The first thing to notice in the updated trend is that a linear trend still best explains the rate of change in species accumulated over time. I have added the R-square value for stats geeks who want to know...in essence, R-square is the amount of variation in the rate of species accumulation explained by time (i.e., days spent birding). The red dots, as previously explained, represents current record holders Alan Davies and Ruth Miller; a fourth-order polynomial trend line explains 99.4% of the variation in their species accumulation over time. The blue dots, which represent Noah's data, indicate that after six months a linear trend explains 98.1% of the variation in Noah's species accumulation trend. Of interest in Noah's trend is that he started out relatively stable, with a small dip as he approached 2,000 species. The pace then resumed up until about 2,500 species when the accumulation became, well, erratic. In fact, there are two particularly steep sections of the accumulation curve between 2,700-3,000 species, and 3,050-3,400 species. Those must have been some amazing birding days! When I read about Noah sitting in New York waiting for a replacement plane however, and adding nothing to the list...well, I think that is self-explanatory.
Date of Estimate
|
Estimated Day of Hitting 5,000
Species
|
Accumulated Number of Species
|
January 15
|
August 16
|
333
|
January 30
|
July 18
|
686
|
February 14
|
July 19
|
1025
|
March 1
|
July 22
|
1404
|
March 16
|
July 30
|
1689
|
March 31
|
August 10
|
1871
|
April 15
|
August 27
|
2029
|
April 30
|
August 27
|
2349
|
May 15
|
September 6
|
2493
|
May 30
|
September 14
|
2699
|
June 14
|
October 3
|
2942
|
June 29
|
October 7
|
3322
|
The next visualization I thought was interesting is illustrated below. It shows the difference in total species between Noah, and Ruth and Allen, for each day they both reported totals. The data are plotted as Noah's position relative to Ruth and Allen; in other words, anything below the zero line indicates that Noah had accumulated fewer species than Ruth and Allen up to that point in time. Conversely, anything above the zero line indicates that Noah has accumulated more species than Ruth and Allen up to that point in time. There was clearly a turning point for Noah at about day 31, when Noah's rate of increase began to close the gap on Ruth and Allen. Perhaps most intriguing though is that the pace of closing (and subsequent exceeding) has been relatively constant over time. At about day 92 Noah surpassed Ruth and Allen, and the gap has only widened ever since to a currently whopping total of 736 species. Another way to look at it is this - Noah's total on June 30 is similar to Ruth and Allen's total on October 7 (within 21 species).
The last figure provided below illustrates a 7-day moving average of new species added to Noah's big year. No doubt there is a lot of day-to-day variation, and there is a reasonably predictable cycle which without testing my own suspicions, probably relates to moving to new geographic areas that cause "releases" from staying in any one area for two long and suffering from diminishing returns. As I indicated in my first post, the key to Noah's success (and for anyone who might try in the future) is to strategically move around the globe at the right time and to the right places so that the species accumulation curve seldom gets a chance to asymptote (i.e., plateau). So far, Noah has done a pretty good job of doing that. Interestingly, out of 181 days of birding, he has had just one day where he added more than 100 species to his year list; 95.6% of the days have resulted in 50 or fewer new daily additions.
Lastly, as I alluded to at the start of the post, is a prediction of how many species Noah will see by the end of the year. I think hitting 5,000 is inevitable, so assuming he keeps going beyond his goal, until December 31, the big question is: How many species will he see? Using the linear regression model that creates the trend line in the top-most figure, I estimated that on December 31 his total will be 6,474 species. The lower and upper 95% confidence intervals of this prediction are 6,455 and 6,492 species respectively. For those who are submitting guesses to Audubon, use this as you will...it will probably change :)
Until next time...happy birding...and number crunching.
I wonder if a more accurate assessment of his total species is the percentage of possible species that were seen. For the Western hemisphere, he reported 2747 species, or 64.41% of total species reported to ebird so far this year, but more importantly 63.74% of the total reported in 2014. In order to keep on the steep part of the accumulation curve he moves on before the yields / day become too low - this can be seen in the three humps on your graph, South America, North America, and Europe.
ReplyDeleteSo expecting him to see around 63.7% of the world's birds (as reported to ebird in 2014), this would be around 6032 species.
This is an interesting consideration. One of the assumptions however is that data (i.e., total species) submitted to e-bird in 2014 is representative of data that will be submitted in 2015. In other words, we have to assume that each year receives a similar amount of effort. If we look at the annual species totals submitted to ebird for the Western Hemisphere from 2010-2014, there is an increasing trend...the total number of species submitted increases every year. For 2015 year-to-date, the total is already well-above the 2012 total, and only 43 species from the 2014 total. We see the same trend for the World, with the 2015 year-to-date total being just 545 species lower than the 2014 year-end total. On the assumption that year-over-year trends are increasing (they will flatten out eventually), it would not be unreasonable to use the 64.4% metric to predict the final total, which in this case, relative to 2014, would be 6,094 species (i.e, 64.4% factors in an assumed year-over-year increase which appears real). Of course, Noah's efforts have made him a statistical outlier, and there in lies the fun. Therefore, using his own data to predict the end result may be the best predictor after-all...only time will tell. Enjoy.
DeleteMy prediction is between 5500 to 5800 species
ReplyDeleteGood luck...another update will be posted in early September.
DeleteInteresting stuff Michael. Thanks for sharing the Davies/Harris accumulation figures as well.
ReplyDeleteI tried to follow up with non-linear model for Noah's year, not entirely successfully...
http://shockingbadhat.net/2015/08/big-year-species-discovery-curve/