Did North America really lose 3 billion birds? What does it mean?

The journal Science released an article entitled “Decline of the North American avifauna” by Rosenberg et al today (Sep 19, 2019), and already disaster laden headlines are appearing in major newspapers (I’m not going to bother to link to them because they’ll probably change by tomorrow but I bet you’ve already seen this in your favorite news source).

Has North America truly lost almost 3,000,000,000 birds over the last 48 years? I don’t know. But I do know that number which is likely to appear in the most headlines is an estimate and one of the less solid numbers and I will argue one of the least important numbers that appears in this publication. I haven’t had time to fully digest the methods yet myself, but these are credible people and on a skim most of the methods look reasonable (fitting a hiearchical Bayesian model is something I am in the middle of writing a paper doing that myself so I’m hardly going to critique it). And the authors put error bars on everything which is too rare and therefore to be commended. A very quick summary of their methods is they make two estimates and combine them. One estimate is of the trend for each species from an index of relative abundance, often from the high quality breeding bird surveys. And second, they estimate the total number of individuals of each species found in North America (excluding Mexico) from an extrapolation from data covering a fraction of a percent of the US. I don’t have a better method but this is clearly a step that introduces a lot of uncertainty. Then they combine the reliable trends in relative abundance and the less accurate total continental abundances and sum it up across 529 species. That’s how they get 3,000,000,000 bird decrease. So these kinds of estimates require a ton of assumptions and extrapolation which makes it hard to do, but that uncertainty and challenge shouldn’t prevent us from doing this. Even if they are very approximate estimates, we need these kinds of numbers to enter our conversation. And I am not brave enough to provide my own alternative estimate! So these are the best out there.

For all the bird apocalypse headlines we will see, what is really in the data and what does it really say? I’ve downloaded their first spreadsheet from supplemental material (I am basically skipping over the NEXRAD radar data because I have less expertise on that kind of data but they got very parallel messages). So I take as a given their estimate of population size “Estimated North American population during range of years between first_year_popest and last_year_popest” which I will call “AvgPop” (even though it is estimated by a more sophisticated method than average and indeed they have to estimate the population sizes over different years so they couldn’t even average over raw data). I am also taking their “Estimated change in number of breeding individuals over the trend period (usually 1970-2017), based on a combination of current population estimates and long-term trajectories. Median of the posterior distribution from the hierarchical Bayesian model” which I call “Abundance Loss”.

From there you can calculate whether a species increased or decreased (Abundance Loss greater or less than zero) as in this Figure

Percent of species showing increases or decreases

We can also look at the percent increase or decrease of populations given by (Abundance Loss)/(Avg Abundance) (although by the vagaries of the methods used this can show declines of more than 100% which initially seems problematic). This is not really a problem since a true percent loss would have (Abundance Loss)/(Initial Abundance) but initial abundances were not estimated (or at least not reported) in this model. Declines of more than 100% are not incorrect nor a sign of errors, but they are a bit awkward and might give a flavor of how approximate some of the numbers are. If we plot a histogram of percent loss we get:

Histogram of species % change (right graph shows only species with between -150% to +150% change)

These two figures suggest that on a species level, many have increased, many have decreased, most are not too terribly far from no change and the data is not terribly far from centered on no change but in fact the average percent loss is -30.9% (this is not that different than their reported 29% loss but I got there by less sophisticated methods). 31% decline over 48 years sounds pretty big but it is over almost 5 decades so less than 1% (0.64%) loss per year on average. We’re on average losing for sure, but not racing towards zero at break neck speed (it would take another 56 years to further decline from 69% to 50% of original numbers by one overly simplistic calculation).

But here is my first question. Is it even useful to say 30.6% (or more accurately 29%) decline AVERAGED across all species? As the 2nd figure shows that average is almost misleading given how much variation it hides. Wouldn’t we be much better off to say: a few species have declined drastically, a few have increased drastically many haven’t changed that much? This is a general pattern, not just in these birds. And then from a conservation view wouldn’t we be better off homing in on those birds showing out of control declines or increases? Probably yes. But its certainly not a loss to know an overall average is a decline, as long as we don’t get too hung up on boiling down an enormous seething pot of complexity and winners and losers to one number.

Here is where things start to get a bit more complicated. The number that I’m sure will be in most headlines is not -29% but -3,000,000,000, the number of individual birds lost. For starters as already discussed that number comes from combining the trends on indices of relative abundance (which I would consider much more robust data) with estimates of total numbers of individuals in each species across North America (which I would consider a much more difficult and therefore approximate number). So right there the 29% number is probably a more accurately estimated number than the 3,000,000,000. But there is something more going on than one has smaller error bars.

The 3,000,000,000 number is a weighted average on abundance. And there is ENORMOUS variation in abundance. The most abundant bird (by their data) is the American Robin with an estimated 366,076,928 individuals living in North America while the rarest bird is the Dovekie (an auk related to puffins) that has 1,500 individuals (the rarest landbird in the datset which might be a better comparison to the Robin is the Snowy Owl with 14,000 individuals, although the Kirtland’s Warbler is often estimated to have around 2,000 individuals and the California Condor has about 500 individuals, but both species were too rare to make their data set – and both have increased in recent years). The Dovekie has a loss of 55% while the Robin has a gain of 5.5%. Up until now every species counted equally and the Dovekie drove down the average trend more than the Robin drove it up (average of the two gives -25.25%). But now the Dovekie trend gets a weight of 1,500 and the Robin gets a weight of 366 million, completely swamping the Dovekie data. This is the right way to get the total number of individuals lost, but it clearly hints that its going to pretty broadly transform the nature of the results and suddenly it really matters which species are increasing and decreasing as only the really common species are going to matter much in that 3,000,000,000 number. If you’re going to weight your data (even in an objective and purpose driven fashion), you want to understand your weights!

Or put another way, when there are winners and losers, it is natural to ask WHO is winning and losing. Every summary analysis to this point in this blog post  has treated a rare species and a common species as of equal weight in the averages (except their 3,000,000,000). It might seem intuitive that the rare species are declining the most and the common species are holding steady or maybe are the increases (and maybe even the increases are invasive species). Indeed the two extreme data points I picked out in the last paragraph seem to confirm this scenario. This would be a worst-case scenario. So lets check (again using their supplemental data). A plot of percent loss versus abundance (with abundance on a log scale) shows:

 

% loss versus population size – common species decline more, rare species decline less

Again there is a lot of noise, but that black line is sloping weakly down (statistically significantly so if “% loss” is log transformed as it probably should be) – common species are on average declining more than rare species! That paints a very different picture! It still doesn’t change the decline of 29% averaged across species, but the big losers in numbers of individuals are really common species. Mathematically that almost has to be true.

The story gets even more interesting if we dive into some of the biggest losers. If we take the 10 species that have lost the most individuals we get:

Loss of 10 species with biggest decline

We see a couple of things. The birds that are losing the most in absolute numbers (# Lost) are also showing very big % Loss (including several with >100% decline), but they are also among the most abundant birds (Avg Pop). Indeed these birds all rank in the top 39 out of over 500 bird species for how abundant they are, and 4 of the top 10 losers are in the top 10 most abundant. Or as the authors of the Science paper note, “While not optimized for species-level analysis, our model indicates 19 widespread and abundant landbirds (including 2 introduced species) each experienced population reductions of >50 million birds”. Yes indeed! That is in the paper but I guarantee you it won’t make it into any headlines and probably will be mentioned in a very small fraction of all the coverage. Two of the 10 biggest losers are actually invasive species (House Sparrow, European Starling). Indeed those two species are widely regarded as among the worst vertebrate pests introduced to continental North America. In case you are feeling bad for the house sparrow, don’t. It is still found on 6 continents (invasive on four of them) and is the most widespread and second most abundant bird in the world. When an invasive species declines is that a good thing or a bad thing? All I will say is that land managers spend a lot of money trying to achieve this outcome usually without much luck.

Eight of these 10 biggest losers are considered “Least Concern” on the IUCN list (farthest from extinction) while the other two (Common Grackle, Blackpoll Warbler) are Non-Threatened (next farthest category from extinction). Although some of these birds may be threatened in individual states or for particular subspecies, yet at a continental scale the conservation concern is almost nil. Several of these birds are not just widespread and abundant in North America but also abundant on other continents. As noted two of these birds are out of control invasives that may just finally be receiving a long overdue adjustment. The Pine Siskin is an irruptive species (huge population outbursts and crashes) which makes it hard to attribute any long term trend. The Blackpoll Warbler, White-throated Sparrow and Dark-eyed Junco breed in primarily boreal or mixed forests and often in edge, early successional or burned habitat whose frequencies may be changing due to human actions, but they are so common we have really done very little research to understand why they are declining.

The remaining four (Savannah Sparrow, Horned Lark, Common Grackle, Red-winged Blackbird plus the already mentioned invasive House Sparrow and European Starling) all love low intensity farmlands and saw large range and abundance expansions as Europeans cleared>90% of the Eastern and Central US for farms. These 6 birds are now likely declining because of agricultural intensification, but it’s important to note that their decline is from an artificially human-induced higher than pre-European level. They may well all still be above pre-European population levels even after huge decline. Is that bad that such species are declining because of agricultural intensification? Probably? But for sure its not the saddest story. The saddest story is the decline of prairie and forest specialist birds (and in a few cases extinctions) 150-200 years ago when European settlers were clearing that >90% of the land. But that story completely predates this (and nearly all other) data sets and is not often told quantitatively. These specialists were the big losers; the birds at the start of this paragraph were the big winners back then 150 years ago but now are in the top 10 losers. So its a case of the tables turning. Unfortunately, though, its not the case that intensification is returning the land to those original losers. Intense agriculture is a biodiversity desert (although it does feed people). But on the other hand, intensification of agriculture can also concentrate agriculture and lead to less land needed for agriculture which at least has the potential for returning land to more natural habitats (and the proportion of land devoted to agriculture is indeed decreasing in countries like the US and UK although some of this is also due to conversion to urban land). In some regions of the US experiencing reforestation, forest birds are gaining while the farmland birds are losing. So while I wouldn’t call it good news, this is hardly the most disastrous story out there. And these species may still be more abundant than they were before Europeans arrived even after large declines (and certainly are in the case of the two invasives). Its really a rather complex story.

So I may or may not have convinced you that declines in at least six of these ten birds are really a return to a more natural order (and three more of them are so abundant we haven’t yet bothered to study very hard why they’re declining but it might also be changes in already human modified lands and the last might not really be declining in a meaningful sense). But I do want you to notice that total on the “# Lost” column. IE 1.6 billion (>half) of the 3 billion decline is found in the 10 biggest losers that are all extremely abundant widespread, and mostly anthrophillic (human habitat loving). In the same vein with a slightly different approach, look at the 40 most abundant species. Of these 40 maximally abundant species, 31 show declines, and the 40 most abundant species together have a combined net decline of over 2,184,000,000 (or over 2/3 of the total decline are in the 40 most abundant species). Bad? Sure. A sign of human impact on birds being large? Sure. But a conservation disaster worthy of words like armageddon? Probably not unless you are selling newspapers.

So is the average bird species declining (in NA)? Probably yes, and probably by about 30% on average, BUT that hides a lot of variation and treats every bird equally. Is that 3,000,000,000 birds lost correct? That is a big extrapolation from the raw data, but it is the best available guess and I don’t have a better one. But are we in a death spiral where common species and invasives are taking over and rare birds are taking it on the chin? Not in any general or on average sense, no. Quite the opposite, more than half the birds lost are from the 10 biggest losers that are all completely safe, often so widespread they are considered a nuisance, two of which are invasive, and most of which are declining because they got so big by exploiting habitat created/modified by humans in the first place which is now changing on them. Or alternatively the 40 most common species (out of 529) make up over 2/3 of the total 3,000,000,000 decline. So that 3,000,000,000 decline probably is being interpreted by people pretty differently than what is really happening on the ground.

So are humans impacting birds in a negative fashion overall or on average. Pretty evidently yes. The Anthropocene is real. Its hard to take away 50% of the total land and 50% of the freshwater each year and 50% of the new plant growth each year globally and not see declines! Given those statistics its actually kind of surprising the declines are not more. And in a small good news embedded in larger bad news fashion, most of those 3 billion in declines are in species that are not at even the slightest risk of extinction.

So do the statistics in this study really directly address questions of conservation or biodiversity preservation? Not so much (at least not in the headlines). For that you need to go look at birds that are rare and declining. Of course that is what conservationists have been doing all along. And there you get a mixed story. Bald Eagles and Wild Turkeys have made pretty spectacular come backs. And the Kirtland Warbler and California Condor are both naturally rare but have actually increased in recent years due to intensive management efforts so fairly positive results but certainly not victory. But we have made several bird species go extinct including the Carolina Parakeet, Heath Hen, Passenger Pigeon, Labrador Duck and probably the Ivory Billed Woodpecker and a few more (many more if we include Hawaii). And a number of other species are on the ropes (a few dozen endangered bird species in North America, again a disproportionate fraction of them in Hawaii). That is a really important story (both the positive and the negative), but it is an almost wholly different story than -3,000,000,000 or -29%.

So should we be worried by the two new headline results? If so, how worried? That’s a subjective question depending on one’s values around biodiversity, balance between human and biodiversity needs, and conservation goals and priorities and how we weight different aspects. Hopefully I’ve given you some facts to help you come to your own answer. I have my answer (which I have not expressed) and I expect you each have your own slightly different answer. And across all the readers I bet there is a pretty wide range.

So what do you think? What are the headlines you are seeing? What is your interpretation? What do you think is the take home message?