 The flooding of 2005 was a catastrophe that will remain in our memories for a long time. After strong precipitation, many Swiss rivers and lakes burst their banks, causing flooding and great damage. The Bernese-Mutter district was also affected by the water masses of the Arre River. Extreme events like 2005 are not unique, but have happened again and again throughout history. There are numerous examples of disastrous floods in Switzerland. One such event, the flooding of 1868, stands apart. The Largo Maggiore reached its highest known water levels and in the Rhine Valley, lakes formed. The event caused tremendous damage and 51 people lost their lives. Coping with the event was a great challenge for the still young nation-state. For the first time in its history, the Federal Council took the initiative in dealing with the natural disaster. To truly understand and to be able to predict such events, we need to study as many as we can. For 2005, we have many observations of the weather, even radar and satellite data, allowing us to analyse such a flood event in detail. In the past, however, quantitative data are scarce, as systematic weather observation systems weren't established yet, and only few observations were made. Those few observations, however, are very valuable, since today, quantitative methods can again make use of this historical information. But first, we need to find the data. Sometimes, you don't need to look very far. In the middle of the 19th century, Daniel Gottlieb Benoit measured and noted meteorological parameters from his house, just a few metres next to the flooded Mutter district. He was not the only one. In fact, meteorological data have been measured since the age of Enlightenment. In Switzerland, the first known meteorological measurement series from Johann Jakob Schroeder in Zurich dates back to the year 1708. Over the years, more and more private observers in natural science societies started to measure different meteorological parameters. Some of them even started their own station network, like the Bernese economical society in the early 1760s, although most of them only lasted a couple of years. We see the same trend all over Europe and North America. With the rise of nation-states in the 19th century, long-term and stable measuring networks were finally established, such as the official network of the Swiss Central Meteorological Society in 1863. Whilst some of these historical series have already been digitised, the majority have not been studied at all. Only many such series together can provide enough information to reconstruct the weather in detail, much like a huge puzzle in which most of the parts are missing. The more pieces we have, the better our reconstruction gets. In Switzerland, the James Project, funded by the Swiss National Science Foundation, aims at excavating these historical meteorological measurements from archives and making them accessible for scientific work. In collaboration with historians, the manuscripts and metadata from these historical series are compiled and photographed. The next step is to digitise the data. This requires carefully typing the handwritten measurements into computer tables. However, it is still a long way from these raw instrument readings to climate reconstructions. First, the measurements need to be processed. Let's take air pressure as an example, which was mostly measured with mercury barometres. What is measured is the length of the mercury column. For scientific applications, however, this length needs to be converted into mean sea level pressure. This conversion requires four steps. Adjustment to a standard temperature, conversion to pressure units, correction for local gravity and reduction to mean sea level. First, because mercury expands or shrinks depending on temperature, we need a temperature measurement near the barometer, as well as the expansion coefficient of mercury to correct this effect. Then, the length of the column, often given in old units such as parry's inches, must be converted to millimetres and then to units of pressure using the hydrostatic equation and the density of mercury. In this equation, standard gravity is used. Local gravity, however, differs from standard gravity, and this must be corrected. Finally, when comparing atmospheric pressure at several stations, the altitude differences between the stations must be considered. Therefore, pressure is often reduced to mean sea level. The next step is quality control. This procedure aims at eliminating errors in the data. There are many sources of errors. The instruments may not work well, or errors may occur during transcription and processing. Simple plots can already visually reveal a lot of these errors. Tests can then be used to find values that are outside of a physically possible or plausible range. A third set of tests then makes use of the full length of the series to check if values are statistically plausible. For instance, values that exceed four standard deviations are flagged. Also, the sequence is analysed. Pairs of temperature observations on consecutive days that differ by more than 25 degrees Celsius are flagged, as are more than three equal values in a row. Further tests can address the physical consistency between two variables and compare the series to data from neighbouring stations. Even after quality control, metrological data series may still reflect non-climatic influences. For instance, the replacement of an instrument or a change of the measurement site may lead to a step change. Such non-climatic effects are called inhomogeneities, and the process of removing them is called homogenisation. To detect artificial changes, they must be separated from the real climate signal. Such changes are not always apparent in the original series, but can be detected, for instance, by subtracting data from a neighbouring station. We plot the cumulative sum of the differences. This is a simple method to detect breakpoints, which lead to changes in the direction of the lines. To find possible explanations for the breakpoints, the metadata must be carefully studied. The result is a homogenised metrological series in standard units, which provides point-wise information on the weather of the past on a local scale. In the next video, we will show you how we get from this data to global weather reconstructions.