An unexpected pattern of distinct weekly periodicities in climatological variables in Germany



[1] Statistical analyses of data from 12 German meteorological stations meeting WMO standards in the period 1991–2005 are presented. These stations represent different local climate conditions in terms of both meteorology and pollution situation. For the average over data of all stations, we identified significant weekly periodicities in many variables such as temperature, daily temperature range, sunshine duration, cloud amount, precipitation, and precipitation frequency. Not only data of stations situated in congested urban areas, but also data of remote stations as e.g. on Mount Zugspitze 2960 m above sea level in the Alps showed significant in-phase weekly cycles. These weekly periodicities cannot be explained completely by local pollution effects or local heat emissions. We tend towards the hypothesis that the anthropogenic weekly emission cycle and the subsequent aerosol cycle interact with the atmospheric dynamics on a larger scale which leads to a forcing of a naturally existing 7-day period among the spectrum of atmospheric periods.

1. Introduction

[2] Today, great effort is undertaken to understand and ideally quantify the anthropogenic impact on the climate in past, present and future [Intergovernmental Panel on Climate Change, 2001]. In opposite to investigations of the most likely current long-term climate change, the study of weekly cycles of meteorological and chemical variables offers an unequally higher degree of statistical reliability and present-day state-of-the-art measurement quality. Since there is no known natural process that creates a periodicity of seven days that prevails for a long time, the existence of such a periodicity is a strong indication of human influence on climate.

[3] Gordon [1994] showed by analysis of satellite microwave sounding data a significant but very small weekly temperature cycle for the northern hemisphere in the years 1979–1992. Simmonds and Keay [1997] found weekly cycles in temperature and precipitation in Melbourne for the period 1960–1994 and attributed these to local heat emissions. Brönnimann and Neu [1997] detected weekend-weekday differences in the near-surface ozone concentration depending on the meteorological conditions in Switzerland. Cerveny and Balling [1998] identified weekly cycles of precipitation and tropical cyclone maximum wind speed over the North–West Atlantic region. These were explained with the help of an air pollution index that also showed a weekly cycle. The difference of day and night wind speed in tropical cyclones also showed a weekly periodicity [Cerveny and Balling, 2005]. In contrast, DeLisi and Cope [2001] did not find a significant weekly precipitation cycle along the Northeast Corridor. Beaney and Gough [2002] found weekend-weekday differences of ozone and temperature in Toronto, but not at a remote station. So they concluded that these differences were a local phenomenon. Weekend-weekday differences of ozone, but also of VOCs and NOx in 1980–1999 were detected at many stations in California by Marr and Harley [2002]. Cerveny and Coakley [2002] identified a weekly cycle of CO2 at Mauna Loa, Hawaii, but not at the South Pole. Weekly periodicities of aerosol optical properties were shown at different locations in North-America by Delene and Ogren [2002] and by Jin et al. [2005]. Beirle et al. [2003] analyzed satellite data and found significant weekly cycles of tropospheric NO2 over many industrialized regions including Germany. Forster and Solomon [2003] detected a weekend effect in the daily temperature range in different regions. They found large scale patterns of areas with a positive or a negative weekend effect especially over the USA, and it is challenging to explain these patterns. Differences of the visibility and the PM10 concentration between weekdays and weekends in Taiwan were shown by Tsai [2005]. Shutters and Balling [2006] reported weekly cycles of various chemical variables, but also of wind speed in Phoenix, Arizona. Recently, Gong et al. [2006] found increasing weekly cycles in various meteorological parameters including temperature and precipitation in China.

[4] In this paper, we show weekly cycles in different climatological variables over Germany for the first time, and we interpret their horizontal and vertical extent together with the previous work in a new way.

2. Data

[5] For this study, we examined 15-year data (1991–2005) from 12 stations of the German Weather Service (DWD), namely Aachen (WMO-No. 10501, 202 m above sea level), Berlin-Tempelhof (10384, 50 m), Düsseldorf (10400, 45 m), Helgoland (10015, 4 m), Hohenpeissenberg (10962, 977 m), Kahler Asten (10427, 839 m), Karlsruhe (10727, 112 m), Konstanz (10929, 443 m), Rostock-Warnemünde (10170, 4 m), Stuttgart-Echterdingen (10738, 396 m), and Zugspitze (10961, 2960 m). The domain in which these stations are situated reaches from 47°25′N to 54°11′N and from 6°05′E to 13°24′E. The number of daily averages is 5479 for each station except for very few missing values, which leads to N = 65748 for the complete dataset. We removed the yearly cycle from all single time series by subtracting a 31-day running mean.

3. Results

[6] The mean temperature anomaly from the mean temperature averaged over all stations is shown in Figure 1a by day of the week, revealing a distinct weekly cycle. Temperature anomaly increases in the first half of the week with a maximum on Wednesday and a minimum on Saturday, the difference between both being 0.19 K. We checked the significance of this finding by applying a one-tailed t-test to all Wednesdays and Saturdays data and obtained α = 0.0001 for this normally distributed data set. We also carried out spectral analyses of the time series and found a peak at a period of exactly 7 days.

Figure 1.

(a) Mean temperature anomaly and (b) mean daily temperature range anomaly averaged over 12 WMO stations in Germany 1991–2005 by day of the week; error bars show standard errors of the mean.

[7] The mean daily temperature range anomaly (Figure 1b) averaged over all analyzed stations in Germany (1991–2005) also shows a clear dependency on the weekday. This parameter is calculated as the difference between maximum and minimum temperature on each day and is regarded as an important indicator of human influence on climate [Forster and Solomon, 2003], as well in respect of long-term climate trends. During the first half of the week, the daily temperature range is significantly greater than in the second half of the week. The one-tailed t-test applied to Monday and Saturday led to α = 0.0152. This means that maximum and minimum temperature do not behave exactly in the same way, although both have a weekly cycle as well (not shown here). This can be interpreted as a consequence of the increase in pollutant concentrations such as aerosol during the week.

[8] Not only for the temperature, but also for the daily sunshine duration and the cloud amount a weekly periodicity could be proven (Figure 2). The sunshine duration is at its maximum at the beginning of the week, and it decreases steadily until Saturday. The difference between Tuesday and Saturday exceeds a quarter of an hour, being 5.6% of the average value of 4.68 hours, and it is significant with α < 0.0001. Correspondingly, cloud amount is on average greater in the second half of the week than in the first one, showing a minimum on Tuesday and a maximum on Saturday (α = 0.0058), which is exactly the opposite behaviour as in the case of sunshine duration. The difference is 0.1 eights, which is approx. 2% of the average of 5.41 eights. This lower range in comparison to the 5% in the case of the sunshine duration is explainable by geometrical considerations, since the shadowed part of the earth's surface increases when the sun elevation decreases at the same cloud amount.

Figure 2.

(a) Mean daily sunshine duration anomaly in hours and (b) mean cloud amount anomaly in eights averaged over 12 WMO stations in Germany 1991–2005 by day of the week; error bars show standard errors of the mean.

[9] Precipitation is as well subject to a weekly cycle in Germany, which is valid both for the amount of precipitation and its frequency. Figure 3a shows the anomaly of the accumulated yearly precipitation on each day of the week from the weekly average (131.6 mm per weekday). The minimum occurs on Monday with a deficit of 9.3 mm (7.1%), and the maximum on Saturday with an excess of 10.3 mm (7.8%). The t-test based on yearly averages indicates significance (α = 0.0180). This is in accordance with lower cloud amount and greater sunshine duration at the beginning of the week. Similarly, the precipitation frequency increases significantly during the week. Figure 3b illustrates the relative anomaly of the number of days with precipitation greater than 0.0, 1.0, 5.0 and 10.0 mm by day of the week, reflecting the increase of precipitation in the course of the week. The corresponding values of α resulting from a t-test are 0.0150, 0.0289, 0.0405, and 0.0319. In addition to temperature (average, maximum, and minimum), sunshine duration, cloud amount and precipitation (amount and frequency), we also found weekly periodicities in relative humidity, air pressure, average wind speed and daily maximum wind speed (not shown).

Figure 3.

(a) Anomaly of the accumulated yearly precipitation on each day of the week from the weekly average and (b) relative anomaly of the number of days with precipitation greater than 0, 1.0, 5.0 and 10.0 mm averaged over 12 WMO stations in Germany 1991–2005 by day of the week; error bars in Figure 3a show standard errors of the mean.

[10] These distinct weekly periodicities are not exclusively caused by stations in congested urban areas where one would expect most likely the occurrence of a weekly cycle in meteorological variables. In fact, the mean temperature anomalies (1991–2005) of all 12 tested stations in Germany show this effect that is significant in all cases (α < 0.05). All stations reach the maximum around Wednesday and the minimum around Saturday. The weekly cycle is smallest on Helgoland (0.06 K), which is an island in the North Sea, but all other stations show a signal between 0.13 K and 0.37 K (Figure 4). There was no big difference between mountainous stations, as Mount Zugspitze, 2960 m above sea level in the Alps, or Mount Hohenpeissenberg, 977 m above sea level, and stations at or near big cities and airports, as Berlin or Frankfurt. The mountainous stations even had ranges in the mean weekly temperature cycle that were among the greatest of all stations, although especially the station on Mount Zugspitze is situated above the polluted planetary boundary layer most of the time.

Figure 4.

Mean temperature anomalies of 12 WMO stations in Germany 1991–2005 by day of the week.

[11] This means that the temperature anomaly is rather independent from the local pollution situation or local heat emissions, and also from the altitude. Instead, there is a horizontal gradient in the intensity of the weekly cycle, showing lower values in the North–West, and greater values in the South–East of Germany. Thus, the horizontal scale of this effect is not local but it is rather at least a mesoscale-α [Randerson, 1976] phenomenon. Its vertical scale exceeds the boundary layer height, and its intensity is not determined by the local population density that can be regarded as a rough measure for anthropogenic emissions.

4. Conclusions and Discussion

[12] We conclude that for the first time the weekly periodicity is significantly proven in many climatological variables in Germany in the years 1991–2005. This is the 15-year period after the German reunification in 1990 that caused significant changes in anthropogenic emissions. The combination of the large amplitude of this effect, its significant occurrence in many variables and its considerable horizontal and vertical extent makes this study unique. Because of the logical connections of e.g. sunshine duration and cloud amount, or precipitation and precipitation frequency over a climatological period of 15 years, it is very unlikely that this effect is purely random. The only 7-day period that is known is the cycle of human activity in western civilizations which offers the possibility to study an anthropogenic influence on climate on short time-scales. Since the effect described here is non-local, local heat emissions cannot be the dominant process. In regard of the horizontal and vertical structure of the effect, we conclude that the direct aerosol effect on solar radiation cannot explain its pattern in a satisfying way. Since also cloud amount and precipitation are modified in the course of a week, we suppose that the indirect aerosol effects on cloud properties and precipitation [Lohmann and Feichter, 2005] play an important role. The prevalent ideas about coherences between an increase in aerosol particle number, an increase in cloud droplets number but a decrease of their radii, and a following decrease of precipitation but longer cloud lifetimes, is not reflected by our results. In contrast, we detected an increasing cloud amount and a decreasing sunshine duration occurring simultaneously with an increase in precipitation in the course of the week on average. Therefore, assuming that mainly anthropogenic aerosols are the probable reason for the effect, in our opinion this study demonstrates that the complete interaction of aerosol and climate can only be understood or modelled comprehensively if the interaction with atmospheric dynamics is taken into account. This interaction takes places on different scales, as we interpret this weekly cycle all over Germany that does not show the structures of typical emission patterns anymore.