First and Last Song Dates

by Carl Strang

I now have 7 years’ data in which I have noted the first and last dates on which I heard each singing insect species. This year was characterized by a mild winter followed by a warm spring and then a summer of drought. The mild winter and spring apparently were responsible for this year’s early phenology. First song dates were the earliest I have recorded in DuPage County for 17 of the 21 species for which I have 7 years of records. The chi-squared value of 77.33 (with an expected value of 3 species per cell for each rank of earliest to latest) is, of course, statistically significant.

The greater angle-wing started earlier and finished earlier this year than in any of the previous 6 years.

The greater angle-wing started earlier and finished earlier this year than in any of the previous 6 years.

As for last song dates, singing insects generally finished early this year. Of the 20 species for which I have 7 years’ data, 15 had their earliest or second-earliest ending dates, and the chi-squared value was a statistically significant 22.68. This was not a particularly cold or dry late summer and autumn, so the implication is that singing insects have a fixed rate of attrition or duration of song season, so that an early start results in an early finish. The 12 species for which I have the best, most reliable records do have differences in observed song season lengths (ranges for the 7 years, and ranked lowest to highest: 18-42 days for Roesel’s katydid, 16-52 days for gladiator meadow katydid, 52-96 days for the greater angle-wing, 58-96 days for snowy tree cricket, 67-91 days for the scissor-grinder cicada, 64-94 days for the greenstriped grasshopper, 62-109 days for Linne’s cicada, 72-105 days for the dog day cicada, 81-107 days for the common true katydid, 107-139 days for Allard’s ground cricket, 111-141 days for striped ground cricket, and 113-143 days for Carolina ground cricket).

The number of clear, cold nights seemed high enough in November that they might partly explain the early conclusion of common ground cricket songs this year, especially given the recent study by MacMillan et al. (2012) indicating that there is a metabolic cost to recovering from cold-temperature paralysis. However, I found no significant relationship between last song dates and the number of November days with low temperatures below 33F over 2006-2012 for any of the three species (Spearman’s r values 0.51 for Allard’s ground cricket, -0.39 for striped ground cricket, and 0.33 for Carolina ground cricket). It is interesting, though, that I have seen a few red-legged grasshoppers active a couple weeks after the last ground cricket.

An early December red-legged grasshopper

An early December red-legged grasshopper

This bigger insect may have larger fat reserves to draw upon and so extend its season.

First and Last Song Dates

by Carl Strang

I have one last topic from 2009 to report in my singing insects study. I now have enough years (4) of data to look for patterns among species in first and last song dates. I am reasonably confident that my records for 20 species are good enough to include in this comparison. The object of this analysis is to see whether this year’s singing insect season is relatively early or late compared to other years.

First song dates at first glance seem to point to 2009 as a relatively late year compared to 2006-2008, with the earliest song date recorded for only one species, second-earliest for 3 species, second-latest for 9 species and latest for 7 species (a random distribution would have 5 in each category). A chi-squared contingency test produces a test statistic value of 8, however, not quite high enough to indicate statistical significance (P > 0.01; it is perhaps worth mentioning that if I were content with a 5% chance of error rather than 1%, statistical significance would have been indicated). The last song dates I noted for the 20 species in 2009 were closer to an average or random pattern within the 4 years, with 5 earliest finishes, 6 second-earliest, 3 second-latest and 6 latest. The test statistic consequently is very small here, at 1.2. This is only the first year of this analysis; I’ll be interested in seeing what emerges in future years.

%d bloggers like this: