Water Air, and Soil Pollution 207 - CREAF - Universitat Autònoma de ...

Water Air Soil Pollut (2010) 207:123–138DOI 10.1007/s11270-009-0124-7Analysis of Decadal Time Series in Wet N Concentrationsat Five Rural Sites in NE SpainAnna Avila & Roberto Molowny-Horas &Benjamín S. Gimeno & Josep PeñuelasReceived: 12 January 2009 /Accepted: 3 June 2009 /Published online: 16 June 2009# Springer Science + Business Media B.V. 2009Abstract Nitrogen emissions have grown in Spainduring the last 15 years. As precipitation scavengesgases and aerosols from the atmosphere, an effect onrainwater concentrations can be expected. However,time-series studies on wet N concentrations in theIberian Peninsula are very scarce. This paper aims tofill this gap by analysing weekly rainfall N concentrationsat a set of rural sites in Catalonia (NE Spain)from 1995/1996 to 2007 and a forest site monitoredfrom 1983 to 2007. The sites encompass a range ofrural environments and climate conditions, from theinland pre-Pyrenees (Sort) to the Mediterranean coast(Begur) and from north (Sort and Begur) to central(Palautordera and La Castanya) and south Catalonia(La Senia). We found a 1-year cycle for concentrationsof NH 4 + and NO 3 − whereby higher valuesA. Avila (*) : R. Molowny-HorasCREAF (Center for Ecological Research and ForestryApplications), Universitat Autònoma de Barcelona,08193 Bellaterra, Spaine-mail: anna.avila@uab.esB. S. GimenoEcotoxicology of Air Pollution, CIEMAT (Ed. 70),Avenida Complutense nº22,28040 Madrid, SpainJ. PeñuelasUnitat d’Ecofisiologia i Canvi Global CSIC–CEAB–CREAF, CREAF (Center for Ecological Researchand Forestry Applications),Universitat Autònoma de Barcelona,08193 Bellaterra, Spainwere reached at the end of spring–early summer,except at the easternmost coastal site of Begur.Weekly NH 4 + concentrations decreased with time atall sites (except at La Senia) whilst NO 3 − concentrationsincreased at all sites during the same period.Rainfall SO 4 2− concentrations decreased with time atall sites. The opposite trends in NO 3 − and SO 42−concentrations determined a shift in the relative acidcontribution of those anions during the 12–13-yearperiod. To interpret the increasing trend, mean annualNO 3 − concentrations were regressed against NO 2Spanish emissions and to some indicators of localanthropogenic activity. The increase at Sort andPalautordera showed good correlation with localanthropogenic indicators. Wet inorganic N depositionranged between 4.2 and 6.7 kg ha −1 year −1 . Whenincluding estimates of dry deposition, total annualdeposition rose up to 10–20 kg ha −1 year −1 , valuesthat have been found to initiate adverse effects onMediterranean-type forest ecosystems.Keywords Nitrate . Ammonium . Inter-annual trends .Seasonal cycle . Rural areas . Mediterranean1 IntroductionNitrogen is a fundamental component of livingorganisms and is a limiting factor for primaryproduction in the biosphere. Humans have alteredthe global N cycle by emitting nitrogenous com-

Water Air Soil Pollut (2010) 207:123–138 1252 Study SitesPrecipitation samples were obtained in four stations ofthe Xarxa de Vigilància i Prevenció de la ContaminacióAtmosfèrica (XVPCA) of the Generalitat deCatalunya. The XVPCA sites were at Sort (So), Begur(Be), Santa Maria de Palautordera (Pa) and La Sènia(LS). The long-term biogeochemical study site of LaCastanya (LC) in the Montseny Mountains was usedas a reference station. Description of the sites isprovided in Table 1 and Fig. 1. Sampling wasperformed from 1996 to 2007 at So, Be and LS (atLS, no data from 1 January 2003 to 3 February 2005),from 1995 to 2007 at Pa and from 1983 to 2007 at LC(no data from 27 September 2000 to 3 April 2002).Weekly wet-only precipitation was collected at theXVPCA stations. At LC, weekly bulk precipitationwas collected from 1983 to 2000 and weekly wetonlydeposition from 2002 onwards.The Sort sampling site is located at the ObservatoriMeteorològic of Sort, in the outskirts of the town.Sort lies in the Pyrenean lower slopes, at 695 m asl,and has 2,264 inhabitants (2007). Recently, animportant tourism industry on snow and nature sportswas developed. In 2003, an industrial area was built ata distance of 300 m from the instrumentation cabin.The main wind direction at Sort is N–NW, with the airblowing from the Pyrenees. Sort can be representativeof a rural site close to wildland areas.The Begur sampling site is placed at the Centred’Estudis del Mar-Nereo, 2 km from the town nucleus.The site is 1.5 km from the Mediterranean Sea at analtitude of 198 m. Aleppo pine (Pinus halepensis Mill.)forests surround the sampling site, except at itssouthern end which is open to the sea. Begur is aresidential village, with 4,086 inhabitants (2007)mainly dedicated to tourism and commercial activities.The main wind direction is from W and NW andsecondarily from SE; therefore, although it is a coastalsite influenced by the sea breeze, the prevailing windsblow from the continental interior areas.La Sènia site was on the outskirts of the town untilJanuary 2003. In February 2005, the station wasmoved 1.5 km from the previous site to the NE of thetown facing open field. La Sènia is a commercialtown focused in the furniture industry, but farmingactivities have increased recently. The town populationwas 6,300 inhabitants in 2007. It is surroundedby olive groves and animal farms. Main winddirections are from the NW and secondarily from S.The sampling cabin in Santa Maria de Palautordera islocated in the Arboretum Park which is about 1 km apartfrom the town centre. Its population was 8,235 personsin 2007. The town has grown steadily in recent yearsTable 1 Study site characteristicsSo Be LS Pa LCCoordinates 1°7′51″ E 3°12′49″ E 0°16′51″ E 2°21′17″ E 2°21′58″ E42°24′24″ N 41°57′35″ N 40°38′3″ N 41°41′18″ N 41°46′39″ NAltitude (m) 685 198 365 215 720Distance to sea (km) 143 1.5 22 21 27Sampling period 1996–2007 1996–2007 1996–2007(no 2003–2004) b 1995–2007 1983–2007(no 2000–2001) cNumber of samples 252 253 277 375 690% increase population a 50.0 49.5 25.5 65.5 –% increase vehicles a 86.5 97.1 118 122 –% increase industrial soil a 25.0 −18.0 4.9 10.4 –% increase farming a −50.0 −65.0 97.5 −18.6 –So Sort, Be Begur, LS La Sènia, Pa Palautordera, LC La Castanyaa Percentage increase relative to year 1991, period 1991–2007 for population and vehicles, 1995–2002 for industrial soil (IDESCAT2008)b No data from 1 January 2003 to 3 February 2005c No data from 27 September 2000 to 3 April 2002

126 Water Air Soil Pollut (2010) 207:123–138Fig. 1 Mean annual precipitationmap and study sites,modified from Ninyerola etal. (2000)SoLCPaPrec. (mm)Prec. (m1400Be42˚130012001100100090080070041˚LS6005004001˚ 2˚ 3˚and has shifted from agriculture and forestry to theservices and industry activity. A major motorway thatruns from SW to NE in Catalonia is located 3.5 kmaway. Two mid-size cities (50,000–100,000 people) areat 15–20 km to the S–SW, and the Barcelona metropolitanarea (3,700,000 habitants) lies 30–40 km to the SW.Main wind direction is from E–NE, therefore from thesea and the valley corridor where most traffic and someindustries are established.La Castanya is placed in the Montseny mountainswhere long-term biogeochemical studies have beenperformed since the late 1970s. The site lies at a lineardistance of 12.5 km from Santa Maria de Palautorderain a valley amidst extensive holm oak (Quercus ilexL.) forests in the Montseny mountains. This site isquite sheltered from the surrounding lowland pollutantatmosphere. More information on precipitationand throughfall chemistry at this site and neighbouringlocations in the Montseny mountains have beenpublished elsewhere (Rodà et al. 1999). Main winddirection is from the N to NW due to topographicallocal features that shelter the site from main windpatterns from the S to SE.The climate in Catalonia is Mediterranean, withdry summer, although orographic intense storms canbe formed in mountain areas, particularly in theeasternmost extreme of the Pyrenees. Winter isusually dry because the Atlantic air masses have tocross the elevated peninsular highlands before reachingCatalonia. In the pre-litoral and litoral regions, therainiest season is autumn, due to the temperaturecontrast between the warm Mediterranean sea air andthe cold upper air masses. Spring can also be rainydue to the formation of cyclones that cross the IberianPeninsula from west to east. At that time of the year,an important African dust activity coinciding with thepassage of Atlantic fronts produces frequent red-rains(Escudero et al. 2005).These general climatic traits are furthermoremodulated by the topography, which in Catalonia isvery diverse ranging in altitudes from the sea to3,400 m in the Pyrenees. As a consequence, annualprecipitation ranges from 1,500 mm in the Pyreneesto 400 mm in the central depression (Fig 1).3 Sampling and Chemical AnalysesAt the XVPCA stations, precipitation was sampledwith a wet-only collector (MCV®, CPH-004, Spain)

Water Air Soil Pollut (2010) 207:123–138 127that opens at the beginning of the rain and closes 20 minafter its end. Precipitation is driven to a polyethylene10-L vessel and rain volume is registered. At LC,precipitation was sampled with four (1983–1993) or two(1993–2000) replicate funnel-type collectors consistingof a 19-cm-diameter polyethylene funnel connectedthrough a looping tygon tube which prevents evaporationto a 10-L polyethylene bottle. In 2002, an Andersensampler (ESM Andersen instruments, G78-1001) dry/wet collector was installed; only the wet samples will beconsidered here.At the XVPCA stations, samples were taken on aweekly basis by site personnel and frozen uponcollection. They were sent to the CREAF laboratorywith a monthly frequency. The LC site was visitedweekly and samples immediately retrieved to CREAF.In the laboratory, samples followed a protocolpublished elsewhere (Avila 1996; Avila and Rodà2002). Participation in the AQUACON project for“Acid Rain” (Mosello et al. 1998) provided an interlaboratorycheck for analytical quality. Data qualitywas also evaluated by (1) an ionic ratio (cation sum/anion sum) and (2) a conductivity ratio (measuredelectric conductivity/conductivity calculated from theconcentration of measured ions and their specificconductivity). A 20% allowance about the centralvalue (1.00) was given, and samples out of this rangewere re-analysed or discarded. Ammonium concentrationdata from 2005 were discarded due to storagecontamination. Sampled precipitation, compared tostandard methods of rain measurement, accounted for96% of annual rainfall at Be and So, 98% at Pa andLC and 98.5% at LS.Concentrations were weighted by volume to giveannual volume weighted mean concentrations(VWM), and deposition was obtained as the productof VWM concentrations by annual precipitation. Inthis study, the median pH values are given to indicatethe central value of the pH distribution.4 Statistical Analyses4.1 Periodogram AnalysisCyclic signals in the time series were explored withLomb–Scargle periodogram techniques (Scargle1982) applied to the concentration series. TheLomb–Scargle periodogram decomposes the originaltime series into a finite linear sum of sine and cosineterms. Unlike well-known fast Fourier algorithms,Lomb–Scargle techniques do not require data valuesto be equally spaced in time, which is here the casedue to frequent weeks without rain. The periodogramis an estimate of the spectral density of the signal,such that a strong periodic signal will show up aslocal maxima at its corresponding frequency. In thecase of the Lomb–Scargle algorithm, the periodogramP(ω) of a data set (X i , i=1,2,…,N) with mean X andvariance σ 2 is defined as:PðwÞ ¼ 12s 28>:P Ni¼1X i X cos wðt i tÞP Ni¼1cos 2 wðt itÞ2þP Ni¼1X i X sin w ðt i tÞP Ni¼1sin 2 w ðt itÞ29>=>;ð1Þwhere τ is defined as follows:Psin 2w t jjtanð2wtÞ ¼ P :cos 2w t jjð2ÞIn the Lomb–Scargle periodogram, the false alarmprobability FAP can be defined as the probability ofrandom noise producing a signal of equal value orlarger than an observed signal z. Here, we calculatethe power level z 0 above which a signal will be foundwith a probability FAP. When the periodogram iscalculated at M frequencies, we have:hiz 0 ¼ ln 1 ð1 FAPÞ 1=M : ð3ÞThere is a probability FAP that the signal from atleast one of those M frequencies will be found abovez 0 by chance alone (Scargle 1982).

128 Water Air Soil Pollut (2010) 207:123–138In order to investigate whether a periodical signalexisted in the precipitation chemistry series, weexplored the periodogram of the data sets from eachstation by computing P(ω) at a total of 100 frequenciesup to 0.1 rad −1 , corresponding to a minimumperiod of ~63 days.4.2 Multiple RegressionTemporal changes in precipitation chemistry datahave been frequently analysed with linear andpolynomial regressions (Butler and Likens 2001;Hedin et al. 1994; Lynch et al. 2000; Kelly et al.2002) or with the Seasonal Kendall trend test(Lehmann et al. 2005) for its advantage in jointlyincorporate seasonal and temporal trends (Hirsch andSlack 1984). Another approach is to include theeffects of rainfall amount along with the cyclic andtrend effects in a multiple regression scheme(Buishand et al. 1988), which is the method we haveused here.In this model, the natural logarithm of theconcentration time series, y i , can be described by: 2py i ¼ a 0 þ a cos365 i f þ b tþ c P i þ e ið4Þwhere a 0 is a constant, a is the coefficient of theperiodic term, φ stand for its phase in radians, b is thetrend coefficient, t is the number of days lapsed(counted from 1 January 1995 onwards), c is theregression coefficient for precipitation, P i stands forthe natural logarithm of the precipitation in millimetresand e i is an error term. The inclusion of acosine term with a 365-day period aims to explain anycyclic annual variation in the data. The trend term b×tis included in Eq. 4 to fit long-term constant drift inthe data set. Finally, the P i term is included to accountfor any remaining dependence between precipitationand concentration. The use of natural log transformationof concentrations improves the fit of theparametric model.The dependent variable y i in Eq. 4 depends nonlinearlyon the coefficient 7 , which in principleprecludes the use of linear regression techniques.However, by making use of the properties of the sineor cosine of the difference of two arguments, the nonlinearEq. 4 becomes: 2py i ¼ a 0 þ a cos365 tð5Þ 2pþ b sin365 t þ b t þ c P i þ e iwhich is then linear in the new coefficients α, β thatrelate to the old coefficients as follows:a ¼ a cosðÞfb ¼ a sinðÞ:fð6Þð7ÞThe statistical significance of the new coefficients wascomputed as in Buishand et al. (1988).The model was applied to wet-only concentrations forthe XVPCA sites from 1995/1996 to 2007. At LC, for thesake of homogeneity of procedure, series analysis wasperformed for 1983–2000 on bulk deposition data.The weekly variation in the natural logarithm of theprecipitation, P i , can be described by a regression modelsimilar to the one shown above (Walker et al. 2000):P i ¼ a 0 þ a cos2p365 i fþ b t þ e ið8Þwhere the coefficients are defined as before. The cosinesterm can be developed into cosine and sine term asabove so that all parameters are explicit and multiplelinear regression techniques can be used.In fitting all the linear regression models, weensured that the following main assumptions werenot violated for each regression (see Ostrom 1991).For linearity, plots of observed versus predictedvalues, as well as residual versus predicted values,were examined; for non-serial correlation, the Durbin–Watsonstatistics was computed and verified to bewithin the interval 1.5–2; for homoscedasticity, plotsof residual versus predicted values and residual valuesversus time were examined and finally, for normality,a normal probability plot of the residuals was drawnand checked. Log-transformed concentrations wereused to fulfil the normality assumption.Analysis of variance (ANOVA) seasonal analysiswas also performed. Seasons were defined as winterfrom January to March, spring from April to June,summer from July to September and autumn fromOctober to December.

Water Air Soil Pollut (2010) 207:123–138 129To examine annual mean N concentrations anddeposition, we have summed up the time seriesannually, which mathematically eliminates all periodicsignals with a frequency of 0.0172 rad −1 (i.e.1 year) period or its multiples. Therefore, VWMconcentrations were calculated annually for thecomplete data set for the common period 1995/1996–2007 and were used for regressions againstemissions and anthropogenic activity.5 Results and Discussion5.1 Cation and Anion Concentrations in PrecipitationPrecipitation chemistry at Catalonia has an alkalinecharacter as the sum of base cations plus NH 4+exceeds the sum of the acid anions (Table 2). MedianpHs were all above 6.0, except at LC, which is a morehumid site at the core of the Montseny massif(Table 2). Higher pH value in the recent period isconsistent with increasing alkalinity trends reportedby Avila (1996) and Avila and Rodà (2002).Begur, with an elevation of 179 m asl and1.5 km from the sea, presented a clear marineinfluence with Na + and Cl − mean concentrations of103 and 115 μeq L −1 , respectively, followed by Ca 2+(69 μeq L −1 ) and SO 4 2− (47 μeq L −1 ). Meanconcentrations for NH 4 + and NO 3 − were 21 and35 μeq L −1 , respectively (Table 2).La Sènia is not far from the Mediterranean Sea(22 km), but the Litoral range acts as a barrier fromthe sea, so it receives a moderate marine influence(Na and Cl concentrations of 34 and 40 μeq L −1 ,respectively). The rain chemistry at LS was dominatedby Ca 2+ (65 μeq L −1 ) and SO 4 2− (48 μeq L −1 ),followed by NH 4 + and NO 3 − (31 μeq L −1 both).Sort is located in the pre-Pyrenean region, at685 m asl and 140 km from the sea, so it presentedvery little marine influence (Na + and Cl − meanconcentrations of 9 and 10 μeq L −1 , respectively).On the other hand, it was influenced by the centralCatalan semiarid depression, as seen by high alkalinityand Ca 2+ concentrations of 54 and 89 μeq L −1 ,respectively. At Sort, NO 3 − was higher than NH 4 + (34versus 24 μeq L −1 ).Santa Maria de Palautordera and La Castanya areonly 12 km apart, but Pa is exposed to the highlyindustrialised Barcelona metropolitan area, whilst LCis in a valley which receives higher precipitation andis more protected from the surrounding pollutedenvironment (Rodrigo et al. 2003). Both sites are20–27 km from the Mediterranean Sea but they areseparated from it by the pre-litoral Catalan range, sothe marine influence is moderate (25–35 μeq L −1 forNa + and 32–43 μeq L −1 for Cl − ). The higher humaninfluence at Pa related to LC is seen in the higherNH 4 + ,NO 3 − and SO 4 + concentrations for the sameperiod 1995–2007 (Table 2). At both sites, NH 4 + andNO 3 − concentrations were similar (39 μeq L −1 at Paand 28–29 μeq L −1 at LC).These contrasting chemical signatures attest for thewide range of conditions recorded at the different sites.5.2 SeasonalityIn the Lomb–Scargle analysis, the only signal abovesignificance was the one corresponding to a period of365 days (ω=2π/365≈0.0172 rad −1 ). All sites showeda 1-year cycle for NH 4 + ,NO 3 − and SO 4 2− , except Be,Table 2 Precipitation volume-weighted mean concentrations (in microequivalents per litre) for major ionsSite Pr (mm year −1 ) (μeq L −1 ) pHNa + K + Ca 2+ Mg 2+ NH 4+NO 3−SO 42−Cl −AlkSo 653 8.4 3.3 89.2 8.2 26.4 35.5 40.6 9.8 54.2 6.36Be 532 103 5.5 69.1 28.7 21.2 35.1 47.0 115 30.6 6.06LS 539 34.4 3.7 64.5 12.4 31.1 30.6 48.0 40.4 39.4 6.53Pa 619 34.5 4.5 70.3 14.1 38.5 39.1 45.7 43.0 35.8 6.32LC 743 25.1 5.0 64.9 9.5 29.0 27.8 37.8 31.1 34.3 5.92Median pH is also shown. Sampling periods are as described in Table 1; for LC, the comparable period (1995–2007) is includedSo Sort, Be Begur, LS La Sènia, Pa Palautordera, LC La Castanya, Pr precipitation amount, Alk alkalinity

130 Water Air Soil Pollut (2010) 207:123–138where only the signal for NO −3 was significant(significance at p=0.01 level).As it is broadly known, concentrations stronglydepend on precipitation amounts in a negative log–logrelationship due to dilution effects (Prado-Fiedler 1990).Therefore, to be able to interpret the variation ofconcentrations, the effect of precipitation must beremoved. In our analysis, this is done by incorporatingthe term cP i in Eq. 4. However, one might be interestedin further analysing the behaviour of precipitation. Thishas been gained here from the study of the multi-linearmodel applied to weekly precipitation data (Eq. 8),whose results are shown in Table 3. The multipleregression model allows to determine when themaximum seasonal cycle is located (i max , in Table 3).A significant seasonal cycle can be seen at Be, So andPa, with maximum precipitation in August–September.This precipitation seasonality agrees with typicalpatterns described for Catalonia, where maximumprecipitation in the pre-Pyrenees and mountainousareas occurs during the end of summer. This isoriginated from convection storms due to the highamounts of water vapour stored from summer evaporation(Martin Vide 1999). At littoral sites, maximumprecipitation usually occurs during autumn (MartinVide 1999) and in accordance, maximum precipitationat Be was found in September.The coefficient b indicates the temporal trend. Itwas not significant at the study sites, except for Pa(Table 3). The non-significance of an inter-annualprecipitation trend gives support to the assumptionthat other factors must be involved in the variation ofconcentrations. Because of the significant decreasingTable 3 Regression results for weekly precipitation at eachsite, calculated from Eq. 8Site adjR 2 a σ a Φ (deg) σ Φ (deg) i max bSo 0.042 0.17* 0.056 86** 19.1 9 -0.019Be 0.052 0.39* 0.107 67** 16.6 9 0.0013LS 0.026 0.32* 0.128 30 20.5 12 0.010Pa 0.036 0.24* 0.105 280** 24.6 8 -0.051*LC 0.003 0.073 0.07 22 51.9 3 0.0069σ a and σ Φ denote, respectively, the standard deviation of thecoefficient or the phase of the cosines term in Eq. 8. i max refersto the month in which the cosines term has its seasonalmaximum. See “Appendix” for description of b*p

Water Air Soil Pollut (2010) 207:123–138 131Table 4 Regression resultsfor each site and for thechemical componentsstudiedColumns are labelledaccording to Eq. 4. σ a andσ Φ denote, respectively, thestandard deviation of thecoefficient or the phase ofthe cosines term in Eq. 4.i max refers to the month inwhich the cosines term hasits seasonal maximum. See“Appendix” for descriptionof bSo Sort, Be Begur, LS LaSènia, Pa Palautordera, LCLa Castanya*p

132 Water Air Soil Pollut (2010) 207:123–138Although the farming activity has increased inCatalonia as a whole (IDESCAT 2008) and so havedone NH 3 emissions (Mulligan et al. 2006), at a locallevel, such increases were not observed in NH + 4 rainconcentrations. This is probably related to the shortresidence time of NH 3 gas in the atmosphere and itshigh dry deposition velocity (Warneck 1988), whichimplies that local influences predominate, which inthis case showed decreasing trends.The variation of NO − 3 concentrations was positiveat all sites (p Pa > Be > LS, with percent changerates from 11.7% at So to 2.4% at LS. The increaserate of Pa (4.7%) can be partly related to a decrease inprecipitation amounts. At LC, for the period 1983–2000, NO − 3 also increased, though with a lower slope(percent change rate 1.7%). This is consistent with itsmore sheltered position from major local pollutionsources and would represent changes in backgroundNO − 3 concentrations.SO 2− 4 tended to decrease, except for LS and Be(Table 4). A general sulphate decline in rainwatercan be attributed to regional and European airpollution control policies, which have been moresuccessful for SO 2 than for nitrogenous compounds.At So, LC and Pa percent change rate decrease wasof −3.6%, −3.2% and −2.5%, respectively (Table 4).Because of these SO 2− 4 decrease and NO − 3 increase,the ratio NO − 3 /SO 2−4 increased steadily at theCatalan sites, indicating the increasing contributionof NO − 3 to the acidic charge. In Fig. 2, the seasonalvariation, the temporal trend and the fitted model forNH + 4 , NO − 3 , SO 2− 4 and the ratio NO − 3 /SO 2− 4 aregiven for the background LC site.5.4 Relationship Between Rain Concentrationsand Anthropogenic ActivityThe relationship between N rain concentrations andemissions were explored by regressing annual volumeweighted means of NO 3 − and NH 4 + against Spanishnational annual emissions reported by EMEP (2007).There was a good relationship (r=0.62; p=0.002;Fig. 3) between the increase in rain NO 3 − concentrationsat LC and the Spanish national NO 2 emissions.The LC site is located in the middle of theMontseny massif and it is separated from the pollutedatmosphere around Barcelona by the elevated ridgesof La Calma (1,350 m asl), so this relationship withnational emissions would represent the evolution ofnational background trends. The regressions of nitraterain concentrations versus NO 2 for Be and LS did notsignificantly differ from the LC one; however, for Paand So, they significantly differed showing muchhigher dispersion (Fig. 3). At Pa and So, we exploredwhether the great excess in NO −3 concentrationsrespect to the background levels was also related tolocal emissions. Changes at a more local scale can beinvoked at these two sites, since they have experiencedimportant increases in population, number ofvehicles and industry, as shown in Table 1. BecauseNO x emission data are lacking at municipal andcounty levels, population numbers, traffic and industrialactivity were used as indicators of generalanthropogenic emissions. When excess NO − 3 concentrationsabove background levels (as represented bythe LC trend) were regressed against local activity atSo and Pa, good correlations were found with thenumber of inhabitants (r=0.87, p

Water Air Soil Pollut (2010) 207:123–138 13365Ln NH4 +4321065Ln NO3 -4321065Ln SO4 2-432102Ln NO3 - /SO4 2-10-1-2-31983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001YearsFig. 2 Observed (open dots), fitted model (black line) and time trend (dashed line) for logarithmic bulk concentrations of NH 4 + ,NO 3 − ,SO 4 2− and NO 3 − /SO 4 2− at La Castanya. Period 1983–2000nitrogen to build up in the soils. This store of nitrogenwould be little available for cycling in the forestecosystems, as the beginning of the wet season inautumn will flush very soluble nitrate forms out of thesoils, therefore depleting the ecosystems of nitrate. Insouthern California, Fenn and Bytnerowicz (1997)and Meixner and Fenn (2004) reported the importanceof the first big rain events in autumn for washing offinorganic nitrogen deposited during summer out ofthe soil profile.

134 Water Air Soil Pollut (2010) 207:123–138aNO 3rain concentrations (microeq L -1 )bNO 3rain concentrations (microeq L -1 )10080604020100LCBeLSBe: n.sLS: n.sLC: r=0.6201000 1100 1200 1300 1400 1500 160080604020LCSoPaLC: r=0.62Spanish NO 2emissions (Gg)Pa: r=0.69So: r=0.7101000 1100 1200 1300 1400 1500 1600Spanish NO 2emissions (Gg)Fig. 3 a Annual mean nitrate rainwater concentrations at LC(open circles), Be (triangles) and LS (grey dots) plotted againstSpanish NO 2 emissions reported to EMEP (2007) for the period1983–2006 and b the same with concentrations for So(triangles) and Pa (grey circles). Regression are LC: y=−5.56(±8.37)+0.023 (±0.0064) x; r=0.62, p=0.002; Be: y=−6.86(±40.5)+0.032 (±0.028) x; r=0.34, p=0.28; LS: y=11.8(±37.9)+0.014 (±0.026) x; r=0.18, p=0.61On the other hand, atmospheric N deposition canhave important effects on the structure and function offorest ecosystems. Our data provide wet N depositionat five contrasted environments in Catalonia. Averagedfor the 12–13-year period, wet deposition rangedbetween 1.6 and 3.3 kg ha −1 year −1 for NH 4 -N andbetween 2.3 to 3.4 kg ha −1 year −1 for NO 3 -N(Table 5). Maximum NO 3 -N and NH 4 -N depositionwas recorded at Pa. N deposition was higher underthe form of nitrate than as NH 4 -N, except at LC. Totalwet inorganic N (t N =NH 4 -N plus NO 3 -N) depositionranged from 4.2 to 6.7 kg ha −1 year −1 (Table 5). Whenregressed against time, annual t N deposition did notshow any significant trend in the study period. Thisresulted from the opposing variation of annual NH 4 -Nand NO 3 -N deposition trends. For NH 4 -N, all sitesshowed decreasing trends though they were onlysignificant for So (r = −0.69, p

Water Air Soil Pollut (2010) 207:123–138 135NO 3residuals (microeq L -1 )806040200y = -0,02+0,11x; r=0,87SortNO 3residuals (microeq L -1 )3020100y = -47+0,014x; r=0,80Sort-201600 1800 2000 2200 2400 2600Number of inhabitants-102500 3000 3500 4000 4500 5000Industrial soil (m 2 )NO 3residuals (microeq L -1 )605040302010y = -65+0,013x; r=0,66Santa Maria PalautorderaNO 3residuals (microeq L -1 )3020100y = -54+0,001x; r=0,84Santa Maria Palautordera04000 5000 6000 7000 8000 9000Number of inhabitants-105 10 4 6 10 4 7 10 4 8 10 4Industrial soil (m 2 )Fig. 4 Nitrate residuals respect to Spanish NO 2 emissions regressed against local anthropogenic activity indicators at So and Pa.Regressions and correlation coefficient (r) are also given6 ConclusionsStrong seasonalities in the concentrations of bothcompounds were found related with agriculturalpractices (i.e. NH 4 + ) or photochemistry (i.e. NO 3 − ).Decreasing trends of NH +4 rainfall concentrationswere found at most of the sites, whilst the oppositewas true regarding NO − 3 concentrations. These resultshave highlighted the influence of regional sources onNH + 4 and NO − 3 rain concentrations and trends. ToTable 5 Summary of N fluxes in wet, dry and total depositionSite NH + 4 -N wet NO − 3 -N wet Sum N wet NH + 4 -N Dry-est NO − 3 -N Dry-est Sum N Dry-est Sum N totalSo 2.41 3.25 5.66 4.23 5.71 9.94 15.6Be 1.58 2.61 4.19 2.77 4.58 7.35 11.5LS 2.35 2.31 4.66 4.13 4.06 8.19 12.9Pa 3.34 3.39 6.73 5.87 5.95 11.82 18.6LC 3.02 2.89 5.91 5.30 5.08 10.8 16.7Dry deposition was estimated with dry/wet ratios from Rodà et al. (2002). Fluxes in kilogramme per hectare per yearDry-est estimated dry deposition

136 Water Air Soil Pollut (2010) 207:123–138interpret the increasing nitrate trend, mean annualNO 3 − concentrations were regressed against NO 2Spanish emissions and to some indicators of localanthropogenic activity. The increase at Sort andPalautordera showed good correlation with localanthropogenic indicators. However, at Pa, the variationof nitrate concentration must be taken withcaution due to possible interaction with precipitation.Wet inorganic N deposition ranged between 4.2 and6.7 kg ha −1 year −1 . When including estimates of drydeposition, total annual deposition rose up to 10–20 kg ha −1 year −1 , values that have been found toinitiate adverse effects on Mediterranean-type forestecosystems. No annual trend for total N was founddue to the opposing directions of the trends ofoxidised nitrogen (increasing, though only significantlyat two sites) and reduced nitrogen (decreasing,though only significantly at one site). The Ndeposition levels experienced at rural sites in Cataloniaapproached N thresholds that have beenreported to induce adverse effects on N cycling ofMediterranean forest ecosystems. Effective air pollutionabatement policies should involve the combinationof continental, national, regional and localmeasurements. More attention must be given to themountainous areas around the Pyrenees.Acknowledgements We thank the finantial support from theSpanish Government (CGL2006-04025/BOS, CGL2005-07543-CLI and Consolider Montes CSD2008-00040 grants),CIEMAT-MARM project on “Critical loads and levels” and theCatalan Government (SGR2005-00312 grant) and Departamentde Medi Ambient i Habitatge grants. Roberto Molowny-Horasacknowledges the finantial support of the Spanish MICINN andthe European Social Fund through the Plan Nacional dePotenciación de Recursos Humanos. Thanks are due to fieldand laboratory personnel, especially Sonia Castillo and RebecaIzquierdo.AppendixThe use of logarithm of concentrations instead ofconcentrations in the regression approach aboveprecludes the direct identification of coefficient bas a trend term of the untransformed concentrationtime series. However, it is easy to show that theso-calculated coefficient b is, in fact, a goodapproximation to the annual change rate of theconcentrations. Let y the natural logarithm of theconcentration of a compound z such that y=Ln (z).Then, taking differences:dLnðzÞ ¼ dz z ¼ a d cos 2p365 t fþ b dt þ c dPwhere each parameter is defined as in Eq. 4 above. Theterm dz/z is the instantaneous change rate inthe concentration of compound z and is a function ofthe summation of three terms: a trigonometric functionwhich cancels if integrated in a 1-year interval, aconstant term b and a term that depends on theprecipitation. We can now readily identify:b dt ¼dz ztrendas the instantaneous change rate in the concentration ofcompound z due to the presence of a trend in the timeseries. Given that, in general, z»dz and that changes inconcentration due to the trend are very small during1 year, we can conclude that:b R1yeardt ¼ b ¼ R dzz trend 1 z $zwhere ∆z is the total change of the concentration in ayear and z is the yearly average of the concentration.Thus, coefficient b clearly represents the approximateannual change rate in concentration.ReferencesAber, J. D. (1992). Nitrogen cycling and nitrogen saturation intemperate forest ecosystems. Trends in Ecology andEvolution, 7, 220–223.Anatolaki, Ch., & Tsitouridou, R. (2007). Atmospheric depositionof nitrogen, sulphur and chloride in Thessaloniki,Greece. Atmospheric Research, 85, 413–428.Anderson, K. A., & Downing, J. A. (2006). Dry and wetatmospheric deposition of nitrogen, phosphorus andsilicon in an agricultural region. Water, Air, and SoilPollution, 176, 351–374.Avila, A. (1996). Time trends in the precipitation chemistry at amontane site in northeastern Spain for the period 1983–1994. Atmospheric Environment, 30, 1363–1373.Avila, A., & Rodà, F. (2002). Assessing decadal changes inrainwater alkalinity at a rural Mediterranean site in theMontseny Mountains (NE Spain). Atmospheric Environment,36, 2881–2890.

Water Air Soil Pollut (2010) 207:123–138 137BOE. (2008). www.boe.es/g/esBoixadera, J., Sió, J., Alamos, M., & Torres, E. (2000). Manualdel codi de bones pràctiques agràries: Nitrogen. DireccióGeneral de Producció Agrària i Innovació Animal.Departament d’Agricultura, Ramaderia i Pesca. Generalitatde Catalunya. p. 59.Buishand, T. A., Kempen, G. T., Frantzen, A. J., Reijnders, H.F. R., & van den Eshof, A. J. (1988). Trend and seasonalvariation of precipitation chemistry data in The Netherlands.Atmospheric Environment, 22, 339–348.Butler, T. J., & Likens, G. E. (2001). The impact of changingregional emissions on precipitation chemistry in theeastern United States. Atmospheric Environment, 25,305–315.Bytnerowicz, A., & Riechers, G. (1995). Nitrogenous air pollutantsin a mixed conifer stand of the western Sierra Nevada,California. Atmospheric Environment, 12, 1369–1377.Bytnerowicz, A., & Fenn, M. E. (1996). Nitrogen deposition inCalifornian forests: A review. Environmental Pollution,92, 127–146.Bytnerowicz, A., Percy, K., Riechers, G., Padgett, P., &Krywult, M. (1998). Nitric acid vapour effects on foresttrees—deposition and cuticular changes. Chemosphere,36, 697–702.Castillo, S. (2006). Impacto de las masas de aire africano sobrelos niveles y composición del material particulado atmosféricoen Canarias y el NE de la Península Ibérica. Ph.D.dissertation, Universitat Politècnica de Catalunya, p. 382.Dana, M. T., & Easter, R. C. (1987). Statistical summary andanalyses of event precipitation chemistry from the MAP3Snetwork, 1976–1983. Atmospheric Environment, 21,113–128.EEA. (2008). Annual European Community LRTAP Conventionemission inventory report 1990–2006. Submission toEMEP through the Executive Secretary of the UNECE.European Environment Agency, Technical report No 7,Copenhagen, Denmark.EMEP. (2007). Transboundary acidification, eutrophication andground level ozone in Europe in 2005. EMEP report 1/2007. Norwegian Meteorological Institute.Escarré, A., Carratalà, A., Avila, A., Bellot, J., Piñol, J., &Millán, M. (1999). Precipitation chemistry and air pollution.In F. Rodà, J. Retana, C. A. Gracia & J. Bellot (Eds.),Ecology of Mediterranean evergreen oak forests (pp. 195–208). Berlin: Springer.Escudero, M., Castillo, S., Querol, X., Avila, A., Alarcón, M.,Viana, M. M., et al. (2005). Wet and dry African dustepisodes over eastern Spain. Journal of GeophysicalResearch, 110, D18S08. doi:10.1029/2004JD004731.Fagerli, H., & Aas, W. (2008). Trends of nitrogen in air andprecipitation: Model results and observations at EMEPsites in Europe, 1980–2003. Environmental Pollution,154, 448–461.Fenn, M. E., & Bytnerowicz, A. (1997). Summer throughfalland winter deposition in the San Bernardino Mountains insouthern California. Atmospheric Environment, 31,673–683.Fenn, M. E., Jovan, S., Yuan, F., Geiser, L., Meixner, T., &Gimeno, B. S. (2008). Empirical and simulated criticalloads for nitrogen deposition in California mixed coniferforests. Environmental Pollution, 155, 492–511.Galloway, J. N., & Cowling, E. B. (2002). Reactivenitrogen and the world: 200 years of change. Ambio,31, 64–71.Galloway, J. N., Townsend, A. R., Erisman, J. W., Bekunda,M., Cai, Z., Freney, J. R., et al. (2008). Transformation ofthe nitrogen cycle: Recent trends, questions, and potentialsolutions. Science, 320, 889–892.Gruber, N., & Galloway, J. N. (2008). An Earth-systemperspective of the global nitrogen cycle. Nature, 451,293–296.Hedin, L. O., Granat, L., Likens, G. E., Buishand, T. A.,Galloway, J. N., Butler, T. J., et al. (1994). Steep declinesin atmospheric base cations in regions of Europe andNorth America. Nature, 367, 351–354.Hirsch, R. M., & Slack, J. R. (1984). A nonparametrical trendtest for seasonal data with serial dependence. WaterResources Research, 20, 727–732.IDESCAT. (2008). www.idescat.netKelly, V. R., Lovett, G. M., Weathers, K. C., & Likens, G. E.(2002). Trends in atmospheric concentration and depositioncompared to regional and local pollutant emissions ata rural site in southeastern New York, USA. AtmosphericEnvironment, 36, 1569–1575.Konovalov, I. B., Beekmann, M., Burrows, J. P., & Richter, A.(2008). Satellite measurement based estimates of decadalchanges in European nitrogen oxides emissions. AtmosphericChemistry and Physics, 8, 2623–2641. www.atmos-chem-phys.net/8/2623/2008/.Lehmann, C. M. B., Bowersox, C. van, & Larson, S. M.(2005). Spatial and temporal trends of precipitationchemistry in the United States, 1985-2002. EnvironmentalPollution, 135, 347–361.Lynch, J. A., van Bowersox, C., & Grimm, J. W. (2000).Changes in sulfate deposition in eastern USA followingimplementation of phase I of title IV of the Clean Air ActAmendments of 1990. Atmospheric Environment, 34,1665–1680.Martin Vide, J. (1999). Els climes del present. Lapluviometria catalana. In J. Calbó, J. Colomer & T.Serra (Eds.), El clima Local. Quaderns 21 (pp. 17–23).Banyoles: CECB.Meixner, T., & Fenn, M. (2004). Biogeochemical budgets in aMediterranean catchment with high rates of atmospheric Ndeposition—importance of scale and temporal asynchrony.Biogeochemistry, 70, 331–356.Mosello, R., Bianchi, M., Brizzio, M. C., Geiss, H., Leydendecker,W., Marchetto, A., et al. (1998). Acid rainanalysis. Environment Institute. Joint Research Centre.EUR 19015 EN. 81pp.Mulligan, D., Bouraoui, F., Grizzetti B., Aloe, A., Dusart, J.(2006). An atlas of pan-European data for investigating thefate of agrochemicals in terrestrial ecosystems. Institute forEnvironment and Sustainability. European Communities,EUR 22334 EN, Luxembourg.Ninyerola, M., Pons, X., & Roure, J. M. (2000). A methodologicalapproach of climatological modelling of airtemperature and precipitation through GIS techniques.International Journal of Climatology, 20, 1823–1841.Ostrom, C. W., Jr. (1991). Time series analysis: Regressiontechniques (quantitative applications in the social sciencesser., no. 9). London: Sage.

138 Water Air Soil Pollut (2010) 207:123–138Padgett, P. E., Allen, E. B., Bytnerowicz, A., & Minnich, R. A.(1999). Changes in soil inorganic nitrogen as related toanthropogenic nitrogenous pollutants in Mediterraneantypeecosystems in southern California. AtmosphericEnvironment, 33, 769–781.Prado-Fiedler, R. (1990). On the relationship between precipitationamount and wet deposition of nitrate and ammonium.Atmospheric Environment, 24, 3061–3065.Puig, R., Avila, A., & Soler, A. (2008). Sulphur isotopes astracers of the influence of a coal-fired power plant on aScots pine forest in Catalonia (NE Spain). AtmosphericEnvironment, 42, 733–745.Querol, X., & 24 authors. (2006). Atmospheric particulatematter in Spain: Levels, composition and source origin (p.40). Madrid: Ministerio de Medio Ambiente.Rodà, F., Retana, J., Gracia, C. A., & Bellot, J. (1999). Ecologyof Mediterranean evergreen oak forests. Ecological studies,vol. 137. Berlin: Springer. 373pp.Rodà, F., Avila, A., & Rodrigo, A. (2002). Nitrogen depositionin Mediterranean forests. Environmental Pollution, 118,205–213.Rodrigo, A., Avila, A., & Rodà, F. (2003). The chemistry ofprecipitation, throughfall and stemflow in two holm oak(Quercus ilex L.) forests under a contrasted pollutionenvironment in NE Spain. Science of the Total Environment,305, 195–205.Scargle, J. D. (1982). Studies in astronomical time seriesanalysis. II. Statistical aspects of spectral analysis ofunevenly spaced data. Astrophysical Journal, 263, 835–853.Seto, S., Nakamura, A., Noguchi, I., Ohizumi, T., Fukazaki,N., Toyama, S., et al. (2002). Annual and seasonaltrends in chemical composition of precipitation in Japanduring 1989–1998. Atmospheric Environment, 26, 3505–3517.Stoddard, J. L. (1994). Long-term changes in watershedretention of nitrogen: its causes and aquatic consequences.In L. A. Baker (Ed.), Environmental chemistry of lakesand reservoirs. Advances in chemistry series, 237 (pp.223–284). Washington D.C.: American Chemical Society.Vitousek, P. M., Aber, J. D., Howarth, R. W., Likens, G. E.,Matson, P. A., Schindler, D. W., et al. (1997). Humanalteration of the global nitrogen cycle: Sources andconsequences. Ecological Applications, 7, 737–750.Walker, J. T., Aneja, V. P., & Dickey, D. A. (2000). Atmospherictransport and wet deposition of ammonium in NorthCarolina. Atmospheric Environment, 34, 3407–3418.Warneck, P. (1988). Chemistry of the Natural Atmosphere. SanDiego: Academic Press. 757 pp.Wolfe, A. H., & Patz, J. A. (2002). Reactive nitrogen andhuman health: Acute and long-term implications. Ambio,31, 120–125.