Twenty four cameras on a grid, all the same model, all set to stamp the date. Eleven takes pictures of the same path in a different month from the one it is in, and it is always ahead.
I run a mammal survey on a forestry estate. Twenty four camera traps on a grid, five hundred metres apart, checked every six weeks for card changes and batteries. I have done it since 2018 and the data goes to a national recording scheme.
The cameras are all the same model bought in one batch, which was deliberate, because a survey with mixed kit is a survey you cannot compare across. Each one has a passive infrared sensor that triggers on a warm thing moving, takes a burst of three stills, and writes a data strip along the bottom of the frame: date, time, temperature, moon phase, battery, and the camera's own ID.
That strip is not added later by software. The camera burns it into the image as it writes the file, from its own clock and its own thermistor, and the file on the card has it.
That matters for everything that follows.
Camera eleven is on a post at the edge of a ride, looking along a badger path into mature beech. It is the dullest station on the grid. Badgers, roe, a lot of grey squirrel, foxes twice a week.
In March 2020 I was doing the card and the images were wrong.
Not corrupted. Beautiful, sharp images of that path with the leaves down and a hard frost on the ground and the beech bare, and the strip along the bottom said the eleventh of March, which was four days earlier, at 04:51, minus one degree.
The wood on the eleventh of March was in leaf burst. I had photographed it myself that week on the other twenty three cameras, and all twenty three show what March looked like.
Camera eleven was showing January.
I changed the camera. New body, same post, same height, same angle.
It did it again in the September.
Images stamped the third of September, 19:20, fourteen degrees, and the wood in them is in full autumn colour with a deep leaf litter, and the third of September on this estate is green.
That is not a clock error. A clock error gives you the right picture with the wrong date on it, and you can tell instantly, because the picture agrees with one of the other cameras and the date does not.
This is the other way round. The date is right. I can check it against the badger, because a badger that went past eleven at 19:20 went past twelve at 19:26 on the same evening, and camera twelve shows a green wood.
Six minutes apart, five hundred metres apart, two different months.
How often: thirty one times in six years, out of something like four thousand trigger events on that camera.
So the overwhelming majority of camera eleven's pictures are ordinary, which is part of why it took me two years to believe it. For six weeks at a time it is the dullest station on the grid and then there are nine frames of somewhere else.
And it is always ahead. Never behind. I have worked out the offset for every one of the thirty one and it runs between nineteen and one hundred and forty days forward, and the median is about seventy.
The wood in those pictures is the same wood. Same beech, same post in the left of the frame, same fallen limb that came down in 2021 and which is in every picture after it fell and in none before, including the ones that are showing a season three months ahead. Whatever this is, it is not looking at a different year. The furniture is right.
The temperature is the thing I find hardest.
The strip carries the camera's own temperature reading, and on all thirty one it agrees with the season in the picture rather than with the date.
The third of September at 19:20 was fourteen degrees and the camera said fourteen degrees, and the picture is October. I thought I had it there: the temperature proves the camera is in September with everybody else.
Then I pulled the eleventh of March one. Minus one degree. The met station four miles away had plus nine that morning. Every other camera on the grid that night read between seven and ten.
So on some of them the thermistor agrees with the calendar and on some it agrees with the picture, and I have been through all thirty one looking for which and the answer is that the ones showing autumn agree with the calendar and the ones showing winter do not.
I cannot make anything out of that and I have stopped trying to.
What is in them.
Mostly nothing. That is the honest answer and it is the one that took the longest to sit with.
Nineteen of the thirty one are the empty path in the wrong season. Something triggered the sensor and the three frames have no animal in them, which happens normally too, all the time, from a branch moving in sun or rain hitting the lens.
Eight have an animal. Roe deer, badger, a fox, all ordinary, all doing ordinary things, and in the wrong season. A roe in winter coat on a date in June.
Four have a person.
I have looked at those four more than I should have.
It is not the same person in all of them, as far as I can tell, and the temptation to make it one is strong enough that I have kept the four files apart and looked at them weeks away from each other. Two are clearly a man in dark clothing, one is smaller and could be anybody, and one is just a leg and a boot at the edge of the frame.
None of them is looking at the camera. That is ordinary too. Most people photographed by a camera trap are walking past and never see it, because the flash is infrared and there is nothing to notice.
The estate is private with public access on two footpaths and camera eleven is nowhere near either of them, but people walk where they like and I have photographed perhaps two hundred of them over six years on the grid as a whole, so four is not a number that means anything on its own.
What I keep going back to is that in all four the leaf litter is deep and the light is low and the ground is white, and all four are stamped on dates in May and June.
I have not taken camera eleven off the grid, and the reason is the survey. Twenty four stations at five hundred metres is a designed grid and pulling one out makes six years of data a different dataset. The scheme would rather have a station with a known fault in it than a hole.
So the images go in, all of them, and the thirty one are marked in my notes with a code I invented, which is Q, for query, and the scheme has never asked what Q means.
I check the cards every six weeks, and at eleven I do the card first now, and I look at the frames on the camera's own little screen standing there at the post before I walk on, which is a thing no sensible surveyor does because the screen is awful and it flattens your battery.
I do it because twice, looking at the last frame on that screen, the picture has been the path in front of me with snow on it, and I have been able to look up from the screen at the same path, in the same second, in the same place, and see that there is no snow.