This new paper shows what appears to be a link between Forbush descreases and terrestrial temperature change shortly afterwards. It is a short time scale demonstration of what Svensmark is positing happens on a longer climate appropriate time scale as the solar magnetic field changes with long periods. I’ve covered the topic of Forbush decreases before, and thus I’ll draw on that for a refresher.
A Forbush decrease is a rapid decrease in the observed galactic cosmic ray intensity following a coronal mass ejection (CME). It occurs due to the magnetic field of the plasma solar wind sweeping some of the galactic cosmic rays away from Earth.
Well we have that going on in a dramatic way right now [Feb 19th, 2011], it’s been going on since late yesterday. See the Oulu neutron monitor (a proxy for cosmic rays) graph:
You can monitor it live on the WUWT solar page here.
Nigel Calder reports of a new peer reviewed paper from the Institute of Physics in Belgrade, Serbia which demonstrates a link between such Forbush events and the increase in the diurnal temperature range averaged across 184 stations in Europe. It is quite compelling to read.
Europe: diurnal temperatures after Forbush decreases
A. Dragić, I. Aničin, R. Banjanac, V. Udovičić, D. Joković´, D. Maletić and J. Puzović, “Forbush decreases – clouds relation in the neutron monitor era”, Astrophysics and Space Sciences Transactions, 7, 315–318, 2011.
It was published on 31 August and the full text is available here http://www.astrophys-space-sci-trans.net/7/315/2011/astra-7-315-2011.pdf It’s typical of the pathetic state of science reporting that I still seem to have the story to myself ten days later.
The focus was on the “natural experiments” in which big puffs of gas from the Sun block some of the cosmic rays coming from the Galaxy towards the Earth. The resulting falls in cosmic ray influx, called Forbush decreases, last for a few days. The game is to look for observable reductions in cloudiness in the aftermath of these events. The results are most clearly favourable to the Svensmark hypothesis for the Forbush decreases with the largest percentage reductions in cosmic rays. Scientists keen to falsify the hypothesis have only to mix in some of the weaker events for the untidiness of the world’s weather to “hide the decline”.
The Serbs avoid that blunder by picking out the strongest Forbush decreases. And by using the simple, reliable and long-provided weather-station measurements of temperature by night and day, they avoid technical, interpretive and data-availability problems that surround more direct observations of clouds and their detailed properties. The temperatures come from 184 stations scattered all across Europe (actually, so I notice, from Greenland to Siberia). A compilation by the Mount Washington Observatory that spans four decades, from 1954 to 1995, supplies the catalogue of Forbush decreases.
The prime results are seen here in Dragić et al.‘s Figure 5. The graphs show the increase in the diurnal temperature range averaged across the continent in the days following the onset of cosmic ray decreases (day 0 on the horizontal scales). The upper panel is the result for 22 Forbush events in the range 7−10%, with a peak at roughly +0.35 oC in the diurnal temperature range. The lower panel is for 13 events greater than 10%. The peak goes to +0.6 oC and the influence lasts longer. It’s very satisfactory for the Svensmark hypothesis that the effect increases like this, with greater reductions in the cosmic rays. The results become hard (impossible?) to explain by any mechanism except an influence of cosmic rays on cloud formation.
To be candid, these results are much better than I’d have expected for observations from a densely populated continent with complex weather patterns, where air pollution and effects of vegetation confuse the picture of available cloud condensation nuclei. Svensmark’s team has emphasised the observable effects over the oceans. Now the approach taken by the Belgrade team opens the door to similar investigations in other continents. Let a march around the world’s land masses begin!
Physicist Luboš Motl also writes about the new paper:
What have they found? If they take all Forbush decreases, the effect is insignificant. However, if they compute the average of the largest Forbush decreases, they find a substantial increase of the day-night temperature difference by as much as a Fahrenheit degree around 3 days after the event [reference to Figure 5 above].
…
A higher day-night temperature difference indicates that the number of clouds is smaller – because clouds cool the days but heat up the nights a little bit, and thus reduce the temperature difference – which is in agreement with the cosmoclimatological expectation: the Forbush decreases makes the galactic cosmic rays disappear for some time (because of some massive, temporarily elevated activity of the Sun).
I think it’s both simple and clever to look at the day-night differences because the overall noise in the temperature is suppressed while the signal caused by the clouds is kept. Just to be sure, it’s obvious that clouds do reduce the day-time differences but that doesn’t mean that they preserve the day-night average. At typical places, they cool the days more than they heat up the nights.
For me, this paper begs replication and confirmation. The problem they have with the European data set is that it is noisy which required the averaging. Here in the USA though, there’s a dataset that may work even better, and that’s from the recently completed U.S. Climate Reference Network operated by the National Climatic Data Center. While that network is too new to be useful yet for long term climate studies, the care that was taken for station siting placement, accuracy of sensors, data resolution, and quality control make it a perfect candidate for use in replication of this effect.
These stations were designed with climate science in mind. Three independent measurements of temperature and precipitation are made at each station, insuring continuity of record and maintenance of well-calibrated and highly accurate observations. The stations are placed in pristine environments expected to be free of development for many decades. Stations are monitored and maintained to high standards, and are calibrated on an annual basis.
The data is of high quality, so any new study looking for this effect may not even need to do the DTR averaging done by Dragić et al. to see the effect if it is real.
The logged USCRN data is now available online here http://www.ncdc.noaa.gov/crn/observations.htm The February Forbush decrease event I highlighted at the beginning of this post might make a good starting point.
I see a paper on this in the near future, maybe even in Dessler record time.
“Leif Svalgaard says:
September 12, 2011 at 7:05 am
The 13 strong FDs cover only about 1/1000 of the time over which the test was made, so can hardly be taken as strong support for an effect.”
Since the increase directly associated with [CO2] as determined by physics in a closed chamber analysis is only ~1/300 C, this would represent 30% of the alledged warming.
Bengt A says:
September 12, 2011 at 8:48 am
What does it matter if there are 5, 13 or 100 FDs as long as we are positively sure that the measured effect is a real effect (and not because of some systematic error etc)?
It has to do with the size of the effect. If it takes a 10% FD to show effect above the noise, while smaller ones don’t make it, then it is harder to argue that the smaller variation of cosmic rays is a significant driver. In addition FDs show up mostly in lower-energy cosmic rays rather than in the high-energy ones that are the only ones that are supposed to have effect.
The result of Dragic et al. is clear enough, or do you have an alternative explanation for the shape of those graphs?
As the FDs are global, the effect should be global too. That is something to look for. So far, it looks like confirmation bias to me. If you look around and only publish what fits your opinion you can find many correlations. One thing that is a problem is their calculation of the error bar. They say it is the standard error of the mean. With 13 FDs the standard deviation would then be the square root of 12 times larger [that is 3.5 times], but the standard deviation of what? Of the 184 station means? which themselves have an error bar. In any event, 13 cases is much too small a number to do statistics on. As I said, I would like to see the individual cases plotted on the graph as well. Perhaps the data is there, but it is up to the believers to supply the evidence. It dosn’t matter what I think. To quote Steven Weinberg: “I have a perfect record of not having anybody changing his mind”. My comments are directed to the lack of evidence to convince me. Other people may have a much lower bar.
Mr. Mosher
You should look at the equatorial region where there is strong magnetic field generated by the night-time equatorial electrojet which in turn is generated by the solar wind. The CR impact may be modulated by the intensity of the electrojet (?), and so affecting change in the cloudiness. Position of the electrojet is governed by the magnetic equator; NASA has shown there is a connection in the storminess and magnetic equator, and there is also some correlation to the global temperature records. For more details see:
http://www.vukcevic.talktalk.net/LFC20.htm
@Mosher,
This sounds like you have some seriously cool data on hand. Let’s see if I can take this research here and make a hypothesis for you to test.
So, if I understand this cosmic ray theory well enough, and the implications for rapid changes in diurnal temperature ranges in response to big comic ray fluctuation events (in theory)…
I would think it doesn’t matter if any one site was cloudless and remains cloudless, but the average cloudiness over all sites. If this event effected cloud formation, we should see less cloud systems coming into the coverage area, and more clouds leaving, for an overall loss of cloudiness. Now, if it also rains prior to the period, that should possibly decrease cloudiness too(? but decrease DTR?); so we need to look for a decrease in cloudiness not associated with prior precipitation, but associated with increasing DTR? Put more strongly, if the cosmic rays are having a reciprocal effect on cloud formation, we should see average cloudiness decrease during the mass ejection event without a prior increase in precipitation, followed by an upswing in clouds a few days after the event (possibly with more precipitation). Thus, there’d be an increase in cloudless and dry during/in response to the event, and increase in DTR? If the hypothesis is correct; so we should see the opposite or not change if has the null to invalidate it.
I know you said the record isn’t long enough for a good DTR average. But, even if that’s so, if looking across all stations, we still might see a response of increased DTR if this hypothesis about cosmic rays forming clouds is correct.
Wind patterns might also be interesting, as it can show the movement of the clouds between stations, and out of/into the whole station coverage area. But gees, this is already such a ton of work. And unfortunately, I don’t know quite enough about cloudiness vs precipitation (other than it has to be cloudy to lead to precipitation, but afterwards effects on cloudiness is where I am unsure) to really put forth a good way to test for the GCR out of all the weather noise.
So to summarize: the hypothesis here is if the GCR are leading to cloud formation, then this event that lowered GCR should lead to a few day burst of decrease cloudiness and precipitation across the whole coverage area of sites (more land cover being investigated the better, as any one site might have no clouds for the duration of the investigation), without an increase in precipitation prior to the cloud loss, and an increase in DTR for the whole coverage area. If we see no change or an increase in cloudiness and precipitation, no change or decrease in DTR (precipitation would decrease DTR due to the wet night?), this would invalidate the hypothesis for this testing area and event. The amount of days to look at this over would be the same as in this paper.
I hope this made sense and sounds good!
@Mosher,
I guess by increase/decreased cloudiness, we’d actually be looking at a decrease/increase in hourly radiation data for that hour verses the average? I hope I am understanding your data sets well enough to put forth useful ideas.
M.A.Vukcevic says:
September 12, 2011 at 10:09 am
You should look at the equatorial region where there is strong magnetic field generated by the night-time equatorial electrojet which in turn is generated by the solar wind.
A little knowledge is a dangerous thing. First of all, any external magnetic fields are hundreds of times smaller than the Earth’s field, so no “strong magnetic fields”. Second, the equatorial electrojet is a day-time phenomenon generated by solar UV [as the ordinary diurnal variation], but magnified by the field lines being horizontal at the equator [charged particle like to move easier along field lines than across], no a night-time thing and not generated by the solar wind. There is a ‘ring current’ due to particles drifting in the Van Allen radiation belts, but that one has the same magnitude all over the Earth [as the Earth is a the center of current that is much larger than the Earth itself.
Leif S
Dragic measures the change of DTR in response to FDs and that’s a parameter totally irrelevant to climate. Thus I find it hard to draw any conclusions from the size of DTR response. Their result is purely indicative, not quantitative. We need other scientific approaches to find out if and to what extent this effect is of importance to climate. Hopefully Kirkby and CLOUD will help out, at least in part.
Dr. Svalgaard,
“In addition FDs show up mostly in lower-energy cosmic rays rather than in the high-energy ones that are the only ones that are supposed to have effect.”
So does that mean that FDs only show up in the higher lattitude data, or that it takes stronger FDs to impact the lower lattitude data?
The sun’s magnetic field would have a greater distance over which to influence even the higher energy GCRs than the earth’s field, but FDs due to directional CMEs would also have less distance in which to deflect GCRs than an active sun’s expanded wind and magnetic influence. Perhaps FD data at lower lattitudes could be compared to active sun background levels at those lattitudes.
~1/300 K
Steven Mosher says:
September 12, 2011 at 8:19 am
“Sorry nice try. The problem is that this dataset doesnt have enough years of data to get a stable DTR average.”
My uneducated guess would be that for looking for abrupt changes, running average or gaussian filter could provide sufficient normal. You’re looking for weather events rather than climate changes so you don’t need to establish climatic normal. Sure you’ll have some outliers because clear sky can’t become any clearer and completely cloudy any cloudier – I just somehow doubt it will be a frequent case on all stations.
Ged:
http://stevemosher.wordpress.com/2011/09/12/forbush-events/
This will give you all a taste of the data that I have. thats HOURLY radiation for every day in feb, 2011 at one station. ( I plotted a bunch of others as well– ignore that it says jan.. its feb )
Charts are not pretty yet and the data strutures are just coming into shape.
Here is what I do not want to do. I do not want to take a bunch of my time and create an analysis of this data only to hear a bunch of nonsense objections. So, I’m asking people who believe in this effect to lay down a testable hypothesis. basically, define the method of analysis you would use to test whether the event on the 19th of febuary lead to a decrease in cloudiness. If people think about this problem clearly they will see some of the challenges…
I’m gunna putter around with this for a day or so and maybe publish the code as an R package.
Hourly data is NOT FUN.
Bengt A says:
September 12, 2011 at 10:27 am
Dragic measures the change of DTR in response to FDs and that’s a parameter totally irrelevant to climate. Thus I find it hard to draw any conclusions from the size of DTR response. Their result is purely indicative, not quantitative. We need other scientific approaches to find out if and to what extent this effect is of importance to climate. Hopefully Kirkby and CLOUD will help out, at least in part.
To the extent that climate is average weather, I would say that the response is relevant. If not, then why are we discussing this in terms of climate, or rather: why are the believers so sure this settles the climate debate [Nobel Prize to Svensmark, e.g.]. Kirkby himself has stated their their result from CLOUD says nothing about any climate effect.
Martin Lewitt says:
September 12, 2011 at 10:41 am
So does that mean that FDs only show up in the higher lattitude data, or that it takes stronger FDs to impact the lower lattitude data?
The Earth’s magnetic field filters out the low-energy cosmic rays at low latitudes, so the FD at high latitudes are about ten times stronger than at the equator.
Perhaps FD data at lower lattitudes could be compared to active sun background levels at those lattitudes.
The size of an FD [and intensity of cosmic rays in general] is determined foremost by the filtering by the Earth’s magnetic field. Not the Sun’s.
Bengt A says:
September 12, 2011 at 10:27 am
Dragic measures the change of DTR in response to FDs and that’s a parameter totally irrelevant to climate. Thus I find it hard to draw any conclusions from the size of DTR response. Their result is purely indicative, not quantitative. We need other scientific approaches to find out if and to what extent this effect is of importance to climate. Hopefully Kirkby and CLOUD will help out, at least in part.
To the extent that climate is average weather, I would say that the response is relevant. If not, then why are we discussing this in terms of climate, or rather: why are the believers so sure this settles the climate debate [Nobel Prize to Svensmark, e.g.]. Kirkby himself has stated their their result from CLOUD says nothing about any climate effect.
@ur momisugly Mosher,
That does not look like fun data to deal with, at all.
I’m assuming that higher amounts of Z are higher levels of solar radiation reaching the station?
I wish I had a better handle on cloud dynamics to give you a more solid hypothesis to attempt to disprove (as I do not want you wasting your time!). But, what I said above is the best I can think of at the moment. This data is so noisy though, I am not sure what tests to do to allow it to be used to test the hypothesis of decreased cloudiness with decreased GCRs. Maybe those “believers” in it you mention will come up with something better than I.
Still, I suppose we should see, some days after the 19th in the same way they saw in this Dragic et. al. paper, an increase in the average Z scores for that span of days, verses the Z scores for the previous span of equal number of days; if this hypothesis is correct, and thus we won’t see this if the hypothesis is invalid. This should be visible for the whole station coverage area, as there is way too much variability for each individual station to see anything but noise, from what your graphs suggest. If this analysis can accurately test the hypothesis fully, in either way, I have no idea; just so much noise.
Leif Svalgaard says:
September 12, 2011 at 10:23 am
……………..
‘night-time’ was lapsus linguae meant ‘day-time’, the rest I will reluctantly agree. Since you have no corrections to my previous post, one would assume you do not ‘disagree’, which may count as a minor miracle.
Ged.
yes. above you will see me talking to tallbloke about the issue of “it cant get sunnier”
Simply: Imagine a station that is bright and sunny on the 19th. Now suppose that a decrease in GCR leads to a decrease in clouds.. WELL, the stations that are already sunny on the 19th
cannot get sunnier! they can either stay the same or get cloudy. simple logic.. But those cases
will play havoc with detection.
I think the approach of looking at DTR is too indirect, its one step away from what you want to prove.. which is that a decrease in GCR leads to a decrease in cloudiness.
I need to think about an optimal detection strategy..
M.A.Vukcevic says:
September 12, 2011 at 11:34 am
Since you have no corrections to my previous post, one would assume you do not ‘disagree’, which may count as a minor miracle.
Didn’t even look, sorry to say…
M.A.Vukcevic says:
September 12, 2011 at 11:34 am
‘night-time’ was lapsus linguae meant ‘day-time’, the rest I will reluctantly agree.
The day-time electrojet is not generated by the solar wind.
Crispin in Waterloo (September 11, 2011 at 10:23 pm)
Thank you again for an interesting discussion. I agree that UV-opaque clouds deserve more consideration.
We can in part thank Dessler and the Team for that with their unidirectional model where clouds of any spectrum opacity are only a product of temperature. They acknowledge that hygroscopic particles, modes and thresholds of nucleation, residence times of various aerosols, and such are components of cloud formation but they contend that such things don’t change, all things being equal. In other words like clouds just exist on their own as you put it. Since clouds can’t vary independent of temperature, any discussion of the mechanics behind how clouds form or their spectrality is interesting but purely academic. From their view, Svensmark, Spencer, the Serbs, and anyone else researching independent mechanisms like GCR’s and FB’s are on a fool’s errand. As such, clouds are treated in the models as if they’re just there doing their positive feedback thing exasperating the whole CO2 vicious cycle.
I’ve ordered a case of pop corn as the I think Svensmark, Spencer, and a few others in the “clouds can vary independent of temperature” bidirectional camp are getting locked and loaded. It has already become interesting and I think one more paper that connects all the dots will have Dessler and the Team behaving like the Keystone Cops as they attempt damage control.
@Mosher,
Very true.
I would think that if we have a big enough coverage area, that should allow us to see the signal. The other issue of course is weather patterns, as clouds usually “flow” in bands. Too small a coverage area (or single station) might get stuck in a band of clouds from say a low pressure system butting up against a high pressure for several days, and potentially overwhelm any small GCR signal as this Forbush should(?) be. We need something really large scale to both exceed weather systems, and average out stations that do not change over the period of time in question: some of the area range cloudy, some of it sunny.
I’m thinking we may need continent scale (i.e. the entire continental USA) to get a good signal for testing, but it might be possible to do something like the entire eastern sea board to the Appalachian mountains. I’m not sure what area these sweet stations are covering, or exactly how much area we’d need.
But yes, I think you’d have to average all the stations together to get an “area” signal for radiation. As clouds cover “area”. Rather than looking at any station directly. That’s just my idea on it.
@ur momisugly Mosher,
Urg, the more I think about this, the more impossible a task to lay on your shoulders this seems. Since, the entire USA can be practically cloudless for long periods of time, or almost completely cloud covered; all based on the fluctuations of global weather patterns and the temperature variation fluxes from the progression of the seasons. I don’t know how we can disprove the hypothesis with this test if any signal we see could just be a global level weather system blowing the clouds off the country. Adding meteorological analysis on top of it all.. is just insane amounts of work.
There still might be a way to scrape the signal out of the data or definitely show it isn’t there, especially with a lot of statistical data sets to work with. But that could mean we’d need to statistically analyze a -lot more- of these Forbush events, and put it all together. I’m not sure…
You are very clever and know R well, so maybe you have some other tricks up your sleeves that can power through this using this single event. Good luck, and don’t burn too much of your time on this!
Keep in mind that a Forbush Decrease is the opposite of the CLOUD experiment at CERN.
With the CLOUD experiment, the team establishes the atmospheric conditions (temperature, pressure, water vapor, and trace constituents), turns on the cosmic rays (protons of relativistic speed), and measures the production of cloud condensation nuclei.
With a FD, we are looking for a decrease in the amount of already formed clouds following a decrease in the production rate of replacement CCN. This also explains the time delay. It takes time for cloud droplets to first loose size and then completely evaporate.
So, Leif.. I should see a modulation of the effect dependent upon latitude
Steven Mosher says:
September 12, 2011 at 12:25 pm
So, Leif.. I should see a modulation of the effect dependent upon latitude
I don’t think so, as the GCRs that are claimed to have effect are the high-energy ones that have little latitude dependence. The Earth’s field lets those through at all latitudes.
Wasn’t a similar effect observed over the USA in the aftermath of 9/11, when all aircraft were grounded for a couple of days?