A September migration morning on the Waddenzee shore

Saturday 14 September, 06.40, dike near Wierum

High water predicted 09.05, height a little above average, wind light from the south-east, mist lying in the fields behind the dike but clear out over the water. I arrived in the dark and sat on the seaward slope with the thermos, which is the whole technique: get there before the birds are pushed, stay low, and do not walk out onto the marsh to get closer. The flats were still wide, and the sound arriving over them was mostly curlew, unhurried and carrying a long way in the cold air.

By half past seven the edge of the water was visibly moving, and the feeding flocks were compressing towards the marsh. Bar-tailed godwits came first, in loose lines, dropping onto the last exposed mud with their bills still working. Knots arrived later and behaved completely differently, a dense grey mass turning in the air with that pulse of light and dark as the flock shows its backs and then its bellies. I gave up counting and wrote down estimates in hundreds, which is honest if not useful.

08.20, the tide takes the last flat

The final ten minutes before a roost forms are the best of the morning. The remaining mud shrinks to a strip, the birds stand shoulder to shoulder on it, and then some threshold is crossed and the whole assembly lifts at once. On this morning they went up in three waves, settled briefly on the outer marsh among the sea aster, and finally packed onto a shell ridge that stays dry through all but the highest tides. Dunlin filled the gaps between the larger waders like gravel between stones.

I counted spoonbills separately because they are easy and I like them: eleven adults and two birds of the year working a creek mouth with that side-to-side sweep, unhurried while everything else panicked. Oystercatchers were noisy on the ridge, redshanks whistled off at the slightest disturbance, and a marsh harrier crossing the plain at nine o'clock lifted perhaps four thousand birds simultaneously. Nothing in my notebook conveys the sound of that, so I stopped writing and listened.

• Bar-tailed godwit: hundreds, arriving in lines before the knots
• Red knot: dense flocks, estimated in hundreds rather than counted
• Dunlin: throughout, filling the roost between larger waders
• Spoonbill: 11 adults, 2 juveniles, feeding through the push
• Marsh harrier: one, 09.00, cleared the entire roost

Where to stand, and why it is usually the dike

Everything I saw that morning depended on the birds not knowing I was there, and the reliable way to achieve that is distance plus a low silhouette. The seaward slope of a dike works because you are above the marsh and can see across it, but your outline is broken by the bank and you are not moving through the roost. Walking out to the shell ridge would have given me a closer view for about four seconds and then no birds at all.

There is a practical reason as well as a courteous one. Birds at a high-tide roost are not resting for pleasure; the flats are gone and they are conserving energy before hours of feeding. A flock flushed repeatedly burns the reserves it came here to build. In autumn, when many of these birds are preparing for a flight of several thousand kilometres, the cost of a careless approach is not abstract, and it is paid somewhere far to the south.

What I carry and what I have stopped carrying

Binoculars and a telescope on a light tripod, because at these distances a scope is the difference between grey birds and identified birds. A woollen hat, gloves I can work the focus wheel through, and a sit mat, since the dike is wet at six in the morning from September onwards. A paper notebook with a pencil, which keeps writing when the pages get damp and does not need charging on a train home.

I have stopped carrying a camera with a long lens. This is a personal choice rather than advice: I found I was spending the whole tide trying for a photograph and coming home with no notes and no memory of the morning. The other thing I stopped bringing is a plan to be anywhere else afterwards. The roost forms when the water says so, and the walk back along the dike into a headwind always takes longer than I think.