Back in January, I expressed my doubt about Skjærvø et al (2015) who reported that children born during solar minima had a higher probability of surviving to adulthood, had higher fertility and higher lifetime reproductive success. Skjærvø et al (2015) argued that this was a response to increased ultra-violet irradiation during sunspot maxima. I was not persuaded. In expressing doubt about this paper, I found myself in unusual company. Dozens of news sites, many of which ought to have known better, promoted it. Only a couple of climate skeptics shared my doubt. Now another sceptic, Willis Eschenbach, has got access to the data set, and begun to look at it. The data set includes birth year, lifespan, sex and solar activity. It does not include fertility, reproductive success, socio-economic status or district which are analysed by Skjærvø et al. The data set also has several thousand fewer observations (4941 births in the period 1676-1878) than reported by Skjærvø et al (8662 births). This discrepancy may affect all the results below. When I first discussed this paper, I was concerned that the Dalton Minima in solar activity (1800-1825) may somehow have unbalanced the otherwise ~regular 11 year cycle and caused an artefact. The figure I wanted to see in Skjærvø et al was a plot of proportion of children that survived to adulthood against birth year. With the data available, I am now able to see this.
#load data url<-"d://Norwegian%20Lifespan%20and%20birthyear.txt" ss<-read.table(url, header=TRUE, as.is=TRUE) #remove births not in study window ss2<-ss[ss$Birth_Year>=1676&ss$Birth_Year<=1878,] nrow(ss2)#No births nrow(ss) #process data n<-with(ss2, tapply(Lifespan, Birth_Year, function(x)length(x))) surv<-with(ss2, tapply(Lifespan, Birth_Year, function(x)mean(x>=20))) sun<-with(ss2, tapply(Solar, Birth_Year, function(x)x)) year<-(as.numeric(names(surv))) #plot x11(width=5, height=6);par(mar=c(3,3,1,3), mgp=c(1.5,.5,0)) plot(year, surv, col=ifelse(sun=="MIN", 2,1), ylab="", xlab="Year CE", ylim=c(-0.4,1), yaxt="n") axis(side=2, at=seq(0,1,.2)) title(ylab="Proportion surviving to age 20", adj=.7) abline(v=year[sun=="MAX"], col="grey80", lwd=2) #highlight solar maxima in grey points(year, surv, col=ifelse(sun=="MIN", 2,1), ylab="Proportion survival to age 20", xlab="", xaxt="n") legend(1676, 0.2, c("Maximum","Minimum"), pch=1, col=1:2, title="Solar") par(new=TRUE) plot(sunspot.year, xaxt="n", xlab="", xlim=range(year), yaxt="n", ylab="",ylim=c(0, max(sunspot.year)*3)) axis(4, at=c(0,50,100)) mtext(4, text="Number Sunspots", adj=0.05, line=1.5)
There are some curious patterns in these data. First no years before 1726 are assigned to solar maxima. This is not unreasonable – the first 50 years of the data fall largely within the Maunder minima with low sunspot numbers. However, unlike the Dalton minima, this was not discussed in the paper. Before 1700, every child survived to adulthood. Survivorship then fell untill about 1740 when it reached 0.6, rose slightly in the Dalton minima, and plummeted after 1860. I could believe the data between ~1740 and 1860. Otherwise not. If the patterns in the full data set resemble those in this partial data set, it is clear that the results of Skjærvø et al are an artefact of limitations of their data. Apparently high survivorship in the solar minima before 1726 and apparently low survivorship in the mixed solar period after 1860 will drive their results. Skjærvø et al’s result was surprising and needed to be backed by strong evidence. These data do not provide that evidence. If I get the full data set, I will update this post.