Nights warm faster

NOAA’s Global Historical Climatology Network (GHCN) publishes 3 different monthly temperature sets for each of the ~7000 weather stations. 1) Average Temperatures 2) Minimum Temperatures 3) Maximum Temperatures. Seasonal variations are averaged out over a 30 year normalisation period to derive so-called ‘temperature anomalies’. This is done by subtracting the 30-year monthly averages from each temperature value which causes temperature trends to zero at the normalisation period. Normally only the average temperature (Tav) anomalies are presented, but I decided to analyse all three. These are the results for global temperatures (including oceans).

Fig 1. Comparison of temperature anomalies normalised to 1961-1990 for a)Tmax b)Tmin c)Tav. All curves pass through zero around 1975 due to the chosen normalisation period. The difference Tmax-Tmin is shown by the blue curve which is plotted versus the right hand Y-scale.

A reduction in Tmax-Tmin of about 0.1C is observed since 1950. Minimum temperatures always occur at night over land areas. This means that nights have been warming faster than days since 1950. The effect is actually much larger than 0.1C because nearly 70% of the earth’s surface is ocean with just single monthly average ‘anomalies’. So nights over land areas have on average warmed ~ 0.3C more than daytime temperatures.

If we assume that average land temperatures have risen by ~1C since 1900, then maximum temperatures have really risen only by 0.85C while minimum temperatures have risen by 1.15C.

This effect may also be apparent in equatorial regions where the night/day and winter/summer temperature differences are much smaller than at high latitudes.

Figure 2. All 117 meridional temperature anomaly profiles from 1990 to 2016. They are coloured blue if the annual global anomalies < -0.2C, Blue,-0.2<grey<0.2, 0.2<yellow<0.4, red > 0.4. Traces are 80% transparent to view them all.

Radiative cooling of the land surface mostly occurs at night. It is much greater when the air is dry such as over desert regions and at the poles. During the day convection and evaporation dominate heat loss. Enhanced CO2 reduces slightly night time cooling efficiency. UHI is also larger at night.

This entry was posted in AGW, Climate Change, climate science, NOAA and tagged , . Bookmark the permalink.

15 Responses to Nights warm faster

  1. plazaeme says:

    There is also an apparent change of “regime” if you look separately sea and land.

    From 1880 – 1900 they go in opposite directions (but the data must be very bad).

    From 1900 to about 1985 land and sea go hand in hand.

    From about 1985 the land warms at double the sea rate.

    Measurements are crazy? Three different things are happening to the system?

    • Clive Best says:

      You’re right. There is something very strange about the land warming faster than oceans. That cannot continue indefinitely or there would be huge onshore winds and there is no evidence for that.

      The apparent ‘cooling’ of land values before 1900 could be due to UHI effects.

      The use of anomalies can cause some strange effects because ‘temperatures’ are normalised to a fixed period.

    • paulski0 says:

      The long term (i.e. 1880-present) land/ocean warming contrast is actually close to model expectations for a ratio of about 1.8 for landSAT/oceanSST in a 60S to 60N “globe” (Things get a bit complicated if you include the Arctic). Over shorter periods the warming ratio can vary due to both internal variability and external forcing factors. For the latter, volcanic forcing in particular will tend to produce a considerably stronger effect on land temperatures. I think the most recent 30-40 year period is on the high side for the observed warming ratio.

      I’m not sure exactly what you’re expecting with regards winds, or if that is actually an expectation associated with long-term warming contrast, but RSS has a satellite wind speed product which indicates a general increase in coastal/off-shore wind speed over the past 30 years.

  2. Ron Graf says:

    If the diurnal temperature range DTR decreases for both enhanced greenhouse effect EGE and urban heat island effect UHIE perhaps the two could be resolved by the differences in the polar regions and oceans where there are no urban effects.

    To clarify, let’s assume that there is a latitudinal trend of DTR from tropical SST to polar SST. One would expect this same DTR trend to hold for tropical land surface temp to polar land temp. So, for example, if the tropical SST DTR trend 1950-present is zero and the polar SST DTR trend is .05C per decade then one would expect the same ratio of trends for land temp DTR from tropics to polar. The difference would be DTR caused by UHIE.

    • Clive Best says:


      Yes it would be worth looking at latitude dependence of Tmax-Tmin. Of course at the poles themselves there is only one day and one night per year.

      Over land UHI warming can only be seen if urbanisation is increasing with time. In other words a warmer city just gets renormalised to zero when doing the seasonal average in the normalisation period. The past appears cooler if the city density grew from 1900 to 1970 and warms faster afterwards if the growth continues past 1990. One of the best examples of this is Sao Paolo which grew from a village to a mega-city in 100 years.

      • Ron Graf says:

        It’s already established that UHIE causes narrowing of DTR and narrowing of winter to summer range. The question is whether this signal shows up in the overall land data wholesale. Certainly the individual UHI studies of Seoul, Brussels, Bankbook, Tokyo and others could be used to validate the effect.

        Strangely, the CS community hold the opinion that UHI does not affect the record since they cannot find it statistically after they do their homogenization. Rural groups rise at the same trend as urban (except in the USA). Peterson was the hero that made UHI disappear just after it was starting to be addressed by Karl and Hansen.

        The history of the handling of UHI by the climate science authorities I found very interesting when I looked into it reading Parker(2010). It was mostly brushed off until Karl 1988. . Karl finds UHI in the USA record in cities of population as small as 2000, with an annual temp elevation of 0.13C from a .12C rise in Tmin and 0.01 fall in Tmax. In cities over 10M the Tmax fell 0.55C and Tmin rose a whopping 5.63 C over its rural neighbors.

        As I understand it now CRUTEM does not adjust for UHI. GISTEMP does but there are as many negative UHI adjustments to as positive so they cancel. BEST, with its computerized check for breakpoints, chops records when a station move is detected but has no logic to find a slowly growing temperature bias like urban sprawl. The result is that weather stations that were moved to the outskirts simply get a chance to bias the record a second time as the urbanization gradually catches up with them.

      • Ron Graf says:

        Actually, there is significant UHI found in Antarctica, of all places. I guess anywhere civilization goes it brings local warming.

        At the beginning, after Scott Base was built in 1957, the temperature difference was about 2.5°C, with McMurdo the warmer. The trend for McMurdo is nearly twice that of Scott Base. In 2009, the last year where we have data from both stations, the difference was more than 3°C.

  3. Christian says:

    @ Clive and plazaeme

    Its a bit tricky and mainly the result of hemispheres. SH has more ocean and NH more land and on NH there is strong amplifcation from snow-ice-albedo-feedback. Without looking, i would say, if you make the same for hemisphere only, the gap would be much lower on both hemisphere in compare to global domain

  4. paulski0 says:

    You’ve largely talked about Tmin/Tmax difference as an enhancement of Tmin. This may seem intuitive, and “nights warm more” is a common belief with regards to how the greenhouse effect works, but from what I’ve seen of the literature it’s generally considered a small to negligible part of the Tmin/Tmax story.

    Instead the difference is more ascribed to a damping of Tmax via the small negative shortwave components of CO2 and water vapor plus the effects of aerosols and their cloud interactions.

    Measurements are tricky. Crutem’s Tmax-Tmin data shows a similar overall magnitude DTR decline, but variance shows differences I think. Berkeley Tmax-Tmin shows a considerably larger decline down to the 1980s, then an uptick to present which isn’t evident in the other datasets. UHI may be a factor but also station movements/changes tend to have a much greater effect on one or the other of Tmax/Tmin. Actually, detected station change effects of opposite sign are not uncommon.

  5. Interesting Clive.

    There is a rule of thumb in meteorology that the lighter the winds, the more important local effects become. Given that Tmin occurs mostly, though not always, near dawn, and that Tmax occurs mostly, though not always in late afternoon. And given the lighter surface winds of dawn compared to the turbulently higher winds of the afternoon, Tmin tends to be more influenced by local effects than Tmax.

    Clive, in a simple but related exercise ( if you are looking for more investigations for your list ), I’ve been looking at variance of Tmax. I’d be curious to see your results for:

    1. For each complete station year ( data for each month ) calculate the standard deviation.

    2. For each year, calculate a spatial analysis ( either latitudinal band averages or your triangular approach )

    3. For the spatial analysis, compute trend(s) of variance.

    • Clive Best says:

      You’re probably right. Tmin is more influenced by radiation cooling whereas Tmax is dominated by heat transport caused by convection. Tavg = (Tmax-Tmin)/2 is a mixture of changes in advection and in CO2/H2O.

      In the next few weeks I will try and look at the station data in more detail.

  6. DrO says:

    I don’t have the time to explain all the issues …

    You really should read the Met/Had and other sites to learn what they actually do to create
    their databases and results. Then read the sites the explain how the instrumentation actually works (or not as the case may be).

    For example, separate from the many “administrative adjustments” (i.e. manipulation) of “surface instruments”, HadCRUT4 is actually CRUTM4 + HadSST3.

    However, the practice seems to be to use water temp (about 1 – 1.5 m or so below the surface)
    for the sea based instrument (i.e. they DON’T USE AIR TEMPS). So, amongst other things, the land portion of the temps is for “air”, while the SST portion is for “water”.

    … THOSE ARE FUNDAMENTALLY DIFFERENT THINGS, and impacted by all manner of complications, such as evaporation/condensation/freeing/thawing plus current, wave, PDO etc etc etc effects.

    The actual data in HadSST3 is from ICOADS, and the “quality/meaning” of those results is highly dubious. Just check any single instrument for down time etc. Moreover, while they claim thousands of instruments in the sea, at any one time you are lucky to have 400 – 1,200 over the entire planet’s oceans (and not 100% continuous in operation), and most of those are closely packed together etc etc, with only a tiny percentage of sea instrumentation in the SH (i.e. if half of the world’s oceans). That is, the actual coverage is rather poor.

    The actual distribution of the SST’s instruments IS NOT a nice uniform 5×5 grid. That is simply an artefact of various grotesque manipulations to square peg a round hole by Had/Met et al.

    The satellite data tends not to have these land/sea “curiosities” in any particular material manner, nor any distributional issues of any sort. These issues, as with so many other issues, is a “surface instrument” matter.

    … I started a note regarding your spherical triangulation illustrating errors there, and connected to these points as well, that will follow in due course.

    In addition, you keep loosing sight of the important things, and that leads to further bush league errors. The important thing is heat balance, not “temperature departure”. Indeed, if you are going to deal with temps, keeping track of the abs temp is important (plus the many “own goals” via “man made serial correlation” in the average of the average of the average etc etc.)

    … for example, the statement in the last para “Radiative cooling of the land surface mostly occurs at night.” is surely wrong (or at best an extremely unfortunate choice of words). Radiative flux is function of T^4 (or at least T^3)

    A simple illustration for Black body calculations is shown below for a selection of temps, plus two rows of “Diff’s” showing the flux differences as a function of temp.

    T(K) 250 260 270 280 290 300 310 320 330
    W/m^2 221 259 301 349 401 459 524 595 672
    Diff+10 38 42 47 53 58 64 71 78
    Diff+2 7 8 9 11 12 13 14 16

    Notice that low temp (e.g. at 250k) has low radiation (221 W/m^2), while high temp has (non-linearly) higher radiation.

    Thus the day time temp of the Sahara might be around 310+ K, which is throwing off 524+ W/m^2, cf. night time temp might be 10 – 20 K or more lower, and thus night time is throwing off around 400 W/m^2 or less.

    So, day time radiative loss is about 124 W/m^2 GREATER compared to this night time comparison. So it is the day time radiation that is much bigger, not the night time.

    … yes, we could discuss instead the “net flux”, and brings us back to examining the heat balance, and away from the “too close to the forest for trees” with temp departures of average of the average of the average from dubious sources.

    … when the day time can throw off massively more heat compared to night, it is not surprising that day time warming would be more difficult for more or less fixed in-flux.

    FYI, I include the “Diff+10” and “Diff+2” fluxes, meaning the difference in flux for a 10K and a 2K increase in temp. Notice that even at very low temp, a 2K (or 2C) increase causes the “Black body planet” to throw off about 8+ W/m^2 more compared to the 2K cooler planet.

    … so, if all this CO2 radiative forcing malarkey is 2 – 3 W/m^2 according to Hansen, Schmidt et al, how is it possible for the planet to warm 2C if the radiative forcing is only 2 – 3 W/m^2, but at 2C warmer the planet throws off 8 – 12 more W/m^2 (i.e. where is the missing 5 – 9 W/m^2 coming from)?

    Obviously this issue becomes much more striking at higher temps due to the highly non-linear relationship, where the “missing heat” is on the order of 50 – 70 W/m^2

    In reality, the planet is not exactly a “black body” and there are other considerations, but surely we can’t simply focus on “temperature anomalies” of averages of averages … in isolation.

    Then, for example, the change of water vapour concentration since 1949 implies a requirement of about 5 W/m^2 via a “back of the envelope” latent heat calculation under the assumption that the entire change in the WV content is due to evaporation. Thus, if that holds, then how does 2 – 3 W/m^2 “forcing” account for this portion of the heat balance. We could expand to this all the other non-temp heat transfer matters (e.g. life, etc etc).

    Curiously, it is possible that the WV increase is actually due, at least in part, to combustion. The by-products of burning carbon fuels are (in a perfect burn) xCO2 + yH2O, where x and y are the respective number of moles/molecules. In virtually all combustion y > x. So combustion is actually producing more WV compared to CO2, and WV is 4 – 15 times more powerful a GHG compared to CO2. One has to wonder why the IPCC et al never highlight this .. my guess is that introducing the truth about heat balances and critical processes would make for a “less clean sound bite” for the UN/IPCC’s wished for scenario, and the truth would complicate (dilute 🙂 their story.

    BTW, all of the “surface based” data sets tend to show “asymmetric” warming during el Nino’s, and they tend minimise the peak effect of el Nino’s, as can be easily seen from the ONI index. The satellite data show symmetry in ONI/temp for el Nino’s. Indeed, the land instrument el Nino asymmetry is much worse compared to the sea el Nino asymmetry. However, it is already possible to see the symmetry forming from the current giant el Nino in the satellite data. While it may require another few months to confirm, it looks as if the satellite data will show a very much extended “pause”, and thus a much increased divergence between the satellite and surface temps anomaly trends. Put differently, if there had not been any or appreciable el Nino’s in the past 20 years, there would have been no material increase in GAT.

    … what will happen then to all this temp averaging nonsense, when two completely independent instrumentations are diverging at an alarming rate, one showing massive warming and one showing essentially no warming (not that GAT is particularly meaningful, but that’s for another day)?

    … I am looking forward with glee to see what spectacular spin the UN/IPCC et al will put on that sort of outcome (i.e. if/when more data further crushes their rhetoric) … perhaps will this see in as little as 3 – 6 months.

    • Ron Graf says:

      …when the day time can throw off massively more heat compared to night, it is not surprising that day time warming would be more difficult for more or less fixed in-flux.

      Dr. O, I’m not an atmospheric physicist but aren’t the daily TOA dynamics somewhat isolated from what is happening at the surface? I agree on the exponential relationship of TOA temperature to its radiative flux but since the TOA temp is ~70K less than the surface this effect is muted.

      On the divergence of the satellite lower troposphere temperature indexes from the surface thermometers, part of that could be due to that the lower troposphere is more correlated with Tmax due to afternoon convection currents and winds kicking up the boundary layer air to the LT. The other part, of course, would be lack of UHI bias and fiddling with historic ship measurements when measuring temp the in LT.

    • Clive Best says:

      Yes. You’re right. Sea surface temperature measurements used to be made by buckets thrown over the side and by engine cooling inflows. So this measured water temperature a meter or more below the surface. Today buoys measure sea surface temperatures. There has been some complicated ‘correction’ algorithm made to make these diverse measurement systems compatible. In any case they end up being sea surface temperatures rather than 2m air temperatures as is the case for weather stations.

      Recently there was an attempt to ‘blend’ the two together but that was based on model differences.

      Satellite data measuring tropospheric emissions does avoid some of these difficulties.

      Your comments regarding IR radiation are correct. The hotter the land the higher the IR loss. However there is an energy imbalance with solar heating. This drives convection, evaporation and thunderstorms which are more efficient at shifting heat up to the tropopause and drive Hadley cells moving heat to higher latitudes.

      Lets wait and see if the satellite data diverges from the surface data. It certainly would prove your point regarding El Nino. Don’t count on RSS though contradicting GISS.

Leave a Reply