Guest Post by Willis Eschenbach
Inspired by a claim made on WUWT that
A new study led by Professor K.M. Hiremath of the Indian Institute of Astrophysics shows the strong, possibly causative correlation between variations in solar activity (red curve) and in monsoon rainfall (blue curve) in Figure 1.
I decided to see what I could find out about the Indian monsoon. I thought I might tell the story by describing the path I walked. I started out with a huge advantage—I knew nothing about the timing, size or pattern of the Indian monsoon rains, other than that they occur during the summer. I knew that when land gets hot in summer, hot air rises, wet air is drawn in from the ocean. Result? Monsoon rainfall. Other than that I had no knowledge, a great advantage in exploration of a new dataset. Preconceptions are the enemy of science …
A Google search led to the Hockey Schtick, which fortunately reproduced two pages from the study. In turn, that identified the data source as the Indian Institute of Tropical Meteorology. Another google search located the IITM data page. I used the “Sontakke” dataset for this analysis, the page is here.
Now, on that page the rainfall data for India is divided into seven regions, three on the peninsula and four in the north. The Hiremath et al. study used all three peninsular datasets, plus the north-west region. Simply because it was higher in the list, I started by analyzing the dataset for North West India (NWI). The Sontakke dataset ends in 2006, unfortunately, but it’s what the authors used. Here are the last two decades of that record:
Figure 1. Rainfall in the North West region of India over the two decades 1987-2006. The monsoon rains come in June, July, August, and September. The top panel is the raw data. The middle panel is the average monthly component of the rainfall. The lower panel is the raw data minus the seasonal component. The trend over this period is not statistically significant. The blue line is the loess average of the data.
Yes, there is a huge difference between the four monsoon months and the dry two-thirds of the year. There is also no statistically significant trend in the two decades of data. Having seen that, I took a look at the same analysis, but for the entire period of the data.
Figure 2. As in Figure 1, but for the longer period of 1844 to 2006.
What stands out in the full dataset is the lack of much long-term variation at all. There is no significant long-term trend to the data, nor any obvious variation over the century and a half of data.
Is there a solar signal in there? Well … perhaps, but if so it’s neither large nor obvious. However, this shows the whole year, not just the monsoon months (JJAS) analyzed by Hiremath et al.
So I then looked at just the monsoon months. Why did I not start with just the monsoon months? Because I first want to see the entire signal before I start sub-setting it.
In any case, here is the total rainfall of just those four months, year by year, compared with the sunspot record.:
Figure 3. Rainfall and sunspots. Upper data is the total of the monsoon rains (JJAS) for that year. Lower data is the total sunspot count by year.
At first glance, that looks kinda hopeful … but a closer examination shows that significant correlation simply doesn’t exist. Perhaps the simplest way to demonstrate this is the cross-correlation function. This shows the correlation at a variety of lags, with positive lags showing rainfall lagging the solar changes.
As you can see, the correlation at all lags is trivially small.
Now, does this show that the paper by Hiremath et al. is wrong? By no means. They looked at an average of four monsoon areas, and I’ve only looked at one of them, NW India. I haven’t examined the rest. However, it isn’t looking good for the solar theory.
I also haven’t taken a close look at their formula relating the solar activity to future rainfall. Why not? Well, I don’t have their paper.
In addition, I couldn’t verify the following from the abstract of the paper:
Those internal forcing variables are parameterized in terms of the combined effect of external forcing as measured by sunspot and coronal hole activities with several well known solar periods (9, 13 and 27 days; 1.3, 5, 11 and 22 years).
Instead, here’s what I found for the cycles inherent in the data.
Figure 5. Periodogram of the NW Indian Rainfall.
Note that there are no long-period cycles that are larger than two percent of the peak-to-peak swings of the data, so we’re way down in the noise. This is trivially small. In addition, there are no peaks at the periods mentioned of 1.3, 5, 11, or 22 years …
So, that’s the investigation to date. Still lots to do. I’ve only looked at one of the four datasets so far. Hiremath et al. looked at the total for the four areas. And I don’t have a copy of the underlying Hiremath paper, so I’m doing my own investigation.
Sadly, my time is quite short these days. I’ve taken a new job, as is my wont, but it has bizarre hours—5:30 AM until 2 PM. Now me, I’m a night owl by nature, so this has played havoc with my available time.
In addition, I’m doing strenuous physical work. We’re doing a rebuild on the lobby and facade of the local movie theater, and there has been a lot of demolition work. The lobby has 11-foot (3.3 metre) ceilings, which we’re ripping sections out of and rebuilding. Much of the work is up in the ceiling, or working up on a two level scaffolding … another part of life’s rich pageant. Now me, I’m sixty-seven, what I call my “middle youth”, so this is, well, somewhat consuming.
Not that I’m complaining, mind you. People sometimes ask me what I do for exercise … I say “I don’t pump iron … I pump wood.” So I’ve been spending my days in the gym, climbing ladders, doing the low crawl in the ceiling, and pumping wood.
However, no matter how cardiovascularly stimulating my work might be, to date it has cut heavily into the time for my climate research and investigations. So, for a while at least, I’ll be contributing less to the discussion.
My best to all,
The Usual: If you disagree with someone, please have the courtesy to quote the exact words that you disagree with. That way all of us can be clear exactly what you are objecting to.
Data and code: The data (including the three datasets I haven’t analyzed yet, plus other indian rainfall datasets) and R code are all in a zipped folder called “Indian Rainfall Folder“. To run the code, set your R workspace to the folder, and it should be pretty turnkey.