Tuesday, January 1, 2019

Happy New Year (and update on shutdown consequences for data).

Wishing all readers the best for 2019, even if it doesn't look so promising right now.

Usually about this time I am gathering data for articles about 2018. But the US government shutdown has now closed down most NOAA sites, and it doesn't look like ending anytime soon. The first affected will be the NCEP/NCAR reanalysis index. But GHCN a,d ERSST are out too, so that cuts out TempLS, on which I would normally report by about the 8th. So fingers crossed.

The last NCEP/NCAR data was from 23rd Dec.AT that stage, there had been a big warm peak, and DEcember was looking to be a very warm month. However, that was passing, so the end result is likely to be cooler, althoug still a lot warmer than November.


  1. The NOAA CDAS is still reporting and the preliminary monthly average global mean surface temperature (GMST) for December 2018 based on daily CDAS averages was 13.046C, which is the lowest December average since December 2014 at 12.893C. The December 2018 global mean surface temperature anomaly (GMSTA) referenced to 1981-2010 was 0.325C compared to 0.246C for November 2018. The 2018 annual average GMST was 14.710C which is the lowest since 14.560C in 2014. The 2018 GMSTA referenced to 1981-2010 was 0.266C compared to 0.380C for 2017 and the lowest since 0.115C in 2014. Monthly trend graphs here.

  2. Thanks, Bryan,
    That is very helpful, including the link, which I have added to the blogroll. Unfortunately, the RSS system here gives it the name of an old post and puts it at the bottom of the list. I'll try to fix that.

    1. Thanks Nick. Interesting that the RSS automatically chose my Holocene Climate post from 2014 for the link. I actually started my blog to post mainly about Paleo Climate and I have summarized those early posts with a page link here. I implemented the blog in WordPress, which allows "posts" as well as "pages". I haven't made a "post" in awhile, but I'm working on one about the recent dearth of buoy temperature measurements in the Arctic Ocean that I hope to post soon. Lately my blog updates have mainly been focused on updating time series graphs of the daily and monthly CFSR/CFSV2 temperature data in the "Daily Updates" and "Monthly Trends" pages, which would not be picked up by RSS.

      If this US government shutdown continues much longer, you might be interested in the information I posted at the bottom of my "Daily Updates" page here, describing where I get the daily and monthly CDAS data and how I process it. My processing is quite slow and tedious using Excel and I suspect with your excellent scripting skills you could probably set up routines to automatically extract and process the temperature data to make graphs and even global spherical maps like you have done with the much lower resolution NCEP/NCAR Reanalysis 1 temperature data.

      I recently started pulling the R1 daily data at the link you provide with your time series plot in order to compare historical R1 vs CFSR/CFSV2 daily temperature data to better understand some of the discrepancies between the two data sets. However, the US government shutdown has stymied that effort for now.

      Oh, and by the way, happy new year to you and yours, and to your blog followers.

  3. This is what I see:

    Yet, I don't really mind this too much as all I require for current research on ENSO are the two measurements at Darwin and Tahiti.

  4. Nick, I hope you will be able to make the TempLS update soon, but while we are waiting, I'd appreciate your thoughts about a simple exercise I performed here.

    I looked at the CERES solar radiation incoming (SRI) at the top of the atmosphere (TOA) versus the CFSR global mean surface temperature (GMST). From this data we can calculate how much temperature change results from the annual change in TOA SRI caused by earth's orbital eccentricity versus the corresponding annual variation in GMST and its implication about how much warming will occur from doubling of CO2. I'm sure others have done this exercise before, but so far I have not run across any discussion of it. I'm hoping you and other readers here can help to enlighten me about the subject. I would be especially be interested to hear about how well the climate models handle this annual cycle. Thanks in advance.

  5. Well, the NOAA ENSO updates are still happening, and it looks like things have been turning away from El Nino levels for a month or so now. In fact, surface temperatures across almost the whole Pacific have cooled since early December. Difficult to see where things will go from here.

    For a longer-term perspective CERES-EBAF has also been updated, with most of 2018 data now available. And it looks like another high TOA imbalance, above 1W/m2. According to CERES-EBAFv4 against an Argo-based offset for 2005-2015 the past 4-5 years have now been persistently around the 1W/m2 mark, compared with average around the 2000-2009 period of a little under 0.6W/m2. If this data is correct it would be the first indication of a strong acceleration in recent warming. I think we would expect a 1W/m2 imbalance to correspond with about a 0.3K/Decade surface trend for the next decade and beyond, assuming forcing keeps increasing as it has.

    Bryan, I'm a bit confused about the SRI-GMST test you propose. The annual cycle in GMST is almost opposite to the annual cycle in SRI because Earth's perihelion is at around the peak of Southern Hemisphere Summer and the effect of that is opposed by the much greater land fraction in the Northern Hemisphere and the faster warming rate of land surface versus ocean surface.

    1. PaulS, yes the SRI-GMST cycles are offset by 6 months, but the fact that these two cycles are so well synced and stable implies to me a causative relationship, but with a 6 month lag time, probably mediated primarily by the oceans.

    2. Perhaps I'm misunderstanding what you're saying but thinking about a 6 month lag or offset really doesn't make any sense to me. The reason the GMST peak occurs when it does is because that's (roughly) the peak of insolation in the Northern Hemisphere, due to the Earth's tilt being a much bigger factor latitudinally than distance from the Sun. Likewise the low of GMST occurs shortly after the low of Northern Hemisphere insolation.

      If you look at the annual cycle for hemispheric surface temperature averages individually the NH peaks in July - a little after the June peak in NH insolation (probably due to mediation by the oceans) - and the SH peaks in January/February - shortly after the December peak of SH insolation.

      So you have two main annual cycle shapes - one peaking in July in the Northern Hemisphere extratropics and one peaking in January/February in the Southern Hemisphere extratropics. And then you have not much in the way of annual cycle in the tropics. But because of the much greater land mass in the Northern Hemisphere, and the greater warming rate of land vs sea, the NH annual cycle is about twice as large as in the SH. Therefore the GMST annual cycle gets dominated by the NH.

  6. ENSO is an oceanic dipole in standing wave parlance, so all that is really required are two oppositely signed anti-nodes. These happen to be fairly close to Darwin and Tahiti so we are safe there not having to rely on the USA government for this data.

    According to the Washington Post weather forecasts by the NWS are performing at a substandard quality, so apparently that's another side-effect of the shutdown.

    I could certainly use daily SOI data stretching as far back as one can go since the LTE model cross-validates both monthly and daily SOI data as I reported at last month's AGU meeting. Good thing that meeting missed the shutdown, as it would have been as sparsely attended as this weeks AMS meeting. As it was, many of the USGS researchers did not attend AGU the last two years.

  7. PaulS, are you trying to say that the observed 3.775K annual cycle in GMST is caused entirely by effects related to earth's axial tilt (seasonal effects)? If so, then what effect does the 22 watts/m^2 annual variation of solar radiation incoming at the top of the atmosphere from eccentricity have? None? My thinking is the annual GMST cycle is the net result of all these effects, but is primarily driven by the incoming solar radiation swing.

    For 2001-2017, the CFSR data shows an average annual range of 12.589K for the northern hemisphere and an average annual range of 5.227K for the southern hemisphere. The southern hemisphere cycle is only one month delayed from the top of the atmosphere incoming solar radiation cycle. Both of the hemispheric cyclical swings are larger than the global swing and include both seasonal earth tilt plus eccentricity effects. The global swing should be the net remainder resulting from the eccentricity induced incoming solar radiation effect.

    If eccentricity was zero, would there still be an annual cycle in GMST from seasonal effects only? If so, then how much? If so, we would need to subtract that amount from 3.775K to estimate the effect from eccentricity, which would reduce the calculated GMST change per 1 watt/m^2 of top of the atmosphere incoming solar radiation change.

    1. Bryan,

      Obviously eccentricity has some effect. The point is that it isn't really discernible in the GMST annual cycle above other factors - namely the combination of Earth's tilt and hemispheric asymmetry in warming rate.

      To put some more numbers on this, check the annual cycle in downwelling TOA shortwave above the hemispheric extratropics (25-90º). In the Northern Hemisphere you see the peak in June at around 480W/m2, with a trough in December at around 120W/m2. This includes the influence of orbital eccentricity whereby the Earth is closest to the Sun in December/January, which has an annual cycle with amplitude of about 20W/m2. The effect of eccentricity on the Extratropical NH solar radiation seasonal cycle is relatively weak, <10% of the influence from tilt, and of the opposite sign, so has a small dampening effect on the tilt-dominated cycle.

      In the Extratropical SH you see the opposite but with very similar magnitudes, with downwelling TOA shortwave peaking in December and hitting the low in June. And here eccentricity effectively enhances the seasonal cycle.

      Now, if we lived on a hemispherically-symmetrical planet what would happen is you'd still see distinctive and oppositional NH Ext and SH Ext annual cycles due to Earth's tilt. However, the amplitude of the SH cycle would be greater due to the enhancing effect of eccentricity there and damping effect in the NH. And averaging together you would get a GMST annual cycle which should largely reflect eccentricity.

      But we don't live on that planet. Earth is highly asymmetrical hemispherically, with the much greater land mass in the Northern Hemisphere meaning much faster warming and cooling. As a consequence the Northern Hemisphere sees a larger amplitude in the annual cycle, so when averaged with the SH annual cycle to give the GMST annual cycle the NH dominates.

      If eccentricity was zero, would there still be an annual cycle in GMST from seasonal effects only?

      Absolutely, yes. In fact, an annual cycle with greater amplitude because eccentricity currently only acts to dampen the prevailing annual cycle.

      If so, then how much? If so, we would need to subtract that amount from 3.775K to estimate the effect from eccentricity

      I suspect the only way you could robustly obtain such an estimate is by using a climate model running in modes with and without the various orbital/axial factors. Trouble is then that the climate model used would necessarily have a climate sensitivity, which means any estimate of climate sensitivity based on that derived adjustment factor couldn't really hold much meaning.

    2. PaulS, thanks. As I pointed out previously, the NH annual surface temperature range (from monthly averages) is about 12.6K whereas the SH range is only about 5.2K. The 7.4K difference in these two ranges is almost double the annual 3.8K global range. The NH range would be higher without the eccentricity effect and the SH range would be lower. Most of the NH-SH difference in ranges is obviously driven by axis tilt seasonal effects related to land/ocean differences. Maybe it would be possible to tease out the eccentricity effect from the seasonal effect by also looking at the range in TOA downwelling shortwave in each hemisphere over the annual cycle? Is TOA downwelling shortwave available from CERES (I'm not very familiar with CERES yet)?

    3. Bryan, the seasonal variation in solar radiation that Earth receives, can be seen in the global 0-2000 m vertically averaged ocean temperature (and heat content).

      Here is Roemmich-Gilson Argo only global 0-2000 dbar data: (dbar and meter are about the same)


      The ocean temperature usually peaks in March, a two month lag after the maximum insolation, which seems reasonable.
      0.01 C change applied to the total 0-2000 m ocean volume, corresponds to about 26 ZJ (10^21) in ocean heat content, so the seasonal variation of OHC in this ocean layer is close to 100 ZJ.

    4. Olof R, thanks. Looks like the annual cycle in the potential temperature of the upper ocean is quite small compared to the 2-meter air temperature cycle, but then I realize the oceans hold much more heat than the atmosphere. Looks like there has been a slight overall upward trend since about 2007 based on the green line which appears to be some kind of running mean.

    5. Yes, the green line is the 12 m running average. The years before 2007 are less reliable because the Argo array wasn't fully deployed until then. There were also widespread pressure sensor problems in the early years that made data less reliable.

      Just for fun, if we assume that the troposphere has a seasonal temperature variation of 3.8 C, and assume that the troposphere is 75% of the atmosphere by weight, the seasonal variation in tropospheric heat content should be about 14 ZJ. So the variation in OHC is six- seven times larger

      Also, I don't know why they use potential temperature. However, those guys are oceanographers, so they may see some advantage in recalculating all temperatures to surface pressure, that ordinary people doesn't understand. Anyway, the difference between potential and in situ temperatures for 0-2000 dbar (picked from Argo marine atlas) is very small, the latter being about 0.002 C cooler.

    6. Olof R, I'm guessing the recent slight rise in temperature of the upper ocean could be caused by an increase in solar radiation reaching the ocean surface (presumably less cloud cover, especially in the tropics) and/or less upwelling of colder water from below.

    7. Olaf,

      The ocean temperature usually peaks in March, a two month lag after the maximum insolation, which seems reasonable.

      I'm not sure why there would be much of a lag. It could be that peak solar absorption by ocean areas occurs near March. If you think about when peak global isolation happens, due to the Earth's tilt the place on the planet receiving the most sunlight at that moment in time is the Antarctic continent. Beyond that there's sea ice and heavy cloud cover over the Southern Ocean. So it could be that the ocean peak occurs when the solar focus moves back North a bit.

    8. PaulS,
      It seems like there is always a lag between the time of maximum insolation and maximum temperature.
      If we look at the diurnal cycle, the peak temperature doesn't occur at noon, but rather 2-3 hours later on average.
      Likewise with the NH summer, the peak land temperature occurs on average around one month after the summer solstice (late July), and the peak SST around two months later (late August).

    9. Olof (Sorry for misspelling your name before),

      In surface and atmospheric temperatures a lag makes sense because of the large heat capacity of the oceans, causing inertia, and the relatively much smaller heat capacity of the surface and atmosphere. With ocean heat content, or average ocean temperature at depth, you're (in theory) already looking at pretty much all the heat in the system at any given moment, so what's supplying the lag?

    10. Yes, the change in the ocean heat content is likely more complex than a simple function of seasonal variation in incoming solar radiation (with a lag).
      I suspect that seasonal change in snow, ice-sheets, sea-ice, ground temperature and frost also can be important heat reservoirs (beside atmospheric temperature and vapour content).
      Strangely, the TOA imbalance (according to CERES-EBAF) has its maximum in February and minimum in June, not January and July as one could expect from solar insolation alone, so it must be modified by cloud and surface albedo, etc, maybe with some hemispheric asymmetry..

      Here's a pic of the TOA imbalance seasonal variation. The unit is W/m2

    11. Paul
      I'm not sure why you think OHC would peak at the same time as solar insolation peaks. The ocean would continue to warm as long as input is greater than output, and those are two distinct processes. (Output is mostly controlled by temperature differences at the ocean/atmosphere boundary, whereas solar radiation gets absorbed by the ocean regardless of those differences.)

  8. Olof
    Fascinating graph. If the 0-2000 m temperature has a yearly fluctuation, the ocean surface temperature should too, right? Obviously with a lag.

    1. Snape,
      I just looked at data from Argo marine atlas. As a global average, almost all temperature fluctuation occur in the mixed layer, 0-100 m depth. There is very little seasonal fluctuation in the 100-300 m layer, and below 300 m there is no seasonal variation at all. This is for global averages, and it might be different for regional, zonal, or hemispheric values..

    2. So the cycle seen in the 0-2000 meter vertical average is mostly the result of what's happening at 0-100 meters?

    3. Yes, I guess that 80-90 % of the seasonal variation is happening at 0-100 meters.
      This is only the seasonal variation though, that can be smoothed away using anomalies or 12 m-average. The long-term trend is that temperature and heat content increase at all depths, but faster closer to the surface.

  9. PaulS, I'm beginning to think now that we might be able to fairly closely approximate the annual effect of eccentricity on the global surface temperature by looking separately at the effect in each hemisphere and then reconstructing the annual cycle with just the portion influenced by eccentricity. For instance, we could assume that eccentricity causes a 20 percent reduction in the NH annual surface temperature range and a 20 percent increase in the SH annual surface temperature range. From this assumption we can calculate the effect on monthly temperature in each hemisphere and then reconstruct the monthly annual temperatures without eccentricity. The resulting difference from hypothetical no eccentricity to actual would give us the effect of eccentricity on the annual global temperature range. The most difficult part would be estimating what percentage to apply.

    1. I thought of a simple analogy that might help explain the OHC lag.

      Fill a tall bucket half way with water and poke a little hole near the bottom, allowing water to squirt out. Then add water to the bucket with a hose, adjusting the flow until the water level is steady. Let's say the hole is draining water at 1 gallon/minute and the hose is adding water at 1 gallon/minute.

      Now turn up the faucet until the flow through the hose has doubled.....2 gallons/min. The water level in the bucket will start to rise.

      Next, adjust the flow down to 1.5 gallons/minute. The water level will continue to rise, just not as fast. Finally, adjust the flow back down to where the water level remains steady.

      So there you have it. The input peaked at 2 gallons/min, but the water level continued to rise long after, until the rate of inflow again matched the rate of outflow.

  10. And now someting else.
    Copernicus ERA-interim for December came in quite cold, only up 0.04 from November, which meant that 2018 lost the rank as "third warmest year" vs 2015 with a tiny amount, 0.0018 C.

    I think that this is a tie, a shared third position, but Copernicus hesitate to do this, and write that 2018 was the fourth warmest.
    The difference is so small that the ranking may easily change when the fast mode reanalysis is replaced with the delayed mode version after 2-3 months. Or when Era-interim is replaced with ERA-5, which will happen soon.

    Era5 has data through October now. I compared 2018 and 2015 month by month, using ERA5 for Jan-Oct and ERA-interim for Nov- Dec. Averaged for all months, 2018 was 0.017 C warmer than 2015, so it looks like 2018 eventually will become the third warmest year.

    All blended temperature dataset will have 2018 as the fourth warmest year. However, in Gistemp dTs, which estimates the global temp using met stations only, 2018 will most likely become the third warmest

    1. I'm surprised 2018 could be as warm 2015, given 2015 was an El Niño year throughout.


    2. SST was warmer in 2015 than in 2018. Land 2 m air temp was warmer in 2015 than in 2018. Oddly enough, it is the air temperature over oceans that gives 2018 the upper hand vs 2015 in Gistemp dTs and ERA5
      (Checked with land and sea masks at KNMI climate explorer. Sea air includes air over sea ice)