To make clear, April Volume residuals not saying anything about extent residuals in september, here for 1979-2016
You have plotted April PIOMAS residuals against September NSIDC SIE residuals, but I'm not sure what physical reality this is meant to represent. If I was looking for some correlation between the two data series, I would use a function such as Excel's CORREL.
There are 38 data pairs in the 1979-2016 range. On 11 occasions, each member of the pair is negative, and on 14 occasions, each pair member is positive.
I know this is an artificial example, but try thinking in terms of tossing two coins, and getting either 2 heads{2H}, or 2 tails{2T}. Using the cumulative binomial distribution, the chances of only getting {1H + 1T} on 13 (or less) occasions is less than 1.7%
The fact that there is this degree of agreement demonstrates that there IS a positive correlation, and the p-value derived from the correlation coefficient and the number of Degrees of Freedom subsequently demonstrates the level of statistical significance.
If you consider the direction of the slope on your diagram, you will see that this actually corresponds to a preponderance of the data pairs being either, both negative, or, both positive.