Wind Tyne mouth
…
Wind at the Tyne mouth, river flow at Bywell and the tide at North Shields. Hour by hour for the next 12 hours, then day by day.
Loading the latest forecast…
…
…
…
| Time | Wind (m/s) | Rain (mm) | River flow (m³/s) | Tide (m) |
|---|---|---|---|---|
| Loading… | ||||
How the last week's forecasts compared with what was then measured. They are checked automatically every hour.
Wind forecasts aren't scored here, because none of our live data feeds measures the wind at the Tyne mouth.
These are research estimates, not warnings. For flood warnings see the Environment Agency's flood service. For marine and weather warnings see the Met Office.
The wind is forecast for a point between the Tyne piers. For the first 54 hours we use the Met Office UK model, which has a 2 km grid. After that we use the ECMWF global model (9 km grid). A dashed line on the chart marks the changeover. Speeds are 10-minute means at 10 m above the surface, in metres per second (1 m/s is about 2.2 mph). Directions are where the wind blows from. We shade winds from the east, meaning anything between north-east and south-east (45° to 135°).
Flow at Bywell is the river's discharge from its 2,176 km² catchment above the tidal limit. The Environment Agency publishes checked flow weeks after the event, so the current flow is calculated from the live river level, using a rating fitted to the Agency's own level and flow data since 2020.
The forecast comes from a neural network (an LSTM, a type of network designed for time series). It was trained on 20 years of hourly flow and rainfall from 18 rain gauges across the catchment. Each hour it reads the past two weeks of rain, air temperature and flow, together with summaries of how wet the catchment has been over the past six months. It then runs forward using the forecast rain. The rain comes from the Met Office UK model for 54 hours, then ECMWF. Both models forecast 15–20% more rain over the catchment than the gauges record, so their rain is scaled down to match the gauges the network was trained on.
Beyond two days, a single rain forecast often gets the timing of rain wrong. From 48 to 72 hours, the forecast line hands over to the middle of the 51 alternative ECMWF forecasts.
The likely range combines two sources of uncertainty. One is the 51 alternative ECMWF ensemble rain forecasts. The other is the model's own error, measured on years it never saw during training. The dark band covers the middle half of outcomes, and the light band 80% of them.
The flow model can't know about unusual releases from Kielder Water, which regulates the North Tyne, or about snowmelt beyond what the air temperature implies.
Tides are predicted from a harmonic analysis of the Environment Agency's North Shields tide gauge: 44 tidal constituents, fitted to the past year of 15-minute readings. Heights are in metres above Ordnance Datum Newlyn (mAOD), which is roughly mean sea level. That datum is not chart datum, so heights differ from those in Admiralty tide tables.
Tidal range is the height difference between a high water and the low waters either side of it. Each day is described by its largest tide. When that tide's range is close to the average spring or neap range, the day is labelled spring or neap. In between, we compare it with the largest tides of the days before and after: growing ranges are heading towards springs, shrinking ones towards neaps. Wind and air pressure can raise or lower the sea by tens of centimetres (storm surge). The measured-minus-predicted difference is shown as "surge" when the gauge is reporting.
The forecast updates every hour. You can download the current forecast data as JSON.