27 June 2017

Satellite Data

We've passed the 50th anniversary of the first meteorological satellite*, on to 60! Even though satellites have been used for decades, it's still far from simple to do so. Or, rather, it is much more involved than I used to think. I suppose it shouldn't really have been a surprise. Automated weather stations on earth have plenty of problems, and they're operating at earthly temperatures, near somebody who can fix minor problems if they occur, and do so before they become major problems.

By way of my day job (again, I do _not_ speak for my employer), I use satellite data and do so heavily enough that I'm on the mailing list for operational updates about the current status of the satellites, data receiving stations, and other parts of the system between my desk (well, the supercomputer at work) and the satellite.

One thing which has really impressed itself on me is that there is far more to satellites than launching them and waiting for lovely data to flow forth. Second is that there must be a large number of very highly skilled people working between the satellite and my desk to ensure that the data flows reliably, at high quality, and in a timely fashion.

Fortunately, though I'm about to name some challenges and problems, they're _known_ problems and challenges, with solutions. Engineers don't like leaving problems standing. Problems are things to be solved! So, a few stories:
* First story goes back to the dawn of weather satellites. A friend's grandfather was on the team that built and prepared for launch the first weather satellite, TIROS. It was launched April 1, 1960. That's just normal engineering and science. The story is: The engineers signed their grandkids' names on the inside of the satellite. In addition to the fact that I know one of the grandkids, I like this because it's a reminder that science and engineering are done by people. We may well have kids and grandkids.

Geostationary satellites, now including the GOES-16 which has recently starting to give us data, have a host of challenges. If life and the universe were simple, geostationary would be easy. Launch your satellite in to an orbit that takes just as long to complete as it takes the earth to rotate. Easy, right? Well, not exactly. One challenge is, the earth is not the only body in the solar system. The sun and moon act to pull the satellite away from that orbit. On average, over the course of a year and month, respectively, those pulls average out to something closer to zero. But we want our observations from a geo_stationary_ satellite, not a geo-more-or-less-close-to-the-more-or-less-standard-position satellite.

That means the satellite has to have thrusters to keep the orbit exactly as desired. From time to time, then, there will be 'station keeping' maneuvers, where the thrusters are fired to keep the satellite in place. You also have to maintain the thrusters. That means that, from time to time, you have to flush them. Since the thrusters send out a gas which has a temperature, and the satellite is mainly observing earth's atmosphere -- a gas with a temperature -- you also have to halt the observations during the station keeping or flush. You also have to warn all your users that data will halt for a time, and name what that time will be.

A geostationary satellite sits over the equator. At the spring and fall equinoxes, so does the sun. That means that twice a year, around the equinoxes, there is an eclipse of the satellite. The satellite is solar-powered, in the main, which means that it has to run on batteries-only for up to 72 minutes a day. It has batteries to carry it over, but it does mean it's a period when the satellite has only one power source. If there's a glitch, no data. The close earth-sun-satellite alignment means that there's a risk of the sun hitting satellite sensors directly. Since they're designed for seeing the much fainter light reflected off the earth, this poses a different sensor hazard. Again, management of the satellite and warn the users.

The geostationary satellites have another challenge to face. They can be in sunlight for many hours at a time. Great for the solar cells and battery charging. But it means that the satellite can get much hotter in some parts than others. And the satellite is trying to observe subtle changes in temperature from the earth, which can be contaminated by the satellite getting hot. One way to help reduce this and related issues is the 'yaw flip maneuver', where the satellite is flipped over.

Lest you think that we should quit using geostationary satellites because they have such awful issues, there are issues for any satellite. For any satellite, there's the sun. The sun emits not just light, but radio waves, solar wind, solar flares, and has other activity. Any and all of them can interfere with satellite operations. By my eyeball estimate from the mailing list, radio frequency interference is the most common of these issues. So, again, users (like me) have to be warned that there are times approaching where data quality may drop, or data receipt could even be interrupted.

Also an any satellite issue is finding out where it is pointing. Either the satellite is far away (about 36,000 km from the earth for geostationary, about 22,000 miles), or it is moving fast (about 7.8 km/sec, 5 miles per second for polar orbiters). Both make it challenging to say where, exactly, the satellite is looking when it takes an observation. Since we want high resolution information -- recent satellites have image/viewing pixels about 500 meters across (about 5 football fields or 3 city blocks) -- we need to know where the satellite is looking to within that much. Preferably we'd know the aim to much finer precision than the size of its viewing pixel. One way to do this is to use star trackers. But, of course, sometimes a star tracker degrades and you have to adapt.

Weather stations on the earth periodically get calibrated. The simplest version is, you take your well-calibrated thermometer out to where the weather station is and compare your well-calibrated thermometer to the thermometer in the weather station. If they differ, you replace the one in the weather station, or at least make a note of how much it is wrong by and send that message out. For satellites ... well, nobody gets to go out and do this kind of calibration. (I'd volunteer!) So there has to be something on-board that can be used for the calibration of the satellite observations. There are two kinds of calibration (that I've been introduced to) -- observe deep space (a 'cold target' calibration), where there should be very nearly zero energy coming in (blackbody temperature of 2.7 K), and observe a target ('hot target') that is near earth temperature, whose temperature you know because you make it be this (say by running a known current through a known resistor and heating it up to somewhere around 288 K). Then, of course, you have to take some time to carry out that calibration. During the calibration period, you can't produce useful data. You're generating data that ensures the other observations mean what you think they do, or that let you decipher the degradation that time, environment, micrometeroids, and so forth have wrought on your satellite instrument.

The second family of satellites I use are polar orbiters. This doesn't mean that they orbit at the earth's poles, which would be very convenient for me as a polar scientist. It merely means that they are in basically circular low earth orbits which approach the pole. It doesn't have to be a close approach -- one never got closer than 55 degrees from the pole (TRMM).

Nevertheless, the sun and moon are still out there, and still affect the orbit. You could just let the satellite's orbit drift. But there is usually a purpose for selecting a particular orbit, and moving away from it means a loss of quality in the data for that purpose. So, as with geostationary satellites, there are maneuvers to keep the orbit right. One satellite last fall (Jason-3) had an orbital maneuver which involved changing the semi-major axis by 0.01113 km, in an orbit about 1300 km from the earth. Since the observations people like me use it for are measured to a precision of 0.00001 km, and 0.001 is considered a huge change, this was a very large, important, orbital maneuver.

All this contributes to why there's a saying: Everyone _except_ the person who collected it believes the data.


William M. Connolley said...

> launching them and waiting for lovely data to flow

There's a useful short-cut for that, in terms of spending. See e.g. http://scienceblogs.com/stoat/2016/08/29/launch-spending/ (me quoting the Economist).

E. Swanson said...

Robert, What are your thoughts on Spencer & Christy's new MSU/AMSU version 6? Did you see my paper on the subject, which was published before their paper?


Robert Grumbine said...

I don't see the short cut. The link is to your note on how the launch costs for all satellites is small compared to the cost of the satellites themselves. I'm not sure, though, that that works out across earth-observing satellites specifically. The radarsat launch in the 1990s was, iirc, about 30% the cost of the satellite itself. But that doesn't get high quality data on my desk any faster, regardless of how cheap the launch is.

I hadn't seen your paper, but will follow up on it. Ironically, there's a fair chance that your editor works a couple desks down from me.

Spencer+Christy v6 ... I wasn't impressed. Fundamentally, more of the same it seemed to me. But I'd appreciate your thoughts.

Entirely coincidentally, RSS (Mears) just released an update to their analysis, accounting for a number of things much like what I mention here. And, also coincidentally, it brings their analysis in to much closer agreement with the surface record.

William M. Connolley said...

What I meant was, looking at the cost is a short cut to seeing that there is vastly more than just launching the satellite to getting the data to flow.

E. Swanson said...

Robert, after my report was on the way to final review, I thought of a slightly different way to visualize the data. I had de-trended the data after finding that the RSS curves had a slight bias compared with UAH, which was the result of different base periods used to calculate the respective series. After calculating the average of each series then subtracting that value, plotting the two series produces a clearer comparison.

HERE's a link to one of the modified graphs, showing the UAH vs RSS difference for the TMT.

pp said...

Read also here for what the recent satellite corrections are about:

Alastair said...


Off topic, but I seem to recall that when you set up your sea ice site, http://polar.ncep.noaa.gov/seaice/sh.html
you were particularly interested in monitoring polynyas in Antarctica. None appeared, until today! Were you aware of that?