Published online 20 August 1998 | Nature | doi:10.1038/news980820-1


Satellite climate record in error

A record of cooling in the lower troposphere, the lowest layer of the atmosphere, seems to have more to do with changes in the orbits of the satellites making the measurements, than any real climatic effect. Once these changes are accounted for, the trend turns from a cooling to a modest warming trend. This conclusion, which is likely to be controversial, is presented in a study in the 13 August 1998 issue of Nature by Frank J. Wentz and Matthias Schabel of Remote Sensing Systems in Santa Rosa, California.

Since 1979, a series of eight satellites put into polar orbit by the United States' National Oceanic and Atmospheric Administration (NOAA) have carried instruments called microwave-sounding units (MSUs). These instruments measure a certain frequency of microwave radiation emitted by atmospheric oxygen. Changes in the strength of this radiation can be related to changes in the internal energy of the oxygen molecule, which in turn provide a guide to the temperature of the atmosphere.

In a report in 1990, Roy Spencer of the NASA Marshall Space Flight Center, and John R. Christy of the University of Alabama, in Huntsville, Alabama, combined MSU observations from the NOAA satellites to produce the first satellite-derived dataset on global atmospheric temperatures [see Science 247 , 1558-1562; 1990]. This report marked a real advance: to get a clear picture of global trends, climate researchers needed a truly global measurement system. Until then, researchers had to rely on combining records from a large number of local sources (such as surface stations and weather balloons.) Because data from these sources may have been gathered by a variety of different instruments and methods, it was not always clear how (or even whether) they could be assembled to give a reliable global picture.

Between 1979 and 1995, the MSU record showed that the lower troposphere - the lowest layer of the atmosphere - was cooling at a rate of 0.05 kelvin per decade. This change, a fall of five hundredths of a degree Centigrade - may seem small, but it ran counter to measurements gathered from the surface that indicated a rise of 0.13 kelvin per decade over the same period. Even though the data sets are not directly comparable, and that the warming trend seen at the surface is expected to diminish with altitude, some have regarded the cooling trend in the lower troposphere as suspiciously excessive. In the meantime, the discrepancy has sparked a lively debate in the climate community about possible instrumental problems, and even the existence of global warming.

Wentz and Schabel now show that the cooling trend seen in the lower troposphere is an artefact caused by a previously neglected effect: that satellites in orbit tend to lose altitude as a result of atmospheric drag. The original analysis used to derive temperature from the MSU data assumed that the satellites, once in orbit, stayed where they were put. But by taking the effect of altitude loss (or 'orbital decay') into account, the corrected estimate shows a temperature rise of 0.07 kelvin per decade in the lower troposphere, rather than a fall of 0.05 kelvin per decade. This is in closer agreement with the trend observed in surface temperatures. The researchers also show that a previously reported - and difficult to explain - cooling of the lower troposphere, relative to the middle troposphere, is another artefact of uncorrected orbital decay effects.

The problem stems from the way in which MSU measurements are made, which is more than a simple matter of flying a satellite and poking a thermometer out of the window. The question is easily stated: how can a satellite in orbit more than 800 kilometres above the surface measure the temperature in the lower troposphere (centred at an altitude of just 3.5 km above the ground) and the middle troposphere (7 kilometres); and tell the two records apart, both from each other and from the radiative effects of the Earth's surface and other atmospheric layers, such as the stratosphere, immediately above the troposphere?

What the MSU does is measure microwave emission from a number of different angles. By pointing the instrument more or less straight down at the ground from orbit (at the 'nadir'), the signal is dominated by emission from the mid- and upper troposphere, and also a part of the stratosphere.

Temperature in the lower troposphere is gauged in a more round-about way, by gathering data from angles closer to the apparent horizon (or nearer the 'limb'), and subtracting it from near-nadir data, according to a prescribed formula. The rationale is that near-limb observations are more sensitive to temperature in the upper troposphere and lower stratosphere, so subtracting near-limb from near-nadir observations gives an estimate of temperature in the lower troposphere.

So far, so good - but only if the MSU, and the satellite carrying it, stay at the same altitude. But the satellites have been slowly falling, their orbital decay being most marked between 1979-83, and 1989-92. These were times of high solar activity, in which increased ultraviolet radiation heated the upper atmosphere, expanding it more into space, and subjecting the satellites to aerodynamic drag. In response to which, the satellites dropped to lower and lower altitudes. Over the period between 1979 and 1995, the satellites experienced a net fall of 20 km, equivalent to an average annual fall of 1.2 km.

Altitude affects MSU measurement by distorting data from near-limb measurements compared with near-nadir measurements. A drop in altitude will not influence the angle at which the instrument looks at the ground directly beneath - the ground will simply appear a few kilometres closer. But it will have a more marked effect on measurements made on the limb, because the angle of incidence on the Earth's surface at the horizon changes with the altitude of the observer. This is a simple consequence of the Earth being round.

Because the temperature of the lower troposphere is not measured directly, but is an estimate derived from a standard formula in which near-nadir and near-limb measurements are combined, a change in one relative to the other will distort the outcome, unless the change is corrected. This is how, by accounting for orbital decay, Wentz and Schabel show that the lower troposphere has experienced a slight warming, rather than a slight cooling.

This finding will force researchers to look much more closely at data derived from satellites, especially if the results are achieved indirectly, with the potential for error to creep in at every step. It could be that other errors, as yet unaccounted for, could swing the balance the other way once more, to give, once again, a cooling, rather than a warming trend in the lower troposphere. Time will tell.

Problems with satellites may prompt researchers to make more use of instruments borne on weather balloons, which are closer to the ideal of sending an instrument to a prescribed height and simply poking a thermometer out of the window. But this raises the problem of the local nature of such records, and their unsuitability for estimating global trends - which is why Spencer and Christy's satellite data set was so important in the first place.

"The crux of the matter" says Dian J. Gaffen of NOAA in a commentary accompanying the Wentz and Schabel paper, "is that climatologists are relying on systems that were never designed for climate monitoring. Various groups have advocated improved observational networks, but they have yet to be realized."