Statistics of drought in eastern Thuringia: How does 2022 fare?

Was the recent July 2022 exceptionally dry? Here’s a graph of the monthly cumulative rainfall recorded at the weather station in Jena (“Sternwarte”). The data are available from the CDC archive of the DWD. Measurements are available since 1827; there is a larger gap in 1870-1877, but else the record is remarkably complete.

July precipitation at Jena weather station (“Sternwarte”)

Obviously, July 2022 (red dot) with 33.9mm was dry but not exceptionally. There were quite a few drier July’s, e.g. in 1827, 1869, 1911, 1943, 1964 and 1969. Indeed, July 2022 ranks as 26th driest July within the entire 187 years with complete monthly measurements. There is also no long-term trend visible.

However, if we take the cumulative precipitation over the entire vegetation period from March 1 to July 31, we get a different picture:

Cumulative March-July precipitation at Jena weather station (“Sternwarte”)

There is no long-term trend visible, but 2022 appears now as an outlier. How exceptional is 2022?

The distribution of the annual cumulative March-July precipitation from all the years of observation can be described with a Gamma distribution:

Histogram of annual cumulative March-July precipitation (PDF) and fitted Gamma distribution.

The minimal value of 127.6 mm recorded in 2022 is quite exceptional: if the fitted distribution is correct, a lower value than 127.6 mm is expected with a probability of less than 0.0007. This would imply that such a dry March-July period would occur only once or less within 1000 years. However this depends very much on the assumed distribution and in particular its tails, which are not constrained given that the record has only 187 years of observations.

Precipitation falls locally very heterogeneous, hence one has to ask to which extent the measurements at the Jena weather station are representative for the region. Since amount and distribution can be quite different, “drought” or “wet” need to be quantified with respect to the local climatology. This can be done using a dryness indicator. One of the various indicators recommended by the WMO is the Standardised Precipitation Index (SPI, for definition see below). The WMO considers SPI values below -1. as “moderately dry”, below -1.5 as “Severly dry” and below -2 as “extremely dry”.

Computing the SPI for the 19 stations within the DWD archive which lie less than 100km around Jena and have measurements at least since 1950 one obtains the following aggregated result (computed from the annual cumulative precipitation from March to July):

SPI aggregated from the 19 weather stations within 100km around Jena providing observations since at least 1950. Central dot shows median, vertical line the range between first and third quartile and the gray dots the minimum, respectively maximum of the SPI of the 19 sites for a particular year.

Thus, also on the regional scale 2022 appears as an “severely dry” year, although there were also severe droughts in 1976 and in 1964. A long-term trend to drier conditions, as often reported in the media, is not visible also on the regional scale.

Given the current dry August, the drought will still worsen.

————–

Standardised Precipitation Index (SPI) is computed by fitting first a statistical distribution to the data, which is then transformed to a normal distribution so that the mean is 0. It can be applied to weekly, monthly or longer cumulative precipitation data. Here it is computed from the annual March to July cumulative precipitation assuming a Gamma distribution.

Tonga volcano shockwave signature in air pressure records in Germany

The explosive eruption of the Tonga volcano on Jan 15, 2022, 04:15 (UTC) created a shockwave which travelled around the world. E.g. it is seen nicely in the air pressure record from Unkeroda (50.934N, 10.269E) in Thuringia. Below is a graph of the record:

Two events are clearly seen (the record has a time resolution of 5′ only): a first at Jan 15, 2022 20:20 CET, and a second at Jan 16, 2022 02:00 CET.

The circular shockwave emanating from the eruption is nicely seen in remote sensing data of midlevel water vapour (GOES-17). After expanding over the hemisphere it converged on the other side of the planet arriving in Germany after 20:00 CET. The speed of the wave can be calculated from the arrival in Unkeroda:

The distance from the Tonga volcano (20.56S, 175.37W) to Unkeroda is 16’598 km (calculated on the reference ellipsoid); the time difference from the eruption until the first event is 14hours 45Minutes, which gives a speed of 312.5 m/s. This is close to a typical speed of sound in the upper troposphere.

Theoretically, assuming a perfectly spherical Earth, the sound wave would have converged at the antipode point of the Tonga volcano (i.e. at 20.56N, 4.53E). This point lies about 3406 km south of Unkeroda. Using the speed calculated above, it should have arrived at the antipode point at about Jan 15, 2022, 23:01 CET. Subsequently, the shock wave should have dispersed again in circular form from the antipode point, reaching Unkeroda again three hours later – this is exactly the time of the second event seen above (i.e. at 02:00 CET on Jan 16). The map below shows the path of the wave crossing Unkeroda (black) and then on to the antipode point and back (pink). The red circle shows the wave at the time of the first and the second event in Unkeroda.

An interesting question is, what happened at the antipode point. Theoretically, assuming a perfect sphere, no attenuation in the atmosphere and no vertical broadening of the wave, the original sound of the eruption could have been heard at the location. Since this not the case, clearly the signal will have been quite distorted. Still, a sizeable pressure signal should have been picked up by local measurements. Unfortunately, the antipode point lies close to the southern border in Algeria, about 240km SSW from the city of Tamanrasset. The METAR data from Tamanrasset airport are unfortunately only available at hourly resolution and thus do not show a convincing signal.

In case you want to ‘hear’ the sound of the eruption as recorded in Estonia: Steffen Noe has resampled the high resolution pressure record from his SMEAR station and turned it into a sound file…

Addition (Jan 24, 2022):

Diego Aliaga from University of Helsinki, Finland, has made an analysis of the shockwave as seen in IR radiation from two geostationary satellites (GEOS16, GEOS17). In his Twitter tweet you can see an animation of the IR brightness anomaly. The circular wave moving from Tonga to the antipode and back, several times, is clearly visible.

Addition (Jan 25, 2022):

Yet another animation of satellite data: this time from EUMETSAT data, made by Mathew Barlow of University of Massachusetts Lowell. It shows nicely the shock wave around the antipode point. As anticipated, it is not converging perfectly, but already somewhat distorted.

Addition (Jan 26, 2022):

More articles on the effects of the Tonga shockwave are appearing: Science has a nice piece on small tsunamis seen at the coast of Japan and also in the Caribbean. In Japan these small tsunamis were seen earlier than the actual ocean surface wave tsunami from Tonga; they were locally generated by the pressure shockwave which travels much faster. EOS contains an interesting commentary with more animations, model simulations and observations of the atmospheric shock wave.

Exceptional recent atmospheric CO2 growth rate?

The New York Times of June 26, 2017 contains an article entitled “Carbon in Atmosphere Is Rising, Even as Emissions Stabilize”, discussing the recent atmospheric CO2 growth rate. It contains statements such as  “The excess carbon dioxide scorching the planet rose at the highest rate on record in 2015 and 2016. A slightly slower but still unusual rate of increase has continued into 2017.”

Is this really that exceptional? Let’s look at the facts:

This bar chart from NOAA’s Earth System Research Laboratory (NOAA-ESRL) displays the annual increase in atmospheric CO2 recorded at the Mauna Loa station on Hawaii. The growth rate in the last two years, 2015 and 2016 both exceed any increase observed in previous years.  How does this fit to the recent finding by the Global Carbon Project (GCP) that the global anthropogenic CO2 emissions have almost levelled over the last few years?

NOAA-ESRL computes the annual CO2 growth rate by averaging the 4 monthly mean observations from November to February and then subtracting the average of the mean observations from the same months one year before. These are the values plotted in the bar chart shown above. If one uses the same recipe but calculates the increase not only over the calendar year, but for 12-month intervals lagged by 1 month, one obtains the following time series:

This graph includes also the CO2 growth rate observed at the South Pole station. (The monthly atmospheric observations were obtained from the Scripps CO2 Program). The two records are remarkably similar, indicating that measurements from either station reflect to a good approximation the change in the global inventory of atmospheric CO2. At both stations the CO2 growth rate peaked at the end of 2015, exceeding any previously recorded 12-month increase. This recent peak coincides with the El Nino from 2015/2016. Since the peak the growth rate shows now a decreasing trend similar to previous El Nino periods (e.g. in 1997/1998, 1988/1989, 1973/1974).

How would we expect the atmospheric CO2 growth rate to evolve given the reported anthropogenic CO2 emissions? The graph below shows in the upper panel the observed CO2 growth rate from Mauna Loa (green curve) together with predictions by two simple global carbon cycle models with prescribed anthropogenic emissions as shown in the lower panel (Technical details about the models and the data can be found at the bottom of this page). The grey background shading in the upper panel indicates El Nino conditions (characterised by the MEI index).

These simple models do not include any climate variability; they simply describe how CO2 emissions are redistributed among atmosphere, ocean and land biosphere, based on our basic understanding of carbon uptake and turnover in these three reservoirs.

Obviously, the year to year variability of the observed CO2 growth rate can not be captured by these simple models but is strongly correlated to the El Nino – Southern Oscillation, discovered already long ago (Bacastow, 1976, Nature). During El Nino phases, south-eastern Asia and also parts of the Amazon region are drier than normal, causing reduced photosynthesis but also fostering vegetation fires. Both processes lead to a temporary negative carbon balance of the terrestrial biosphere, which is mirrored as an anomalous increase in the atmospheric CO2 growth rate.

On the other hand, the simple global carbon cycle models reproduce quite faithfully the longer term trend including decadal variations of the atmospheric CO2 growth rate. This can be taken as an indication, that, at least under present conditions, the first order dynamics of the global carbon cycle are well represented in these models.

Why is atmospheric CO2 not yet reflecting the reported “stabilisation” of the anthropogenic emissions during the recent years?

The emissions from fossil fuel burning have indeed flattened over the last ~3 years (reported emissions are only available up to 2015; the grey point for 2016 in the emission graph above is an extrapolation to the year 2016 made by the GCP).  On the other hand, emissions from changes in land use (a.o. deforestation) have picked up over the last few years so that the total anthropogenic emissions (brown curve) are still increasing with a similar trend as over the last 15 years. One has to realise, however, that the land use emissions for the years 2011-2015 are also only an estimate, made by the GCP based on scaled biomass burning emissions derived from satellite measurements and are thus quite uncertain.

A second effect to consider is the dynamics of the global carbon cycle. Keeping constant total emissions beyond 2015 (grey points), the simple carbon cycle models predict only a very slow reduction of the atmospheric CO2 growth rate (see the model curves in the upper panel). Given the large inter annual variations, it would take many years to detect such a small declining trend in the atmospheric CO2 growth rate. Of course, if the anthropogenic emissions were to be reduced from their present values, as required to meet e.g. the Paris agreement to limit climate change in this century, then a declining trend in the atmospheric CO2 growth rate should become larger and eventually easier to detect.

Thus, given the available information, the global carbon cycle behaves as expected: the recent El Nino left it’s imprint on the atmospheric CO2 growth rate, but it was not exceptional compared to previous El Nino periods. Furthermore, the longer-term increase of the growth rate is consistent with the reported anthropogenic emissions.  But this does not mean that we do not have to worry: We are discussing here changes in the atmospheric CO2 growth rate, not the atmospheric CO2 concentration! As long as the growth rate is positive, atmospheric CO2 continues to increase. Stabilising emissions does not stabilise atmospheric CO2! – a fact that is long known in the carbon cycle science community, but often forgotten in the public debate.

---

Technical details: The two simple global carbon cycle models used to calculate the expected atmospheric CO2 growth rate are the “box-diffusion model” of Oeschger et al. (1975, Tellus), and the “Bern-SAR model” as used by IPCC in the second assessment report (1995) for the calculation of the global warming potentials of various climate forcing agents.  The Bern-SAR model is linear in the impulse response variant used here (see Joos et al., 2013, ACP), the box-diffusion model includes the non-linear ocean carbonate chemistry. Neither of these models include any climate feedback effects and are thus of limited use beyond the present conditions. Fossil fuel and land use CO2 emissions are obtained from the GCP (Le Quéré et al., 2016, ESSD).

 

Summer prediction challenge 2016 – results

How was the summer 2016? Here are the results.

Note: This post update contains additional information regarding the temperature offsets between Jena city (Schillergässchen) and the institute weather station. Temperature maxima are 1.66°C higher in the city than at the institute; likewise temperature minima are 1.06°C higher in the city than at the institute. The statistics shown in the post of June 19, 2016 were calculated with the “city” temperatures (in order to compare them with the climatology or the long-term time series). However, the metrics as given clearly refer to the “institute weather station”… The info given below corrects for this difference where appropriate.

Metric 1: Maximum temperature recorded and the number of “tropical nights”

Temperatures recorded every 10′ on the institute roof during the summer of 2016 are shown in Figure 1. The maximum was recorded at 34.35°C on June 24, 2016 at 15:20 (red dot).

Figure 1. Temperatures recorded at the MPI BGC weather station in summer of 2016.

There was no tropical night in 2016. The updated graph of tropical nights in recent years is shown in Figure 2. Note that this is calculated with the institute temperature values; if they were corrected to the city values, there would have been two tropical nights: on June 23/24, and on August 27/28.

Figure 2: Number of tropical nights in summer in Jena.

The maximum temperature recorded on the institute roof on June 24, 2016 corresponds to a temperature in the city of 36.01°C. This is lower than last year, when an all-time record of 38.94° was recorded. It is close to the expected trend of the summer maxima (0.023°/yr) over the last 150 years.

Figure 3: Maximum summer temperatures recorded in Jena since 1820. The value for 2016 is indicated with the red circle. Data prior to 2003 are from the old weather station in the Schillergässchen. Data after 2003 are based on the institute roof weather station corrected for the mean offset between the two stations.

A view of the daily summer temperatures in Jena is provided in Figure 4. The summer was characterised by two heat waves: one in late June and one in late August. In both, maxima exceeded more than 2 sigmas above the climate normal (1961-1990).

Figure 4. Daily temperatures recorded in 2016 in Jena (city temperatures). Solid lines: Minima (blue), mean (green), maxima (red).  The horizontal straight lines and the shaded areas show the statistical expectation (mean and 1-sigma) for the three summer months based on the reference period 1961-1990.

Compared to the climate normal (1961-1990) the summer of 2016 was substantially warmer. The distribution of the daily temperatures (mean, min, max) are shown in  Figure 5 and compared to the climate normal. The mean was 2.36° warmer.

Figure 5: Histograms of daily temperatures in summer (June-August), for the climatology (1961-1990, green) and as observed in 2016 (red).

The entire year of 2016 was very dry compared to the climate normal (Figure 6). Monthly precipitation sums reached the median of the climate normal only in February, May and July. In all the other months, the precipitation sum was lying in the lowermost quartile of the climate normal.  Notice however, that the climate normal was computed from the observations at the Schillergässchen station, while the data for 2016 were from the MPI BGC weather station. It is possible that there is a systematic bias between the two sites.

Figure 6: Box-whisker plot (blue) of the monthly precipitation sums computed from the climate normal (1961-1990). Red circles denote monthly precipitation sums recorded at the MPI BGC weather station.

Metric 2: Minimum of the 15-day median of the daily lowermost 5% quantile of the recorded CO2 concentration

The 15-day median of the recorded daily lowermost 5% quantile of the CO2 concentration at the institute weather station follows pretty much the minimum monthly temperature recorded at Mauna Loa (Figure 7, 8).

Figure 7: Daily lowermost 5% quantile of the CO2 concentration recorded at the institute weather station (blue dots). Black line: 15-day median. Orange dots: monthly CO2 concentration measurements at the Mauna Loa observatory. The vertical black lines delineate the 3 summer months in each year.

Figure 8: Minimum of the 15-day median of the daily lowermost 5% quantile of the CO2 concentration on the institute roof (black circles). Orange squares: monthly CO2 concentration measured at the Mauna Loa station.

The  minimum 15-day median in 2016 was 389.4 ppm.

Summer Prediction Challenge 2016

How will the summer 2016 be in Jena? What is your forecast? Submit your forecast and you can win a bottle of good wine!  Below are the two metrics described for which you can submit your forecasts. Both are related to the meteorological “summer”, i.e. are based on the data observed at the weather station of our institute roof from the time period June 1, 2016 until August 31, 2016. Send your entry by email to me (martin.heimann@bgc-jena.mpg.de). It should contain your forecast for one or the other metric or for both and a few words/lines explaining how you got to your forecast (statistical model, grandma’s guess, reading of coffee grounds, whatever).

The deadline for entries is: June 30, 2016, 23:59:59 GMT

Note that the deadline is already one month in summer, hence you can potentially profit from the observations that are made during June…

Metric 1: Maximum temperature recorded and the number of “tropical nights”

This metric consists of two numbers: (1) the maximum 10′ temperature recorded on the institute weather station and (2) the number of tropical nights during the summer of 2016. Tropical nights are defined as nights with a minimum temperature larger than 20°C. Statistics for both metrics are given below.

Maximum temperature recorded on the institute weather station

The following figure shows the summer maximum temperature recorded in Jena since 1833. The data before 2004 are from the weather station in the Schillergässchen, the more recent data are from the institute weather station with a small adjustment explained here.

There are some caveats regarding this graphic. Daily maximum temperatures were recorded from analogue Tmin/Tmax thermometers until 2003. Since 2004 the maximum temperature is determined from the 10′ digital recordings of our institute weather station. The red linear trend line has a slope of 2.3 °C/century, which is above the long-term mean summer temperature trend in Jena (1.2 °C/century).   If we take the distribution of the individual years around the trend curve and project it with the trend to the year 2006, we obtain:

Thus you can choose your guess based on this statistic…

The annual number of tropical nights in summer since the year 2000 is shown below:

Prior to 2000 there is practically no year with a tropical night. It is likely that this partially reflects a measurement bias; I think that the continuous measurements on our institute weather station are more accurate to pick the daily minimum temperature than the analogue minimum thermometers used at the Schillergässchen weather station. The different local conditions might also introduce some bias.

Metric 2: Minimum of the 15-day median of the daily lowermost 5% quantile of the recorded CO2 concentration

This metric tries to capture the seasonal trough of the CO2 concentration. The figure below shows the lowermost 5% quantile measurements of each day (dots) and the running 15-day median (black curve). The vertical lines in each year bracket the meteorological summer. Also shown for comparison is the monthly measurement from Mauna Loa (orange dots).

A table with the values of the annual summer minima of the black curve and the annual monthly minimum recorded at Mauna Loa is shown here:

and in graphical form:

The required metric more or less follows the Mauna Loa record. The value for 2016 is based on the observations until June 15, 2016.

Winter prediction challenge 2015/16

How will the winter 2015/2016 be? Can you predict it accurately? For those of you who did not attended my brief introduction talk in the colloquium on October 8, 2015, here are the talk slides:

Winter Challenge Slides

If you want to participate, here are the rules:

Metric 1: Coldest 2-week (14-days) mean daily minimum temperature
in meteorological winter: December 2015 – February 2016

A single cold night does not make a “cold winter” – hence a frost period  of 2 weeks. A graph of the last 65 years of this metric is given below:

The statistics of this metric is shown in the histograms below, both for the last 30 years and also for the entire record of the Jena weather station since 1833 (minimum temperatures were recorded only from 1833).

Statistics:

If you want to play with the data for improving your statistical forecast, you can download the daily mean, max and min temperatures from the Jena weather station here. The data displayed above in the graph of T_dmin14 are available as a text file (cvs) here.

Daily updates of how Metric 1 develops during this winter are displayed here.

Interestingly: the entries by challenge participants exhibit two clusters: a “warm” cluster around -1.5°C and a “cold” cluster around -5°C. Similar to the phases of the NAO…

Metric 2: Maximum 10-minute CO2 concentration measured on institute roof and minimum of lowermost 5% quantile of daily 10′ CO2 measurements during the meteorological winter: December 2015 – February 2016.

Since we are a biogeochemical institute, Christian Rödenbeck suggested also a prediction of a biogeochemical quantity. This metric contains two numbers. A graph of these numbers for the winters since measurements began is shown below, as well as a table of the numbers. The full suite of measurements is displayed and updated daily on Institute Roof CO2.

A typical winter day in CO2 mixing ratio (concentration) looks like this:The maximum is reached in the morning under stable meteorological conditions, reflecting the build-up of local respiration and fossil CO2 sources. The minimum in the afternoon is caused by enhanced vertical mixing and presumably is close to mixed planetary boundary layer concentrations for this part of Europe.

Daily maxima in winter time over the last seven years are shown below. The large symbols indicate the respective winter maximum, which is to be forecast for the winter 2015/2016. The red dot show the maximum recorded so far in December 2015.

(Addendum: found some spikes in the time series which are clearly artefacts. They were removed by applying a 30′-median running average. Indeed, the maximum in 2009 shown below is an artefact. The all-time maximum recorded in 2010 is however a real CO2 peak.)

Daily lowermost 5% quantile values (C_0.05) of the 10′ measurements are displayed below, together with monthly values from Mauna Loa.

The lowermost C_0.05 values closely follow the MLO time series with an offset of about 2-4 ppm. The histogram of the difference to the MLO curve shows a sharp edge on the left hand side:

Thus forecasting C_0.05 for the winter of 2015/2016 could be based forecasting MLO and the offset…

 

 

 

Daily updates of how Metrics 2 develop during this winter are displayed here.

 

Entries:

Your entry must contain:

1. Names of the team members
2. Estimation method (1 to a few paragraphs explaining how you came to your prediction). Everything is permitted: Secret phone call to your grandfather, website, fancy statistical estimate, GCM simulation, horoscope etc. But document it!
3. Challenge entry: Your prediction either for Metric 1 or Metric 2 or for both.

Send your entry to me by email for now. I will then make a life website showing the entries and the actual development as the winter proceeds. A link to the life website will be added here.

Deadline:  December 15, 2015, 23:59 CET

A bit longer than originally planned. Gives you an opportunity to watch the first two weeks of the winter before sending in your forecast…

Prize: a bottle of good wine for the winners of each metric challenge

Feel free to comment on this challenge – use the comment link below.

 

A little blog for science discussions

This is still a beta test – ultimately I hope this blog will be hosted by our own institute in order to get all ads to disappear…

Use this blog for science discussions. For a new topic, please contact me directly and send me the material and I’ll post it (at the moment only I can do this). For discussion items anyone can comment and contribute.