How do newly rescued pressure observations help reconstruct the weather?

The pressure observations that are being rescued will have a large demonstrable effect on our knowledge of past weather.

In the graphic below, the left hand side always shows our current estimates of the locations of the isobars (lines of equal pressure) for 9am on 26th February 1903. There are 80 different sets of lines, each of which are considered equally likely given the information currently available. There is not much agreement over western Europe, so it looks very messy!

The animation on the right hand side shows how confident we become when individual observations (red dots) are added one by one. Starting in southern Europe the isobars gradually begin to become better defined until a much clearer picture emerges when all 55 newly rescued observations have been added.

We will be repeating this process for the whole 1900-1910 period that we are rescuing data for so that we can see the improvements in detail. This will demonstrate the value of this data for answering interesting science questions about how the circulation of the atmosphere varies over time.

An analogy is that we currently have blurred ‘polaroid’ pictures of the weather at this time, but gradually we are sharpening the image. As we add observations the picture becomes more like a digital photo.

[Very technical details: this animation uses all 80 members from v3 of the 20th Century Reanalysis and an offline version of the Ensemble Kalman Filter to assimilate the new observations.]


1 million observations!

1644 volunteers have now helped rescue over 1 million daily weather observations from 61 locations all over Europe during 1900-1906! Thank you!


Grey regions indicate where there is missing or uncertain data. Often this will be due to a station joining or leaving the Daily Weather Report logs. Some of these gaps will be filled in during the quality control phase of the project.

Examining the rescued pressure data

The Weather Rescue volunteers have now rescued all the observations written in the Daily Weather Reports for 1900-1903, so it’s time to have a closer look at some of the data. We’ve picked October 1903 as this is the wettest month ever recorded in the UK.

The video below, made by Philip Brohan, looks complicated at first. The left hand side shows an animation of our previous understanding of the atmospheric pressure, for every hour during October 1903. This is produced using a modern weather forecasting model, but the simulations do not include the information in the new rescued observations, which are shown as the red dots.

There are several blue lines on the map (looking a bit like spaghetti!) which show the pressure in different simulations of the past. Because we don’t currently have that many observations to constrain the weather patterns there is a lot of uncertainty about where some of the pressure features are. The thicker black lines show the pressure contours where we are more confident.

The right hand side lists the weather stations in the Daily Weather Reports, and the red bars show the rescued pressure observations. The blue dots show the different simulations for that location and time.

There are lots of interesting features to note. You may notice that the maps often have a sharp adjustment – these are the times at which the existing data is ‘assimilated’ into the simulations. Often the blue dots in the right hand panel show a wide range – when the new observations are included then this uncertainty will reduce substantially. Sometimes the existing simulations (blue dots) are close to the new observations (red bars), but often not. It is these occasions where the new data has most value, and we expect the blue ‘spaghetti’ lines to become much closer to each other.

Importantly, the improvements will be largest when low pressure storms are passing near the stations. We want to learn about the frequency of intense storms in the past to compare with now and these rescued observations will significantly improve our understanding.

We have also made an animation of the rainfall data rescued so far in 1903. Note October in particular as being a very wet month for the UK!

February 1903 – the Ulysses storm

In late February 1903, a large storm came across Ireland and Scotland. It is believed to have inspired a small passage in the novel Ulysses by James Joyce:

O yes, J.J. O’Molloy said eagerly. Lady Dudley was walking home through the park to see all the trees that were blown down by that cyclone last year and thought she’d buy a view of Dublin.

The storm caused structural damage to many regions, and uprooted trees.

Our current best view of the storm comes from the 20th Century Reanalysis – this is essentially a modern weather forecast model used to simulate past weather by filling in the gaps between the fairly limited observations available. Continue reading