27 November 2020
My latest blog, written on 26 October, speculated that "if seasonal patterns hold, rain should be even more plentiful in November and water levels should be supported." Well, that turned out to be completely false.
With the benefit of hindsight and as there remain just a few days until the end of the month, November will enter the history books as the driest in 10 years.
It has rained just 0.8 mm/day on average at stations along the Rhine. This is less than half the level recorded in November 2019 and is actually less than was recorded over the summer in Europe - in a month that is supposed to be the third or fourth wettest of the year!
Needless to say, this has had a huge impact on the Rhine, with water levels falling steadily in the last few weeks. At Kaub, water levels are just above 1 metre at the time of writing and are likely to fall to around 80 cm in a few days, thus nearly breaching the official low water threshold of 78 cm for the location. This means more pain for barge operators, who must incorporate these restrictions into their operational plans.
Rain is likely to remain elusive in the early part of December, thus prolonging the pain for everybody. And, though highly speculative at this stage (given that weather forecasts become unreliable after 10 days), the ECMWF 10-42 day forecasts do not point towards a very wet December either.
26 October 2020
Rhine water levels are surging amid rainy weather, the fourth such bounce registered since the end of summer in Europe.
Rain has kept water levels reasonably high this autumn and mostly above the five-year average. Soil moisture readings along the river are back to normal levels, after being consistently in drought territory.
This is in stark contrast with the spring and summer periods when water levels were constantly low due to hot and dry weather. It has rained around 2.4 mm per day on average so far during October, above the five-year average of 1.8 mm.
If seasonal patterns hold, rain should be even more plentiful in November and water levels should be supported.
For now, we expect Kaub water levels to increase to above 2 metres by the end of October, allowing normal navigation all along the river. We’ll need more rain in November to maintain these levels, but what’s clear is that a rainy autumn has mostly staved off the threat of very low water levels for the rest of this year.
15 September 2020
Long Short Term Memory (LSTM) models are a powerful type of neural network ideally suited to predict time-dependent data. Rhine water levels fit right into this category: they vary over time, depending on a range of variables such as rain, temperatures and snow cover in the Alps.
The Rhine is Europe’s lifeblood. For centuries it has been used as a major artery for shipping goods into Germany, France, Switzerland and Central Europe. However, with climate change, water levels on the river are likely to become more variable. Forecasting the river's level accurately is therefore a primary concern for a whole range of actors, from shipping companies to commodity traders and industrial conglomerates.
Unlike classical regression-based models, LSTMs are able to capture non-linear relationships between different variables; more precisely, the sequence dependence among these variables. This blog focuses on the problem of Rhine river forecasting using LSTMs, rather than the theory behind these models.
The problem we are looking to solve here is the following: we would like to forecast next-day water levels at Kaub, a key chokepoint in western Germany, with the highest possible accuracy.
We have historical daily data from 2 January 2000 to 27 July 2020, equivalent to 7513 observations. The dataset includes 15 different categories, displayed as columns:
Note that the choice of variables is entirely mine and is based on my experience dealing with Rhine analysis. Selecting the right inputs is one of the most important steps in time series analysis, whether you are using classical regression models or neural networks. If you select too few variables, the model may not capture the full complexity of the data (this is called underfitting). By contrast, if you choose too many inputs, the model is likely to overfit the training set. This is bad as well as it could mean the model struggles to generalise to a new dataset, which is essential for predictions.
First, let’s load all the libraries we will need for this exercice:
import datetime import matplotlib as mpl import matplotlib.pyplot as plt import numpy as np import pandas as pd import seaborn as sns import tensorflow as tf from sklearn.preprocessing import LabelEncoder from sklearn.preprocessing import StandardScaler from sklearn.metrics import mean_squared_error import joblib
Below is a sample of the first few lines of the dataset. We can load it easily with the Pandas library:
# first, we import data from excel using the read_excel function df = pd.read_excel('RhineLSTM.xlsx') # then, we set the date of the observation as the index of the entire dataframe df.set_index('date', inplace=True) df.head()
Once loaded, we can plot the dataset using the Matplotlib library:
# specify columns to plot columns = [0, 1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14] i = 1 values = df.values # define figure object and size plt.figure(figsize=(9,40)) # plot each column with a for loop for variable in columns: plt.subplot(len(columns), 1, i) plt.plot(values[:, variable]) plt.title(df.columns[variable], y=0.5, loc='right') i += 1 plt.show()
It’s also generally a good idea to plot histograms of the variables:
# histograms of the variables df.hist(figsize=(9,18)) plt.show()
Using the Seaborn library, you can create a violin plot to understand the distribution of each variable:
# calculate dataset mean and standard deviation mean = df.mean() std = df.std() # normalise dataset with previously calculated values df_std = (df - mean) / std # create violin plot df_std = df_std.melt(var_name='Column', value_name='Normalised') plt.figure(figsize=(12, 6)) ax = sns.violinplot(x='Column', y='Normalised', data=df_std) _ = ax.set_xticklabels(df.keys(), rotation=90)
I have trained different types of models on the Rhine dataset to establish which one fits best:
The Python code required to create this chart is too long for this blog, but you can access it here, applied to a different dataset.
LSTM networks are a form of recurrent neural network that can learn long sequences of data. Instead of neurons, they are made of memory blocks connected to each other via layers. A memory block contains gates (input, forget, output) that manage its state and output, and enable it to be smarter than a typical neuron.
They have been used extensively in academic circles to forecast river height and have been proven to outperform classical hydrological models in certain situations.
Let’s start all over again with the code from the beginning of this article to load the Rhine database:
import datetime import matplotlib as mpl import matplotlib.pyplot as plt import numpy as np import pandas as pd import seaborn as sns import tensorflow as tf from sklearn.preprocessing import LabelEncoder from sklearn.preprocessing import StandardScaler from sklearn.metrics import mean_squared_error import joblib # first, we import data from excel using the read_excel function df = pd.read_excel('RhineLSTM.xlsx', sheet_name='Detailed4_MAIN’) # then, we set the date of the observation as the index of the entire dataframe df.set_index('date', inplace=True)
To build a functioning LSTM network, the first (and most difficult) step is to prepare the data.
We will frame the problem as predicting today’s rate of change in Kaub water level (t) given the weather and Swiss upstream flows of today and the previous 6 days (backward_steps = 7).
The dataset is standardised using the StandardScaler() function in the Scikit-Learn library. For each column of the dataframe, each value in the column has the mean value subtracted, and then divided by the standard deviation of the whole column. This is a pretty ordinary step for most machine learning models and allows the whole network to learn faster (more on this below).
Then, the dataframe is passed through a transformation function. For each column, we create a copy of each of the previous 7 days’ values (15 * 7 = 120 columns). The shape of the resulting dataframe is 7506 rows x 120 columns.
# load dataset values = df.values # ensure all data is float values = values.astype('float32') # normalise each feature variable using Scikit-Learn scaler = StandardScaler() scaled = scaler.fit_transform(values) # save scaler for later use joblib.dump(scaler, 'scaler.gz') # specify the number of lagged steps and features backward_steps = 7 n_features = df.shape # convert series to supervised learning def series_to_supervised(data, n_in=1, n_out=1, dropnan=True): n_vars = 1 if type(data) is list else data.shape df = pd.DataFrame(data) cols, names = list(), list() # input sequence (t-n, ... t-1) for i in range(n_in, 0, -1): cols.append(df.shift(i)) names += [('var%d(t-%d)' % (j+1, i)) for j in range(n_vars)] # forecast sequence (t, t+1, ... t+n) for i in range(0, n_out): cols.append(df.shift(-i)) if i == 0: names += [('var%d(t)' % (j+1)) for j in range(n_vars)] else: names += [('var%d(t+%d)' % (j+1, i)) for j in range(n_vars)] # put it all together agg = pd.concat(cols, axis=1) agg.columns = names # drop rows with NaN values if dropnan: agg.dropna(inplace=True) return agg # frame as supervised learning reframed = series_to_supervised(scaled, backward_steps, 1)
We must split the prepared dataframe into training and test datasets to allow a fair evaluation of our results. The training dataset represents 80% of our values and we will use the remaining 20% for evaluation. As we are dealing with data ordered through time, it is a very bad idea to shuffle the dataset, so we keep it as is. Next, we reshape our training and test datasets into three dimensions for later use.
# split into train and test sets values = reframed.values threshold = int(0.8 * len(reframed)) train = values[:threshold, :] test = values[threshold:, :] # split into input and outputs n_obs = backward_steps * n_features train_X, train_y = train[:, :n_obs], train[:, -n_features] test_X, test_y = test[:, :n_obs], test[:, -n_features] print(train_X.shape, len(train_X), train_y.shape) # reshape input to be 3D [samples, timesteps, features] train_X = train_X.reshape((train_X.shape, backward_steps, n_features)) test_X = test_X.reshape((test_X.shape, backward_steps, n_features)) print(train_X.shape, train_y.shape, test_X.shape, test_y.shape)
Finally, we are able to fit our LSTM network. Thanks to the TensorFlow/Keras library, this only requires a few lines of code. I have chosen to fit 64 memory blocks in batch sizes of 72. I use the Adam optimisation algorithm, which is more efficient than the classical gradient descent procedure.
# design network model = tf.keras.models.Sequential() model.add(tf.keras.layers.LSTM(64, input_shape=(train_X.shape, train_X.shape))) model.add(tf.keras.layers.Dense(1)) model.compile(loss='mae', optimizer='adam') # define early stopping parameter callback = tf.keras.callbacks.EarlyStopping(monitor='loss', patience=3) # fit network history = model.fit(train_X, train_y, epochs=25, callbacks=[callback], batch_size=72, validation_data=(test_X, test_y), verbose=2, shuffle=False) # plot history plt.figure(figsize=(12, 6)) plt.plot(history.history['loss'], label='train') plt.plot(history.history['val_loss'], label='test') plt.ylabel('mean absolute error [Kaub, normalised]') plt.legend() plt.show()
After the model is fit, we can launch the forecast and invert the scaling to obtain our final results. We can then calculate an error score for the model. Here, the model achieved a root mean squared error (RMSE) of 6.2 centimetres, which is good but can probably be improved on.
# make a prediction yhat = model.predict(test_X) test_X = test_X.reshape((test_X.shape, backward_steps*n_features)) # invert scaling for forecast inv_yhat = np.concatenate((yhat, test_X[:, -(n_features - 1):]), axis=1) inv_yhat = scaler.inverse_transform(inv_yhat) inv_yhat = inv_yhat[:,0] # invert scaling for actual test_y = test_y.reshape((len(test_y), 1)) inv_y = np.concatenate((test_y, test_X[:, -(n_features - 1):]), axis=1) inv_y = scaler.inverse_transform(inv_y) inv_y = inv_y[:,0] # calculate RMSE rmse = np.sqrt(mean_squared_error(inv_y, inv_yhat)) print('Test RMSE: %.3f' % rmse)
I have also checked the model's performance during extreme weather events, such as the extensive floods recorded in Europe in May-June 2016 which caused more than Eur1 billion of damage in Bavaria alone. The model was generally able to track the rise in water levels, however during two particular peaks (one in mid-April and the other in early June) it gave low figures.
The trained LSTM network also performed well during storm Axel, in May 2019, which caused a very rapid rise in water height of more than 1 metre on two consecutive days. Once again, however, it gave slightly lower estimates than actual figures when floods peaked.
In both cases, I suspect the low figures are due to one major Rhine tributary, the Moselle, which was left out of the analysis for lack of reliable meteorological data. The Moselle joins the Rhine at Koblenz, just north of Kaub. This is a possible area of improvement in future.
Finding the best hyperparameters is one of the most important (and time-consuming) tasks in machine learning. I ran many different versions of the network in order to find the best possible setup:
I have deployed a version of this model online. Its performance can be monitored here.
I’ll try to improve this model in future by adding/removing variables and tweaking hyperparameters further. If you have any thoughts, don’t hesitate to get in touch with me.
31 August 2020
Water levels on the Rhine river are likely to more than double in the space of just a few days in early September as a result of higher upstream flows from the Swiss Alps and higher rainfall.
Kaub levels fell below the 1 metre threshold in late August to their lowest level since the 2018 drought. This followed a scorching summer in Western Europe that saw temperatures reach 40 degrees Celsius on some days.
Water levels on the Rhine river are likely to more than double in the space of just a few days in early September as a result of higher upstream flows from the Swiss Alps and higher rainfall.
We forecast Kaub water levels to reach nearly 2.4 metres on 3-4 September. The WSV is even more bullish, predicting 2.5 metres on 2-3 September. Heavy rainfall in Switzerland is to blame.
Flows at Rheinfelden, near the Swiss/German border, have gone up from 800 cubic metres/second on 29 August to nearly 1,200 cubic metres on 30 August and will reach 1,800 cubic metres today, on 31 August.
However, this surge will be just that without more rainfall over the next few weeks. The WSV expects Kaub levels to fall back to 1.5 metres on 9 September, although this forecast is still subject to weather changes.
28 May 2020
Water levels on the Rhine are hovering around their lowest level in five years for the time of year. They’ve been at the bottom of the five-year range since the start of April and do not show any signs of recovering meaningfully in the short-term.
The main reason has been the persistent lack of rain along with higher-than-normal temperatures in Switzerland and southern Germany, which has kept underground reservoirs dry. Soil moisture along the southern section of the Rhine is well below both last year’s figure and the five-year average. And this is before the summer season begins in Europe.
Another contributing factor has been the relatively poor snow season in the Alps. Warm temperatures over April and May have depleted what was already a shallow snow cover going into the spring. Snow cover below 2,000 metres in the Swiss Alps was all gone in mid-April, around two weeks earlier than usual. And snow cover at higher altitudes is just above 1 metre at the time of writing, compared with 4 metres at the same time last year.
In fact, glacier melt has been the main source of upstream run flow since the start of May, overtaking both snow melt and rainfall.
All this means that the Rhine is living on borrowed times and could fall below the critical 1-metre threshold if no rain comes. This website forecasts Rhine water levels to rise a little in early June together with rain, but that is still subject to uncertainty and could easily be reversed later in June.
28 April 2020
It rained about 3 mm today in Northeastern France, Western Germany and Switzerland, and more is coming over the next few days. As the forecasts stand, it could rain substantially over the weekend of 2-3 May, with about 13-14 mm on each day.
This should be enough to bolster Rhine water levels, which have hovered around the low end of the five year range since the beginning of April. At Kaub, a key chokepoint, they were just 1.2 metres today, not enough for barges to circulate fully loaded.
This blog forecasts Kaub water levels to reach nearly 2 metres by 7 May, helped by rising Swiss upstream flows (themselves the result of a combination of precipitation and snow melt) and rain in Germany and France. Germany’s Federal Waterways and Shipping Administration forecasts water levels of 1.5-1.6 metres by that date.
I don’t know for sure why our figures differ by 50 centimetres. It could be because the WSV uses weather forecasts that show less rain than me, and/or because our coefficients differ for April and May (ie, the same amount of rain triggers a higher water level in my model than in theirs).
But, in any case, the pattern is clear and water levels are going to rise. May is typically 60% rainier than April in southern Germany, even if it is far too early to know if that will be sustained later in May. In addition, as I explained in my blog published last week, more snow from the Alps is likely to melt, which will feed the river.
All in all, we are therefore unlikely to test the recent lows on the river, barring a drastic change in the weather.
21 April 2020
The coronavirus is upsetting the world around us, but amid the furore it can be easy to miss important phenomena.
Where I’m sitting in Paris, we’ve had almost no rain so far in April. This is borne out by weather data showing that, at 0.4 mm per day on average so far this month, April is proving to be the driest month along the Rhine since April 2007.
This has had important implications for river levels, which are at their lowest (for the time of year) in five years. At Kaub, a key chokepoint, the river was just above 1 meter at the time of writing, not enough for barges to circulate fully loaded.
With many of Europe’s key industries closed and energy demand in freefall amid the coronavirus pandemic, these restrictions may not be so visible. But if they last, they will be a thorn in the foot when businesses reopen in May.
I’m not too worried about the next few weeks, however, as we’re right at the beginning of the snow melting season in the Alps. As you can see from the chart below, snow melt became the biggest provider of water to the river on 7 April and will no doubt continue to increase seasonally in line with temperatures.
However, given the relatively poor snow season during the winter of 2019-20 (a consequence of very mild temperatures), snow levels are not high enough to sustain the river much beyond the first half of June.
For this reason, we need rain, or it will be a scorching summer on the Rhine.
5 March 2020
The Rhine river goes through annual peaks and troughs. In the chart below, historical daily values at Rheinfelden, in Switzerland, have been averaged over a monthly period to better highlight seasonal trends. In this blog post, I’ve chosen to focus on upstream river cycles, which are largely influenced by what’s going on in the Alps. Other sections of the river will be treated in future posts.
The peak in water flow occurs in the spring, when snow from the Alps melts down. In the chart, annual peaks in flow can be seen very clearly. The stream eases through the summer and autumn as snow reservoirs become progressively depleted, even if the melting of hard ice continues to feed the river for some time.
Low water is typically reached in the winter. During the season, precipitation falls as snow at high altitude and can also be stored as ice in glaciers. At lower altitude, in the Jura and the Swiss Plateau, precipitation will typically fall as rain (unless the weather is very cold) and flows directly into rivers. In the chart above, rain is the reason for the small peaks we see in between the large spring peaks.
The seasonal maximum in snow cover tends to happen right before the peak in water flow at Rheinfelden. Abundant snowfall during the winter is necessary to ensure water supplies in the spring. On that count, the 2019-20 winter season has proved disappointing with snow levels below previous years due to mild weather (See blog update from January 2020: Snow levels in the Alps are well below normal – since then, the situation has improved but not materially so).
Seasonality cycles are somewhat irregular, however. The annual peak in water flow is normally in June, but in the last few years it has occurred as early as May and as late as August, depending on A) temperatures in the spring and summer and B) how much snow and ice is stored in the Alps at the end of winter. A warm spring, such as in 2015, can lead to high flows early in the season, but is also likely to result in lower runoff in late summer and in the fall, with knock-on impacts on river navigation and hydropower production.
In some years, such as in 2017 and 2018, one can hardly notice the annual spring peak. In 2017, snow levels were extremely low at the end of winter, but this was compensated by lots of rainfall in the spring and later in the year, hence the lack of an obvious peak. In 2018, snow levels were high at the end of winter, but then weather turned hot from April onwards, ensuring steady snow melt until the end of June. The summer and fall saw record low water levels as there was no more snow left in the Alps and practically no rain.
A simple technique to correct seasonality in a time series is to use differencing. We can remove the annual seasonal component by subtracting the average river flow in any given month from the same month in the previous year. This is clearly imperfect, as the peaks and troughs in snow-related water runoff are not the same every year. But this allows us to better understand what’s going on. When comparing the result with monthly rainfall averages, it becomes clear that some of the difference between years is attributable to rain. In the chart below, surges in river flow often run in parallel with increases in rain, though it is by no means the only factor.
In the 1930s, French hydrologist Maurice Pardé worked on the seasonality of Alpine water runoff and much of what he discovered still holds today. The flow of the Rhine and its major Swiss tributaries (the Aar, the Limmat and the Reuss) can be reconstructed by monitoring the flow of numerous smaller rivers feeding from glaciers, snow peaks or fed by rain. In total, 16 water regimes exist in Switzerland and 12 apply to the Rhine. Rain-dominated runoff shows regular peaks and troughs marked by rain episodes. Flow is slightly elevated in the winter (November to March) relative to the rest of the year.
Snow-dominated runoff is marked by higher flow between April and July, the period when snow melts.
Finally, glacier water runoff starts to rise in late April and peaks in June, slightly later than snow runoff. But it is close to zero during the rest of the year.
We’ve identified a list of around 50 representative river monitoring stations dotted around Switzerland. Once a day, this blog updates their latest flow values, using live data from the BAFU website. This enables us to determine if the current upstream Rhine flow volume is mainly driven by rain or ice/snow melt, and to compare current values for each runoff regime against its respective history.
3 February 2020
Rhine water levels are likely to surpass high water flood marks by Wednesday 5 February or, at the latest, Thursday 6 February. However, for now, it looks like only the first level of high water – set at 4.60 metres in Kaub – will be breached, so the worst disruptions may be avoided.
Rainfall has been intense in the last few days, in turn helping the Rhine and its tributaries swell. In the middle section of the river, the last two days have both seen more than 1 cm of rain (an exceptional level) and the next two are likely to bring the same volume of precipitation.
On top of it, temperatures have been unusually warm, above 10 degrees Celsius on average. This means some snow has melted in the Alps, thus feeding the river further. I know this, because this blog monitors the daily share of rain, snow and glacier melt in upstream water flows.
January was unusually warm and dry, but February is proving very wet so far.
24 January 2020
Snow levels in the Swiss Alps are below normal for the time of year. It’s not just bad for the ski lovers amongst you but also potentially anyone who depends on the Rhine, as snow provides a crucial reservoir of water for the river in the spring and summer.
Snow thickness at high altitude stations (those situated above 2,000 meters) was 1.6 meters on average on Thursday 23 January, around half last year’s level and well below the five-year average. The situation is concerning because January is normally a month when snow levels build up quickly in the Alps due to a combination of precipitation and cold temperatures. However, this year, snow levels have been mostly stable compared with December amid warm weather.
And, while temperatures in Western Europe have turned noticeably colder over the last couple of days, they are due to warm up going into the end of January and the beginning of February.
It’s too early to write off this year’s snow season, as a cold and wet February could reverse the trend. However, if things continue as they are, a lack of snow is likely to weigh on Rhine water levels in the spring.
Water levels at Kaub, in Western Germany, were just above 1 metre at the time of writing, but are likely to rise significantly in early February if current forecasts of significant rainfall in Germany and Northeast France materialise. On this blog, I expect Kaub levels to surge to around 3 metres by early February, meaning water levels are likely to be plentiful for now.
4 November 2019
Water levels on the Rhine are at their highest in five years – for the time of year, that is.
October and November are typically low flow months as there is no more snow and little ice left to melt in the Alps following the summer, while rainfall in Continental Europe remains low. In 2018, following an exceptionally hot and dry summer, and an even drier fall, water levels at Kaub reached an historic low of just 30 centimetres, bringing barge traffic on the river to a standstill.
But what a difference a year makes. Kaub levels are currently at 2 metres and are likely to rise to 2.4 metres later this week. There is no concern about barge traffic. The 2019 summer was less hot than in 2018, but more importantly, it was also far more rainy, as shown by statistics I’ve gathered from the German weather service DWD, which are now updated on a daily basis and accessible on this blog.
Average precipitation at 20 weather stations situated close to the river** has been consistently higher in 2019 than in 2018 and, at more than 3 millimetres per day in October 2019, was more than three times as much as during October 2018. As a result, soil moisture along the river is now well above both 2018 levels and the five-year average: we’re now at levels typically reached in late November or early December. This, in turn, provides a reservoir of underground water for the river to continue to flow at robust levels over the coming weeks.
Meanwhile, upstream flow from the Alps has been consistently high in the last few months due to a similar pattern of plentiful rain and strong snow melt in Switzerland. The river’s flow at Rheinfelden***, on the Swiss side of the border, is due to hit 1,300 cubic metres per second on 9 November, its highest in a month and well above the five-year average flow rate.
**I have picked the 20 most relevant stations based on their location in the Rhine basin. They are situated in between Germany’s Rheinfelden and Cologne.
***There are two cities named Rheinfelden, one Swiss and one German.
9 August 2019
The situation on the Rhine is not as dire as it looked in July. You can thank rainfall, which has thrown the river a lifeline.
There have been brief but intense episodes of rain over Southern Germany and Switzerland over the last two weeks, which have fed the river. Water levels at the critical hub of Kaub, near many of Germany's industries, rose briefly to 2 metres in late July and look set to rise once again over the next few days.
Rain is forecast in Switzerland on Saturday 9 August and then again in the early part of the week starting Monday 12 August. Water flows at Domat/Ems, in the Swiss canton of Graubünden near the origin of the Rhine, are set to reach 250 cubic metres per second on 13 August, their highest since June, according to the Swiss Federal Office for the Environment (BAFU). In Rheinfelden, close to the German border, they'll reach nearly 1,300 cubic metres per second.
This has meant that the Rhine has maintained itself above 2018 levels. This time last year, Kaub levels were already well below 1 metre, but they are now heading above 2 metres.
However, I doubt these episodes of rain will have more than a temporary impact on the river. Soil remains exceptionally dry in southwest Germany, so rain struggles to penetrate deep into the ground. This, in turn, means that the river isn't really fed by underground reserves when it most needs it. This could all change of course with sustained rainfall over Germany and Switzerland, but it's unlikely before the autumn.
18 July 2019
Water levels on the Rhine are now firmly below historical averages due to hot weather and the seasonal reduction in river flows from the Alps. Levels are likely to reduce even more over the next few days and could break historical lows for the time of year.
For example, Kaub water levels are now below 2 metres - the level at which authorities begin to trigger navigation restrictions - and could fall below 1.5 metre by 21 July, according to Germany's Federal Waterways and Shipping Administration (WSV). Last year, during the drought that lasted for several months, Kaub levels fell to around 1 metre by the end of July.
There is likely to be some rain in Southern Germany on 19 July, but it will be limited and is unlikely to feed the river much. The weather is expected to remain hot (ie, at or above seasonal norms) over the next 10 days in Western Europe.
Meanwhile, river flows in Switzerland have fallen sharply since June's peak in line with seasonal averages. In Rheinfelden, close to the German border, the river currently flows at a little more than 1,000 cubic metres per second, down from 2,000 cubic metres per second a month ago but still more than in 2018.
4 July 2019
Water levels at the key German hub of Kaub, situated near many of the country's industries, will fall below 2 metres on Sunday 7 July, according to a forecast by the country's Federal Waterways and Shipping Administration (WSV), due to hot weather.
At that level, authorities begin to trigger navigation restrictions on the river and operators may need to reduce loads in order to sit higher on the water and reduce the risk of running aground.
Water levels in Duisburg, Cologne and Kaub are roughly in line with the five-year average and remain higher than at this time last year. However, the recent hot weather in Western Europe, combined with an extended drought in Germany, mean that water levels are falling quickly. No rainfall is currently forecast over the next 10 days in Western Germany, so the situation could quickly worsen.
25 June 2019
River levels on the Rhine are likely to fall sharply over the next few days in line with warm temperatures across Western Europe, which favour water evaporation. Water levels at the key measuring hub of Kaub, situated in Central Germany, will fall from 3 metres currently to 2.6 metres at the end of June, according to Germany’s Federal Waterways and Shipping Administration (WSV). At that point, they will just be 60 centimetres above the 2-metre threshold that triggers navigation restrictions for many ships using the river. This, in turn, could hamper barge movements from the Netherlands for key commodities such as diesel, jet fuel and coal.
For now, water levels in Duisburg, Cologne and Kaub remain above their respective five-year averages due to strong snow melt in the Alps as well as the cold and rainy weather registered during May and the beginning of June. However, the hot weather wave now enveloping much of Western Europe, if it continues, will quickly bring water height below seasonal averages.
The Rhine is still suffering from the exceptional drought that took place during the second half of 2018. The ground around the river is very dry, even after the winter/spring season. This means that the river cannot be fed by ground reservoirs and is thus entirely reliant on rainfall. Without rain over the coming two weeks, German authorities may have no choice but to impose navigation restrictions. The German Drought Monitor, published by Germany's Centre for Environmental Research, currently pegs the ground around the river at the "severe drought" level.
Further upstream, in the Alps, water flows on the Rhine and its major tributaries (the Aare, the Limmat and the Reuss) have fallen for 10-12 days now, indicating that snow melt - which typically occurs between May and July - may have already peaked. River flows are likely to ease over the coming 9 days.
This web app seeks to gather data on water levels from several sources, including Germany’s WSV and the Swiss Federal Office for the Environment, and to present them in a digestible format. The Rhine is Europe’s lifeblood. For centuries it has been used as one of the main arteries for shipping goods into Germany, France, Switzerland and Central Europe. However, with climate change, water levels on the river are likely to become more variable.