COBISS: 1.01 ANALySIS OF LONG-TERM (1878-2004) MEAN ANNUAL DISCHARGES OF THE KARST SPRING FONTAINE DE VAUCLUSE (FRANCE) ANALIZA DOLGOČASOVNEGA (1878-2004) POVPREČNEGA LETNEGA PRETOKA KRAŠKEGA IZVIRA FONTAINE DE VAUCLUSE (FRANCIJA) Ognjen BONACCI1 Abstract UDC 556.36(44) Ognjen Bonacci: Analysis of long-term (1878-2004) mean annual discharges of the karst spring Fontaine de Vaucluse (France) Statistical analyses have been carried out on a long-term (1878-2004) series of mean annual discharges of the famous karst spring Fontaine de Vaucluse (France) and the mean annual rainfall in its catchment. Te Fontaine de Vaucluse is a typical ascending karst spring situated in the south-eastern region of France. Te spring has an average discharge of 23.3 m3/s. Te average annual rainfall is 1096 mm. Its catchment area covers 1130 km2. Using the rescaled adjusted partial sums (RAPS) method the existence of next fve statistically signifcant difer-ent sub-series was established: 1) 1878-1910: 2) 1911-1941; 3) 1942-1959: 4) 1960-1964; 5) 1965-2004. Te diferent spring discharge characteristics during this long period (1878-2004) can be caused by natural climatic variations, by anthropogenic infuences, and possibly by climate changes. At this moment it should be stressed that objective and scientifcally based rea-sons for diferent hydrological behaviour in fve time sub-peri-ods could not be found. Keywords: karst hydrology, mean annual discharges, annual catchment rainfall, karst spring, Fontaine de Vaucluse, France. Izvleček UDK 556.36(44) Ognjen Bonacci: Analiza dolgočasovnega (1878-2004) povprečnega letnega pretoka kraškega izvira Fontaine de Vaucluse (Francija) V prispevku predstavim statistično analizo časovne vrste povprečnega letnega pretoka in letnih padavin v zaledju slavnega izvira Fontaine de Vaucluse v Franciji. Fontaine de Vaucluse je tipični kraški izvir pri katerem voda priteka iz velike globine. Nahaja se v jugovzhodni Franciji. Povprečni pretok izvira je 23,4 m3/s. Povprečna količina letnih padavin v zaledju, ki meri 1130 km2, je 1096 mm. Z uporabo metode umerjenih delnih vsot (RAPS) smo določili pet statistično pomembnih različnih podobdobij: 1) 1878-1910: 2) 1911-1941; 3) 1942-1959: 4) 1960-1964; 5) 1965-2004. Vzrokov za različne pretoke preko celotnega obdobja (1878-2004) je lahko več; npr. klimatske spremembe in antropogeni vplivi. V tem trenutku moramo poudariti, da objektivne znanstvene razlage za različne hidrološke značilnosti v petih podobdobjih še ne poznamo. Ključne besede: hidrologija krasa, povprečni letni pretok, količina letnih padavin, kraških izvir, Fontaine de Vaucluse, Francija. INTRODUCTION Te Fontaine de Vaucluse represents one of the most famous and most important karst springs on the Earth. It is located in the south-eastern karst region of France (Figure 1), about 30 km eastward of the town of Avignon. It represents the only fow exit from the 1500 m thick karst aquifer of Lower Cretaceous limestone (Blavoux et al., 1992b). Te karst system of the Fontaine de Vauc-luse is characterised by an approximately 800 m unsatu-rated zone. Emblanch et al., (1998) and Emblanch et al., (2003) stressed important role of this zone for the transformation of rainfall into runof. Te Fontaine de Vauc-luse karst spring catchment area is estimated to be 1130 1 Faculty of Civil Engineering and Architecture, University of Split, 21000 Split, Matice hrvatske 15, Croatia, E-mail: obonacci@ gradst.hr Received/Prejeto: 27.11.2006 TIME in KARST, POSTOJNA 2007, 151–156 OGNJEN BONACCI km2 (Cognard-Plancq et al., 2006a; 2006b). Te average catchment altitude is 870 m a. s. l. Te average annual air temperature of the catchment is 9,6 °C. Fig. 1: Location map of karst spring Fontaine de vaucluse. Te Fontaine de Vaucluse is typical ascending karst spring (Michelot & Mudry 1985; Blavoux et al., 1991/1992; 1992a). Its limestone channel ranges in diameter from 8 to 30 m (Mudry & Puig, 1991). Te lowest depth reached by diver was -308 m below the gauging station datum of 84.45 m a. s. l. Tis depth is still not at the bottom of the ascending karst channel. Te maximum water level measured at the gauging station was 24.10 m above the datum, the minimum was a few centimetres below the datum. Te rate of the maximum discharge of the spring has never been pre-cisely measured, but it is estimated that it cannot exceed 100 m3/s (Blavoux et al., 1991/1992; 1992a). Cognard-Plancq et al. (2006b) state that maximum spring discharge varies between 100 and 120 m3/s. Tis surmise identifes a karst spring with limited discharge capacity (Bonacci 2001). Te historical minimum discharge is 3.7 m3/s (Blavoux et al., 1991/1992). Every karst aquifer has complex hydrodynamic be-haviour. Te Fontaine de Vaucluse karst system respons-es to rainfall quite rapid in comparison with the large recharge area. Te peak of hydrograph occurred 24 to 72 hours afer the rainfall events. Te spring water level and discharge recessions are slow, which can be explained by the existence of a large storage capacity of the aquifer (Cognard-Plancq 2006b). Te primary objective of the investigation was to defne sub-periods with diferent hydrological behav-iour of the Fontaine de Vaucluse karst spring during 127 years period (1878-2004), analysing time series of mean annual spring discharges. It should be the frst step in explanation of this extremely important and interesting phenomenon. ANALySIS OF CATCHMENT ANNUAL RAINFALL TIME SERIES Te climate in the catchment is Mediterranean. Rainfall distribution over the year as well as over the large spring catchment is irregular. Intensive and signifcant rainfall events occurred during autumn and spring, while summer and winter are generally dry. Interannual fuctua-tions of rainfall on the catchment are very high. In order to defne an historical homogeneous catch-ment rainfall database Cognard-Plancq et al., (2006b) used six rainfall gauging stations. Te mean elevation of these stations is 445 m a. s. l., while the mean elevation of the spring catchment is 870 m a. s. l. Transformation of the measured monthly rainfall to the altitude of 870 m a. s. l. was made. Te average annual catchment rainfall in the 1878-2004 period is 1096 mm, while the minimum and maximum observed values were 641 mm (1953) and 1740 mm (1977) respectively. Data series with linear trend line of the annual rainfall on the Fontaine de Vaucluse catchment for the period 1878-2004 are presented in Figure 2. Te increasing 152 TIME in KARST – 2007 trend of the catchment rainfall of 1.045 mm per year is not statistically signifcant but should not be neglected in further analyses. Fig. 2: time data series of annual rainfall P at the Fontaine de vaucluse catchment with trend line for the period 1878-2004. ANALySIS OF LONG-TERM (1878-2004) MEAN ANNUAL DISCHARGES OF THE KARST SPRING FONTAINE DE VAUCLUSE ANALISIS OF MEAN ANNUAL DISCHARGES TIME SERIES Data series with linear trend line of the mean annual spring discharges q for the period 1878-2004 are pre-sented in Figure 3. Te decreasing trend of the mean an-nual discharges of 0.0468 m3/s per year is not statistically signifcant. Te average annual catchment discharge in the 1878-2004 period was 23.3 m3/s, while the minimum and maximum observed values were 7.61 m3/s (1990) and 53.4 m3/s (1915) respectively. Fig. 3: time data series of mean annual discharges Q at the Fontaine de vaucluse karst spring with trend line for the period 1878-2004. It should be stressed that annual catchment rainfall during the same period has an increasing trend. In Figure 4 linear regression between the mean annual the Fontaine de Vaucluse discharges q and the Fontaine de Vaucluse catchment annual rainfall P is shown. Te linear correlation coefcient is only 0.713, which is rela-tively low. A special problem is that the regression line cut abscissa line at 222 mm of annual rainfall P, which is relatively low value. Explanation of so unusual rainfall-runof relationship can be found in fact that accuracy of discharges and rainfalls are not very high, and maybe the value of catchment area of 1130 km2 is not precisely defned. It should be stressed that determination of exact catchment area in karst is one of the greatest and very ofen unsolved problems. Tis may be the case with the catchment of the Fontaine de Vaucluse spring. Te weak relationship between runof and rainfall means that some others factors (probably: air temperature, groundwater level, interannual rainfall distribution, changes of catchment area during the time, preceding soil wetness, anthropological infuences, climate change etc) have infuence on it. A time series analysis can detect and quantify trends and fuctuations in records. In this paper the Rescaled Adjusted Partial Sums (RAPS) method (Garbrecht & Fernandez 1994) was used for this purpose. A visualisation approach based on the RAPS overcomes small systematic changes in records and variability of the data values themselves. Te RAPS visualisation highlights trends, shifs, data clustering, irregular fuctuations, and periodicities in the record (Garbrecht & Fernandez 1994). It should be stressed that the RAPS method is not without shortcomings. Te values of RAPS are defned by equation: t-l ^Y where Y is sample mean; is standard deviation; n is number of values in the time series; (k=1, 2…,n) is counter limit of the current summation. Te plot of the RAPS versus time is the visualisation of the trends and fuctuations of yt. Time data series of Rescaled Adjusted Partial Sums (RAPS) for mean annual spring discharges in the period 1878-2004 are given in Figure 5. Terefore, the total data Fig. 5: time data series of the Rescaled Adjusted Partial Sums Fig. 4: Linear regression between the mean annual the Fontaine (RAPS) for mean annual discharges Q for the period 1878-2004 de vaucluse discharges Q and annual the Fontaine de vaucluse with designated next fve sub-periods: 1) 1878-1910; 2) 1911-catchment rainfall P. 1941; 3) 1942-1959; 4) 1960-1964; 5) 1965-2004. TIME in KARST – 2007 153 OGNJEN BONACCI series was divided into next fve subsets: 1) 1878-1910; 2) 1911-1941; 3) 1942-1959; 4) 1960-1964; 5) 1965-2004. Cognard-Plancq et al., (2006a; 2006b) defned the same fve stationary sub-periods using diferent methodology. Five time data sub-series of the Fontaine de Vau-cluse karst spring mean annual discharges q with trend lines for fve defned sub-periods are shown in Figure 6. In order to investigate statistically signifcant diferences between the averages of fve time sub-series for q and P the t-test was used. Te neighbouring averages of dis-charges for all fve sub-series are statistically signifcant at Fig. 6: Five time data sub-series of the Fontaine de vaucluse karst spring mean annual discharges Q with trend line for fve defned sub-periods. Te rescaled adjusted partial sums (RAPS) method es-tablished existence of next fve statistical, and hydrologi-cal signifcant diferent time sub-series: 1) 1878-1910: 2) 1911-1941; 3) 1942-1959: 4) 1960-1964; 5) 1965-2004. Variations in the Fontaine de Vaucluse karst spring hy-drological regime during relatively short period of 127 years are very strong and cannot be neglected. Anthro-pogenic impacts are probably the main cause of such be-haviour of the mean annual spring discharges time series analysed, but the natural pattern of drought and wet years is also possible. Land-use changes and overexploitation of surface water and groundwater at the spring catch-ment on hydrological regime of the Fontaine de Vaucluse spring certainly exists. Teir exact quantifcation during analysed period is extremely questionable due to miss-ing of many parameters. Strict division of natural and anthropogenic infuences on the hydrological regime is hardly possible. The significant changes of spring discharge characteristics during 127 years long period (1878-2004) can be caused by natural climatic variations, Fig. 7: Linear regressions between mean annual discharges Q and annual catchment rainfall P defned for fve diferent sub-periods. the 5 % and even more 1 %. At the same time the neigh-bouring sub-series averages of the catchment rainfall are not statistically signifcant. Figure 7 shows fve linear regressions between mean annual discharges q and annual catchment rainfall P de-fned for fve diferent sub-periods. It can be seen that linear correlation coefcients for all sub-series, except for third (1942-1959) and fourth (1960-1964) ones are higher than the linear correlation coefcient for whole time series. by anthropogenic influences, and possibly by climate changes. It is extremely hard, but at the same time extremely practically and theoretically important, to find correct and scientifically based explanation of this phenomenon. Cognard-Plancq et al., (2006a) consider that rain-fall-runof data have shown the large impact of clima-tologic variations on the hydrogeological system. Tey conclude that the underground storage zone is an impor-tant infuence on karst spring outfow, which depends on rainfall amount over 2 or 3 previous years. Investigations made in this paper do not confrm this statement. Correct answers on many questions dealing with changes in hydrological-hydrogeological regime of the Fontaine de Vaucluse karst spring cannot be done using only annual data. Some processes can be explained mea-suring and analysing climatologic, hydrologic, hydrogeo-logical and geochemical interactions in shorter as well as larger time increments. Te problem is that most of parameters required for these analyses were not monitored in the past. 154 TIME in KARST – 2007 ANALySIS OF LONG-TERM (1878-2004) MEAN ANNUAL DISCHARGES OF THE KARST SPRING FONTAINE DE VAUCLUSE More accurate and precise delineation and defni-tion of the Fontaine de Vaucluse spring catchment should be done. It is possible that its catchment area changes as a function of groundwater level. Tis means that ground-water level measurements in deep piezometers should be organized across the catchment. Te second task which should be considered in further analyses is detailed anal-ysis of infuence of rainfall distribution during the year on the spring runof. Tis can have very strong infuence on the relationship between rainfall and runof, especial-ly in karst areas. Te author thanks to Anne-Laure Cognard-Plancq and Christophe Emblanch from Laboratoire d’Hydrogéologie, Faculté des Sciences, Université d’Avignon et Pays de Blavoux, B., Mudry, J. & Puig, J.-M., 1991/1992: Bilan, fonctionnement et protection du systeme karstique de la Fontaine de Vaucluse (sud-est de la France). Geodinamica Acta, 5 (3), 153-172., Paris. Blavoux, B., Mudry, J. & Puig, J.-M., 1992a: Te karst system of the Fontaine de Vaucluse (Southeastern France). Environ. Geol. water. Sci., 19 (3), 215-225. Blavoux, B., Mudry, J., & Puig, J.-M., 1992b: Role du con-texte geologique et climatique dans la genese et le fonctionnement du karst de Vaucluse (Sud-Est de la France). In: H Paine, w Back (eds.) Hydrogeology of Selected Karst Regions. IAH International Con-tributions to Hydrogeology, Vol. 13, 115-131. Bonacci, O., 2001: Analysis of the maximum discharge of karst springs. Hydrogeol. J., 9, 328-338. Cognard-Plancq, A.-L., Gévaudan, C. & Emblanch, C., 2006a: Apports conjoints de suivis climatologique et hydrochimique sur le rôle de fltre des aquiferes karstiques dans l’étude de la problématique de changement climatique; Application au systeme de la Fontaine de Vaucluse. Proceedings of the 8th Conference on Limestone Hydrogeology. Neucha-tel, Sep. 21-23, 2006, 67-70. It can be stated that main dilemmas about variations of mean annual discharges of the Fontaine de Vaucluse karst spring during 127 years long period have not been solved. Tey should be explained using number of dif-ferent procedures and climatic as well as other indica-tors, and performing further detailed measurements and analyses. Te paper presents the need for interdisciplin-ary analyses incorporating several approaches and tech-niques. For the sustainable development and the protec-tion of such valuable water resource it is very important to establish prerequisites for the defnition of a causes and consequences of its hydrological changes. Vaucluse 84000 Avignon, 33 Rue Louis Pasteur, France, which kindly provide me with data analysed in this paper. Cognard-Plancq, A.-L., Gévaudan, C., &, Emblanch, C., 2006b: Historical monthly rainfall-runof database on Fontaine de Vaucluse karst system: review and lessons. IIIéme Symposium International Sur le Karst „Groundwater in the Mediterranean Coun-tries“, Malaga, Spain. In: J J Duran, B Andreo, F y Carrasco (eds.) Karst, Cambio Climatico y Aguas Subterraneas. Publicaciones des Instituto Geological y Minero de Espana. Serie: Hidrogeologia y Aguas Subterrraneas, N°18: 465-475. Emblanch, C., Puig, J. M., Zuppi, G. M., Mudry, J., & Bla-voux, B., 1998: Comportement particulier lors des montées de crues dans les aquiferes karstiques, mise en évidence d’une double fracturation et/ou de cir-culation profonde: Example de la Fontaine de Vau-cluse. Ecologae Geol. Helv., 92: 251-257. Emblanch, C., Zuppi, G. M., Mudry, J., Blavoux, B. & Batitot, C., 2003: Carbon 13 of TDIC to quantify the role of the unsaturated zone: Te example of the Vaucluse karst systems (Southeastern France). J. of Hydrol., 279 (1-4): 262-274. ACKNOwLEDGEMENT REFERENCE TIME in KARST – 2007 155 OGNJEN BONACCI Garbrecht, J. & Fernandez, G. P. , 1994. Visualization of trends and fuctuations in climatic records. water Resources Bulletin, 30 (2): 297-306. Michelot, C. & Mudry, J., 1985: Remarques sur les exu-toires de l’aquifere karstique de la Fontaine de Vau-cluse. Karstologia, 6(2): 11-14. Mudry, J., Puig, J.-M., 1991: Le karst de la Fontaine de Vaucluse (Vaucluse, Alpes de Haute-Provence, Drôme). Karstologia, 18 (2): 29-38. 156 TIME in KARST – 2007