|Global annual mean temperature anomalies as measured by surface data (black bars) and the MSU satellite unit (gray bars) using the MSU-2R technique. Temperatures on the vertical axis are times 100 degrees C. (Illustration from Hurrell and Trenberth, "Satellite versus surface estimates of air temperature since 1979," Journal of Climate 9(9), 1996.)|
A flurry of papers over the past several years in the Journal of Climate, Climatic Change, and other literature has tried to pin down the causes of the MSU/surface disagreement. At last month's meeting of the American Meteorological Society (AMS) in Long Beach, California, an afternoon session was devoted to the discrepancies. The lively session provided a thorough airing of the issues at hand, but no clear resolution.
Two key players in the debate are NCAR's James Hurrell and Kevin Trenberth. They have collaborated on a series of recent papers (including two presented at AMS) aimed at reconciling the patterns of difference between the surface and the MSU data. "We need both surface and MSU records to get a proper perspective on what's going on," says Trenberth. "Although the MSU data are excellent for many purposes, we think there are some substantive problems with the trends as depicted by MSU."
John Christy (University of Alabama at Huntsville) and Roy Spencer (NASA) have led the development of temperature-trend retrieval from MSU data. Both presented papers at AMS on their most recent work. Christy maintains that the MSU data are solid. "Since the MSU measures temperatures through a deep layer of the atmosphere, the only proper comparison is with temperature profiles from radiosondes. Our extensive comparisons show no significant disagreements." Christy believes the main difference between surface and MSU readings lies in the very real distinction between surface and tropospheric temperature, an area now receiving increased scrutiny.
In the MSUs' favor is their strong correlation with global radiosonde records. Christy notes the minimal difference in decadal temperature trends between MSU-2R and radiosonde data across North America, the Arctic, and the tropics. In all three regions, Christy has found a difference of 0.027degreesC or less per decade between the radiosonde and MSU trends.
Although it has longevity on its side, the earth's surface temperature record has well-documented flaws. Tom Karl (National Climatic Data Center) found in 1994 that about 10% of the half-degree C warming seen at the surface this century could be due to error induced by the uneven distribution of weather stations across land and by their absence over the oceans, where sea-surface temperatures (SSTs) are routinely used as a substitute for surface air temperature. Karl believes the error could be proportionately greater over shorter intervals of a decade or two.
Adding to the murkiness, a number of recent earth-system upheavals have had global impact; in particular, the major El Niño/Southern Oscillation (ENSO) event of 1982-83 and volcanoes El Chich—n (1982) and Mt. Pinatubo (1991). Philip Jones (University of East Anglia) estimates that the effect of these transitory events could be twice as large in the troposphere as at the surface, with El Niño warming and volcanic cooling both amplified aloft.
Correction factors must be applied to each MSU to account for different satellite orbits and orbital shifts over time. Moreover, as much as 20% of the MSU-2R readings above land areas, and up to 10% above oceans, come from surface-based emissions. Hurrell and Trenberth argue that unusual land characteristics at the time of satellite transition, such as those caused by persistent droughts or floods, may have altered the surface emissions and biased the MSU-2R trends.
Christy's MSU-to-radiosonde comparisons show little disagreement in the tropics, although Hurrell and Trenberth question the reliability of radiosonde records there, where site changes and other confounding events are frequent. Instead of satellite transitions, Christy suspects natural features are responsible for the 1981 and 1991 temperature drops. He cites a 1981 African volcano, Nyamuragira, along with 1982's El Chich—n and 1991's Mt. Pinatubo. However, MSU temperatures have been slow to return to their pre-Pinatubo values, while surface temperatures have rebounded completely. Indeed, even with the influence of ENSO and volcanoes factored out, the MSU and surface readings continue to reflect increasing disparity over time.
Improved computer models may help clarify things. Hurrell and Trenberth are now using NCAR's latest community climate model, CCM3 (part of the NCAR climate system model), and new global reanalyses produced by the National Center for Environmental Prediction and NCAR's Scientific Computing Division. Christy has reservations about using climate models to project global temperature trends: "These vertical variations being modeled are very small, and they are affected by such factors as volcanic aerosols, which are not included explicitly in CCM3 and similar models." Trenberth argues that volcanic effects are reflected in the sea-surface data used as input to study the 1979-95 period. "CCM3 is indeed suitable for this research, and it in fact accounts for the vast majority of the temperature variability we've found."
Some of the work by Hurrell and Trenberth has focused on differences in the vertical profiles between land and ocean. The strongest surface warming of the past few years has been found over Northern Hemisphere continents, particularly in winter, while MSU's slight cooling trend has been centered above the oceans. Hurrell points out that wintertime cold fronts sweeping from land to sea are modulated near the relatively warm ocean yet maintain their strength higher up. In contrast, over land, inversions often trap cold air near the surface while the troposphere above remains warmer. The prevalence of these inversions means that a reduction in their number or strength (through increased low-level water vapor, for example) could produce a significant warming of average surface temperature with little change at higher levels. Moreover, the inversions are often too shallow to be diagnosed or tracked by computer models.
Those involved in the global temperature game, and others on the sidelines, are paying keen attention as the latest numbers roll in. Says Spencer: "The surface and satellite records cannot continue to diverge indefinitely. They should gradually come closer to one another." Both records are now seen as critical to unraveling global change, especially with an atmosphere acting contrary to some modeling results and theoretical assumptions. As Christy puts it: "Do we have an atmosphere that's less rigid than we think?"