Mauna Loa Daily Meteorology

Guest Post by Willis Eschenbach

As a confirmed data junkie, I’m fond of hourly data. The interesting processes in the climate system unfold on the scale of minutes and hours, not years. So I picked up a project I’d started a while ago, but as is too often the case I’d gotten sidetractored by … oooh, shiny … and I’d forgotten about it until I stumbled across my code again.

This project was looking at the hourly averages of various meteorological variables measured at the observatory on Mauna Loa, Hawaii. This is the same place that the CO2 data has been measured since 1959. The data is available here.

To start with, here is the daily temperature at three different altitudes—2 metres, 10 metres, and 35 metres.

average daily cycle temperature mauna loa observatoryFigure 1. Daily cycle of average temperatures at three different altitudes above the ground.

There were some interesting parts of this to me. One is that the surface temperature peaks at about 1 PM … but as you go up in altitude, the peak occurs earlier. Hmmm …

Also, I was surprised that ten metres up in the air the daily variation is less than half of that down at two metres.

Because the atmosphere is heated from the bottom, it is unstable during the day, and overturns. During the night, on the other hand, the atmosphere is coolest at the bottom, so it stabilizes and stratifies. You can see the timing of the onset and the end of the daytime period of turbulence, which starts just before nine am, and lasts until just after dark.

Next, here is the average precipitation rate hour by hour:

average daily cycle precip mauna loa observatoryFigure 2. Daily cycle of average precipitation rates, millimetres per hour.

Here, we see the typical sequence of weather around a tropical island. The big peak in thunderstorms occurs in the afternoon around three or four o’clock. You also get a much smaller number of early morning thunderstorms.

Next, I looked at the winds:

average daily cycle wind mauna loa observatoryFigure 3. Daily cycle of average wind speed, metres per second.

This shows something interesting. The “terminator”, in addition to being a series of increasingly bad movies, is the name for the line between light and dark on the surface of the planet. On one side of the terminator, the light heats the air near the surface. This makes the air rise on the lighted side of the terminator. The existence of warm lighter air on the lighted side, plus cool heavier air on the dark side, leads to the “terminator wind”. This is a wind created by the temperature difference across the terminator.

This plot shows the difference between the dawn terminator wind and the dusk terminator wind. The terminator wind always blows from dark to light, which means it always blows toward the sun. Now, the trade winds in the tropics always come from the east and blow towards the west. So at dawn, the terminator wind opposes the trade winds, because it is blowing out of the darkness in the west towards the sun rising in the east. This leads to the drop in wind speed after dawn that you can see in Figure 3.

But at dusk, the terminator wind blows in the same direction as the trade winds, and this increases the average wind speed after the end of the day. Can’t say I understand the rest of the variation, though. I do note that the wind picking up and dying down occurs at the same time as the onset and dying out of the daytime overturning.

(Curiously, I found out about terminator winds by spending lots of time at sea. The sweetest terminator wind is on a dead calm night, not a breath of air … and then the moon rises, and if you are lucky, you can feel the moon wind sweep across the ocean, always blowing towards the moon … but I digress.)

I next looked at the absolute humidity. This one was a surprise.

average daily cycle ah mauna loa observatoryFigure 4. Daily cycle of absolute humidity, in grams per cubic metre.

The reason that this was a surprise to me was that I had not expected it to vary that much. From a low of two grams per cubic metre at dawn, it more than doubles when it rises to a peak of five grams per cubic metre at three pm. Why is this important?

Water is the dominant greenhouse gas. Because it is an “L-shaped” molecule, water vapor has many ways to absorb radiation. The molecule can flex and twist and stretch in various combinations, so it absorbs thermal radiation (longwave infrared) of a wide variety of frequencies. The important point is this:

The change in the amount of longwave infrared absorbed by atmospheric water vapor is approximately proportional to the log of the change in the amount of water vapor. 

And the amount of water vapor in the air varies during the day by a factor of about two and a half to one … I’d never realized how much greater the afternoon longwave absorption is compared to the absorption at dawn. Who knew? Well, I’m sure some folks knew, but I didn’t.

So I fell to considering the effect of this daily variation. The increase in atmospheric absorption will warm the afternoons, and decreased absorption will cool the early mornings as compared to the average. Now, one corollary of Murphy’s Law can be stated as:

Nature always sides with the hidden flaw.

In terms of the climate system, the poorly-named “greenhouse effect” works to increase the surface temperature. Murphy’s Law means that all related emergent, parasitic, and other losses in response to that surface warming will tend to oppose this effect. In other words, we expect the natural response to elevated surface temperature to be one of cooling of the surface.

For one example among many, when the desert surface gets hot, “dust devils” emerge out of nowhere to cool the surface by means of increased evaporation and convection. They pipe the warm surface air aloft, increasing surface heat loss. But there are no “anti-dust-devils” that act to decrease surface heat loss … Murphy’s Law in action.

Now, the radiative loss varies as the fourth power of the temperature. This means that if the temperature varies around some average value, the radiative losses will be larger than if the temperature were steady. As a result, since the variation in absolute humidity warms the afternoons and cools the early mornings, to that extent it will increase the overall surface radiative losses … Murphy at work again.

Anyhow, we’ve now finally gotten to my reason for writing this post. The figure below shows one more meteorological variable measured at Mauna Loa—the daily cycle in air pressures.

average daily cycle pressure mauna loa observatoryFigure 5. Daily cycle of atmospheric pressure, hectopascals. Note that because of the high altitude of the observatory, the pressure is much lower than the ~1000 hPa pressure at sea level.

I was, and I remain, puzzled by this variation. Why should the pressure peak at both eleven o’clock in the morning and eleven at night, and be at its lowest just before both sunrise and sunset? And why would the two peaks and the two valleys be about the same amplitude? That question is why I’m publishing this post.

All contributions gratefully accepted …


Further Info: The procedure used for the Mauna Loa CO2 measurements is here. For those who think Mauna Loa is a bad choice for CO2 measurements because it is an active volcano, give it a read.

My Usual Request: If you disagree with me or anyone, please quote the exact words you disagree with. I can defend my own words. I cannot defend someone’s interpretation of my words.

My Other Request: If you think that e.g. I’m using the wrong method on the wrong dataset, please educate me and others by demonstrating the proper use of the right method on the right dataset. Simply claiming I’m wrong doesn’t advance the discussion.


newest oldest most voted
Notify of

So I’m no expert, but isn’t the explanation just atmospheric tides?

Willis Eschenbach

If you mean lunar tides, they don’t happen at the same time every day.

The largest component in atmospheric tides are due to diurnal solar heating, not lunar gravity effects. While there are both solar and lunar gravity signals in atmospheric pressure data, most of the variation is due to diurnal heating. At least, that’s how I currently understand it. Hope I’ve got that right…

See this article in ScienceDaily… for more info on solar driven density waves in the Earth’s atmosphere relating to Mauna Loa measurements.

don k

“If you mean lunar tides, they don’t happen at the same time every day.”
If you’re looking at a long term record, wouldn’t the 24.8 hour lunar component of gravity induced tides tend to average out to zero while the 24.0 hour solar component reinforced itself with every additonal day of data?


Yes – atmospheric tides.
Wikipedia has a good explanation
“Atmospheric tides are primarily excited by the Sun’s heating of the atmosphere… This means that most atmospheric tides have periods of oscillation related to the 24-hour length of the solar day”
“At ground level, atmospheric tides can be detected as regular but small oscillations in surface pressure with periods of 24 and 12 hours.”
(Sorry for duplicate post below – better as a reply)

Fred Harwood

Instrument siting relative to prevailing winds? Wild guess.

Tidal forces work on a liquid medium so why not a gaseous one? A comparison with local tides would be a test of this hypothesis.

Lewis P Buckingham

Perhaps at 11 am the water vapour plus other gas column above the island is becoming saturated with water vapour and precipitation follows when it rains reducing the weight of the column hence its pressure.
At night the relative humidity remains high as the air water vapour column cools and becomes more dense. Falling more dense air/water vapur column draws in more air so the total column above the island rises in mass.

Med Bennett

Could it be an atmospheric tide?

Med Bennett

Solar tide I meant. There was another piece I read on WIWT a little while ago about correlation of precipitation with solar tidal atmospheric forcing.

Mauna Loa Observatory is located at 3.4 km on the north slope of Mauna Loa, which is nearly 4.3 km high. The meteorology at MLO is strongly influenced by the mountain. For example, during early morning a thin layer of air flows down from the summit, which causes the wind vanes to point accordingly. Later the air may be perfectly still or from other directions. During mid-morning the inversion layer rises from below. If it reaches MLO, the ambient water vapor increases dramatically and cumulus clouds appear overhead. On some days the clouds form an overcast. Before sunset, the clouds usually dissipate. I have calibrated instruments at MLO at least once a year since 1992 and will do so again this summer. I typically stay at MLO for 2 weeks during these calibration sessions. In addition to calibrations, I compare my twilight measurements that provide elevation profiles of aerosols with results from the MLO lidar. Various papers cover the meteorology at MLO. So does “Hawaii’s Mauna Loa Observatory: Fifty Years of Monitoring the Atmosphere” (Univ. of Hawaii Press, 2012). (Disclosure: I wrote this book with partial support from NOAA.)

Hocus Locus

We are in the presence of greatness! Another fine post from Willis Eschenbach, one of the greatest data-tinkerers of our era, and here! And here is one of the world’s finest engineer-tinkerers. I am grateful to thank you Mr. Mims for all those articles in Popular Electronics and your books, especially the precious little hand-drawn graph paper ones I used to carry around as a kid, the same way some people carried bibles or poetry in times past. You have helped inspire in me the joy of building things, and a life-long dream of becoming an electrical engineer. Though I’m over 50 now and time is running out and the economy has gone south and yet anything is possible and it’s a beautiful world. Thank you for being a part of it, sir!

Lee Osburn

I agree 100%. Mr. Mims also is a regular provider for scientific items in the Express-News here in Texas. About 6 month ago a friend suggested I contact him to show him my hand drawn graphs. Unfortunately, the articles were not appearing anymore so I have been waiting to see his name in print again. This is my chance to have him contact me.
Mod. please provide him with my e-mail address.
LeeO – Concan

Willis Eschenbach

I can only agree. Forrest Mims is one of my scientific heroes. He used to write the “Amateur Scientist” column for the once-great Scientific American magazine. I always read his contributions with great interest.

Barry L.

Bernoulli’s principle?
Wind speed drops, pressure goes up.


Yes. Both caused by different responses of sea and land, causing wind from sea during the day and wind towards the sea during the night.


Yes, Willis.
Since wind speed it the only other variable you have which shows a semi-diurnal variation, I would say the clue is there.
Timing of the minima in wind are not quite the same but not far off either.


And vice versa.. pressure difference causes wind.

just an off the wall idea of a solar tide

Don V

Interesting data! I know this is stating the obvious: but if the old PV=nRT still holds true on Mauna Loa, and if P is varying with such regularity, then either n or T or V or some combination of the three is/are varying to create the P change. What does the data look like when you superimpose the wind velocity (driven by delta P), T and Humidity over the changing P plot?

Richard Keen

Don V says: February 20, 2016 at 6:08 pm … if the old PV=nRT still holds true …
Only if there’s no external heat entering or leaving the “cube” of air.
But there’s sunlight during the day, radiative cooling at night, and bulk motion of air up and down the hill.

Those are, I believe simply the atmospheric tides. They are about the right amplitude for that latitude I think. If you want to learn more, see here:
Atmospheric Tides, Richard S. Lindzen, Sydney Chapman, Space Science Reviews 10, 1969.
I am fairly certain you can find this on line for no charge.

Just for “fun”, I downloaded a year’s worth of pressure readings from the Hilo airport (available on MesoWest). The link below displays a graph of average daily pressure for the year 2015. Hilo is at sea level so I scaled the readings by 0.67 so they would have the same average as Willis’ data to make comparison easier.
What would one conclude from this?

Willis Eschenbach

Nice … I do appreciate a man who does his own investigations.


Brisbane ( somewhat similar, but I am not one to help explain…………….
High 1018.0hpa @ 9:44
Low 1014.7hpa @ 2:44

“But at dusk, the terminator wind blows in opposition to the trade winds…”
Don’t you mean the opposite of opposition with the opposite direction of the previous example being in opposition as well?

Willis Eschenbach

Too right, fixed.


Sfc pressure is the weight of the atmospheric column of air above the sensor. At night surface radiative cooling sets up downslope Katabatic winds which remove air from the air column thus lowering air pressure. When the sun comes up this process is reversed with surface warming induced Anabatic upslope winds. However if free convection takes place (clouds/thunderstorms) air rises into the upper atmosphere and disperses away from the top of the column near the tropopause. This removes air from the column over the sensor lowering pressure but only after mature convection is established with upper outflow. Before this air is accumulating over the sensor so hence rising pressure.


This process is similar to other mountain density wave events like Morning Glory in Northern Queensland Australia which is truly an amazing sight. I suspect there is a regular density wave spike in pressures that moves away from Mauna loa most days.

Worth looking at. Simply stunning weather phenomenon at its best.

@ pbweather,6:32 pm Feb 20. Thanks for the information and the stunning video, never knew the phenomena existed, truly spectacular!

Nice work Willis. I agree with Forrest Mimms that the diurnal patterns on the volcano are likely to be strongly influenced by the orographic aspects of the volcano.
I notice that the peak in absolute humidity corresponds with the peak in precipitation. It would be nice to have incoming solar radiation data to verify, which I suspect may be available, but most likely it is very cloudy when the absolute humidity is high and the clouds would reflect most of the incoming sunshine before it could interact with water molecules within and below the clouds.
I was busy today and have not had a chance to work with the data yet, but I plan to compile daily, monthly, and annual stats to look at long term trends. Maybe you have already done this for a future post?

I think you’ll find it cycles with temp, at night cooling near 100% rh condenses water out, and then a lot evaporates during the warming cycle during the day.

Managed to get the data compiled and graphed the annual averages:
There does seem to be a slow-down in the temperature rise since 1995:comment image

Oops. That first image is the USCRN station at Mauna Loa. Forgot to clear it off the clipboard. Here’s the full graph of annual temperature averages 1977-2015:comment image


Those graphs are what you would expect with Katabatic and Anabatic winds. Untill you add a wind direction graph to that set, they wont make much sence

Where I live, the terminator moves east in the morning because of the Alouette Mountains :


I thought the original Terminator movie was good, I guess they went downhill since then.
So, anybody got better stuff ??
I’ll bet you don’t.


Lol. Haven’t seen “The Exterminator (1980)”.
Your second link was to a collection of ways to exterminate Daleks of Dr. Who fame.
Were you intending on linking to:



I wont let the wife see that……..just in case.


Re: “In terms of the climate system, the poorly-named “greenhouse effect” works to increase the surface temperature”
The moon is the same distance from the sun as Earth. Moons day time temperatures are way hotter than earths (even in hot deserts in summer) and way colder at night (even than Antarctica in winter). So greenhouse gases warm the lows and cool the highs, so greatly moderating Earth’s temperatures and creating a Goldilocks temperature range for life as we know it.
Not sure how this affects your Murphy response to hottest part of day, probably not at all, but I suspect greenhouse gases are themselves a Murphy response?


Yes, greenhouse gases do that but liquid water does it more because liquid water is a much better heat sink than land.
Indeed, the temperature smoothing effect of liquid water is why the Earth’s Northern Hemisphere (NH) has greater temperature variation than the Earth’s Southern Hemisphere(SH); more of the NH is covered by land than the SH. See here.


A daily air pressure cycle exists at sea level as noted on ships’ barographs. So probably not to do with wind variations. Regarding why peaks occur at 1100 and 2300, suggest this is due to Hawaii’s time zone being 15 degrees from its longitude zone. Peaks should be at 1200 and 2400, and they are due to solar gravity effects. Lunar effects can occur at any time of day so lunar tidal effects cancel out.


Yes – atmospheric tides.
Wikipedia has a good explanation
“Atmospheric tides are primarily excited by the Sun’s heating of the atmosphere… This means that most atmospheric tides have periods of oscillation related to the 24-hour length of the solar day”
“At ground level, atmospheric tides can be detected as regular but small oscillations in surface pressure with periods of 24 and 12 hours.”

bit chilly

very interesting paper louis . a very telling quote near the end regarding ohms law that could be applied to other areas of climate science practice. has this research been taken further to the suggestions in the paper ? i would think the possibility for at least one proxy relating to atmospheric heat content lies within the realms of that research .
congratulations on yet another fantastic set of observations converted to an enjoyable and informed post willis ,including the linked posts,as someone that spends way too much time in and around the sea (according to my wife) you are continually able to put a smile on my face. classy as ever.

Lee Osburn

Willis, your figure 1
As the sun begins to light up the sky before the sun comes up, the pressure begins to rise and peaks out around 10:30. During this period temperature increases at the highest rate for the day.
Your graphs seem to disagree with the theory of the earth heating before atmosphere.
LeeO – Concan, Texas

I suspect it pressure from the terminator winds, which would start early.

Lee Osburn

Another piece of the puzzle would be insolation data. And maybe irradiance data to see the brightness of the sky.

Peter Sable

The problem with being on a mountain is that given those differences in swing vs. altitude, any long term trend in “average” temperature is likely from local variation in winds on the decadal scale.. So that article you linked to the other day that “confirms” global temperature changes from MLO is just coincidence.

Glad to see you digging into this.
Might I suggest two things, first, stop looking at calendar days, and look at the daily temperature cycle, day to day min temp (or clock time if that floats your boat), and also look at rel humidity, it shows why the absolute humidity changes, and highlights how the change in state of water vapor greatly impacts the limits of the daily thermal cycle.

Dr. S. Jeevananda Reddy

The daiurnal variation of meteorological parameters depends upon several factors, such as:
inland area, island area, hill area, forest areas, water reservoirs areas, coastal areas, etc
winter, summer, rainy season, etc
Urban areas, rural areas, etc
If look at India — humidity shows high in the morning observations [0830 IST] and low in afternoon observations [1730 IST]; maximum temperature reaches around 3 pm and minimum around 6 am [just before sunrise]. The exact time shifts around these two timings with season, sun’s moment.
So, it is not a good idea to look at seasonal and diurnal variations of stations to infer something.
Hawii is an island, surrounded by water.
Dr. S. Jeevananda Reddy

” humidity shows high in the morning observations [0830 IST] and low in afternoon observations [1730 IST]; maximum temperature reaches around 3 pm and minimum around 6 am [just before sunrise”
This is about what you’d expect with the temperature dependence of rel humidity.


That’s true of relative humidity (rh%). But what Willis graphed was absolute humidity (g/m3).

That’s true of relative humidity (rh%). But what Willis graphed was absolute humidity (g/m3).

Right, but the amount of absolute water vapor is limited by temp, at night as the temps fall, rel humidity get high and it has to condense water to cool, opposite during the day, takes a lot of energy to evaporate all that water. So this will show up as big changes in absolute humidity, because the air temp has changed. The graph show how night time cooling limits the maximum water vapor.

Dr. S. Jeevananda Reddy

At Mauna Loa, the thermal heat maxima was pushed backwards [from around 1500 hr to around 1200 hr] with the thermal precipitation peak as well humidity and wind speed condition at around 1500 hr. This is the general character of the islands.
Dr. S. Jeevananda Reddy

David Chappell

Hong Kong Observatory publishes a 24-hr time series for its reporting stations. I have just checked the sites that measure atmospheric pressure and all show the same phenomenon, pressure peaking at around 1030 and 2300. So it’s not just because Mauna Loa is up a mountain.
Here’s one example

Lee Osburn

Thank you for that example. But, in order to see the diurnal pressure produced by the sun, the plot needs to begin at midnight. Then the solar forcing on the pressure will fit exactly in the middle of the graph.

David Chappell

Why does it have to begin at midnight and to have the peak in the middle? It’s a rolling plot but the time is there on the X-axis and quite clearly aligned with the peaks.

Lee Osburn

In order to see the forcing wave that occurs at noon, it is best to plot the days so all the noons align. The sun transits at the high point (thus high noon) each day. The pattern it produces when graphed is always going down (oops, sorry but maybe 1 out of 50 does not). This pattern is repeated daily. It is distinctly different from the midnight diurnal which seems to be floating between days according to the background barometric pressure. The word “oscillation” is used in the science papers to describe this pattern.
I am using the voltage from a solar battery charger to monitor the sunlight (different than insolation). The plots clearly show the sky to be the brightest that peaks at 9am. Over the period of sunlight this brightness appears to be causing the pressure to rise and the temperature to quickly rise. Insolation meets the brightness around 10 to 10:30 when the noon forcing begins. They appear to be separate functioning intities.
My best detail plot is using the Barograph app. I plot it on the quarter hour to see the details ,however to see the “real details” of the data I plot it every time the digit changes. There is noise in this amount of detail but it is well below the signal level. I am trying to establish “markers” that show changes that occur when forcing changes the character of the plot. They jump around alot so it takes waiting and plotting a long time to even see them occur.
Of course clouds don’t help. They cause the sunlight to instantly drop until the cloud passes. It then overshoots where it previously was with spikes and then settles back to the original value. For this reason I use a mirror to watch the sky when recording so I can tell the difference between clouds and something else.

Lee Osburn

I must add another feature I see on the plots. After plotting for two years and marking the plots for high noon (always around 12:30 – 12:50 here in central Texas), I draw a line between those marks. The line intersects the floating midnight pattern. After playing with more data, am able to “predict” what the next high noon diurnal mark should be. Of course when a front passes, everything changes. Chaos at work.

Why should the pressure peak at both eleven o’clock in the morning and eleven at night, and be at its lowest just before both sunrise and sunset?

Sometimes it is advantageous to think of extreme examples to try to figure out why things happen as they do. In this case, suppose all temperatures were the same for all heights and then the 2 m temperature shot up by 20 C for a certain reason. This hotter air would push in all directions including into the ground as well as trying to push the colder air above it out of the way. That is why there is higher pressure at 11:00 AM. I know it is not the highest peak, but changes are starting to slow down at this point and the atmosphere is constantly trying to achieve equilibrium.
What about at 11:00 PM? Suppose things were at equilibrium and the air at the 10 m point suddenly shot up by 20 C. This air would expand in all directions and push harder down on the air at the surface, increasing the pressure.
What about before sunrise? Things have reached equilibrium and very little changes have occurred in the preceding hours.
What about before sunset? All temperatures are close to the same so there is relatively little extra pressure due to the hot 2 m air expanding or the 10 m air pushing down. And with the 2 m air rapidly cooling, pressures would be expected to decrease.

@Willis…While it’s interesting to see daily data on precipitation, daily cycles in temperature, the daily cycle of precipitation, average daily wind speed, absolute humidity, and daily atmospheric pressure, I don’t see what that has to do with the overall temperature which is essentially flat since 1977. I think that is the most important thing to focus on, since CO2 has risen steadily since 1977 and the temperature has not risen during the same period at the same measuring site. To focus on CAGW.
\Willis I couldn’t do what you do. But I think the focus should be elsewhere.(AGW).

Regarding Mauna Loa measurements…

” I think that is the most important thing to focus on, since CO2 has risen steadily since 1977 and the temperature has not risen during the same period at the same measuring site. To focus on CAGW.”
Might I strongly disagree, the proof Co2 isn’t the cause of anything, is that the rate of night time cooling hasn’t changed over that time.
There are many things that can cause areas to become warmer, but if the rate of night time cooling isn’t slowing, it isn’t co2.


According to the Mauna Loa site.
the annual average temperature has not risen. See figs1, 5, 6 & table 2 at
in fact show a fall
(in fig 1) a fall of midday temperatures of 0.014°C
but also show of reduction in night time cooling of 0.039°C (indicating a blanket effect ?)
They say (pg 982)
“Based on these values, we hypothesize that the influence of CO2 increase is primarily a
night-time effect.”
but no mention of water vapor !!
“In this paper we have discussed the observed temperature
trends at Mauna Loa b>in the context of global changes in CO2.<b
[ If you only test for X then X will always be the culprit. ]
"There are other potential causes for these temperature trends,
including local changes in wind velocity and cloud cover,
and variations in the occurrences and intensity of regional
synoptic systems. However, because of feedbacks, it is hard
to separate cause from effect. For example, we would expect
changes in temperature to impact and be impacted by cloud
cover, wind, etc.,"

” but also show of reduction in night time cooling of 0.039°C (indicating a blanket effect ?)
They say (pg 982)
“Based on these values, we hypothesize that the influence of CO2 increase is primarily a
night-time effect.”
I’ll have to look at the pdf later, but most of the ones I’ve seen look at calendar days, so they are seeing the cooling from the prior days warming, and not looking at the response from today’s warming, which happens tomorrow.
Also it will depend on what area they are looking at, as the ocean cycles impact this regionally. Collectively about 78 million sample since 1940 show slightly more cooling than it warmed the prior day.
The sample set are all stations in the GSOD data set that record a full year of data.
Did you look at the link I provided? More details are there.

” in fact show a fall
(in fig 1) a fall of midday temperatures of 0.014°C
but also show of reduction in night time cooling of 0.039°C (indicating a blanket effect ?)
They say (pg 982)
“Based on these values, we hypothesize that the influence of CO2 increase is primarily a
night-time effect.”
I took a at the beginning of the papge (they are hard to read on a phone)
But they take cooling separately from warming, especially near the ocean, it will be hard to cool below that floor, so cooling will be at least in part based on how much it warmed the day before.
So if you subtract the drop in warming from the drop in cooling, it still shows about half, but that could easily be from a slightly warmer ocean.
This is off the top of my head, I’d want to go see exactly what my process on the same area says, it too might show the same thing. But when you look at all of the stations it shows a slight cooling over all.
It also shows no accumulated warming, only regional swings in min temp.
And again, go look at my link, there’s a lot of data there, plus links to even more data as well as my code.


Lot of good stuff on your site,… you are now bookmarked.
a few sites you my find interesting


It would be interesting and useful to see the hourly averages also by month.

“I am curious about the co2 data, which the author stated was available since 1976…however, I can’t figure out where that is…any help would be appreciated…Abby” You’ll probably want to click on “Full Record”


There is a note about the adjustments,
“The adjustment increases the likelihood that concentrations will remain above 400 ppm permanently after 2015.”
3 people asked questions, no answer was given, but comments were closed; so much for the advancement of science !!!


Willis, as someone mentioned, it could be interesting to use “longitude time” instead of “timezone time”.


Or solar time, as measured as the longitude at which the Sun shines from zenith.


Yeah, but don´t hold your breath for any change from w. He hann´t amended his post re:”What Powers The Electricity?” yet where he claims hydro contributes 0,4% to the energy mix despite the data he uses says 6,7%. It would be interesting to se his aggreation code that givs him 0,4%.


The dataset probably does not provide that accurate information.


“I was, and I remain, puzzled by this variation. Why should the pressure peak at both eleven o’clock in the morning and eleven at night, and be at its lowest just before both sunrise and sunset? And why would the two peaks and the two valleys be about the same amplitude? That question is why I’m publishing this post.”
Willis, as always, I follow your posts very closely as seem to always present in a manner that allows me to “think” as you do. When I got to your last graph (Figure 5) it seemed to be counter-intuitive to my understanding. After several minutes I scrolled down and read that you were also perplexed. That made me feel better. 😉
As a side note, when stalking/hunting deer in close cover one needs to be prepared for the change in wind direction that will occur around 8:00 a.m. The wind drift will shift to the direction of the rising sun. When that happens the deer will move and it is then time to re-position your stand. A couple of hours later the thermals will change it to up the mountain. Time to move again. For those who don’t know, a deer can smell you for a couple of hundred yards depending on conditions.


It seems the air pressure relates to the temperature, which in turn relates to the humidity. It might be confirmation of thermostatic processes relating to temperature and humidity/rainfall.
Air pressure starts rising with air temperature just about dawn. Then as the air heats up and becomes more humid, the air pressure declines around midday as rainfall and humidity begins to decrease both the temperature and the pressure towards 3pm. Once the air has lost some of its moisture, pressure begins rising again after around 3pm.
But this doesnt seem to explain why pressure falls from around 11pm. This might be from reaching an equilibrium or tipping point once the 2 opposing processes of air pressure rising due to loss of humidity from 3pm begins to cancels out with temperature falling at a point around 11pm.
The equivalence of amplitude of the peaks might relate to how nicely temperature and humidity relate in the tropics, also reflected in the air pressure.

Smokey (can't do much about wildfires)

ONSHORE BREEZE: During the day, as coastal/island land masses heat, air rises which results in winds which flow from above the surrounding (cooler) water over the (warmer) land mass.
OFF-SHORE BREEZE: During the night, the air over land cools to temperatures below the atmosphere over the surrounding water. This results in airflow from the (cooler) shore to (warmer) sea.
I imagine this contribute to your wind speed peaks & troughs as well.




Could the pressure be regulated by wind speeds and cloud formation?


“I was, and I remain, puzzled by this variation. Why should the pressure peak at both eleven o’clock in the morning and eleven at night, and be at its lowest just before both sunrise and sunset? And why would the two peaks and the two valleys be about the same amplitude? That question is why I’m publishing this post.”
It’s a SOLAR tide:

Pat Smith

‘In terms of the climate system, the poorly-named “greenhouse effect” works to increase the surface temperature. Murphy’s Law means that all related emergent, parasitic, and other losses in response to that surface warming will tend to oppose this effect. In other words, we expect the natural response to elevated surface temperature to be one of cooling of the surface.’
Willis, I think you are saying that you would expect the earth’s feedback to greenhouse-forced warming to be less than one. That, if a doubling of CO2 led to an increase in temperature of, say, 1.2C, the effects above would tend to reduce this as would appear logical, not increase it greatly, as some people believe. This seems to me to be the principal point of difference in this debate. Can anyone suggest how this might be resolved?

I don’t believe there’s been any increase due to the in crease in co2


Well butter me on both sides! From my personal AWS. (Belmont, NSW. Australia. Alt ~15m). Around midnight every day.


Around midnight and midday, sorry.

Tom Moran

Here’s something shiny:
Hypsographic demography: The distribution of human population by altitude
The global distribution of the human population by elevation is quantified here. As of 1994, an estimated 1.88 × 109 people, or 33.5% of the world’s population, lived within 100 vertical meters of sea level, but only 15.6% of all inhabited land lies below 100 m elevation. The median person lived at an elevation of 194 m above sea level. Numbers of people decreased faster than exponentially with increasing elevation. The integrated population density (IPD, the number of people divided by the land area) within 100 vertical meters of sea level was significantly larger than that of any other range of elevations and represented far more people. A significant percentage of the low-elevation population lived at moderate population densities rather than at the highest densities of central large cities. Assessments of coastal hazards that focus only on large cities may substantially underestimate the number of people who could be affected.

The Pressure changes are caused by changing directions of the sun interacting with the mountainous landscape, basically the sun is moving the pressure pattern around the mountain where the pressure is being recorded, it seems like a fairly simple explanation imo. Very interesting!


From the data source given by Willis:
Field 5: [HOUR]
The hour is from 0 to 23, and signifies the beginning of the hour.
For example, hour 05 is from time 05:00 to 05:59.
legal time in Hawaii is Hawaii-Aleutian Time Zone ‎(UTC-10)‎
No daylight saving it would seem.


Lat/Long: 21°19’N / 157°51’W ie roughly 160W = 108W -2h
Looks pretty close to solar time.


oops, 160W = 180W -2h ie UTC -10h

James Arathoon

This is my personal take from reading some of the available literature on semi-diurnal atmospheric pressure waves….
The semi-diurnal atmospheric pressure wave gives rise to barometer variations seen most clearly in the tropics have two maxima at around 10 am and 10 pm local solar time. This is the same whether the measurements are made over land or at sea. It is one of the most regular, and thus easily predictable meteorological phenomena that exists (almost a part of astrophysics rather than meteorology). However this level of predictability does not at the same time imply that the phenomena is completely understood.
At high latitudes the semi-diurnal pressure wave becomes much weaker and once the larger swings in pressure brought about by anti-cyclones and cyclones etc., the semi-diurnal pressure wave is lost in the noise, I am told that the semi-diurnal variations in pressure can still be extracted from the data at high latitudes if the hourly time series are long enough and the weather tranquil enough.
Richard Lindzen and Sydney Chapman’s theory (outlined in their book Atmospheric Tides) relies on the existence of gravitationally and thermally forced excitations being most strongly felt in the high troposphere and above. Correlated winds found in the high atmosphere need to be of the right magnitude and speed to give the required surface level pressure variations, that then remain in sync with the local solar time at any given place air pressure measurements are made. Pressure measured at the base of the atmosphere effectively measures the total weight of the atmosphere above our heads at any given time. Thus large pressure variations can only arise via tropospheric winds and associated air mass movements, but smaller changes in pressure like the semi-diurnal pressure wave can still conceivably arise from highly correlated air mass movements and associated winds across a large enough range of levels in the high atmosphere.
As usual the science suffers from insufficient data of the right quality, with seasonal variations that occur in the high atmosphere not well measured or understood. Being able to account properly for the seasonal variations in the semi-diurnal pressure wave (and other linked phenomena) is probably key to a much better understanding overall.
Earlier theories including a development of Lord Kelvins resonance theory are described in M.V. Wilkes (1949), Oscillations of the Earths Atmosphere, Cambridge Uni. Press

You are on the mountain but a kind commenter has shown that the Hilo airport pattern is substantially similar with a lower amplitude afternoon cycle.
Most intriguing suggestion is solar tides. The following animation shows about the right periods, but seemingly less equal.comment image
If it could be slowed down a bit it would be way more fun to convert from UT for Hawaii.
At 100 km in the thermosphere the mass of molecules is so small it is hard to believe they could manage a ~1.25 hPa near surface variation, but if these are the tops of lower amplitude motions throughout the atmosphere…

JJM Gommers

The atmospheric pressure is a result of two flow directions of the wind, the uprising and the normal wind direction. Velocity differencies are converted in pressure. Later in the day vica versa

A bit of a drive by, wrong!!


I don’t want to be an odd bod here but I have flown microlight trikes for over 25 years in the tropics now. In the early morning, around sunrise, the wind always seems to come from the direction of the sunrise, East. In the evenings the wind is almost always calm. In fact sunset was the best time to fly because the wind was almost zero.
My activities were in the Southern hemisphere at around 18 deg South. The speed of my plane is about 55-65 mph and during the daytime I might experience winds of up to 20 mph and in the early morning about but in the late afternoon almost zero.
Travelling East was always a drag in the morning. My understanding was that the air expanded ahead of the sunrise and so caused a headwind when travelling East in the early morning.

Just south of Lake Erie, frequently the winds start up in the mornings, only to stop at sunset.

James Arathoon

For those interested the more recent book by David Edgar Cartwright (1999) Tides A Scientific History, CUP, has a short section on Atmospheric tides in Chapter 10, plus an Appendix D outlining a simplified theory of internal tides in a stratified fluid. He references H. Volland (1988), Atmospheric Tidal and Planetary Waves, Kluwer Academic Pub, Holland as the latest monograph on the subject circa 1999.

Oh, aren’t diurnal air pressure variations the result of tides? The figures I’ve always known were surface pressure highs at 10, AM and PM LST and lows at 4, AM and PM LST, with variations occasionally as much as 4 mb. It seems that your figures are at most an hour different from that rule of thumb.


I worry about the Mauna Loa CO2 measurement method, and whether they have taken account of the humidity and temperature effects on the CO2 fraction. Whilst they freeze their samples, they are taken from air that has varying levels of humidity, temperature and pressure. And AFAIK the hourly data is used only when the rate-of-change of CO2 is at a minimum!
The following datasheet from a Gas Analyser company gives some useful tables as to the quite large changes in CO2 fraction with varying temperature, pressure and humidity:
But slightly off-topic, I also wonder about the accuracy and resolution of the A to D converters in the earlier generations of instruments, plus the use of a polynomial to flatten out the non-linearity of the sensors. And, as the original data-processing was presumably done with by early computing techniques, whether there could be scope for a build-up of assymetric errors [from e.g. roundings]….. that would integrate over time to give a nice smooth upward bias.
The AGW hypothesis is critically dependent on the Keeling Curve, and to my mind it is crying out for a thorough independent analysis into all aspects of the measurements and data-processing, including the history of the process.
Anyway, power to your elbow, as they say!

Measurements at Mauna Loa are hourly calibrated with three calibration gases and every 25th hour with a fourth out of the range calibration gas. The full protocol is here:
Humidity is kept below detection limits by freezing water vapor out over a cold trap (-70°C) and temperature and pressure at the inlet are kept within limits, but are less important as the calibration happens within the same limits. Independent flask samples taken by third parties are in general within +/- 0.2 ppmv.
The use of in-line calibration gases from the early days on was the magic idea of Keeling Sr., which made that the normal drift of the instrument and pressure/temperature changes were largely compensated for. In the early days the result of the measurements was on analog paper rolls, where afterwards the (voltage) measurements of air and calibration gases was measured up and manually calculated… Early measurements were therefore +/- 1 ppmv.


Thanks for your response. I am interested in the concept of finding fault with the Keeling Curve as an instance of the scientific method in action. If there is no discoverable flaw in the method over the history of the series, then it will stand as an exquisite example of science. If there are flaws, then science will have learned something. But as the AGW hypothesis is utterly dependent on this one curve, if it is found to be flawed then the whole metaphysical infrastructure built on AGW will also be flawed.
I understand that the measurement process has gone through several technologies, together with identified problems such as corrosion in the pressure vessels which would have caused changes in the composition of the mixtures. I also understand that the increase in CO2 emissions from increasing building work, and traffic, at the observatory have been identified. But, with no details as to how these corrections were made, and whether they were applied to the historical data, for me there is an open question.
As you say the early measurements were to the nearest +/- 1 ppmv ( as opposed to a % basis), and implies a rounding. AFAIK any subsequent calculation should be done twice; one at the reading minus 1 ppmv, and one at the reading plus 1 ppmv, This needs to be done for every subsequent step in the calculation, to trap out roundings that are also inherent within arithmetical operations (and also in digital computing!). So while this means a proliferation of the numbers that need to be reported, such is the nature of arithmetic, as Goedel tells us. But sticking to only one number, and one calculation per step risks the proliferation of a cumulative bias. As an example taking the number 100 +/- 1, the assymetry between 99 * 99 and 101* 101 when compared to the nominal 100*100 could lead to a subtle bias in subsequent manipulations.
You say that CO2 does not get absorbed much in pure water (or in the water in humid air?) … but I look at those bottles of fizzy water and wonder about the workings of Henry’s law … on a mass as opposed to a volume basis,
Finally, I provided a link to a fascinating PDF which contains a useful set of tables showing what happens to CO2 concentrations at varying levels of humidity, temperature and pressure. So whilst as you say the Scripps system prepares the air ( and presumably the reference gases) at -70 deg C, one wonders what happens to the CO2 levels in samples that start with varying levels of humidity, as they are cooled down.

There was one problem in the history of the CO2 measurements, which influenced all measurements before a certain date (around 1975 if I remember well): In the early days they used CO2 in N2 mixtures to calibrate the measurements, out of fear of rust formation in the gas cylinders. The accuracy of the manometric method to calibrate the mixture was about 1:40,000.
What they didn’t know was that the NDIR equipment shows a slightly different result for CO2/N2 mixtures than for a CO2/air mixtures. After detecting that problem, all equipment was recalibrated with CO2 in bone dry air mixtures and as all raw data still were available, the CO2 levels were recalculated.
After months of use, the calibration gases are tested again to see if there is was a change over time and the fourth calibration gas used every 25th hour is to test for any deviation in the other three…
I don’t think there is a real problem with the calculations themselves, as that is quite straightforward: air is passing during a certain period (20 minutes of one channel, 20 minutes from another channel), where the first minutes are not used to give the time to equilibrate with the ice surface of the cold trap and the rest of the line. CO2 doesn’t enter solid ice, but some may stick to the ice surface. The same for the steel tubes, but within a few minutes that is all in equilibrium. The same for the calibration gases: the first minutes are not used.
The three calibration gases of very accurate known CO2 level then are used to make the (slightly non-linear) calibration curve of the instrument voltage readings where the 10-second snapshots from passing outside air are compared with. In the early days it thus was measuring the voltage on paper, nowadays it is all automatic…
The only error introduced is how much the change in pressure and/or temperature at the entrance of the device changed during the hour from one calibration to the next. As these changes are kept to a minimum, the absolute pressure and temperature at a certain moment is less important, as the calibration gases are supplied at the same pressure and temperature… They measure a ratio, not absolute levels…
As they use averages of 10-second snapshots over 2×20 minutes, the average error in general should be smaller than the absolute error of any individual snapshot (as far as I remember of that kind of math… too long ago). And they keep the standard deviation of all the calculated CO2 levels within that hour.
For daily, monthly and yearly averages, the data are marked for known outliers (downwind from the volcanic vents, upwind from the valleys) and these outliers are not used for averaging. Not that it makes much difference: including or excluding the outliers gives less than 0.1 ppmv change over a year. Again, the error should get less for the averaged data.
Thus there are no subsequent calculation besides averaging the individual results into daily, etc. averages.
I have received a few hours of raw data from Mauna Loa and recalculated what they had done:
The fascinating autobiography of C.D. Keeling contains a lot of information about the way he came to this kind of data quality control, including some mentioning of traffic interference:
In my opinion, the CO2 measurements are an example of how to measure something in nature with the best available techniques and the most rigorous quality control. One can only hope that one day the temperature measurements get the same treatment…


Thanks for the sample data and the reconstruction of the method of initial calculation.
I’d still like to know, for creating the yearly figures they used the method of summing the differences, which if there was any assymetry in the data caused by e.g rounding or inaccuracies in the polynomial correction, would show up as a nice steady increase or decrease.
But there is something you said that bothers me;
“In the early days…….The accuracy of the manometric method to calibrate the mixture was about 1:40,000.”
Now 1 in 40k is 25 ppm!
IF what you say is true, and 1975 was the date they switched from the manometric method, then there is the question that anything prior may need to be discounted, which would have the effect of reducing the curve by 17 years.

The 1:40,000 is on the measurements itself, around 310 ppmv in 1959. That means that a calibration mixture of e.g. 350 ppmv at that time, the measurement error was maximum 0.01 ppmv. Not bad, as Keeling Sr. made his own equipment: he was an experienced glass blower. That instrument still was in use at Scripps until a few years ago and is now in a museum… It was only used to calibrate both any new NDIR equipment and the calibration gases used to continuously calibrate the measurements. It wasn’t used for direct measurements in the field (but it was for flask samples) as it was too much time and labor consuming
As far as I can see, they only use the (“clean”) hourly averages (based on the calibration curve) to calculate the daily to yearly averages. If the variability of the errors is random the error of the new average would be less. If the errors are systematic (by a bad calibration gas or a defect instrument), the error of the new average would be near the same as for the individual samples. I don’t see any accumulation of the errors themselves, but I can be wrong (it is too long ago for me…)
For the calculation, I used a linear response of the instrument to the CO2 levels compared to the calibration gases, that didn’t give more than a few hundredths of a ppmv difference for hourly averages and 0.01 ppmv in the average over the two days of raw data I received, compared to what was published…
That calculation (with a macro, calculating the calibration curve for each hour) is here:
If I remember well, when they detected the CO2 in N2/air problem, they recalibrated all equipment with the new mixtures, that made that they had the change in voltage for CO2 in air vs, CO2 in N2 for each instrument and calibration gas. As they had the raw voltages for every sample and calibration gas in the past, there was no problem to recalculate the real CO2 levels from the raw data.
The same happens today, if one of the calibration gases shows a deviation when recalibrated after use, then all values where that mixture was used are recalculated out of the raw voltage data…

Svend Ferdinandsen

These observations would be a valuable investigation for weather models/climate models.