Hydrological Sciences-Joumal-des Sciences Hydrologiques, 45(3) June 2000 477

Natural climatic changes and solar cycles: an analysis of hydrological time series

MARIO TOMASINO & FRANCESCO DALLA VALLE Hydrological Unit, Production Division, EN EL, Corso del Popolo 245, 1-30172 Mestre VE, Italy e-mail: tomasino.mario@enel.it, dallavaile.francesco@enel.it

Abstract The effects induced on the climate by human activity have become a major issue for the new millennium. In order to arrive at sustainable conclusions it is necessary, first of all, to assess and quantify natural climatic changes. In general this is done by analysing available time series. In the case of historical hydrometeorological data sets, a comparative analysis with solar cycles is not usually conducted. This work, however, demonstrates that the effect of solar cycles observed at the Equator is also visible at middle and high latitudes with multiple periodicity of the basic solar frequency (roughly 11 years). This could well be due to the interaction between solar forcing and circulation mechanisms within the atmosphere, i.e. water-air-soil interac­tions coupled with anthropogenic forcing. This theory has been tested by comparing different types of historical data series with the River Po discharges and cyclic appearance of slime bloom in the Adriatic Sea.

La variation climatique naturelle et le cycle solaire : une analyse de séries hydrologiques Résumé Les effets sur le climat de l'action de l'homme sont devenues le problème du nouveau millénaire. Pour arriver à de conclusion crédibles il est nécessaire d'estimer avant tout la variabilité climatique naturelle. Normalement, pour évaluer cette variabilité on étudie des séries de données climatiques. Habituellement pour les séries hydrologiques on ne fait pas d'analyse comparative avec le cycle solaire. Dans cette note on va montrer que l'effet du cycle solaire, trouvé à l'équateur, est également visible aux moyennes et hautes latitudes avec une périodicité multiple de la période solaire de base (11 ans). La cause de cette variation de périodicité sera recherchée en étudiant l'interaction entre le forçage solaire et le mécanisme de circulation à l'intérieur de l'atmosphère (interaction eau-air-terre et effet de serre). Cette hypothèse à été évaluée en comparant différentes séries hydrométéorologiques avec la série des débits naturels du Po et les apparitions de mucilage dans la Mer Adriatique.

INTRODUCTION

Over the last decades, there has been a great increase in scientific inquiry and public debate regarding the effects of human activities on the climate. In order to reach credible conclusions, it is of vital importance to be able to estimate natural climatic changes. The climate of the Earth is the manifestation of how solar irradiance is absorbed and redistributed in the atmosphere, in the oceans and reradiated into space.

A variation in the energy that reaches the Earth's surface, or that is reflected by it, may affect the climate. If it were possible to accurately define the "natural" variability of a series of data (temperature, rainfall, discharge etc.), it would be reasonably possible to estimate the part induced by human activity (the greenhouse effect with the concomitant causes triggered by the energy industry, car exhaust emissions, forest fires etc.).

Open for discussion until 1 December 2000

478 Mario Tomasino & Francesco Dalla Valle

It is in this context that an investigation carried out on the historical data set of the discharge records from a river and the comparison with other historical data sets, such as surface-air temperature anomalies in the Northern Hemisphere, field of atmospheric pressure at sea level over North Atlantic region (North Atlantic Oscillation, NAO index), rainfall and hydrometric levels in a lake, are discussed herein. It was observed that these data sets follow, with oscillations, the decadal Sun cycles and the longer-term variations of solar activity (solar energy fluctuations), but with time lags and some differences in phase. In order to assess the variability of natural data sets, it is necessary to take the solar forcing into consideration.

ANALYSIS OF HISTORICAL DATA SETS

The idea of determining the climatic variability by analysing hydrometeorological data is not new. At the end of the nineteenth-century, Bruckner (1890) created a hypo­thetical sequence of climatic fluctuations based on a comparison of similar meteoro­logical data recorded in the same periods in different parts of the Northern Hemisphere. In this way, he identified large-scale climatic fluctuations of the order of a few hundred years, such as the cold-wet period between 1600 and 1850, now referred to as the Little Ice Age, and the present warm-dry period, 1850 to date. Within this sequence he identified alternating shorter cycles: cold-wet or warm-dry, varying in length from 10 to 35 years, and averaging 20 years (Bruckner cycles, Fig. 1).

A study of discharges recorded in the main European rivers and in the Yellow River in China (Tardy, 1986) also reveals analogous patterns during the same periods, confirming that climatic fluctuations produce similar effects in different natural environments, even if they are a great distance from each other.

Of particular interest is an analysis of a water level data set recorded in Lake Victoria. The peculiarity of this equatorial lake is that roughly 80% of its water balance is directly based on rainfall and evaporation, while only a small percentage depends on river inflow and outflow (Yin & Nicholson, 1998); it therefore behaves like a huge evaporimetric tank.

It has been observed that a good correlation between sunspot and lake levels has been maintained from 1885 to 1927 and from 1968 to 1997 (Mason, 1998). The comprehension of the effect of solar activity on Earth climate has been enhanced by the work of Friis-Christensen & Lassen (1991). They advanced the theory that variations in the global temperature may be correlated to the Sun's activity, i.e. that the

1691-1715 173S-17SS 1771-17S0 1806-1825 1841-18SS 1871-1885 (1911-1930) (1950-1970)

1716-1735 1756-1770 1781-1805 1826-1840 1856-1870 (1886-1910) (1931-1949)

Fig. 1 Schematic diagram of the Bruckner cycles with 20-year climatic fluctuations from 1691 up to 1970 (from Veggiani, 1986).

Natural climatic changes and solar cycles: an analysis of hydrological time series 479

9,65 1 ; ; : ; ; ; ; ~ p 0.3

12.05 \ 4 ! i ! i— , , 1 i ! r i ! \ 1- -0.5 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000

year

Fig. 2 A comparison between the mean length of the solar cycle and temperature anomalies in the Northern Hemisphere (developing an idea advanced by Friis-Christensen & Lassen, 1991).

global warming of the Earth's surface temperature which has occurred over the last 100 years in the Northern Hemisphere shows a good correlation with the length of the solar cycle. In fact, solar activity over the past 100 years has generally increased, since the length of the cycle has dropped by about 11.5 years to less than 10 years in length, i.e. the shorter the interval, the more active the Sun (Fig. 2). The cycle length is obtained during maximum and minimum periods from the difference between the subsequent and previous maximum/minimum points. The obtained lengths were then filtered using a low pass with coefficient 1,2,2,2,1. The last of the points were obtained by fitting the lengths of the two subsequent cycles to the one from the last cycle to be observed. The temperature anomaly curve represents the monthly value mean of temperature anomalies during the course of the reference solar cycle. This work was viewed with a degree of scepticism since in the event of an increase or a decrease in sunspots, the variation in solar energy reaching the Earth rises or falls by not more than 0.1%. Critics maintained that this fraction was far too low to account for the variations in the surface temperature.

In another more recent article, Svensmark & Friis-Christensen (1997) provided a possible explanation. Satellite data have shown that during high solar activity, there is a reduction in the level of cosmic radiation reaching the Earth and this in turn leads to a reduction in cloud cover. A reduction in cloud cover entails more solar radiation reaching the Earth. The effect of cosmic rays on cloud formation is under discussion, but it may be linked to the ionizing effect they have on the water particles suspended in the atmosphere. Barlow & Latham (1983) have shown experimentally that charged drops become more effective attractors of aerosols than uncharged ones.

As far as Lake Victoria is concerned, there appears to be a good correlation between water levels and number of sunspots (Fig. 3), and this correlation seems to become more evident when dust levels in the atmosphere are high. This occurred after severe volcanic eruptions: Krakatoa in 1883 and Monte Agung in 1963 (Mason, 1998).

To clarify, mention should be made of the study of Pudovkin & Veretenenko (1995 and 1996), concerning the response of cloud cover to the flow of cosmic rays.

480 Mario Tomasino & Francesco Dalla Valle

(/) o n r t / j

O

n t-

rti 0)

200

180

1HU

140

1?0

100

80

60

40

20

0

u

# i'

1»

" Sj \

# /

"ki

t~ I I ' \

« » Ht

.

;-| -

,

--t . 1 *

,

1

j

. .

-

s

r -

-.

>

- . '

-J— ; -t O i - l O r - C D i - t D r - <0 -r- CO T ~ t o -r- C D - i — tO t ~ tO -t— tD c n o o - < - T - c g c \ i c o c o ' ^ - ' 3 - L O L o c o t D r ~ - r - a D c o c n a > O D C 7 i œ a i c n o > c r ) C D c n c n c 3 î C T ) C ï ) a ) c n Q j a î c n c n œ o ï

13.5

(m)

m

c

t-

u X

1 3 -

12.5 -

12 -11.5 -

11 -10.5 -

10 -C O T - < 0 T - C O T — C D * — CO T— C O ^ — C O ^ t D T — 10 T - CD *— t0 0 ) o o T - T - c ^ c M C 5 c o ^ r ^ r m L O ( o t D f ^ [ ^ c D c o œ a > c o c 7 > a > c n c » a ) a > a > a > c r c a i a > a > c n o c o a > c 7 > a > c r j a j

Fig. 3 A comparison between the annual mean number of sunspots and the end-of-year hydrometric level in Lake Victoria.

They discovered that the amount of cloud cover decreases when the solar activity increases and cosmic ray flux decreases.

Svensmark & Friis-Christensen (1997) argued that if the flow of cosmic rays really is the cause of variations in cloud cover, the effect should be minimum close to the geomagnetic Equator, where the lines of the Earth's magnetic field are horizontal and they therefore have a greater protective effect with respect to the ionizing particles. Svensmark & Friis-Christensen (1997) concluded that the topic needs further study regarding latitude and altitude distribution of the cosmic rays effect. In any case, the Equator and the surrounding areas appear to be very promising in this respect, since the solar cycle's signal, which comes from space, affects the cycles of the stratosphere-troposphere and penetrates the atmosphere to descend to the Earth's surface (Labitzke & van Loon, 1993): for the summer period, it was observed that the year-to-year variations in the vertical flows of the tropics and subtropics contain the components of the solar cycle (Labitzke & van Loon, 1997). Since the equatorial zones are the least disturbed (Svensmark & Friis-Christensen, 1997), they seem to be the most suitable areas for observing phenomena linked to solar cycles.

On the other hand, the signal variation of the natural forcing generated by the Sun is barely visible at the middle and high latitudes of the Earth. This is probably due to forcing factors in the atmosphere (greenhouse effect gases, anthropogenic forcing and natural forcing) and the fact that the mutual interactions between the oceans and the Earth create a prevailing internal atmospheric variability so that it is difficult to distinguish the effects of the basic parameters. Nevertheless, it could be interesting to present an example of how the signal could be detected at middle latitudes. This hypothesis has not been proved beyond reasonable doubt and it could still be used as a basis for further studies.

Natural climatic changes and solar cycles: an analysis of hydrological time series 481

DISCHARGE RECORDS FROM THE RIVER PO AND THE CYCLIC APPEARANCES OF SLIME BLOOMS IN THE ADRIATIC SEA

The River Po, which meanders its way around 45°N, has a catchment surface area of 70 000 km" and the characteristic of being surrounded on three sides by high mountains (the Alps and the Apennines, see Fig. 4). The river flows into the upper (northern) part of the Adriatic Sea. This rather shallow sea is occasionally subject to episodes in which unpleasant-looking slime blooms appear (Tomasino, 1996). These slime blooms are composed of a gelatinous, threadlike substance, which floats in the sea and can engulf fish and plankton, provoking their death. As the latter putrefy, they give off a fetid smell. Since there is a flourishing tourist industry along the sandy coastline of the Adriatic Sea, the appearance of mucilage is of great concern to tour and hotel operators.

The freshwater discharge entering the sea from the River Po plays an important role in the dynamics of the northern Adriatic since it generates density currents which, coupled with the density currents generated naturally by the seasonal cycle, represent one of the main driving forces in the sea's circulation. Further circulation components are induced by the wind and the tide (i.e. atmospheric and astronomical components).

An investigation into the slime bloom phenomenon has shown that a single year with a river discharge deficit is not enough to upset the system, but more than one year is. In mathematical terms, this means that river discharges should be analysed with a numerical filter. In temporal terms, the cause (the monitored discharges) cannot be too far away from the effect (slime blooms appear in late spring-summer) and therefore only river discharges recorded within the period preceding the appearance of slime blooms, i.e. October-March, were analysed. By trial and error the 3-year numerical filter was determined to be the most suitable for detecting the appearance or absence of slime blooms (on-off model). The filter values Ok are defined as follows:

Qk l *

D ; - f - „ 9 (1)

Alps /x<

Fig. 4 Location of the River Po catchment in North Italy.

482 Mario Tomasino & Francesco Dalla Valle

where g, is the mean discharge of the River Po at Pontelagoscuro during the October 0' - 1) to March (i) period.

The investigation period is from 1921 up to the present day, since discharges have been monitored on a daily basis on the River Po at Pontelagoscuro since that date and records are available for the entire period.

By processing the data, a sequence of annual values was obtained that assume a sinusoidal trend with clearly identifiable maxima and minima. It is possible to establish the threshold above which the function departs from its natural variability by computing the standard deviation of the phenomenon. Figure 5 demonstrates that when the function has come close to, or crossed, the lower threshold, suitable conditions existed; in fact slime blooms may appear either in massive quantities or sporadically (Tomasino, 1996). This phenomenon could be relieved by the presence of strong winds, which reactivate the circulation of the basin and also by potential flood conditions in the river in late spring/early summer. Once the concurrence of the signal and the appearance of the phenomenon had been established, it was assumed that the phenomenon was due to the lack of nutrients transported by the river; these are indispensable for aquatic life (episodes had already occurred in the 1800s at a time when pollution was not supposed to have existed).

More recent océanographie studies earned out in the Mediterranean have confirmed the existence of a three-year inter-annual cycle in its waters (Artegiani et al., 1997). The biochemical inertia in the upper Adriatic would justify this assumption, but if water residency times were considered, this hypothesis would no longer be valid in view of the limited depth of the basin. In April, when October-March Po River discharge data become available from the Pontelagoscuro station, the model was used to predict the state of the upper Adriatic Sea in the May-August period. The model, which has been available since 1990 and has proved capable of predicting the large-scale appearances of slime bloom in 1988 and 1989, was accurate in its predictions for the years 1990, 1991 and 1992. In all, there were five years in which the situation was critical for a longer time (Fig. 5). Fortunately, the last three years were relieved by concomitant favourable phenomena, i.e. the presence of wind and flood conditions in the river in the late spring; appearances of slime blooms were therefore limited.

Fig. 5 Comparison between filtered records from winter (October-March) River Po discharges and mucilage appearance in the northern Adriatic Sea.

Natural climatic changes and solar cycles: an analysis of hydrological time series 483

As the years have gone by it has been realized that the discharge patterns in the river follow the change from a dry period to a wet period. During the period 1993-1996, catastrophic flood conditions reappeared in Italy causing over 200 deaths.

This processed discharge data set described above was compared to other climatic data sets.

ANALYSIS OF THE WINTER DISCHARGE DATA SET AND COMPARISON WITH TEMPERATURE RECORDS

By comparing the patterns of annual anomalies in the air temperature in the Northern Hemisphere, recorded at stations between 44 and 64°N over the period 1880-1985 (106 years) (Hansen & Lebedeff, 1987), with the filter of the discharges from the River Po, a good agreement is revealed (see Fig. 6) in terms of trends and irregularities. The linear correlation calculated between the 5-year running mean of the anomalies in air temperature and the filter (equation (1)) of the River Po discharges is equal to 0.32, with a significance level of 94%. The significance level relative to correlation coefficient is evaluated using resampling methods as described by Edginton (1995) using 100 000 random permutations of the time series. To evaluate the significance level for the correlation between filtered series, the permutations are applied to unfiltered data. The resampling value of the correlation coefficient is computed as follows: (a) the original time series are permuted, (b) the permuted series are filtered, and (c) the resampling value of the correlation coefficient is evaluated between

^Temperature anomalies 44°-64° N. (Hansen, Lebedeff, 1987)

Mean discharge in the Po at Pontelagoscuro from the period October-March

1.00

0,80

0.60

0.40

0.20

0.00

-0.20

-0.40

Fig. 6 Comparison between the filter of winter (October-March) discharges of River Po and temperature anomalies for the area between 44° and 64°N (Hansen & Lebedeff, 1987). The continuous lines superimposed on the two sets represents the 5-year running mean for the temperature anomalies and the mean for the last 3 years of discharges in the River Po. The linear correlation between the two filtered sets is equal to 0.32 (with a 94% significance level).

484 Mario Tomasino & Francesco Dalla Valle

permuted-filtered time series. However, the same type of analysis performed on Jones' data set (Climatic Research Unit, School of Environmental Sciences, University of East Anglia, UK—data set started in 1859, updated until January 1999) did not give the same results. This is probably due to the nonlinear rise in the average temperature in the Northern Hemisphere, which has been recorded over the last 10 years, and also to the fact that Jones' data, unlike Hansen's, refer to the whole of the Northern Hemisphere.

An examination of the data set regarding the filter of the discharges (see Fig. 6) reveals peaks that recur at an interval of between 10 and 30 years, with an average of roughly 20 years. A harmonic analysis of the data set confirms the presence of a component in the spectrum with a period equal to roughly 20 years (with a significance level of 95%; see Fig. 8(b)). The significance level for the frequency components is evaluated using resampling methods as described by Manly (1997) with 5000 random permutation of the time series. These facts could confirm Bruckner's conclusions (Bruckner, 1890), but his hot-dry and cold-wet sequences seem to be too simplistic. From a physics point of view, it follows that, as the temperature of the atmosphere rises, evaporation also increases, along with the vapour content in the air and the likelihood of precipitation. A combined analysis of the anomaly sets shows that, as the temperature increases, so does the rainfall and, in turn, the discharte it generates.

Considering that the river discharge is the result of the transformation process earned out by the River Po basin on the meteorological variables, and that the Mediterranean is capable of generating considerable quantities of moisture over this basin, these predicted increases in precipitation were indeed observed in the periods indicated. By calculating the number of days on which the 99th percentile was exceeded by the distribution of the daily mean discharges, the figure obtained ranges from a total of zero to a few days above the 99th percentile per year during minimum discharges, and over 25 days per year during maximum discharges.

Another study (Mann et al., 1995), which used a longer data sample of air temperatures overland monitored over 400 years, covering the entire hemisphere, showed that such cycles do exist, e.g. a cycle of between 15 and 20 years has a signifi­cance level of over 99%. These cycles can be partly explained (Chierici et al., 1996) by deterministic components such as sunspot cycles and the Earth's nutation axes cycles. One should also bear in mind that, during the Earth's rotation, its position in relation to the Sun does not remain constant (Milankovitch cycles).

Continuing the analysis of historical data sets, the rise in water levels of Lake Victoria (Equator) which occurred at the beginning of the 1960s, explained by rainfall variation in the catchment area (Sene & Plinston, 1994), can be seen to coincide with a marked increase in the discharges of the River Po recorded over the winter period (45°N) and with an increase in the hydrological variability in rainfall and discharges (runoff) in north central England (55°N) (Burt et al., 1998). A linear correlation was found equal to 0.39 with a significance level of 98% between the series of River Po discharges and the ratios between winter rainfalls (DJF) and summer rainfalls (JJA) at Durham, UK (Fig. 7).

Proceeding in chronological order, the next cycle is evident in Lake Victoria from the beginning of the 1970s, but is only hinted at in records from the River Po and from north central England, whereas the 1980s cycle appears in all three sites, though its appearance is slightly delayed in Lake Victoria.

Natural climatic changes and solar cycles: an analysis of hydrological time series 485

3020

2740

2460

"g2180

$ 1900

•§ 1620 <n

Q 1340

1060

780

500

M Mean discharge in the Po at Pontelagoscuro from the period October-March

*> Variability in precipitation at Durham - Rainfall DJF/Raînfal! JJA

0 0 ~ O

•ù

o s

I I i l ! * i >

J \ x| 1

\ / " V / : x

1 " 1 X

; s

\

71 °~^; o t>

! "

- ^ ,

* O

i ' 1

0 " =>n 0 ,

1 / V À

X >? K r '

--

a

j »

j

1

—GO- 8

- 1 1.

1

à A f \

- -

o

© Cig

j

1

"^VjA

•j X

;

>

•

o

«N,

*

1 x

]/\

I i X \

" 1 X

- 3-50 E

- 3.00 |

• 2.50 »

" 200 I •1-50 i - 1.00 1

- 0.50 I

- 0.00 |

-0.50 o

-1.00 S

Fig. 7 Comparison between the filter of winter (October-March) discharges of the River Po and the ratio between winter rainfall (DJF) and summer rainfall (JJA) at Durham, UK. The continuous lines superimposed on the two sets represent the 5-year running mean for the ratios between rainfalls and the mean for the last three years of discharges in the River Po. The linear correlation between the two filtered sets is equal to 0.39 (with a 94% significance level).

A harmonic analysis of the three data sets (Fig. 8) reveals congruent behaviour, despite the different nature of the three sets, i.e. the water levels of a lake, the winter discharges of a river and the ratio between winter and summer rainfall. An analysis of the end-of-year water levels in Lake Victoria (Fig. 8(a)) conducted on the period 1899-1976 (Yin & Nicholson, 1998), reveals the presence of a significant com­ponent with an 11-year period, corresponding to the periodicity of sunspots. The analysed period is from 1899 to 1976 and its length is equal to that of the other time series analysed to have the same frequency component. The spectral analysis conducted on the entire period 1899-1994 has revealed that the component of 12-year period has a 97% significance level. As one moves up towards the middle latitudes (see Fig. 8(b)), the 11-year component diminishes as the component with a period of 19.5 years (roughly double the period observed for Lake Victoria) broadens considerably, becoming significant with a probability equal to 95%. On assessing the components present in the Durham data set further north (Fig. 8(c)), the component with an 11-year period appears to be even more reduced, while the component with a 19.5-year period has a 97% significance level. What now needs to be understood is why the cycles that can be detected at the Equator (Fig.8(a)-(c)) differ at the mid-and high latitudes, even if they basically remain decadal or multiples thereof, as in the case of sunspots. The most plausible answer to the skipping of cycles, or to their cover, may be sought in the atmosphere's circulation variability and in the related observational indicators, e.g. turbulence, clouds, snow at ground level, etc.

486 Mario Tomasino & Francesco Dalla Valle

Fig. 8 Comparison between the spectra of three different hydrological data sets referring to different latitudes: (a) hydrometric levels in Lake Victoria (Equator), (b) mean winter discharges (October-March) in the River Po at Pontelagoscuro (45 °N), (c) ratios between mean winter rainfall (December-February) and mean summer rainfall (June-August) for the Durham station, UK (55°N).

COMPARISON BETWEEN THE WINTER DISCHARGES IN THE RIVER PO AND NAO

The simplest possible indicator of circulation fluctuations in the atmosphere, and in particular of the low frequency variability components, is the field of mean sea level pressure (MSLP). It has been demonstrated (Hurrel, 1995) that, in the winter period, the atmospheric pressure in the North Atlantic oscillates around characteristic values (North Atlantic Oscillation—NAO). The difference between the pressure measured at sea level in Iceland and in the Azores may assume an index value greater than 1.0 or less than -1.0. Different types of atmospheric circulation are linked to these index values. If the index number is positive (the pressure over the Azores is higher than

Natural climatic changes and solar cycles: an analysis of hydrological time series 487

over Iceland), Europe and the Mediterranean will have mild and drier winters. If the index number is negative, the same areas (Europe and the Mediterranean) will experience harsh and wetter winters, due to the different paths of the winds and clouds. Oscillations present an inter-annual variability and there are periods in which circulation patterns persist for several years, almost as if they "remember" the behaviour of the.previous year.

An analysis was performed on the NAO index set from the winter period (December-March: DJFM), since the signal is more marked in that period, reporting it with the temperature trend (5-year filter), and the discharge data set from the River Po (3-year filter) (see Fig. 9). What is already known about NAO and temperature trends is that from 1920 to 1938, when the NAO index was basically positive, there were winters that were somewhat milder than the ones in the previous period; from 1940 to 1970, with a generally negative NAO index, the winters were frequently colder than normal, while for the last 27 years, with a NAO index that has been mainly positive, the winters have been warmer and drier (cf. Hurrel, 1995).

An analysis of the NAO data and of filtered river discharge records shows that for 1940-1997, the periods of change from a dry period to a wet period, occur in the downward phase of the NAO data and around the minima. In fact, during the downward phase of the NAO data (from positive to negative), the discharges are low around the maximum NAO index value, and they then gradually increase towards the minimum value. The set concerning the filter of the River Po discharges presents a linear correlation equal to -0.30 (94% significance level) with the filtered set

3 NAO index for the months DJFM Mean discharges in the Fo at Fbntelagoscuro from the period October-March

J

r ic --4V ,, ,

. Jl l l lLiuJi iËjJ l'y ii i "

Fig. 9 Comparison between the filter of winter (October-March) discharges of the River Po and the NAO index for the months (DJFM). The continuous lines superimposed on the two sets represent the centred filter with weights (1,3,5,6,5,3,1) for the NAO index and the mean for the last three years of discharges in the River Po. The linear correlation between the two filtered sets is equal to -0.30 (with a 94% significance level).

488 Mario Toinasino & Francesco Dalla Valle

(centred running mean with weights 1,3,5,6,5,3,1) of Jones' NAO index. It should also be noted that, when the river discharge maximum was passed (i.e. with a change in cycle), it always occurred with the NAO of the DJFM period descending in the positive values and it is marked by an abrupt change to negative, as in the years 1928-1929, 1935-1936, 1961-1962, 1978-1979, 1995-1996 (see Fig. 9). Moreover the presence of two consecutive year negative, or close to zero, NAO index values coincides with the covering of a decadal cycle, as can be seen in years 1955-1956 and 1956-1957; 1969-1970 and 1970-1971; 1984-1985 and 1985-1986. These behaviour patterns, which are repeated in sign and frequency, underline the importance of the atmosphere's internal energy, which may obscure the effects of the exogenous contributions of solar cycles.

CONCLUSIONS

In order to understand the natural variability observed in hydrometeorological records, an analysis using non-traditional methods could lead to the detection of cyclicities that have not been identified before. The validity of these cycles should therefore be sought in other data sets of different disciplines. For instance, this methodology has enabled the cyclic appearance of slime blooms in the Adriatic Sea to be explained.

The identification of cyclicities could become the starting point for establishing seasonal forecast models, which would be invaluable for optimizing the economic value of water resources. Seasonal forecasts, which could provide general indications on the coming season, would also prove invaluable for agricultural purposes, as well as for the optimum usage of water in natural and manmade reservoirs, especially from a competitive viewpoint when water has to be shared between domestic, irrigational, flood control, navigation, hydroelectric, leisure and other demands.

An analysis of historical data sets from around the Equator (end-of-year water levels in Lake Victoria), the northern part of Italy (45°N—the filter of river discharges from the October-March period) and north central England (~ 55°N—filter of the ratio between winter rainfall (DJF) and summer rainfall (JJA)) led to the detection of analogous cyclic behaviour patterns, but in order to explain them, it is necessary to seek a common external cause: the Sun. Therefore it may not be sufficient to limit these studies to atmospheric conditions alone.

It was further observed that there are decadal cyclicities at the Equator and the mid- and high latitudes and the cycles established by Bruckner were put into a modern context. These cyclicities allow one to perceive the "average tendencies" of the weather in advance. To these tendencies one could add the results of the classical meteorological models in order to assess future "transitory periods", i.e. potential weather extremes above (flood conditions) or below (drought conditions) the norm. In this time of high positive anomaly in the hemisphere's surface temperature, this study would appear to be promising and timely.

In conclusion, it could be pointed out that nature seems to delight in revealing certain aspects of its character, those that can be reasonably predicted. On the other hand nature is quite capable of changing course unexpectedly: the study of interaction between solar activity and atmosphere seems to be promising to understand why and when.

Natural climatic changes and solar cycles: an analysis of hydrological time series 489

Acknowledgements The authors wish to thank F. Allodi for his kind cooperation in furnishing data for River Po discharges, Yin Xungang and S. E. Nicholson who provided the Lake Victoria data set, and T. P. Burt for the Durham rain data set. Special thanks to Jim Hurrel for his useful comments to a previous version of the paper. Thank are also due to the anonymous referees for useful comments and to Helen Thomas for help in the translation of the text.

REFERENCES

Artegiani, A., Paschini, E., Russo, A., Bregant, D., Raicich, F. & Pinardi, P. (1997) The Adriatic Sea general circulation. Part I, air-sea interactions and water mass structure. / . Phys. Oceanogr. 27, 1492-1514.

Barlow, A. K. & Latham, J. (1993) A laboratory study of the scavenging of sub-micron aerosol by charged raindrops. Quart. J. Roy. Met. Soc. 109, 763-770.

Bruckner, E. (1890) Klimaschwankungen seit 1700 nebenst Bemerkungen fiber die Klimaschwankungen der Diluvialzeit. (Climatic fluctuations since 1700 in comparison to climatic fluctuations of the Diluval era, in German). Geographische Abhandlungen IV(2), 153-184. Vienna, Austria.

Burt, T. P., Adamson, J. K. & Lane, A. M. J. (1998) Long-term rainfall and stream flow records for north central England: putting the Environmental Change Network site at Moor House, Upper Teesdale, in context. Hydrol, Sci. J. 45(5), 775-787.

Chierici, F., Gasperini, M., Gasperini, L. & Marabini, F. (1996) Short-term periodic climatic change: a combined effect between the sunspot cycle and the lunar nutation. Tech. Report no. 43, CNR Istituto di Geologia Marina, Bologna, Italy.

Edginton, E. (1985) Randomization Tests. Marcel Dekker, New York, USA. Friis-Christensen, E. & Lassen, K. (1991) Length of the solar cycle: an indicator of solar activity closely associated with

climate. Science 254, 698-700. Hansen, J. & Lebedeff, S. (1987) Global trends of measured surface air temperature. J. Geophys. Res. 92, 13345-13372. Hurrel, J. (1995) Decadal trends in the North Atlantic Oscillation: regional temperatures and precipitation. Science 269,

676-679. Labitzke, K. & van Loon, H. (1993) Some recent studies of probable connections between solar and atmospheric

variability. Ann. Geophys. 11, 1084-1094. Labitzke, K. & van Loon, H. (1997) The signal of the 11-year sunspot cycle in the upper troposphere-lower stratosphere.

Space Sci. Rev. 80, 393-410. Manly, B. F. J. (1997) Randomization, Bootstrap and Monte Carlo Methods in Biology (second edn). Chapman & Hall,

London, UK. Mann, M. E., Park, J. & Bradley, R. S. (1995) Global interdecadal and century-scale climate oscillations during the past

five centuries. Nature 378, 266-270. Mason, P. (1998) Levels of lake Victoria. World Water and Environ. Engng October 1998, 9. Pudovkin, M. I. & Veretenenko, S. (1995) Cloudiness decreases associated with Forbush-decreases of galactic cosmic rays

J. Atmos. Terr. Phys. 57, 1349-1355. Pudovkin, M. I. & Veretenenko, S. (1996) Variations of the cosmic rays as one of the possible links between the solar

activity and the lower atmosphere. Adv. Space Res. 17, 161-164. Sene, K. J. & Plinston, D. T. (1994) A review and update of hydrology of Lake Victoria in East Africa. Hydrol. Sci. J.

39(l) ,47-63. Svensmark, H. & Friis-Christensen, E. (1997) Variation of cosmic ray flux and global cloud coverage—a missing link in

solar-climate relationships, j . Atmos. Solar-Terrest. Phys. 59, 1125-1232. Tardy, Y. (1986) Le Cycle de L'Eau. Climats, Paleoclimats et Geochimie Globale, 169-209. Masson, Paris, France. Tomasino, M. (1996) Is it feasible to predict "slime bloom" or "mucilage" in the northern Adriatic Sea? Ecol. Modelling

84, 189-198. Veggiani, A. (1986) Le fluttuazioni del clima dal XVIII al XX secolo. I cicli di Bruckner. (Fluctuations in climate from the

eighteenth to the twentieth centuries. I. Bruckner cycles, in Italian. Torricelliana 37, 107-159. Yin, X. & Nicholson, S. E. ( 1998) The water balance of lake Victoria. Hydrol. Sci. J. 43(5), 789-811. Received 13 August 1999, Accepted 21 February 2000