The water level on Lake Neusiedl in 2022 — viewpoints of a sailor and statistician

13 min readMay 3, 2023

(this article is also available in German, an update for 2023 is available)

Version 1.0 | 2022 May 09 | Initial Version in German, Translated to English

I have been sailing on Lake Neusiedl, in Austria, since I was a child, in all different kinds of boats. Sailing is relaxation, exercise, and social life for me.

In recent years, however, I have noticed a worrying trend in the water levels in the lake. Imagine yourself back in May this year (2022), and planning your sailing over the summer months. If you went to the harbour or the lido on Lake Neusiedl at that point, you might have thought that the level didn’t look too bad. There perhaps wasn’t as much water as you would have liked, but there was some. On average, however, the lake loses 18 cm of water through evaporation from May to the end of September. In 2021, it lost more than 32 cm during this period.

It seems likely that the water level will drop during the 2022 summer season. In only two summers in the past 23 years did the water level in the lake not drop by 1 September. In almost two thirds of years, we have seen a drop in the water level of 15 cm or more (see figure below).

In 2003, 2013 and 2017 the lake lost 30 cm or more.

Surely this doesn’t matter in the long term? Surely the winter rain and snow simply fill the lake up again? Unfortunately not. Perhaps that was the case in the past, but it certainly is not reliably true now. The water level on 9 May 2022 was 24 cm less than on the same date the previous year.

Why does the water level matter, though? For me as a sailor, it matters because the lake provides me with exercise, relaxation and a social life. If I can’t sail, all these are gone. But the lake is important as more than just a sailing venue, and to more than just me. It is an important part of the landscape and the microclimate in the local area, not to mention crucial for local industries from fishing to tourism.

In my main job, I am a consultant for analytical solutions and statistics at a large Analytics Software Company. I have always enjoyed analysing data. I like analysing the data on the water level in Lake Neusiedl, but in recent years, I have not been increasingly unhappy about the results.

Modelling the water level change in the summer months

Elevations in Austria are typically given as metres above the Adriatic Sea (mAS). On 9 May 2022, Lake Neusiedl had a water level of 115.22 metres above sea level.

To make life easier, I use the term water depth in this article and calculate it as the water level in mAS minus 114m. The lake is obviously not the same depth everywhere, but in this model, I assume that the lake bottom is at 114mAS and the difference is the noticeable water depth of the lake. At 115.22mAS, this would be a water depth of 1.22m on 9 May 2022.

Elevations in Austria are typically given as metres above the Adriatic Sea (mAS). On 9 May 2022, Lake Neusiedl had a water level of 115.22 metres above sea level.

I also show the depth in front of the harbor exit at Weiden am See. In 2020, I measured at one point that the lake bottom was at 114.18mAS. A water level of 115.22mAS therefore corresponds to a water depth of 1.04 m at the harbor exit.

Simulation for 2022 — What would happen if the summer were like 2021, 2019, 2010, 2003, …?

In the first analysis, I used data for all the years for which it was available: 1999 to 2021, a total of 23 years. For these years, we know what happened to the water level after 9 May. In my simulation, I linked the current water level in 2022 with the water level in the past. I shifted the curve for past years so that it would also have a depth of 115.22mAS on 9 May.

Based on this shift, we can now deduce how the year 2022 would proceed if the water level were to develop as it did in a selected past year.

Expected water depth in graph and table

This is shown graphically in the following diagram. The black line shows the water level trend in 2022 up to 9 May. From 9 May onwards, the last 23 years are shown so that we can see the range and what scenarios are possible here. This is more informative than talking exclusively about mean values or mean value trends, because it also gives us an impression of the variability and minima and maxima of the last 23 years.

*Scenarios for water depth in 2022 (water level minus 114m)*

In the graph above we see that over the last 23 years there have been:

Two years (2010 and 2014, 8.7 %) in which the water level was roughly maintained or slightly improved between the beginning of May and the beginning of September. These years are shown in dark green in the graph above.
Six years (1999, 2005, 2006, 2008, 2009, 2020, 26.1%) in which the water level had dropped by only 4 cm to 11 cm by 1 September. These years are shown in green.
Six years (26.1%) in which the water level fell by 15cm to 22cm by 1 September (years 2004, 2011, 2012, 2016, 2018, 2019, shown in amber).
Nine years (39.1%) in which the water level fell by between 22 cm and 34.5 cm (2000, 2001, 2002, 2003, 2007, 2013, 2015, 2017, 2021, shown in brown). The worst year was 2003, where the water level fell by 34.5cm, followed by 2013 and 2017 with about 30 cm each.

The following table shows the water level at the beginning of each month in June, July, August, September and October, assuming that the year 2022 follows the same pattern as the respective reference year.

Additional effect in the event of wind

Water loss of 10 cm or more can also occur during (sustained) strong winds. For example, the harbours in Breitenbrunn, Neusiedl and Weiden lose water in strong north-west and north winds, and those in Rust, Mörbisch and Illmitz in strong south winds. We therefore also need to correct the table for wind.

Reduced depth in the harbor exit

The following graph shows the same data as in the graph on water depth, but with the depth in front of the harbor exit at Weiden am See on the Y-axis. This is 18 cm less than in the calculation of the water depth with 114mAS as the basis. In more than a third of the scenarios, the water level there is 80 cm or less.

*Scenarios for the water depth in front of the Weiden harbour exit in 2022*

The monthly water level change in summer — A statistical explanatory model

Back in 2020, I built a statistical explanatory model for the monthly water level changes in the summer months of June, July, August and September and also published it in a Youtube webinar presentation.

The aim of this model was to quantify the influence of rainfall, temperature and wind on the monthly change in water level. I contemplated using complex machine learning algorithms, but in the end, chose a simple linear regression model because of its better explanatory power and the limited data available.

Predictor variables for the model

A model like this quantifies the influence or “predictive power” of characteristics (e.g., average temperature, precipitation) on a target variable, “water level change on the 1st of the month compared to the 1st of the previous month”.

I “offered” the following characteristics to the regression model; these were calculated per month:

Precipitation in mm
Number of rainy days
Longest time interval (in days) in which it did not rain
Average temperature
Number of days with a maximum temperature > 25°C
Number of days with a maximum temperature > 30°C
Water load (wind loss) for each of Neusiedl, Breitenbrunn, Podersdorf and Illmitz

The software calculates a weighting for each of the influencing variables. The statistician’s task is to select the influencing variables so that the model:

• Has a high predictive power: it can predict the water level well from the variables used.

• Is stable: the results should not fluctuate greatly from month to month and should also apply to future time periods.

• Is simple to interpret, a common reason for using a linear regression model rather than a complex machine learning model.

Depending on how many features were entered, the explanatory power of the regression model varied between 85.5% and 88.1%. This means that ~85%-88% of the water level change can be explained by the model. (For statisticians: I am talking about the non-adjusted R2 here).

A simple model with only two features

I finally decided on a simple model with two features:

Precipitation in mm
Number of days with a maximum temperature > 25°C

It is easier for us to think of the “number of hot days per month” than to decide whether a monthly average temperature of 19 degrees means a hot July. The model has an explanatory power of 85.2% with these two features only. The parameters of this model are as follows:

Constant: −58.2
Number of days > 25°C: −3.33
Precipitation: +1.097

We can therefore interpret the model equation as follows

• In the summer months (June, July, August, and September), the lake loses an average of 58.2 mm of water (just under 6 cm).

• Every day where the temperature is more than 25°C, the lake loses another 3.33 mm (meaning that for every 10 days at this temperature, the lake loses about another 3.3 cm).

• Rainfall (precipitation) of 1 cm increases the water level by about 1.097 cm.

Suppose July has 12 days with more than 25°C and 50mm of rain. The predicted change in water level is a decrease of 43mm (−58mm−(3.33*12)+(50*1.097)) (see figure below).

Predictions for the future

The next question, of course, is whether we can predict whether the lake will dry up or the water will fall below a certain depth. It should be clear that this model can make that prediction with quite high accuracy as soon as you know the amount of precipitation or the number of hot days.

Of course, we cannot know these precisely for future years. However, we can use the model to carry out what-if analyses and estimate future developments. In my Video I show this from 14 min 50 sec onwards as an example for values in selected summer months.

We must all make our own predictions about what we expect to see in terms of temperatures and precipitation in times of climate change. However, just as there are expected values in statistics, there are also outliers. The data show us that we could recover a lot of the water with two cooler years and a higher than average annual precipitation. Could we also recover it manually, by ‘refilling’ the lake?

Filling the “bathtub”

I wouldn’t exactly call the ecosystem of Lake Neusiedl a bathtub, but it is a broadly closed system. Once the water is in, it doesn’t flow out again. The only loss is therefore evaporation. Let’s therefore assume that we could fill it by turning on a supply line. How long would that take?

First, you would need permits, financing, and an actual water supply line to be built. I am no expert, but I doubt this could happen before the end of 2025. Beyond that, however, how much water could the supply line bring in? Will it solve the problem of low water levels?

Reference value: The river Wulka

On the western shore of Lake Neusiedl, near Donnerskirchen, the River Wulka flows into Lake Neusiedl. According to Wikipedia, it has a water inflow of 1.2 cubic metres per second.

*Source: Hydro Burgenland 2019 —* https://wasser.bgld.gv.at/hydrographie/die-fluesse/trausdorf

It seems reasonable to use this figure as a reference value for the calculation.

Raising the water level by 10 cm

Lake Neusiedl has an area of 320 square kilometres. We therefore need to look at how long it would take the supply line to raise the water level of the lake by 10 cm. If we want to add a depth of 10cm to an area of 320 km2 ( 320,000,000m2), we will need 32,000,000 cubic metres of water. This is equivalent to either:

A cuboid with a floor area of 1000 x 1000 m and a height of 32 metres, which I have drawn on the satellite picture below.

Or a cube with edge length 317 m.
This volume divided by our comparative value of 1.2 cubic metres per second = ~26,600,000 seconds = ~7407 hours = ~308 days, or roughly 10 months. If the supply line brings in 1.2 cubic metres per second, it will therefore take about 308 days until the water level has increased by 10cm. Evaporation and precipitation would take place both now and with the supply line and can therefore be omitted.

308 days to raise the water level by 10 cm

A water supply that brings in as much water as the Wulka would therefore supply an increase of about 1cm in the water level in one month. However, the mean water level in Lake Neusiedl on 13 April 2022 was 25 cm less than on the same date last year (115.47 vs 115.22mAS). The water supply would have to flow for a little more than 2 years to bring back what has been lost since last year.

Let’s put this in perspective. On a day where it rains heavily, a good 20 mm of precipitation can fall. One day of rain can therefore bring as much water as the water supply system can provide in 2 months.

From a water level perspective, an increase in water level of 12cm per year is of course better than nothing. These calculations show very clearly, however, that it will take some time for the water level to become noticeably higher — and we must not forget that any water supply system would need to be approved and built. This is not a short-term solution. Of course the answers would be different if we used a different flow rate. However, this gives an idea.

Conclusion and a few personal views

Will the lake will dry up? We are certainly closer to this now than we were 10 years ago. We already have very little water in the lake and therefore have no buffer for future hot summers, especially if the winter rainfall does not increase. The climate forecasts are unfortunately pointing in the wrong direction too.

Even if the lake does not dry up completely, dropping levels will have a massive impact on water sports, tourism, ferry traffic and fishing.

Will a possible water supply line save us? Certainly not in the short term, because it will take time to approve, build, and then raise the water level. Would it ever be approved or gain funding? I don’t know — and nor do I know the ecological impact. I merely show you a statistical model, and suggest that its main conclusion is that there are no easy answers.

Updates

I will keep updating and expanding this article. I plan to update the simulation each year if possible. I would be happy to receive feedback, suggestions and ideas. This article was also published by me on LinkedIn and XING in a post. If you have a user profile there, you can leave comments on the post there. I can also be reached by email for these topics at sastools.by.gerhard@gmx.net.

2023 Simulation (in German)