Para-sight recently wrote a post at DSN urging caution in assigning blame to the BP oilspill as the cause of recent cetacean deaths along the Gulf Coast. To try and address one of the issues he raised, I wrote a guest post there using the raw data from NOAA to quickly try and get an idea of how bad the dolphin situation in the gulf really is when examined in the light of the historical data available. Unfortunately the graphs NOAA created comparing 2010 and 2011 to historic data made it hard to truly evaluate as they used the average value for each month for the historic data. Based solely on their presentation of the data all that can really be determined with any confidence is, as Para-sight noted:
…dolphin deaths are not that rare to start with and that this is the time of year for it (i.e. it’s not actually “unusual”, it’s “seasonally-appropriate”). Secondly, it means that it’s not much worse in magnitude than last year, albeit peaking a little earlier.
and
But in these cases, the devil IS so often in the details. I am not saying oil wasn’t involved; I’m just saying we can’t infer causality from correlation without a closer examination of the data.
Running the raw data through excel (to subtotal & clean) and R and I could see more of the information Para-sight warned was needed to get a better feel for the situation. And sure enough I was able to add to his observations that:
the pattern of the event does strongly suggest that the oil spill is related, but looking at the prior events, it is not outside the realm of possibility that this is not directly related to the oil spill. The best evidence to determine the involvement (or lack thereof) of the oil spill is going to be found in necropsy and toxicology reports of the stranded animals.
In the meantime however several people commenting on the original post pointed out that the biggest issue was with the a specific age class of dolphins. As “Dr. Ed Cake” commented:
…In the meantime, this year’s neonates continue to wash ashore dead in the BP spill zone and all potential mortality factors can be dismissed except the impacts of BP’s oil and Nalso’s Corexit dispersants.
and “Don” stated:
NOAA provides data specifically for “baby dolphins.” It shows a five-fold increase in mortalities in 2011 compared to 2010.
So that last bit especially really made me wonder. The more detailed data showed there was an unusual mortality event (UME) happening with cetaceans in general, though it is not clearly linked to the spill. But a five fold increase for one key age class??
So tonight I journeyed forth again to the NOAA website, and found on scanning down to the bottom of the page, there is a table of “baby dolphin” deaths as Don phrased it. It is more accurately “premature, stillborn, or neonatal dolphins”. NOAA used the cutoff of 115cm total length to split out these deaths and presented them as both a table and graph:
neonatal, stillborn and premature dolphin deaths
Graph of all neonatal, stillborn and premature bottlenose dolphins
As both commenters pointed out there seems to be a major issue going on here! Just look at February! a 36x increase from 2010 (pre spill) and an almost 18x increase in deaths over historic averages. And March is better, but still a 2x increase over last year and 3x increase over historic levels.
However, here again we are given historic numbers solely as the average. As Para-sight warned, and we saw with the complete cetacean data set, that is a dangerous way to look at data like this.
But what about the “baby dolphin” data?
From the truncated graph NOAA presented it appears there may be a seasonal pattern (not unexpected) but it is hard to tell with out the complete year graphed. It is also not possible (again) to tell what the range of values is for each month. The numbers 100 and 2 average out to the exact same value as 52 and 50 do, but the range of values, and the story it tells, is quite different! Back to the data set, with definition in hand I extracted all bottlenose dolphin records with lengths of 115cm or less and duplicated the analysis I did on the full set over at DSN. The resulting table:
Gulf Coast Strandings of Bottlenose Dolphin <115cm
First thing to notice comparing the two tables is that yes February deaths were an order of magnitude higher than the historic range.
The second thing to note is — they don’t match. Not just that I included the entire year in the table, but the numbers for the averages in each month they did include don’t match. They only averaged 2002-2007 for their historic baseline. So I redid the table using only 2002-2007, still the numbers were slightly off. Using their dataset and the definition for premature / neonatal / stillborn they gave, I get different numbers than they do. Aggravating. I am comparing historic data to the 2011 data, I double checked all the subtotals are correct and redid the summary table creation three times each time with the same result. I have verified that the data in the table is true to the definition given by NOAA and the raw data NOAA has provided. I really don’t like omitting data without good reason (which was not provided by NOAA) so I finished the evaluation keeping the 2008-2009 data in:
Box and Whisker plot of Stranded Bottlenose Dolphins less than 115cm. 2011 data blue diamonds, 2002-2010 data boxplots with outliers as open circles.
So what does this one tell us? And is it different from the conclusions inferred from the NOAA graph?
Based on the NOAA graph and table one would assume that all months except April were UME months, and April, one could easily argue, is on it’s way to being a UME month as well, after all, that is only 10 days of April data. However, looking at the box plots only January and February really stand out as a UME(s). February is clearly an outlier, in fact the most deaths and the most radical outlier of the entire data set. January is an outlier but is it a UME? I guess that depends on the definition, but it is two standard deviations above the mean January number of deaths, so I would say it is a UME. While March is high, it is still within the range of expected numbers from the historical data. Granted March is right at the upper edge of that range, but it is within that range nonetheless. If we were using Para-sight’s suggestion of using values greater than two standard deviation above the mean as the definition of a UME then only January and February would be as well. However when taken with the previous two months I would suggest that March is a continuation of the same UME.
As for April, with only 10 days of data for the analysis, it is right at the edge of the expected range of values from the historic April data. One could argue, I suppose, that it will climb much higher, but without the actual numbers any discussion of that is pure speculation. I expect that it will climb as well, but we’ll just have to wait for the rest of the data to come in to really say anything about it conclusively.
What is more interesting to me though than the month by month numbers is the pattern, both historic and for this year. Historically, just as with the larger population, bottlenose dolphins < 115cm have a clear pattern of strandings. The late winter and spring (January through May) is the period of strandings, with most strandings clustered in the February through April period with the sharp peak in March. The rest of the year have much lower numbers. The box plot and the full table also reveal that each year it would be unexpected to find no strandings under 115cm but in the summer through early winter (June through December) it is actually an unusual event to find a stranding, let alone more than one.
This is a direct clue to the reproductive biology pattern of the Bottlenose Dolphins in the Gulf of Mexico. I have very little knowledge of dolphins so had to turn to some local experts for help (contrary to popular belief all ocean scientists DO NOT work on dolphins). Turns out that Bottlenose Dolphins generally mate from late March to May and have ~11-12 month gestation period. At birth a bottlenose dolphin is anywhere from 85-110 cm long (the reason for the 115cm cutoff). Looking at the length records the smallest animal found was a January 2006 stranding at 40cm. Fully 83% of the strandings were of individuals between 85 and 110cm which combined with the seasonal pattern would corresponds essentially to very late term neonatal or newborn aged individuals. Examining the length vs. month of death did not provide any real patterns except that March had the widest range of sizes in the strandings, which is not too surprising as March had significantly more deaths than any other month in the aggregated data.
This year however the peak is clearly in February. Why? Unfortunately I did not see any length data for this year, just summary numbers. Were these individuals mostly less than 85cm? That would infer prematurely born babies or late term miscarriage. If they were mostly in the 85-110cm range? That would suggest they were from the early mating season, likely conceived before the oilspill, and at full term.
So, yes , there is a major unexplained mortality event that is causing a significantly higher number of neonatal, stillborn and premature Bottlenose Dolphin strandings. February’s mortality is the worst event for individuals under 115cm on record, by a fair margin. I expect the number for April will go higher, quite possibly being an outlier. From the timing, the oil spill is certainly a suspect for being a cause. That does not however mean it IS responsible. There is still no way to know the oilspill’s relation to these deaths without detailed necropsy and toxicity study results.
