Feb 26 - days and numbers
Still thinking about survey and the number of observations to represent species. A different look is to start with individual days and plot number of species against number of observations on those individual days. Using recent years data (2021-2023), for our busier months (Apr-Sep), this is 524 days. There is data for every day for all three year in Jun-Sep, some missing days in Apr-May. Here's the chart:
The dots are color coded: Blue is Apr-May, Green is Jun-Jul, Orange is Aug-Sep. Remember that this is the same number of days (183) for Jun-Jul and Aug-Sep. Slightly less for Apr-May (158).
As expected, Jun-Jul show the highest numbers. Good diversity on the wing and people looking.
Apr-May is clustered on the lower left. While not much to see early on, there are some higher numbers in late May.
Aug-Sep is clustered generally higher than Apr-May, but well below Jun-Jul. No days above 50 species, even with higher numbers of observations.
Comments
Isn't Apr-May clustered on lower left?
right - everyone needs an editor! I'll correct.
My pleasure to be of service, Sir ;-)
Some thoughts on this plot follow:
The very general trend (curve shape) for each of the three time periods seem to be very similar, but of course there is quite a bit of vertical spread for each period (I think this is due mostly to the observation variability noted in the 23 Feb note).
There seems to be an obvious diminishing returns pattern (the increase in species doesn't keep up with the increase in obs...a general linear relationship near the beginning becomes asymptotic as obs increase). This makes sense...there is a limited number of species, but almost no limit to the number of observations that could theoretically be made.
Instead of using obs/day, it might be interesting to normalize obs/day by the number of observers...in other words calculate obs per observer per day...this would reduce the impact of outliers (observers that underachieve and those that overachieve)
Instead of using species/day, it might be interesting to normalize species/day by the number of possible species for that time of year...in other words calculate percent possible species per day and plot those against obs/day or obs/observer/day...One could use the data from the 23 Feb curve to do this and it would be good to use the finer time divisions (10 days vice 2 months).
I agree with jheiser that there is an important seasonal variation in observing behavior, but I don't think it is possible to effectively account for that in a dataset that heavily depends on volunteers (with all levels of skill and commitment). Maybe the observers need more guidance on what to report as observations. For instance, I don't record every different eastern pond hawk I see at a specific location in a day...don't really see the point. On any given day I do try to record as many species as I can. Also there is the fliers vs. perchers issue. Fliers take a lot of extra effort and some skill, although they do land at times of course.
The 23 Feb species curve is interesting. The front end looks like a normal/gaussian curve to me, but the back side is significantly different. The front side represents birth/arrival of the species and the back side represents their demise or departure. If each species had the same average life span I think we would see a complete typical normal/gaussian curve, but since that isn't true, the back side of the curve is significantly different. Of course I have no idea why the front side of the species curve should be shaped like a normal/gaussian curve.
Add a Comment