Meanwhile, our instruments record. One hundred parametric data points per bird per millisecond, on average, or upwards of fifty gigabytes of data for the whole murmuration. Since the murmuration may persist for several tens of minutes, our total data cube for the whole observation may contain more than thirty terabytes of data, and a petabyte is not exceptional. We use some of the same data-handling and compression routines as the particle physicists in CERN, with their need to track millions of microscopic interaction events. They are tracking tiny bundles of energy, mass, spin and charge. We are tracking warm, feathery bodies with hearts and wings and twitchy central nervous systems!
All of it is physics, though, whether you are studying starlings or quarks.
On my workstation I sift through slices of the data with tracker-wheels and mouse glides.
I graph up a diagram of the murmuration at a moment in time, from an arbitrary viewing angle. It is a smear-shaped mass of tiny dots, like a pixelated thumbprint. On the edges of the murmuration the birds are easily distinguishable. Closer to the core the dots crowd over each other, forming gradients of increasing concentration, the birds packing together with an almost Escher-like density. Confronted with those black folds and ridges, it is hard not to think of the birds as blending together, clotting into a suspended, gravity-defying whole.
I mouse click and each dot becomes a line. Now the smear is a bristly mass, like the pattern formed by iron filings in the presence of a magnetic field. These are the instantaneous vectors for each bird – the direction and speed in which they are moving.
We know from previous studies that each starling has a direct influence – and is in turn influenced by – about seven neighbours. We can verify this with the vector plots, tracking the change in direction of a particular bird, and then noting the immediate response of its neighbours. But if that were the limit of the bird’s influence, the murmuration would be sluggish to respond to an outside factor, such as the arrival of a sparrowhawk.
In fact the entity responds as a whole, dividing and twisting to outfox the intruder. It turns out the there is a correlation distance much greater than the separation between immediate neighbours. Indeed, that correlation between distant birds may be as wide as the entire murmuration. It is as if they are bound together by invisible threads, each feeling the tug of the other – a kind of rubbery net, stretching and compressing.
In fact, the murmuration may contain several distinct ‘domains’ of influence, where the flight patterns of groups of birds are highly correlated. In the plot on my workstation, these show up as sub-smears of strongly aligned vectors. They come and go as the murmuration proceeds, blending and dissipating – crowds within the larger crowd.
This is where the focus of our recent research lies. What causes these domains to form? What causes them to break up? Can we trace correlation patterns between the domains, or are they causally distinct? How sharp are the boundaries – how permeable?
This paper, the one that is bouncing back and forth between us and the referee, was only intended to set out the elements of our methodology – demonstrating that we had the physical and mathematical tools to study the murmuration at any granularity we chose. Beyond that, we had plans for a series of papers which would build on this preliminary work with increasingly complex experiments. So far we have been no more than passive observers. But if we have any claim to understand the murmuration, then we should be able to predict its response to an external stimulus.
I am starting to sense an impasse. Can we honestly go through this all over again with our next publication, and the one after? The thought of that leaves me drained. We have the tools for the next phase of our work, so why not push ahead with the follow-up study, and fold the results of that back into the present paper? Steal a march on our competitors, and dazzle our referee with the sheer effortless audacity of our work?
I think so.
THE NEXT DAY I set up the sparrowhawk.
I need hardly add that it is not a real sparrowhawk. Designed for us by our colleagues at the robotics laboratory, it is a clever, swift-moving drone. It has wings and a tail and its flight characteristics are similar to those of a real bird. It has synthetic feathers, a plastic bill, large glassy eyes containing swivel-mounted cameras. To the human eye, it looks a little crude and toy-like – surely too caricatured to pass muster. But the sparrowhawk’s visual cues have been exaggerated very carefully. From a starling’s point of view, it is maximally effective, maximally terrifying. It lights up all the right fear responses.
Come the roost, I set down a folding deck chair, balance the laptop in my lap, stub my gloved fingers onto the scuffed old keyboard, with half the letters worn away, and I watch the spectacle. The sparrowhawk whirrs from the roof of the 4WD, soars into the air, darts forward almost too quickly for my own eyes to track.
It picks a spot in the murmuration and arcs in like a guided missile. The murmuration cleaves, twists, recombines.
The sparrowhawk executes a hairpin turn and returns for the attack. It skewers through the core of the flock, jack-knifes its scissor wings, zig-zags back. It makes a low electric hum. Some birds scatter from the periphery, but the murmuration as a whole turns out to be doggedly persistent, recognising on some collective level that the sparrowhawk cannot do it any real damage, only picking off its individual units in trifling numbers.
The sparrowhawk maintains its bloodless attack. The murmuration pulses, distends, contracts, its fluctuations on the edge of chaos, like a fibrillating heart. I think of the sparrowhawk as a surgeon, drawing a scalpel through a vital organ, but the tissue healing faster than the blade can cut.
Never mind – the point is not to do harm, but to study the threat response. And by the time the sparrowhawk’s batteries start to fade, I know that our data haul will be prodigious.
I can barely sleep with anticipation.
BUT OVERNIGHT,THERE’S a power-outage. The computers crash, the data crunching fails. We run on the emergency generator for a little while, then the batteries. Come morning I drive out in the 4WD, open the little door at the base of the tower, and climb the clattery metal ladder up the inside. Inside there are battery-operated lights, but no windows in the tower itself. The ladder goes up through platforms, each a little landing, before swapping over the other side. Heights are not my thing, but it’s just about within my capabilities to go all the way to the top without getting seriously sweaty palms or stomach butterflies.
At the top, I come out inside the housing of the turbine. It’s a rectangular enclosure about the size of our generator shed. I can just about stand up in it, moving around the heavy electrical machinery occupying most of the interior space. At one end, a thick shaft goes out through the housing to connect to the blades.
The turbine is complicated, but fortunately only a few things tend to go wrong with it. There are electrical components, similar to fuses, which tend to burn out more often than they should. We keep a supply of them up in the housing, knowing how likely it is that they will need swapping out. I am actually slightly glad to see that it is one of the fuses that has gone, because at least there is no mystery about what needs to be done. I have fixed them so many times, I could do it in my sleep.
I open the spares box. Only three left in it, and I take one of them out now. I swap the fuse, then reset the safety switches. After a few moments, the blades unlock and begin to grind back into motion. The electrical gauges twitch, showing that power is being sent back to our equipment. Not much wind today, but we only need a few kilowatts.