Flora of the Borrego Desert: Voucher Collection Dates
Andy Sanders and I were having a discussion about whether voucher collection dates might reveal some information as to years in which the bloom in the lower elevations of the Colorado Desert occurred later than the normal peak bloom months of February and March. As a result, I searched the Consortium for vouchers with coordinates in a rectangular box that included the lower elevation Borrego Desert plus the portion of Imperial County to the east. Specifically, I searched for vouchers between latitudes of 33.084 and 33.345 and longitudes between -116.395 and -115.570.
I eliminated 30 vouchers that were from counties other than San Diego and Imperial, which appeared in this search since they had erroneous coordinates. I eliminated another 62 vouchers that were in the higher-elevation Pinyon Mountains in the southwest corner of the search box.
This left 2145 vouchers. The locations of those vouchers are shown in the following map:
Many of the vouchers do not have accurate coordinates, so the plotted locations should not be take literally. The vouchers inside the Salton Sea were collected in 1920, 1929, 1941, with the coordinates being supplied by a computer program (Terrain Navigator) interpretation of the locality, and typically having an error distance of 3 miles.
Of those 2145 vouchers, 2114 have a date of collection precise to the day. The rest of this page gives results for those vouchers.
One thing to keep in mind throughout this analysis is that there is a selection effect introduced by only analyzing vouchers with coordinates. This gives a bias toward recent years, except for vouchers from institutions where all vouchers have been geolocated, like UCR, with 235 vouchers in this list.
Fortunately, SD has also geolocated many of their vouchers as well, with 1505 vouchers in this list. Since those two institutions dominate this voucher list, with a combined 82% of the list, this bias may not be very strong, and probably does not influence most of the plots below.
First, I did a sanity check to see if there was anything funny about the days. The following plot is a histogram of the day of the month, to see if collectors tend to just put down the first, middle, or last day of the month:
There is nothing unusual about the first, middle, or last day of the month, but this is definitely an unusual plot. The pink line gives the average number of vouchers per day of the month, and there seem to be statistically-significant excursions about that line. For example, the peak at day 6 turns out to be due to 88 vouchers collected by Frank Gander on 6 March 1935.
This histogram also indicates that except for Gander's trip, fewer vouchers are collected per day in the first 11 days of each month than in the remaining part of the month. I have no idea why that might be the case! However, since the number of vouchers is dominated by a few collectors who often collect many vouchers in a single day, as demonstrated by the Gander point, this trend may not actually be statistically significant. Note that this domination makes it difficult to calculate the statistical significance of this trend.
This strong outlier due to one day's worth of vouchers indicates that one must be very cautious about binning the vouchers versus day of the year, since a large bin size might obscure this collector effect.
The following plot shows the number of vouchers versus each day of the year:
This plot is a mess! Outliers are everywhere.
The top three outliers with the largest number of vouchers are:
- Day 65, due to 88 vouchers by Frank Gander on 6 March 1935.
- Day 84, due to 54 vouchers by Edith Purer on 25 March 1933
- Day 117, due to 64 vouchers by Larry Hendrickson and L. Louise Jee on 27 April 2005.
Clearly, it is dangerous to try to draw conclusions about anything other than very broad-scale patterns in the above plot.
The following plot shows the same data, with each data point now being the average of ten days. To show the effect of the outliers, an additional curve is plotted that replaced the three data points identified above with the ten day average, which in effect removes most of the extra vouchers from the plot for those three points.
A comparison of those two curves shows how one day of collection by a single collector can produce sharp features in the plot. Anything other than the broad-scale pattern is thus suspicious.
Thus the answer to the question Andy and I were discussing is: there is no information in the voucher dates as to years in which the bloom in the lower elevations of the Colorado Desert occurred later than normal. Breaking the vouchers down by year for May and June is pointless, due to the very small number of vouchers then, and the very large effect of individual collectors.
Even trying to study vouchers collected in March and April is perilous, as shown by the very large dip in the number of vouchers taken in late March.
The only clear picture from this histogram is that most vouchers are taken in March and April, with February being the next most-collected month.
For completeness, here is the histogram of the vouchers by year:
The 1930s bump, first noticed by Andy Sanders and Steve Boyd ten years ago, with at least the highest point due to Mr. Gander, is evident, as are the efforts of the San Diego County Plant Atlas, beginning in 2003. The severe drought years of 2002 and 2007 are also seen as the two lowest years by far since 2000. Other than those conclusions, there are too few vouchers per year to infer anything further.
I thank Andy Sanders for stimulating this page, and for helpful comments. Data provided by the participants of the Consortium of California Herbaria (ucjeps.berkeley.edu/consortium/).