Last month there was some buzz regarding an article in the journal Environmental Research Letters showing a correlation between lightning frequency and the solar wind. The solar wind carries enormous quantities of charged particles from the sun in all directions. Since lightning is an atmospheric process that brings the electromagnetic state into local equilibrium, it stands to reason that the addition of charged particles into the earth system might have an influence on the rate of lightning strikes.
The British researchers looked at data from the Advanced Composition Explorer (ACE) satellite and lightning data in Great Britain. Their study used the median value of the "Vy" variable as a proxy measure of solar wind intensity. They did a lot of work to glean the peak occurrences and to determine intervals on either side of the peak. Make no mistake, they did a robust study. What I present here is not nearly as robust.
There are actually a number of sensors on the ACE satellite so what I did was to select days with at least 1,000 lightning strikes in Alaska during the months of June and July between 1998 and 2011 and to see how the solar wind variables compared to the number of lightning strikes. Those years were selected because that is when the satellite was fully operational. June and July were selected because they are the peak lightning months in Alaska. For ease of analysis, the number of strikes was put into one of 6 categories. (Note: 40% of June and July days saw over 1,000 lightning strikes).
Four variables are mapped. 1) plasma speed, 2) plasma temperature, 3) Vy and 4) Vx. Items 1 and 2 are pretty straightforward. The sun emits charged particles that are one phase hotter than gasses; i.e., plasma. Items 3 and 4 are a little more confusing. They represent how much the ACE sensor was deflected in the X and Y direction by the solar wind using local (gsm) coordinates. The X direction is how much it is deflected toward the earth; hence, large negative values of Vx for a strong solar wind. The Y direction is perpendicular to the X direction. Figures 1 through 4 below correspond to the four variables described in the previous paragraph.
Unsurprisingly, the intensity of the solar variables corresponds to the rotation of the sun about its axis. At the sun's equator, it rotates once every 25 days and at near the poles it rotates once every 36 days. The variables used in this study have a period of approximately 27.7 days.
Clearly, a relationship exists. The trends are unmistakable but the magnitude may be just background noise. This is a cursory study and is certainly not authoritative. Mainly it is just food for thought.
Figure 1. Plasma speed compared to lightning strike groups.
Figure 2. Plasma temperature compared to lightning strike groups.
Figure 3. Vy compared to lightning strike groups.
Note: This is a cross-posing from my super secret FB page.
Interesting notes Brian. We follow things like this (http://www.solarham.net) to understand and predict amateur radio propagation. Lightning and the Auroral intensity can either help or hinder long distance communication. The solar wind plays a role in it all, as do CME's and the SSN during the Solar Cycle.
ReplyDeleteThe paper authors chose a declining portion of Cycle 23 for their analysis (2002-5), and admit an anti-correlation can exist during an unmentioned period during Solar peaks via modulation of the HMF. Viewing data for a whole Cycle, or at least low to low points may have been a worthwhile bonus.
They also note the requirement for environmental conditions capable of supporting lightning events during supportive levels of the solar wind. In concert they infer a relationship can exist. I suppose a WX forecaster could use both parameters if there were time, or better yet, a program that triggered a warning light somewhere in the office.
Gary
I was curious as to the rationale of the British paper's choice of years to use. One this I only briefly alluded to was how significant the effects are. Even if the relationship exists (and is statistically significant) it may be so small as to not have practical implications. Still, it is interesting nonetheless.
DeleteThanks Brian for the topic. It is a good reason to think about the possible connections.
DeleteMaybe looking at parallel Solar Cycles-solar wind-lightning cycles-CAPE (or whatever best mirrors TRW potential)-etc., may offer more clues as Eric suggests.
I didn't explore their bibliography, still should have a look. This is all what I didn't learn in school. But back then, not many knew what they know now, so maybe that's a pass.
Gary
Any kind of change in the electrical conditions of a thunderstorm would have to be created by changes in the Earth's magnetic field. A faster solar wind creates that agitation via magnetic reconnection. I bet that if you plotted number of lightnings vs Bz (the z component of Earth's magnetic field B) you would see similar results. And as a corollary of all this - wouldn't we see a 11 lightning cycle following the sun?
ReplyDeleteYes Eric, the Bz chart looks almost identical to the Vy chart (Figure 3) except that the bulge in 4,000 to 5,000 category is not as pronounced. The geophysics of all this is outside of my area of expertise so I will defer to others on the mechanisms. Thank you for the insight.
DeleteVery interesting, Brian - thanks for bringing this to our attention. How difficult would it be to put error bars on the graphs? I don't doubt the relationships are statistically significant, but the changes are still a small fraction of the overall ranges and it would be interesting to see the distribution for each category.
ReplyDeleteCloud electrification has an influence on precipitation processes, so it would be interesting to search for a signal in storm precipitation totals or daily rain amounts.
The charts have been updated with percentile lines. I'm interested to know how those lines affect the interpretation of the (possible) relationship, if any.
DeleteBrian, thanks for adding the extra information to the figures. The reason that I see the percentile lines as valuable is that the larger the difference between categories compared to the range within each category, then the more significant the result would be. For example, if the percentile lines were very tightly clustered about the average, then the difference between e.g. 1001-2000 and >6000 would be more significant and robust.
DeleteIt certainly looks like the results are significant... perhaps a t-test for difference of means would formalize this.