Portal de Periódicos da CAPES

Reply to comment by H. J. Hendricks Franssen on “An unexpected pattern of distinct weekly periodicities in climatological variables in Germany”

2008; American Geophysical Union; Volume: 35; Issue: 5 Linguagem: Inglês

10.1029/2007gl032432

1944-8007

D. Bäumer, Bernhard Vogel,

Climate variability and models

[1] The comment by Hendricks Franssen [2008] (hereinafter referred to as HF) presents a data analysis for two Swiss stations, Zurich and Lugano, with respect to weekly cycles in precipitation and sunshine duration. The main conclusion in the comment is that the weekly cycles observed in Germany were probably random. Neither this data analysis nor the Monte Carlo experiment applied to Zurich data allows any straight conclusions in regard to the data presented for Germany. The assertion of all the German time series of precipitation and sunshine duration being highly temporally correlated is not correct. The limitation on the period 1991–2005 is intentional because of the near-constant and relatively well-known aerosol conditions. Those conditions were not prevailing in former times. For example, a decrease in aerosol mass density but an increase in aerosol number density is reported in the literature. [2] We thank H. J. Hendricks Franssen for his comment on our paper about weekly periodicities in climatological parameters in Germany [Bäumer and Vogel, 2007] (hereinafter referred to as BV). In this comment, measurements of precipitation (1864–2005) and sunshine duration (1901–2005) of the Swiss stations Zurich and Lugano are analyzed with regard to the dependence from the day of the week. [3] The results for precipitation and sunshine duration in Zurich, that is situated north of the Alps as Germany, corroborate our findings for the observation period 1991–2005. The precipitation maximum on Saturday in this period is stated to be significant by HF. The mean deviations on different days of the week have continuously increased in the past four decades and have never been as large as in 1991–2005 since 1864 according to Table 1 of HF. In BV, we solely analysed the period 1991–2005 because of the high quality of measurements and the non-constant composition of the ambient air (including the aerosol load) on longer time scales. In addition, a significant weekly cycle in sunshine duration occurred in Lugano in 1991–2005 according to HF. Although HF presented these significant weekly patterns, he asserted therein that the significant deviations reported for Germany in 1991–2005 by BV are only an artefact that is caused by merging multiple time series with a high spatial auto-correlation. Apart from this is a contradiction in terms in our opinion, the spatial correlation of precipitation data among different stations on a length scale of the order of Germany is not very high, especially not in summer and not in mountainous terrain. The applied Monte Carlo experiment on Zurich precipitation data is not sufficient to assess the German conditions, and its interpretation is also slightly imprecise with respect to Zurich. Furthermore, HF ignore the manifold possible interactions of aerosols and precipitation [e.g., Rosenfeld, 2007]. [4] We respond on the comment in detail as stated below, and these reasons explain why we do not follow the main conclusions drawn by HF. [5] 1. In Zurich, the same weekly pattern of precipitation as in Germany (positive anomalies from Friday to Sunday with a significant maximum on Saturday, but negative anomalies otherwise, BV) was found for the period 1991–2005 according to HF. [6] 2. In Zurich as in Germany, a weekly cycle of the daily sunshine duration with a maximum in the beginning of the working week and a subsequent decrease towards the end of the working week has been detected in the period 1991–2005 (BV; HF). Both in Zurich and in Germany, its relative amplitude is lower than in the case of precipitation, and it is not significant in Zurich according to HF. [7] 3. In HF, the weekday precipitation anomalies of 1991–2005 for Zurich are also compared to anomalies since 1864. It is shown that the largest range and the largest average deviation occurred in 1991–2005, which indicates that the largest precipitation variability among different weekdays since 1864 occurred in 1991–2005. Especially after 1968, the intra-weekly variability increased considerably. Some of the 15-year periods before 1984 show different cycles or probably random patterns. Assuming that our hypothesis of an aerosol effect is valid, we would not expect that there is a persistent weekly periodicity in the precipitation over centuries. In fact, we expect that the anthropogenically induced weekly periodicity has become visible in the past few decades. From aerosol number particle measurements in the diameter range between 4 nm and 3 μm on Mount Hohenpeissenberg [Global Atmosphere Watch, 2004], for instance, it can be seen that there was an increase in particle number from 2500 to 4000 cm−3 in the period 1995–2004. Additionally, the situation in Eastern Germany is unique, where the German Reunification in 1990 led to a rapid change in emission and subsequent aerosol concentration patterns, e.g., as shown by Spindler et al. [2004]. Kreyling et al. [2003] found a strong increase in particle number, but a decrease in particle mass in urban regions in Eastern Germany between 1991 and 2001. [8] 4. As well HF compare the deviations of the sunshine duration by day of the week for Zurich in 1991–2005 to a previous series of 15-year periods that date back until 1901. The period 1976–1990 has a rather similar weekly pattern as the period 1991–2005 with a minimum on Saturday and positive anomalies on Sunday and Monday. Previous data show some greater deviations on single days, e.g. the Wednesday in 1961–1975, or the Friday in 1916–1930. In addition to what we stated above with regard to the historical precipitation data, the quality of the sunshine duration measurement was not constantly high throughout the 20th century. Older sunshine data was measured by the Campbell-Stokes sunshine recorder with relatively high errors. It detects sunshine if the beam of solar energy concentrated by a special lens is able to burn a dark paper card. Later on, the World Meteorological Organization defined the sunshine duration in a period as the sum of that sub-period for which the direct solar irradiance exceeds 120 W m−2 [World Meteorological Organization, 2003]. In addition to what is stated on that by HF, we conclude that measurements that are based on this exact definition are much more appropriate to detect the small differences that can be caused by the aerosol effects. Furthermore, also operational precipitation and temperature measurements, the guidelines for locating stations, and the locations of stations and their environment had been subject to changes at some stations which led to a continuous improvement of the data quality in many cases, but not necessarily of their homogeneity. Together with the non-constant aerosol properties, these are the reasons why we confined ourselves to the period 1991–2005 (BV) when we could rely on high quality measurement technique, e.g. for sunshine duration. [9] 5. Commenting on the results for Lugano presented by HF, we therefore also concentrate on the period 1991–2005. The precipitation analysis for Lugano did not show any weekly periodicity, but for the sunshine duration a weekly cycle with even a significant maximum on Saturday and Sunday and a significant minimum on Thursday has been detected by HF. That means that there is a phase shift in the maximum of two days in comparison to Zurich or Germany (BV). We can not tell whether this is random or not, but it is remarkable that in this period 1991–2005 such a significant signature appeared in Lugano. Besides, as mentioned by HF, the meteorological situation south in comparison to north of the Alps is usually rather different, since one of the stations, either Zurich or Lugano is in the Lee of the Alps when the other one is situated upstream. Therefore, the different conditions in Lugano in comparison to Zurich might be caused by a systematic anti-correlation of the data. Furthermore, we would not necessarily expect that the same signal is existent at such an extremely located station as Lugano that, for instance, frequently shows heavy precipitation events of more than 50 mm per day. [10] 6. In HF, the argument of spatial correlations of temperature, sunshine duration, and precipitation is mentioned to explain the significance that was followed by BV for the German data, although a significant precipitation maximum on Saturday for Zurich in 1991–2005 is presented in the comment (together with a significant weekly cycle in sunshine duration for Lugano and a relatively weak one for Zurich). However, this obviously demonstrates that significant anomalies on different days of the week are very well found without merging data from different stations. There is no temperature analysis given for neither Zurich nor Lugano. In BV, Figure 4, it is shown that the German stations show similar weekly temperature cycles all over the country. We agree that the temperature data of nearby situated stations are usually correlated, but the correlation of precipitation and sunshine duration decreases rapidly when the distance between stations increases. Especially convective precipitation enormously varies even on very small spatial scales. For the spatial correlation of precipitation in Frankfurt with other stations in Germany, Schönwiese and Rapp [1997] showed that the correlation rapidly decreases with increasing distance in summer. For instance, the correlation coefficient is 0.7 in a distance of about 50 km to Frankfurt, and it decreases to below 0.5 in a distance of 100 to 200 km depending on the direction. The atmospheric variables humidity and lapse rate as mentioned by HF probably do not completely trigger convective precipitation as can be seen from the difficulties in formulating universal convective indices. In winter, the correlation coefficient relative to Frankfurt is higher but still reaches values of below 0.5 in Germany. The difference arises from the predominantly convective character of precipitation in summer. In the Great Plains, United States [Groisman and Legates, 1994], a similar pattern of correlations with also very rapidly decreasing coefficients in summer was found. [11] 7. We do not fully agree with the interpretation of the Monte Carlo experiment applied to the Zurich precipitation data 1991–2005 as described by HF. Only 9% of the realizations showed a deviation on one weekday that is as large as the measured one on Saturday. Beside the fact that it is not sufficient just to look for deviations on single weekdays but also for a weekly cycle in general in our opinion, we disagree with the conclusion by HF that this probability of 9% justifies the assertion that the German weekly cycles (based on 12 stations) were probably purely random. According to HF, the Monte Carlo experiment yields a 21% probability of a weekly cycle which he defines as the occurrence of at least 4 consecutive weekdays with above or below average precipitation regardless of the amplitude of deviations. In our opinion, both requirements rather should be combined to best reproduce reality which would lead to an estimated probability of clearly below 5%. Besides, we emphasize that it is somewhat problematic to draw conclusions for complete Germany by just analysing data from Zurich which moreover is not located in Germany. Furthermore, we wonder about the statement by HF that the higher Saturday precipitation in Zurich was “probably due to higher extreme precipitation events”. The author easily could have checked this assertion. He also could have analysed the precipitation frequency as in our work (BV) to address this issue. [12] 8. We hypothesized (see BV) that there could be an interaction of a weekly aerosol cycle and the atmospheric dynamics in Germany, which could explain such a widespread weekly temperature cycle. In the meantime, we could demonstrate the existence of a significant weekly cycle of the aerosol optical thickness over Germany and adjacent areas [Bäumer et al., 2008] which supports our hypothesis. On other continents, for instance, a systematic pattern of weekday-weekend differences of the daily temperature range could be found. Forster and Solomon [2003] demonstrated this for the daily temperature range weekend effect in the Unites States that showed a large-scale pattern. They concluded that the pattern is not random but significant. Its intensity increased remarkably after 1959. Also Canada, Mexico, East China, and Japan showed significant patterns according to their work. For China, the analysis was repeated but for many variables [Gong et al., 2006]. Significant anomalies could be found with changes in their sign from winter to summer for many stations and variables, which was interpreted by the authors as an interaction of different aerosol effects and seasonal circulation. Both in winter and in summer, the weekend effect increased since the late 1970s. These studies (and further work cited by BV) also demonstrate in our opinion that a systematic interaction of a weekly aerosol cycle and meteorological variables is possible or even probable in various regions with high anthropogenic activity.