including trends in the lapse rates from 700 millibars to 500 millibars. This layer stability estimate was determined using temperature data alone. Simply stated,

DLR = DT5 - DT7

where positive values indicate decreasing lapse rates (stabilization) and negative values indicate increasing lapse rates (destabilization). As a result of these calculations, the initial quarter-million plus observations were reduced to 352 final values, four for each of the final 88 sites.

At 700 MB, warming was indicated at 71 of the 88 sites, or 80.7% of the sites. The 71 stations warmed by an average of +0.611 degrees Celsius. The remaining 17 stations cooled by an average of -0.198 degrees Celsius. The average 700 MB North American temperature change was +0.454 degrees Celsius. See Table One.

At 500 MB, warming was indicated at 83 of the 88 sites, or 94.3% of the sites. The 83 stations warmed by an average of +0.716 degrees Celsius. The remaining 5 stations cooled by an average of -0.094 degrees Celsius. The average 500 MB North American temperature change was +0.670 degrees Celsius. See Table One.

The 700/500 MB thickness data indicated increases at 79 of the 88 sites, or 89.8% of the sites. The average increase was +5.990 meters. The average decrease was -1.501 meters. Overall, the average North American 700/500 MB thickness change was +5.224 meters. See Table One and Figure One.

The lapse-rate trend estimates reveal that stabilization occurred at 69 of the 88 sites, or 78.4% of the sites. Stabilization is indicated when temperatures at 500 MB either warm more or cool less than those at 700 MB. Of the 69 stabilizing sites, 51 indicated a stabilization greater than or equal to two-tenths of a degree Celsius. Of those 51, ten stabilized by more than one half of a degree Celsius.

Considering the remaining sites, three showed no change while sixteen sites (18.2% of the total sites) trended toward destabilization. The largest area of apparent destabilization (shown by increasing lapse rates) extended from the Caribbean Sea through the central Gulf of Mexico to Louisiana and eastern Texas, then northward through Arkansas into Illinois. A second area was indicated over the coast of the Carolinas. A third area of destabilization was indicated over the coastal regions of Alaska. See Table One and Figure Two.

Thickness trends and temperature trends correlated strongly, as one would expect. Warming was indicated at most sites. The overall warming trend, along with the significant warming indicated over Alaska, seemingly corresponds with climate model forecasts and glacial retreat observations which were presented by Mitchell, et al (1990).

Nearly 80% of the sites involved in this study demonstrated stabilization similar to Gutzler’s findings for the western tropical Pacific (1992). Furthermore, the sixteen destabilized sites were notably contained within distinct regions rather than being randomly distributed about the hemisphere. Hence, we are provided with evidence that a stabilizing phenomenon characterized by decreasing mid-level lapse rates has occurred on a more globally spatial scale. However, a question for further study is whether or not this larger area (predominantly continental in climate) has experienced an offsetting increase in near-surface humidity similar to that of the tropical western Pacific.

Acknowledgments. My thanks go to Deirdre Kann (Science and Operations Officer of NWSFO Albuquerque), who provided the CD-ROMs, FORTRAN programming, and proofreading. I also thank David S. Gutzler (UNM) for providing copies of his research for reference.

Gutzler, D.S., Climatic Variability of Temperature and Humidity Over the Tropical Western Pacific, Geophysical Research Letters, vol. 19, no. 15, 1595-1598, 1992.

Mitchell, J.F.B., et al, Equilibrium Climate Change - and its Implications for the Future, Climate Change: The IPCC Scientific Assessment, Cambridge University Press, 131-172, 1990.

Quattro Pro, Version 5.5, Novell, Inc., 1995.

Radiosonde Data of North America, 1946-1995, August 1996 update, Version 1.0. Forecast Systems Laboratory, Boulder CO, and National Climatic Data Center, Asheville NC.