Duck Ponds

Different duck ponds host different populations of birds. So don't just walk past them, thinking 'just another duck pond'. It's worth spending enough time at a pond to see which birds use it and which don't, and then to guess why.

Three different duck ponds

To establish this point, I have chosen three duck ponds very convenient to me. After all, the main advantage bird-watching in Chippenham has for me over bird-watching in Slimbridge, or Martin Mere, is that Chippenham is an awful lot closer. I have interpreted the phrase "duck pond" very loosely, too: two of my ponds are actually small rivers. Because my three examples are all local to me, they are also all quite close to each other. The longest distance between the two is Chippenham - Devizes, which is about 10 miles - an easy journey for any of the birds mentioned here, should they choose to take it (even e.g. Moorhen and Coots can make much longer journeys if they want; "Birds in England", by Brown and Grice, notes that the English population of both birds increases in Winter due to migrants from countries in Europe with harsher climates).

Chippenham

The river Avon flows through the centre of Chippenham; there's quite a nice bridge there, and a park abutting one side of the bridge, complete with an area of grass expressly designated for duck feeding (which is allowed, as opposed to pigeon feeding, which isn't, though nobody seems to have convinced the pigeons of that). Even the river on the other side of the bridge isn't completely inhospitable. There is another area of park and rough ground within a hundred yards or so, and there is even a spot at the side of the river where there is enough vegetation for a Coot to rear a couple of broods of young last year. I have just looked at what can be seen from the bridge.

Calne

Calne is crossed by a number of very small rivers and streams, most of which eventually flow into the river Marden. I have looked at a stream that flows not far from Somerfield's Supermarket. It's never more than six feet wide, and is crossed by a number of foot bridges. I have walked from the supermarket car park along a path by the stream, which gains some grass around it once it leaves the car park. This widens out and the stream gains a small pond as you walk along. Eventually you walk out on a path with grass on either side, and some housing scattered amongst the grass. There's usually quite a few Starlings around here (which I haven't noted down) as well as the ducks that you might expect.

Devizes

There is a very nice canal that runs by Devizes, with a wharf, and an large number of locks down Caen hill, but I haven't looked at that here; that would be cheating. Instead, I have noted down the birds seen at a duck pond between a park and a church on the other side of the town. I've looked on a street map and an Ordnance Survey map to see if there is anything special around there, but haven't found anything. The largest area of open ground nearby, apart from the park, seems to be playing fields for Devizes Comprehensive School.

Observation Design

With the exception of my first trips to Devizes and Chippenham, I have made two observations on each day I went out for this, one in Chippenham and one in either Devizes or Calne. What I should have done is to toss a coin to decide what time I would go to Chippenham, and what time I would go to Devizes or Calne. Unfortunately I haven't done this; I have just chosen whatever time was convenient to me. Partly this is because when I go to Calne I park in either Sainsbury's or Somerfield's car park, so I like to combine it with my weekly shop. I can't rule out the theory that what I believe is a difference between different duck ponds is actually a difference between the time of day I happen to visit them. I don't really believe that, though, because some of the birds seem pretty sedentary (e.g. Coots in Chippenham with obvious nests) and some other things seem to be determined by the nature of the site (Swans and Canada Geese in Calne would look a bit silly on a stream you could nearly jump across or wade through).

Observations

The following tables are in time order within the tables. Each row corresponds to a visit (all dates are in 2006), and each column contains the counts for a particular bird, with blank meaning not seen. Just in case it isn't completely obvious, BH Gull is Black-Headed Gull, C Goose is (Greater) Canada Goose, LBB Gull is Lesser Black-Blacked Gull, and Swan is Mute Swan.

Features in the Data

Some features leap out from the data. Mallards are almost guaranteed everywhere (although Devizes was Mallardless on the 14th of April). Chippenham always provides Coots, but Calne never has them, although it does at least provide Mallards and Moorhens (which can't be taken for granted: Devizes pond never had them).

Coots and Moorhens use ponds in different ways. Coots will dive to feed on vegetation on the bottom of reasonably deep ponds, whereas Moorhens use the pond more for refuge than anything else, often feeding on the land around it. The Concise Birds of the Western Palearctic (Snow and Perrins) says that Moorhen are "mainly concentrated in habitat with smaller open water surfaces than acceptable to Coot".

Devizes pond is the only site to offer Canada Geese. In fact, they aren't regular even at the canal by Devizes, and some of the time I do see them there I only see them flying over. If I had to guess, it would be that they were grazing on the school playing fields; "Birds in England" says "Many flocks feed on amenity grasslands and have become very tame, readily accepting food from humans." However I have no plans to invade the school playing fields to verify this.

Statistics: is there a difference or was it just chance?

Because I didn't choose the relative order of my two visits on each day at random, I can't really distinguish between effects due to timing and effects due to place, but I'll ignore this here. As a simple introduction, let's concentrate on the Coot numbers in the Chippenham vs Calne table. This table splits into 7 comparisons, with the first being the 11th of March: 4 Coot in Chippenham, but none in Calne. If there was no difference between Chippenham and Calne, it would be just as likely that we had 4 Coot in Calne but none in Chippenham. No big deal on its own, just as tossing a coin and getting a single head is no big deal - but we saw the same thing for all seven comparisons: Chippenham had Coots, but Calne didn't, so we have seen our coin come up 7 times in a row, which happens with a probability of only 1 time in 128. Since I didn't predict that Chippenham would be the Coot-haven I need to cut that down to 1 in 64 - the chance of either getting 7 heads in a row or 7 tails in a row, but that still suggests that something is going on.

The catch with that simple approach is that I have picked out Coot as an example after looking at the data. I can get round that objection by using a computer program that pools the information from all the columns. It takes each comparison and (when only two sites are involved) boils it down to a list of numbers, where each number shows which site had more of a given species (reducing it to comparisons and discarding raw numbers makes me less worried about the effect of extreme observations, such as the 5th of March, when Devizes pond was host to a huge flock of Black-Headed Gulls). Once the program has reduced the comparisons for each day to a single list of numbers, it forms totals per-species for each site. For example, Calne scored in total 0 for Coot, 3 for LBB Gull, 4.5 for Mallard, 4.5 for Moorhen, and 1.0 for Swan. It always lost on Coot, but it drew twice on Swans (when neither site had Swans), and so on. The next step is to subtract the average score (between the two sites being compared) and then square the differences and sum the result. This gives a score for a site that shows how far away it is from the average of the two sites. With only two sites, both sites will always be the same distance from their joint average, so we only need to look at the results for one site.

The computer program then plays with the data, swapping each comparison around at random. From its output, I can tell that for Calne vs Chippenham only 2 of the 128 possible swaps produced an output as extreme as the one that actually happened (these are the arrangement of Calne and Chippenham that actually occurred, and the rearrangement that labels each Calne observation as Chippenham and vice versa). So again we have worked out that, if there was no difference between the two sites, the probablity of seeing a result at least as extreme as this at random is only 1/64. The Devizes vs Chippenham comparison comes up with exactly the same result: 1/64.

We can combine the two 1/64s by multiplying them together to get 1/4096 and then asking the question: if I picked two numbers at random between 0 and 1, and then multiplied them together, what is the chance that I would get something as small as 1/4096, or smaller? It turns out that it is about 1/440. This, together with the evidence from bird books that helps to explain the difference between the different sites, convinces me that we are really seeing a difference between the different duck ponds, and not just a fluke.