Science —

Satellite temperature record update closes gap with surface records

Improvements to algorithms raise calculated warming trend a bit.

Satellite temperature record update closes gap with surface records

Satellites seem like an obvious technological solution to the considerable challenge of tracking changes in Earth’s climate. But Earth-observing ain’t easy. A single instrument can zoom over the locations of thousands of stationary thermometers—but that puts thousands of eggs in one instrumental basket. Measuring temperatures from space takes a lot more than some mercury in a tube, and you can’t fix your instrument if something goes wrong.

Illustrating that fact is a new update to one of the major satellite temperature datasets, which ends up changing the recent part of the record in a subtle but significant way.

As we’ve explained before, satellite measurements of atmospheric temperature are actually more dependent on adjustments than measurements done using weather stations or ships. The datasets are based on measurements of microwave radiation on a series of relatively short-lived satellite instruments going back to 1978.

Measuring is hard

The microwave-sensing instrument itself has to be constantly calibrated by making measurements of deep space and of a small target on the satellite that’s hooked up to standard thermometers. Stitching together the measurements from each satellite in the series requires careful analysis of its overlapping time span. Any change to the orbiting altitude of the satellite also has to be precisely tracked, as it affects the temperature calculation.

The orbits of most of the satellites gradually change over time, so they don’t consistently fly over a spot at precisely the same time of day—they can drift to earlier and earlier or later and later times. This has been particularly problematic. If a satellite starts out measuring temperature above your area at 3:00pm each day but drifts to 5:00pm over a few years, it will show that area as cooling over time—even if the area hasn’t. Adjustments are required to remove that artificial drift.

There are still other complications. Satellite measurements can be tuned to focus on different layers of the atmosphere, like the mid-to-upper troposphere or the stratosphere above. The measurements you’re most likely to hear about cover the lower troposphere—closest to the Earth’s surface—but these are also the trickiest measurements to produce. Instead of simply using one set of microwave measurements pointed directly downward at the Earth, angled measurements from multiple passes have to be combined to bring out the lowest part of the atmosphere and subtract out the influence of everything else. The daily temperature swing is also largest near the surface, so that time-of-day drift is greatest for this layer.

The net result of all of this is that it’s extremely easy to get things a bit off. As we’ve gotten a better handle on the problems, changes to the data-processing techniques have significantly altered the satellite temperature record over the years—sometimes increasing the warming trend over one time period, sometimes weakening the trend over another. Some of the datasets produced by Remote Sensing Systems (RSS) got an update last year, but the lower troposphere dataset had to wait for additional testing. Last week, the new version was released.

New corrections

Among the changes was the addition of two new satellites and some improvements to the calibration algorithms for other troublesome hardware. By looking for apples-to-apples opportunities where measurements from at least three satellites could be compared, Carl Mears and Frank Wentz of RSS threw out some periods in which one satellite didn’t match up with the others. They stopped using data from NASA’s AQUA satellite at the end of 2009, for example, because it started acting up.

The most important part of the update, however, is a tweak to their method for correcting the time-of-day drift problem. The solution they had been using was to run a climate model to figure out how big the daily temperature cycle is around the world. That allows you to calculate the difference between 3:00pm measurements and 5:00pm measurements, for example, and subtract it from the actual data.

The new method simply adds in the fact that the size of the daily temperature cycle itself changes with the seasons. So instead of using a single daily cycle correction for each location, the correction varies depending on the time of year. And this no longer comes out of that same climate model—it’s actually based on measurements from satellites flying over at different times of day.

They tested this correction against a couple other methods that wouldn’t work for the entire dataset, but could at least form a comparison over limited time periods. One method tried to only use satellites during the part of their life that time-of-day drift was the smallest. The other method used several satellites that do not drift at all as a baseline to correct drifting satellites that were operational at the same time. Both showed that their new correction method is about as good as they can possibly do, the researchers say.

So what do these changes do to the overall record of global temperature change? If you’re eyeballing the graph, there is a visible shift over the last couple decades that raises temperatures a smidgen. That’s because the new time-of-day correction had the biggest impact on a rapidly drifting satellite in the early 2000s, the periods when one satellite was misbehaving were recent and because of the new satellites that were added in.

A comparison between the previous (blue line) and a new (black line) version of Remote Sensing Systems’ lower troposphere global temperature record.
Enlarge / A comparison between the previous (blue line) and a new (black line) version of Remote Sensing Systems’ lower troposphere global temperature record.

Taken over the entire time period, that increases the warming trend from 0.134 degrees Celsius per decade to 0.174 degrees Celsius per decade. (Yes, we are diving into decimals here.) That change actually makes the satellite record’s warming a tiny bit greater than the warming trend in NASA’s surface-temperature record.

Previously, the RSS satellite record has been a favorite of politicians who reject the conclusions of climate science because it showed less warming—particularly if you ignore everything before the warm El Niño in 1998, which looks even stronger in the upper air records.

Those politicians will probably switch to the competing University of Alabama at Huntsville satellite dataset run by Roy Spencer and John Christy—two of the most prominent scientists who still reject the evidence that humans are responsible for climate change. Their dataset was also recently updated, and while the previous version had shown slightly more warming than the old RSS dataset, it now shows even less. With RSS now falling in line with the major surface temperature records, the new Alabama dataset stands apart as a bit of an outlier.

Then again, even that new Alabama dataset shows that the world continues to warm.

Journal of Climate, 2017. DOI: 10.1175/JCLI-D-16-0768.1  (About DOIs).

Channel Ars Technica