TOM VELDKAMP

QUATERNARY RIVER TERRACE FORMATION IN THE ALLIER BASIN, FRANCE. a reconstruction based on sand bulk geochemistry and 3-D modelling.

CENTRALE LANOBOUWCATALOGUS

0000 0456 8479, B1BU0THEEK LANDBOUWUNTVERSITEE WAGENINGEN

Promotor: Dr. S.B.Kroonenber g Hoogleraar in degeologi e enmineralogi e ,%Sj »\J^ oiM

A. Veldkamp QUATERNARY RIVERTERRAC E FORMATION INTH EALLIE R BASIN,FRANCE . a reconstruction based on sand bulk geochemistry and 3-D modelling.

Proefschrift Terverkrijgin gva n degraa dva n doctor ind e landbouw- en milieuwetenschappen opgeza gva n derecto rmagnificus , dr.H.C .va n der Plas, inhe topenbaa r te verdedigen opwoensda g 18decembe r 1991 desmiddag st ehal f tweei nd eAul a van deLandbouwuniversitei t teWageningen .

|50'. 55'' U ^7 si "Nietall etijdgenote nbewone ndezelfd etijd .He tverlede n verandertvoortdurend ,maa rslecht sweinige nbeseffe ndat. " "In mijn hele heelal heb ik nooit een onverbiddelijke onveranderlijkenatuurwe taangetroffen .Di theela lvertoon t uitsluitend veranderende verhoudingen die SOBS door een kort-levendebewustzij n alswette nworde ngezien. " De gestolen verslagen van Leto II Atreides, in: God emperoro fDune ,F .Herbert ,1981 .

Opd evoorkan tva ndi tproefschrif tstaa tee nkopi eva nd eoudst e kaartva nhe t studiegebied (Limagne).Dez ewer dgemaak t in156 0 door Florentin Simeoni.

Aanmij n ouders enMarg a NM0^lOl> '1^

STELLINGEN

Met bulkgeochemische karakterisering van fluviatiele sedimenten kunnen geochemische en fluviatiele processen kwantitatief en statistisch verantwoord worden gereconstrueerd. Dit Proefschrift Kroonenberg, S.B, Moura, M.L. & Jonker, A.T.J., 1988, Geochemistry of the sands of the Allier river terraces, French Massif Central, Geologie en Mijnbouw 66:297-311.

Geomorfologische processen kunnen beschrijvend worden gesimuleerdme tbehul pva n 'finitestat emodelling' . Dit Proefschrift

De aanwezigheid van vruchtbare landbouwgronden ('terre noire1 Gachon, 1963) in de Limagne is het indirekte gevolgva n deglaciatie gedurendehe t LaatPleniglaciaal . Dit Proefschrift Gachon, L. , 1963, Contribution a l'etude du Quaternaire recent de la Grande Limagne marnocalcaire: morphogenese et pedogenese, Annales agronomiques, vol 14, no hors serie 1.

Het alluviaal-stratigrafisch model van Bridge & Leeder (1979) zou realistischer zijn als het rekening hield met fluviatiele versnijding als gevolg van opheffing. Dit Proefschrift Bridge, J.S. & Leeder M.R., 1979, A simulation model of al­ luvial stratigraphy, Sedimentology, 26:617-644

Aangezien de kennis omtrent de Kwartaire dynamiek van het klimaat veel nauwkeuriger is dan die van de tektoniek, vormt de laatste de meest beperkende factor voor realis- tische modellering van terrasvorming. Dit Proefschrift

De uitgestrekte Keniaanse planatievlakken zijn niet gevormd door pediplanatie-processen zoals beschreven door L.C. King, (1962). King, L.C, 1962, The morphology of the earth. Oliver and Boyd, London.

Doordat er steeds meer bodemkundig onderzoek plaatsvindt zonder of met te weinig veldwerk, bestaat het gevaar dat bodemkundigen steeds meer moeite hebben om op verschil- lende schaalnivo's 4-D te denken. 8 Zowel de informatica als de statistiek zijn slechts werktuigen binnen de geowetenschappen, iets wat steeds meercollega' slijke nt evergeten .

9 Miall's omschrijving (1983) van de ideale fluviatiele specialist van de jaren tachtig als: "A Quaternary sedimentologist with experience in petroleum geology and river engineering, a passion for hydraulics, statistics and scuba diving, a more than passing interest in tec­ tonics and a lot of money for coring equipment, combined with a good knowledge of biostratigraphy, palaeomag- netism, geochemistry, petrology, pedology, palaeoclimatology and structural geology" is toch nog incompleet voor de jaren negentig door de opkomst van modellering als onderzoekstechniek. Hiall, A.D., 1983, Basin analysis of fluvial sediments. Spec. Pubis Int. Ass. Sediment. 6: 279-286.

10 Judo is de weg naar het doeltreffendst gebruik van zowel lichamelijke alsgeestelijk ekracht . Jigoro Kano

11 Op een geologische tijdschaal gezien is het menselijk gedrag en handelen net zovoorspelbaa r als dat van iedere andere levensvorm, omdathe tvoortbestaa n van de soort de belangrijkstedrijfvee ris .

12 Lood zal een nog groter milieuprobleem worden in Frankrijk zolang de Fransen hun uitgesproken voorkeur behoudenvoo rwijndrinken ,jage ne nautorijden .

13 Het is onwaarschijnlijk dat het huidige plateau van Gergovie ook daadwerkelijk de plaats is waar Julius Ceasar in5 2A Ddoo rd eGallier s isverslagen .

Stellingen behorendebi jhe tproefschrif tva nA .Veldkamp : Quaternary river terrace formation in the Allier basin, France: a reconstruction based on sand bulk geochemistry and 3-D modelling. VOORWOORD

Met dit proefschrift worden vier interessante jaren bijna vollediggewij daa nonderzoe kbekroond .Natuurlij kstaa ndi evie r jaarnie tlo sva nd evoorafgaand e 24jare ni nmij nleven .I kwi l dan ook allereerst mijn ouders bedanken voor nun stimulering gedurendemij nstudie .Me tnam emij nmiddelbar eschooltij dheef t hunhelaa smee rda nallee nvoldoenin ggeschonken .Vervolgen swi l ik al die mensen bedanken die een rol hebben gespeeld inmij n wetenschappelijkeontwikkelin ge nnie thieronde rworde ngenoemd . Heti ndi tproefschrif tbeschreve nonderzoe ki she tresultaa tva n eengezamelijk e inspanningwa tto tuitdrukkin g komt ind ename n vand eco-auteur sva nd eopgenome npublicatie s enmanuscripten . SalleKroonenber gwi l ikbedanke nvoo rd egrot evrijhei d die hij mij gaf bij de invulling en uitvoering van het onderzoek. Zijn aanstekelijk enthousiasme en optimisme heb ik altijd als stimulerend ervaren. Ikbe nMari oMour a zeererkentelij k wanthe twa shi jdi emi j heeft ingewerkt op het gebied van de zandgeochemie. Verder heb ik met veel plezier samen gewerkt met Toine Jongmans, Tom Feijtel,Igo rStaritsky ,A bJonker ,Pie tOosterom ,Alfre dStein , EdMeije re nNic ova nBreeme nbinne nhe tkade rva nhe tV Fprojec t Frankrijk (VF86-09 en VF89-71). Binnen dit project hebben vele studenten ind eLimagn ehu nveldwer kuitgevoer d ikwi lme tnam e PaulRomkens ,Loe sJansse n enMargo tSpreuwenber g bedankenvoo r hun bijdrage aan het geochemische onderzoek. Tijdens mijn veldwerkhe bnuttig eassistenti egeha dva nmij nbroe rE de nnie t te vergeten mijn ouders welke niet te beroerd waren om vele zandmonsters tezeven . Ton Engelsma wil ik bedanken voor het gedogen van mijn labatoriumaktiviteite n enBra mKuijpe r enJa nHuitin gvoo rhe t uitvoerenva nd eXR Fanalyse sva nmij n50 0zandmonsters .Ja nva n Doesburg was altijd bereid om van een onbekende mineraal een rontgendiffractogramt emaken . Inbe nJ .va nde r Plichte nHen k Heijnisva nd eRijksuniversitei tGroninge n ergerkentelij kvoo r het zorgdragenvoo rd eU C enTh/ Udateringen .Antj eWeit zheef t zowel de Zusammenfassung als enige tekeningen gemaakt. Piet Versteeg en andere medewerkers van de tekenkamer Biotechnion wordenhartelij kbedank tvoo rhe tmake n van devel e figuren in dit proefschrift. Verder wil ik mijn (ex)kamergenoten, medekoffiedrinkerse nander ecollega' sbedanke nvoo rd eprettig e en informelesfeer . StephanVermeule nbe ni kvee ldan kverschuldig dvoo rhe tmed e ontwikkelen van het eerste rivierterrassen model gedurende de eindfaseva nonz estudie .Waarbi jw ezee rplezierig ebegeleidin g hebbengeha dva n Prof.Dr .A.J .Udin k tenCat ee nProf .I rM.S . Elzasva n de vakgroep informatica. Gerda Lenselink en Meindert vande nBer g zijnvoo rmi j zeerplezierig e collega's diealtij d openstaa nvoo rdiepgaand evakinhoudelijk ediscussies . Jannick Legeat is thanked for his enthusiastic support and interest inm y research inth eLimagne . Het Sportinstituut Wageningen wil ik danken voor de mogelijkheden ommij n agressie op een gekontroleerde maniert e kunnenafreageren . De Vakgroep Bodemkunde en Geologie wordt bedankt voor het beschikbaar stellen van een stationeringsplaats, ondanks beginproblemenhe bi kmij nstationeringsperiod etoc hal spretti g ervaren. De gepresenteerde onderzoekingen werden gesteund door de Stichting Aardwetenschappelijk Onderzoek Nederland (AWON) met subsidie van de Nederlandse organisatie voor zuiver Wetenschappelijk Onderzoek (NWO), (Project no 751.358.009, Geochemie en sedimentpetrografie van rivierafzettingen in de LimagneslenkCentraa l Massief, Frankrijk). Totslotwi l ikMarg amij nvrou wbedanke nvoo rhe tregelmati g willen lezen,meedenke ne naanhore nva nmij nhersenspinsels . PUBLICATIONS

The chapters 2.2 through 4.2 were of will be published as separatepapers ,wit hsom emodifications .Thes epaper shav ebee n adapted and slightly revised to fitth e format of thisthesis . The followingpaper s formth ebasi so fthi sthesis : 2.2: Veldkamp, A. & Kroonenberg, S.B., (subm). The Late Quaternary terrace chronology of theAllie r (Limagne, France). Quaternaire. 2.3: Veldkamp, A., & Jongmans, A.G., (subm). Trachytic pumice clasts inMiddl e Pleistocene Allier terrace deposits (Limagne, France):A chronostratigraphical marker.Quaternaire . 3.1:Veldkamp ,A. ,& Kroonenber gS.B. , (inpress )Th eapplicatio n of bulk sand geochemistry in Quaternary research. A methodological study of the Allier and Dore terrace sands (Limagne, France). Applied Geochemistry. 3.2:Veldkamp ,A. , &Feijtel ,T.C. , (inpress ) Parent material controlled subsoil weathering inth eAllie r terraces (Limagne, France). Catena. 3.3: Veldkamp, A. & Staritsky, I.G., (in press) Spatial variability in fluvial terrace sand composition at. the Allier/Doreconfluenc e (Limagne,France) .Geomorphology . 3.4:Veldkamp ,A. , (subm)Climat econtrolle d sediment fluxesi n the Allier basin (Limagne,France ) during the Late Quaternary. Quaternary Research. 4.1: Veldkamp, A., & Vermeulen, S.E.J.W., 1989,Rive r terrace formation,modellin gan d3- Dgraphica lsimulation .Eart hSurfac e Processes and Landforms.Vo l 14,641-654 . 4.2: Veldkamp, A., (in press). A 3-D model of fluvial terrace developmenti nth eAllie rbasi n (Limagne,France) .Eart hSurfac e Processes andLandforms . SUMMARY

The research presented in this thesis is focussed on a quantitative reconstruction of the effects ofpas t environmen­ tal dynamics within a fluvial system. The study area is part of theAllie rbasi n (Limagne) inth eAuvergne ,Massi f central, France. The research was carried out in several stages. At first field work was carried out to determine the terrace stratigraphy and chronology inmor e detail.A new age estimate of the Fva (65 m above present river bed) is based on pumice clasts found in the terrace sediments. Younger terraces were dated with 14C and Th/U disequilibrium methods. Fx terrace sediments (15 and 10 m above present river bed) were mainly deposited during the Late Weichselian, while the Fwb terrace sediments (25m above present river bed) have most probably a Late Saalian age.Du e to these new age estimates a revision of the existing Allier terrace chronology is necessary. This new chronology shows a large time gap between the deposition of the Fv and Fwterrac esediments . Next, sands of various terrace units were collected and bulk geochemically measured with XRF. This bulk geochemical research allowed a statistically significant discrimination of different terrace levels.Th e processes which shape and shaped the actual sand geochemistry were successfully quantified. It was found that grain size has only a very limited effect on bulk geochemical variability while longitudinal sorting processes and weathering have a stronger impact on actual sediment composition. Although the effects of parent material controlled weathering in the Allier sands were successfully modelled, the older terrace sediments are unsuitable for paleoenvironmental reconstructions. Such a reconstruction was done for the Weichselian and Holocene terrace deposits at the Allier/Dore confluence. The sediment mixing behaviour of these rivers is estimated by calculating sediment mixing ratios. This reconstruction started with an investigation of spatial mixing effects of Allier and Dore sediments in time by means of mapping and geostatistics. Results suggest an environmental control over the spatial variability of sediment mixing at this confluence in time. The reconstructed relative sediment fluxes of the Allier in time show a good correspondence with known past environments. Relative Allier sediment fluxes seem mainly climate controlled whereby large fluvio-glacial fluxes at the end of a glacial played a dominant role in the Allier system. These large sediment fluxes in the Allier system caused a strong rise inth eAllie r riverbed level contributing toth edevelopmen t of lakebasin s (Marais) inGrand eLimagne . Further a large scale and long term model of terrace for­ mation was constructed using finite state modelling. This methodology allows the construction of a general 3-D terrace formation model containing aswel l quantitative as qualitative knowledge on fluvial systems. Finally, an adapted version for the Allier (LIMTER) is made incorporating all present know­ ledge on this system. LIMTER allows the formulation and evaluation of long term terrace formation scenarios for the Allier system. Simulation results suggest that terrace stratigraphy in the study area is mainly the result of the internal Allier dynamics and climatic change. Local tectonism caused the development of unpaired terraces while the general regional uplift played a dominant role in terrace formation andpreservatio n ingeneral . The terrace research aspresente d inthi sthesi s shows that it is well possible for any fluvial system to simulate the interaction climate/tectonism and fluvial dynamics. The for the Allier simulated dynamics,ne t sedimentation during and at the end of aGlacia l and netdissectio n during an Interglacial hasn ogenera l validity. SAMENVATTING

Het onderzoek dat in dit proefschrift wordt beschreven heeft als hoofddoel on een kwantitatieve reconstructie te maken van de gevolgen van veranderende milieu-omstandigheden gedurende het Kwartair in een riviersysteem. Het studiegebied ishe t Allierstroomgebied (deLimagne )gelege n ind eAuvergne , Massif Central teFrankrijk . Het onderzoek is gefaseerd uitgevoerd. Allereerst is er voornamelijk veldwerk uitgevoerd om de terrassenstratigrafie en hun ontstaansgeschiedenis beter uit te zoeken. Zo werd een nieuwe leeftijdsschatting voor het Fv-terras (65 m boven huidige rivier) verkregen door de vondst van een bepaald type puimsteen in de terrassedimenten. De jongere terrassen zijn radiometrisch gedateerd met behulp van de koolstof 14 methode en de Thorium/Uranium niet- evenwichtmethode. Zo weten we nu dat de Fx-terrassedimenten gelegen tussen de 15 en 10m boven de huidige rivier, vooral tijdens het Laat Weichelien zijn afgezet,terwij l het Fwb terras (25m boven dehuidig e rivier) zeer waarschijnlijk een Laat Saalien leeftijd heeft. Als gevolg van deze nieuwe dateringen werd een revisie van de bestaande Allier terrassenchronologie noodzakelijk. Deze nieuwe chronologie heeft een groot tijdshiaat tussen de afzet- tingva n deF ve nFw-terrassedimenten . Vervolgens zijn de zanden van deverschillend e terrassen en terraseenheden bemonsterd en geochemisch gemeten. Dit bulkgeochemisch onderzoek laat een statistisch significant onderscheid zien tussen de verschillende terrasnivo's. Boven- dien zijn hiermee de processen die de actuele terras- zandsamenstelling bepalen of bepaalden goed te kwantificeren. Het blijkt dat zandkorrelgrootte slechts een beperkte invloed heeft op de bulkgeochemische samenstelling terwijl stroomaf- waardse sortering en verwering na afzetting een groter effect hebben of hadden op de huidige zandsamenstelling. Ondanks dat de verwering van de Allier terrassedimenten goed gemodelleerd konworden , zijn de sedimentenva n de oudere verweerdeterras - sen ongeschikt om milieuomstandigheden uit het verleden te reconstrueren. Een dergelijke reconstructie is wel gedaan aan de Weichselien en Holocene sedimentenbi j de samenvloeiingva n Allier en Dore. Het sedimentmenggedracht van deze rivieren in het verleden is onderzocht door een zogenaamde sediment- mengratio te berekenen. Om deze reconstructie mogelijk te maken is eerst een onderzoek uitgevoerd naar de ruimtelijke variabiliteit van het sedimentmenggedrag met behulp van karteren en geostatistiek. De resultaten suggereren sterk dat ook de ruimtelijke variabiliteit van riviersedimenten gerelateerd is aan de rivierdynamiek. De gereconstrueerde relatieve sedimentfluxen van de Allier correleren erg sterk met milieuveranderingen in het verleden. Vooral klimaatsveranderingen lijken de sedimentstromen in de Allier sterk te hebben bepaald. Grote fluvio-glaciale sedimentstromen aan het eind van een lange koude periode (Glaciaal) zijn erg dominant geweest in het Allier systeem. Een indirekt gevolg van deze grote sedimentstromen ishe t ontstaan vanmeerbekken s in de Limagne door een sterke stijging in beddinghoogte als gevolgva nd egrot esedimentaanvoer . Vervolgens is een grootschalig lange termijn model gemaakt met behulp van 'finite state1 modelleren dat terrasvorming driedimensionaal simuleert. Door 'finite state1 modelleren kan een model worden gemaakt dat zowel beschrijvende als gemeten kennis bevat. Dit algemene model isvervolgen s voord e Allier aangepast (LIMTER). LIMTER is een conceptueel model voor het opstellen en evalueren van terrasvormings-scenario's binnen het Allier systeem. Simulatieresultaten laten zien dat de terrasstratigrafie in het studiegebied voornamelijk het gevolg is van interne fluviatiele dynamiek en klimatologische veranderingen in de tijd. Lokale tektoniek veroorzaakte het ontstaan van ongepaarde terrassen terwijl de algemene regionale opheffing een dominante rol heeft gespeeld bij de conservering vanterrasse n ind etijd . Het in dit proefschrift gepresenteerde terrasonderzoek laat zien dat het goed mogelijk is in een fluviatiel systeem de interaktie klimaat/tektoniek en fluviatiele dynamiek semi- kwantitatief te reconstrueren. De voor de Allier gesimuleerde dynamiek van netto sedimentatie gedurende en aan het eind van een glaciaal gevolgd door netto interglaciale versnijding heeftgee n algemene geldigheid. RESUME

L'etude decrite dans cette these a pour objectif principal d'etablirun e reconstitution quantitative desconsequence s dues auxchangement sclimatique sduran tl eQuaternair edan su nbassi n versant.L aregio netudie ees tl ebassi nversan td el'Allier ,l a Limagne, situee enAuvergne ,dan s leMassi fCentral . L'etudea et eeffectue ee nplusieur sphases .L apremier ephas e a eteprincipalemen t une phase de travail sur leterrain , pour determiner la stratigraphie des terrasses et la chronologie de leurformation .Un enouvell e estimation de1'ag ed e laterrass e Fv (65mau-dessu sd el arivier eactuelle )a p uetr eobtenu egrac e a la decouverte d'un certain type de pierre ponce dans les sediments de la terrasse. Les terrasses plus recentes ont ete datees par radiometrie a l'aided e lamethod e du Carbone 14 et decell ed udesequilibr e Thorium/Uranium. Nous savons ainsiqu e les sediments de terrasses Fx qui setrouven t entre 10 et 15m au-dessusd e larivier eactuell eon tet edepose s principalement vers la find e laglaciatio nd eWurm ,alor squ el aterrass eFw b (25m au-dessus de la riviere actuelle) date tres probablement de la fin de la glaciation du Riss. Cette nouvelle datation a necessairemententrain eun erevisio nd el achronologi eexistant e des terrasses de l'Allier. La nouvelle chronologie revele un grandintervall eentr ele sformation sde ssediment sde sterrasse s Fve tFw .

Aucour sd el adeuxiem ephase ,de sechantillon sde ssable sd e differentsgroupe sd eterrasse son tet epreleves ,e ton tensuit e etesoumi sa un eanalys egeochimique .Cett eanalys ed e1'ensembl e desechantillon smontr eun edifferenc estatistiqu esignificativ e entrele sdifferent sniveau xd eterrasses .E noutre ,ell eperme t de quantifier correctement lesprocede s qui determinent ou ont determine la composition actuelle des sables des terrasses.I I s'averequ el agrosseu rde sgrain sd esabl en' aqu'un einfluenc e limitee sur la composition geochimique de 1'ensemble des echantillons, alors que le tri et 1'alteration en aval qui se sontproduit s apres la deposition, ont ou avaient plusd'effe t surl acompositio n actuelle.S'i letai ttou t afai tpossibl ed e simulerl 1alterationde ssediment sde sterrasse sd el'Allier ,i l s'averequ e lessediment sd eterrasse sdegradee splu s anciennes ne sont pas adaptes a la reconstitution des circonstances environnementalesd upasse .Un etell ereconstitutio na toutefoi s ete executee pour les sediments de laGlaciatio n deWur m etd e l'Holocene deposes au confluent de l'Allier et de la Dore. Le schema du melange des sediments de ces deux rivieres dans le passe a ete etudie en calculant ce qu'on appelle le taux de melange des sediments. Pour pouvoir effectuer cette reconstitution, il a d'abord fallu realiser une etude de la variabilite spatiale du schema de melange des sediments en s1aidant de la cartographie et de la geostatistique. Les resultatssuggeren tqu el avariabilit espatial ede ssediment sde s riviereses tlie ea l adynamiqu ed el ariviere .L areconstitutio n descourant sd esediment srelatif sd el'Allie rmontr eun egrand e concordance avec les changements environnementaux du passe.I I semble que ce soient les changements climatiques qui aient principalementdetermin ele scourant sd esediment sdan sl'Allier . Dans le bassin de l'Allier, ce sont les grands courants de sediments fluvio-glaciaires de la fin d'une longue periode glaciairequ ion tet epredominants .Un econsequenc e indirected e cesgrand scourant sd esediment ses t1'apparitio nd edepression s lacustres en Limagne due a une forte elevation du lit de la riviere resultant de 1'importantappor td esediments .

Ensuite un modele a long terme et a grande echelle a ete elabore al'aid ed umodel e 'FiniteState 'pou r lasimulatio ne n trois dimensions de la formation des terrasses. Grace a cette methode,i la et epossibl ed ecree ru nmodel equ ipouvai tinclur e aussi bien les donnees descriptives que les donnees quantitatives. Ce modele general a ensuite ete adapte aux conditionsd el'Allie r (LIMTER).LIMTE Rperme tl aformulatio ne t 1'evaluation des scenarios de formation des terrasses de l'Allier. Les resultats de simulation montrent que la stratigraphie des terrasses de la region etudiee sont principalementl aconsequenc ed'un edynamiqu efluviatil eintern e etd echangement sclimatiques .L atectoniqu elocal ea occasionn e 1'apparition de terrasses non-couplees alors que le relevement regionalgenera la jou eu nrol edominan tdan sl aconservatio nde s terrasses atraver s lesages .

La presente etude montre qu'il est tout a fait possible de reconstituer de fagon semi-quantitative 1'interaction entre climat/tectonique et dynamique fluviatile dans un systeme fluviatile. La dynamique, simulee ici pour l'Allier, de la sedimentationnett ependan te ta l afi nd'un eperiod eglaciaire , suivie de la dissection nette interglaciaire n'est pas valable danstou s lescas . ZUSAMMENFASSUNG

Die mit dieser Promotionsschrift vorgestellte Untersuchung beschaftigt sich mit der quantitativen Rekonstruktion der Auswirkungen wahrend des Quartars veranderter Umweltbedingun- gen auf ein fluviales System. DasUntersuchungsgebie t istTei l des in der Auvergne gelegenen Allierbeckens (Limagne), Massif Central,Frankreich . Die Untersuchung wurde in verschiedenen Phasen durchgefiihrt . Zuerst erfolgten Gelandearbeiten, die eine detaillierte Bestimmung der Terassenstratigraphie und chronologie ermoglichten. Eine neue Alterseinstufung der Fva(Fv)-Terasse (65 m iiber dem heutigen Flussniveau gelegen) konnte aus dem Fund eines bestimmten Bimssteintypes in den Terassensedimenten abgeleitet werden. Jungere Terassen wurden radiometrisch mit Hilfe der 14C- und der Thorium/Uranium- Ungleichgewichtsmethode datiert. Es konnte nachgewiesen wer­ den, dass die Fx-Terassensedimente, die zwischen 10 und 15 m iiberde m heutigen Flussniveau liegen, hauptsachlich imVerlau f der spaten Weichselvereisung abgelagert wurden. Dagegen lassen sich die Fwb-Terassensedimente, die 25 m iiber dem heutigen Flussniveau liegen, mit grosster Wahrscheinlichkeit als spat saaleeiszeitlich datieren. Auf Grund dieser neuen Datierungen wurde eine Revision der bestehenden Allierterassenchronologie notwendig. Die neue Chronologie zeigt eine grossere Zeitdif- ferenz zwischen der Absetzung der Fv- und der Fw-Terassen- sedimente. Desweiteren wurden Sandproben verschiedener Terassen und Terasseneinheiten gesammelt und mittels Rontgenfluoreszen- sanalyse (RFA)geochemisc h untersucht. Die Kenntnis der gesam- ten geochemischen Zusammenstellung der Sedimente erlaubt eine statistisch signifikante Unterscheidung der verschiedenen Terassenniveaus. Ausserdem lassen sich hiermit die Prozesse, die die aktuelle Terassensandzusammenstellung bestimmen bzw. bestimmten, gut quantifizieren. Es zeigte sich, dass die Korngrosse der Sande nur begrenzten Einfluss auf deren gesamte geochemische Komposition hat, wogegen longitudinale Sor- tierungsprozesse und Verwitterung nach der Sedimentation grosseren Einfluss auf die heutige Sandzusammenstellung haben bzw. hatten. Trotzdem der Verwitterungsverlauf der Al- lierterassensedimente gut modelliert werden konnte, eignen sich die Sedimente der alteren, verwitterten Terassen nicht, um die Palaomilieubedingungen zu rekonstruieren. Eine derartige Rekonstruktion erfolgte fur die weichselzeitlichen und holozanen Terassenablagerungen des Allier/Dore Zusam- menflusses. Das in der Vergangenheit gezeigte Sedimentmis- chungsverhalten dieser Fliisse wurde mittels Schatzung der Sedimentmischungsverhaltnisse berechnet. Diese Rekonstruktion wurde durch eine Untersuchung der raumlichen Variabilitat der Sedimentmischungseffekte der Allier und der Dore im Zeitver- lauf ermoglicht. Dabei wurden Kartierungen und geostatistische Arbeitstechniken benutzt. Die Ergebnisse suggerieren, dass die raumliche Variabilitat der Sedimentmischung im Zeitverlauf am Zusammenfluss stark von der zeitlichen Flussdynamik abhangig ist. Die rekonstruierten relativen Sedimentstrome der Allier im Verlauf der Zeit korrespondieren gut mit bekannten Um- weltdynamikenen der Vergangenheit. Der relative Sedimentstrom in der Allier scheint stark von klimatischen Veranderungen bestimmt zu sein. Die am Ende langer Vereisungsperioden auftretenden, grossen fluvio-glazialen Sedimentstrome spielen eine dominante Rolle im Alliersystem. Eine indirekte Folge dieser grossen Sedimentstrome ist das Entstehen der Seenbecken (Marais) in der Limagne, die durch starke Steigung der Strom- betthohe infolge starker Sedimentanfuhrgebilde twerden . Weiterhin wurde ein grossmasstabiges Langzeitmodell der Terassenbildung mit Hilfe des 'finite state1 Modellierungsan- satzes erstellt. Diese Methode erlaubte es, die Terassenbil­ dung dreidimensional zu simulieren, wobei sowohl quantitative als auch qualitative Informationen iiber fluviale Systeme verwendet werden. Im Anschluss daran wurde eine angepasste Version fur das Alliersystem erstellt (LIMTER), dass alle verfiigbare Information iiber dieses spezielle, fluviatile System enthalt.Mi t Hilfe von LIMTER konnen furda sAlliersys ­ tem fur lange Zeitspannen Terassenbildungsscenarios formuliert und evaluiert werden. Die Ergebnisse der Simulationslaufe zeigen, dass die Terassenstratigraphie im Untersuchungsgebiet hauptsachlich das Produkt interner Dynamiken im fluvialen Alliersystem sowie klimatischer Veranderungen ist. Lokale Tektonik verursachte das Entstehen ungepaarter Terassen, wogegen allgemeine regionale Hebung eine dominante Rolle bei der generellen Erhaltung der Terassen bis in die rezente Zeit gespielthat. Die in dieser Promotionsschrift vorgestellte Untersuchung zeigt deutlich, dass es gut moglich ist, in einem fluviatilen System die Interaktion zwischen Klima und Tektonik einerseits und fluviatiler Dynamiken andererseits semiguantitativ zu rekonstruieren. Die fur das Alliersystem simulierte Dynamik, als deren Charakteristiken iiberwiegende Sedimentation wahrend und am Ende einer Vereisung und iiberwiegende Relief- verschneidung wahrend der Interglaziale definiert werden, hat keineAllgemeingiiltigkeit . CONTENTS

Page: Chapter 1 INTRODUCTION 1

Chapter 2 STUDY AREA AND THE ALLIER TERRACE CHRONOLOGY 5

2.1 Study area 5 2.1.1. The Allier and Dore basins 7 2.1.2. Fluvial terrace formation in the Allier basin 10

2.2 The Late Quaternary terrace chronology of the Allier. 12 2.2.1. Introduction 12 2.2.2. Materials and methods 13 2.2.3. Terrace litho-stratigraphy 13 2.2.4. A regional reconstruction of the Late Quaternary Allier dynamics. 23 2.2.5. Climatic and fluvial dynamics in the Allier basin. 26

2.3 Trachytic pumice clasts in Middle Pleistocene Allier terrace deposits A chrono-stratigraphical marker. 28 2.3.1. Introduction 28 2.3.2. A stratigraphic marker 31 Pumice 32 2.3.4. Correlation between pumice and dated Sancy eruptions 32 2.3.5. Paieoenvironment 35 2•3•6. Terrace chronology Va 36 2.3.7. Tectonic implications 37 2.3.8. Conclusions 38

Chapter 3 SAND BULK GEOCHEMISTRY 39

3.1 The application of bulk sand geochemistry in Quaternary research. A methodological study of the Allier and Dore terrace sands. 39 .1.1. Introduction 39 .1.2. Material and methods 40 .1.3. Results 44 ,1.4. Regression with grain size data 49 ,1.5. Regression modelling of weathering effects 51 ,1.6. Discussion 53 .1.7. Conclusions 58

3.2 Parent material controlled subsoil weathering in the Allier terraces. 60 3.2.1. Introduction 60 3.2.2. Materials and methods 61 3.2.3. Results and discussion 63 3.2.4. Conclusions 75 3.3 Spatial variability in fluvial terrace sand composition at the Allier/Dore confluence. 76 3.3.1. Introduction 76 Study Area 78 3.3.3. Material and methods 78 3.3.4. Results and discussion 80 3.3.5. Conclusions 88 3.4 Climate controlled sediment fluxes in the Allier basin during the Late Quaternary. 90 Introduction 90 Materials and methods 94 Results and discussion 95 Sediment mixing ratios at the Allier/Dore confluence 105 Sediment fluxes and climate 110 Methodological evaluation 111 Conclusions 112

Chapter 4 LONG TERM MODELLING OF RIVER TERRACE FORMATION. 113 4.1 River terrace formation, modelling and 3-D graphical simulation. 118 4.1.1. Introduction 118 4.1.2. Materialsan dmethod s 119 4.1.3. Modelconstructio n 120 4.1.4. Modeloperatio n 123 4.1.5. Results 127 4.1.6. Discussion and Conclusions 134

4.2 A 3-D model of fluvial terrace development in the Allier basin. 136 4.2.1. Introduction 136 4.2.2. Modelcharacteristic s 139 4.2.3. Modelinpu t 142 4.2.4. Modeloutpu t 145 4.2.5. Simulationresult s 145 4.2.6. Evaluationo fth esimulate dRanda nterrac esequenc e 152 4.2.7. Conclusions 155

Chapter 5 SYNTHESIS 157

REFERENCES 165

APPENDICES Appendix I Geochemicaldat aAllier ,Dor e Appendix II Programlistin gGraphica l3- Dmode l Allier (LTMTER). Chapter1 INTRODUCTION

Thegloba lenvironmen ti ssubjec tt oconstan tchanges .Latel y thesegloba l climatic changeshav ereceive dmuc hattentio n from themedia .Ther e isgrea t and growing concern about the global environment, such as the accelerating increase in global greenhouse gases (C02, CH4, CFC's etc) which may lead to a new global climate which will be warmer and more turbulent with associated sea level rise, etc. Other concerns are less well understood, irritantproblem s likeozon eholes ,increas egroun d levelu.v . radiation,aci drain ,al lo fwhic hma yhav eprofoun d influence onecosystem s andbiomas sproductivity . Inorde rt ounderstan d the futurew ehav et ostud yth epast . Awa yt ounderstan d potential impactso fsuc hgloba l changesi s tostud yanalogu esituation s inth epast .Whe non ei sintereste d inth e nearby future oneha s to study short term variations in thepas ta spropose ddurin gth elates tINQU Acongres si nBeijing . If one wants to get a more complete picture of the overall mechanisms ofth egloba lenvironmenta l changeson eha st ostud y paleoenvironmental dynamics on a geological time scale. Such overallstudie sca ngiv eimportan tclue sabou tth epossibl ehuma n impact onth eeart henvironment .

Thebes tstudie dpas tenvironment su pt ono war eth edee psea , the polar environment (ice cores) and the eolian continental situation (losssection s inChina) . Thesepas tenvironment s are stored in long continuous sediments which represent long time seriesu pt omor etha n2. 4millio nyear sago .Th edirec teffect s of these global changes on the fluvial system are still poorly understood.Th efluvia lsyste mi ndynami cequilibriu m isabl et o adjust itself to changes of external variables by changing its internalvariable slik echanne ldept han dwidth ,rive rroughness , mean velocity, channel form, and slope (Schumm, 1977;Dawso n & Gardiner, 1987). River terraces are an essential part of the fluvial system and are former abandoned floodplains which are founda selongate dplateau salon gvalle yslope sabov eth epresen t river bed. Their formation can be looked upon as a result of changesi nequilibrium ,cause db yvariatio ni nexterna lvariable s (Dury, 1970;Leger , 1983). Three driving forces behind terrace formationar efrequentl ycited :change si nclimate ,tectonis man d base level.Bot h climate and tectonism play a significant role in terrace formation as almost no terraced valley is known without any change inbot h factors during theQuaternary . Base level changes have only importance in fluvial reaches between actual coastsan dth eshel fedges . River terraces provide long, but fragmentary, continental recordso fchangin ggeo-environments .Fro ma morphometrica lpoin t of view terraces are quite simple features but from a sedimentological point of view they are very complex. Each terrace unit is made up of several stacked often incomplete sedimentary cycles, representing alternating depositional and erosional stages.Th eresultin gterrac estratigraph y providesa relativechronolog y towhic hothe rgeological ,geomorphologica l orpalaeohydrologica l eventsca nb erelate d (Dawson &Gardiner , 1987). Unfortunately the character of the fluvial record makes itimpossibl et ogiv ea continuou slon gter mregistratio no fth e paleoenvironment.A multi-disciplinar yapproac hi sthu snecessar y tostud yth epas tdynamic s ofa fluvial system. During the last ten years the number of palaeohydrological investigations increased considerably. Fluvial dynamics during theLat eQuaternar y havebee nstudie db y Starkel (1983), Dawson &Gardine r (1987),Gregor ye tal . (1987)etc .Thes estudie suse d terracesedimentology ,stratigraphy ,morphology ,an dradiometri c datings as research methodologies. During geochemical investigations of fluvial sands in France (Kroonenberg etal. , 1988) and The Netherlands (Moura & Kroonenberg, 1990) it was found that bulk geochemical sand composition can also serve as an excellent indicator ofbot h sedimentary processes and long- termchange si nsedimen tcompositio na sa resul to fclimati can d uplifthistory . Anotheraspec to ffluvia lsystem si stha tthe yar es ocomplex , anddevelo po nsuc hlon gtim espans ,tha tlaborator y experiments andrea lsyste mmeasurement s (e.g.Schum me tal. ,1977 )ca nonl y partly revealpar to fthei r functioning. Computer simulation is increasingly recognized as a novel way to understand the way geomorphicsystem swor k (e.g.Anderson ,1988 ). Especiall yproces s orientedgeomorphologist sten dt omode lthei rmeasure dprocesse s morean dmore .O nth eothe rhan d thenee d formodel sdescribin g lesswel lknow nan ddefine dsystem si sals ogrowing .A simulatio n purely focussed on riverterrac e formation was doneb y Boll et al. (1988). Their qualitative 2-D model is unfortunately unsuitable forquantitativ eapplications .

The research presented in this thesis is focussed on a quantitativereconstructio no fth eeffect so fpas tenvironmenta l dynamics within a fluvial system. At first, river terrace sedimentsar einvestigate dwit hbul ksan dgeochemistr yan dthes e quantitative results areuse d as inputs for amode l simulating river terrace formation in a fluvial system as a function of climatic changesan dvertica l crustalmovements .

Thestud yare ai sa par to fth eAllie rbasi n (Limagne)i nth e Auvergne, Massif Central, France. In this area many basic geomorphologicaldat awer ealread ycollecte dfo ra projec to fth e AgriculturalUniversit yWageninge n (VF86-90).Th estud yare awa s thought asver y suitable forthi s kind of research because the Allieran di ttributarie shav eman yterrace san drelativel ymuc h ofth eQuaternar yhistor y isknow ndu et oth eeventfu lvolcani c history.

The research was carried out in three stages.A stage with mainly field work to sample and determine the terrace stratigraphy and chronology inmor edetail .A secon d phasewit h analyzingan dprocessin gsan dbul kgeochemica ldata ,an da thir d and final stage dedicated to modelling the Allier Quaternary terrace formation.

This thesis contains eight paperswhic h are all revised and edited toon e format.Th eorigina lpaper sar eliste dbelow . Theorde ro fchapter si saccordin gt oth ethre ename dstages . Stage one. Inchapte r2 ,th estud yare ai sbriefl yintroduce d (2.1), followed by new observations and insights of theAllie r Quaternary terrace stratigraphy and chronology (2.2, 2.3). Stagetwo . Chapter3 ,describe sth eapplicatio no fbul ksan d geochemistry as a tool in Quaternary terrace research (3.1), followed by a model of the impact ofweatherin g in the Allier terracesand s (3.2). Fluvialdynamic san d sandgeochemistr y are linked in section 3.3 and 3.4 where the spatial, longitudinal andtim erelate ddynamic so fsedimen tfluxe si nth eAllie rdurin g the LateWeichselia nar e reconstructed. Stacrethree . Thecomplet eintegratio no fth ecurren tknowledg e on long term dynamics of fluvial systems in general and the Allier systemparticularl y ismad e inchapte r 4.I nsectio n 4.1 amodellin gmethodolog yan dth eevolvin gconceptua l3- Dmode lo n fluvial terrace formation are discussed. In section 4.2 an adaptedan dextende dmode lfo rth eAllie rsyste mi spresente dan d evaluated. Finally,i nchapte r5 a synthesi si smade .Th eusefulnes san d validity of the research methodologies are discussed and evaluated followed by aregiona l evaluation. chapter 2 STUDY AREA AND THE ALLIER TERRACE CHRONOLOGY

Thestud yare ai ssituate di nCentra lFranc ean dcomprise sth e Allierdrainag ebasin .Th eAllie rdrain sth eLimagn erif tvalle y and the surrounding Hercynian crystalline Massif Central. The Allierbasi nha sman ydifferen tterrac elevel switnessin g former flood plain levels. One of the earliest papers on the Allier terraces in the Limagne graben dates back to 1917 (E.Chaput) . Detailedinformatio no nal lpreviou sinvestigation so nth eAllie r terraces canb e found inJ- F Pastre'sthesi s (Pastre, 1987).

2.1 STUDYARE A

Mostinformatio ni nthi schapte ri sderive dfro mrevie wbooks . More details on the general geological setting canb e found in Autran& Peterlong o (1980)an di nJun g (1971). Th egeolog yo fth e studyare a isexcellentl ymappe da ta 1:50.00 0scal eb yth eBRG M (Bureau de Recherches Geologiques et Minieres). The following ratherdetaile dmap sserve da sbasi so fth efiel d investigations Vichy XXVI-29 , Maringues XXVI-30 , Thiers XXVI-31 , Clermont- Ferrand XXV-31. The general geological setting isshow n inFi g 2.1.1.an d thestudie d terraces are shown inFig .2.1.2 .

The oldest parts in theMassi f Central are Late Precambrian and Caledonian high grade metamorphic rocks as gneisses and granulites.Th emajorit y ofth emetamorphi c rocks inth eMassi f Central were metamorphosed during the Hercynian orogeny. The majority of the rocks are gneisses and granites, but also Palaeozoicsedimentar yrock san daci dvolcani crock sar elocall y found. Afterth eHercynia norogen yprolonge ddenudatio nseem st ohav e dominated. Some residual flint occurrences suggest Mesozoic sedimentationi nth eMassi fCentra lbu tmor esubstantia levidenc e is still lacking. During the Tertiary when the Alpine orogeny started,man y grabens opened up inth eMassi f central like the Rhone,Loir e and Limagnegraben . Figure2.1.1 . Generalgeologica lsettin go fth estud yare a

6 Fault line Metamorphic and intrusive rocks Paleozoic sediments Cenozoic volcanics Oligocene sediments EH Pliocene sediments Quaternary fluvial sediments Holocene fluvial sediments Holocene lake and marshsediment s

7 142 1 28 35K m

Thesedimentar y infillo fthes etectoni cdepression s started immediately. The Limagne rift valley which mainly developed during the Oligocene,wa s filled with marls, chalks,sand s and clays during the Oligocene and Miocene. Meanwhile volcanism started alongth eactiv e faulting zones fromMiocen e onwardst o recent times.Canta l volcanism reached itclima x between 9an d 6 million years ago,whil e theMon t Dore volcano had its most activephas ebetwee n 3an d 1millio nyear sago .Lat eQuaternar y and Holocene volcanism tookplac e inth e Chainede s Puys.

2.1.1. TheAllie r and Dorebasin s This study ismainl y focussed on the terraces of the Allier and Dore rivers near their confluence (Fig 2.1.2). The Dore, draining the crystalline Forez, is a major Allier tributary. Othermajo rtributarie sar eth eAllagnon ,drainin gth eCanta lan d Cezallier, and the Couze Pavin and Couze Chambon draining the Mont Doreregion . TheAllie rbasi ncover sa surfac eo f1431 0km 2an dha sa mea n annual discharge of 147m 3/s.A tth eAllie rDor econfluenc e the upstream Allier basin covers approximately 9000 km2 while the Dorebasi nsurfac ecomprise s125 0km 2. Themea nAllie rdischarg e atthi sconfluenc e is9 8m 3/san dth emea nDor edischarg e1 8m 3/s (Pastre, 1987). The Allier basin is underlain by both volcanic (22%) and crystallinebasemen t rocks (58%),th eDor ebasi n predominantly by crystalline basement (72%). Additionally, mainly basement- derived Oligocene sediments occur in both basins, 20%an d 28% respectively.

* V V * V * u

,\\\y,\;.\\y/Randanl ;Chqmev.;.y fdesyX;/".;:

\l'[-l\ Hercynian basement Volcanics Oligocene sediments terrace deposits

Figure2.1.2 . •Hiestudie dAllie ran dDor eterraces .

The two basins differ also in geological history. Whereas the higher parts of the Allier basin, notably the Cantal and Mont Dore volcanoes, were subjected to severe glaciation (Veyret, 1980; Kieffer, 1971), theDor ebasi nha sbee n largely freefro m glaciers,excep tfo ra ver ysmal lpar to fth eFore z (Etlichere t al. 1987). Both rivers have about eight main terrace levels, numbered fromZ (presen trive rbed )t oS (oldestterrac elevel) . Including thedifferen tsu blevel sa tleas t 14differen tlevel s (Z,ZY ,Y , YX, Xb,Xa ,Wb ,Wa ,Vb ,Va ,Ub ,Ua ,T ,S )ca nb edistinguished . TheAllie rterrac edeposit s aregravell y and sandy sediments poor in clay. The gravel composition reflects the different lithologies within the Allier basin. This composition is not linear correlatedwit hbasi n lithology,th evolcani c components usuallypredominat ewhil eth eOligocen e rocksar eusuall yrare . Common heavy minerals of theAllie r terraces are augite,gree n andbrow nhornblende ,olivine ,mica' san d opaques (VanDorsser , 1969; Rudel,1963 ;Pelletier ,1971 ;Pastr e1986 ;Tourenq , 1986). Theopaqu ecomponen twhic hca ncompris emor etha n50 %o fth efin e sand fraction, is predominantly composed of basaltic rock fragments (Kroonenberg et al., 1988). Larue (1977) studied the downstreamchange si ngrave lpetrolog yan dTouren q (1986)studie d changes inheav y mineral content of theAllie r terraces.Thei r resultsd ono tmatc hwel lan dar eno tcomparabl ea sthe ystudie d different specific fractions ofth eAllie rsediments . The heavy fraction of the Dore terrace sands (Van Dorsser, 1969; Pelletier, 1971;Tourenq , 1986;Va nWijck , 1985)consist s essentially of mica's,tourmaline , zircon, opaques and augite. Theoccurrenc e ofaugit e indicates somevolcani c influence,bu t asvolcani cbedroc k isabsen t inth eDor ebasin ,thes emineral s must originate fromwind-blow nashes .

alti ude(m )

350-

300

250JW 1km E

H terrace deposits

Figure2.1.3 . The schematical cross section of terraces at Randan. 2.1.2. Fluvial terrace formation inth eAllie r basin As this study focuses on the fluvial Allier dynamics during theQuaternar ya shor tovervie wo fth ecurren tknowledg ewil lb e presented. Two major investigations were carried out on the fluvial sediments ofth eAllie rterraces .Laru e (1979)mad eth e first thorough investigation on chrono-stratigraphy based on sediment compositionwhil e Pastre (1987) focussed on theheav y sand mineral assemblies in order to correlate and date the different terracedeposit sb yrelatin g sandmineralog y withth e mineralogyo fdate deruptions .Bot hLaru ean dPastr estudie dth e wholeAllie rbasin . Oneo fth emos tcomplet eterrac esequence si nth eAllie rbasi n isfoun dnea rRanda n (Cloziere tal. ,1980) . Thischronosequenc e displaysa tleas tte nterrac elevels ,show nschematicall y inth e cross-section of Fig. 2.1.3. The oldest terrace level in this sequence,T isthough tt ob eabou t 2millio nyear sol d (Pastre, 1987). Therear etw oexistin gterrac echronologie s (Fig.2.1.4. )on e establishedb yLaru e (1979)an don eb yPastr e (1986,1987) .Bot h authorsuse dindirec tway so fcorrelatin gterrace swit ha certai n age. AGE in Larue (19791 Pastre(1986 1 Kycsr"; BP |FX |10-15ml 100 . I |Fx(0-5m) 200 . |Fwd(10-15m) |Fw(25-35m) 300 . 1 |Fwc(25ml 400 .

|Fwb(25-30m> 500 . • |Fwa(30-35nO 600 . |Fv(60-70m)

700 _ | Figure2.1.4 . Thetw oexistin gterrac e to .

chronologies 900 _ |Fvb(60-70ml

1000 . |Fva(60-70m> 1100 _

10 Thetypica lsequenc eo fterrace spresen ti nth eAllie rvalle y shows that accumulation and vertical erosion alternated repeatedly during the general valley deepening. Very similar terrace sequences are found along the Rhine,Meus e and Thames (Vande n Berg, 1989;Va n Straaten, 1946;Brunnacke r &Boenigk , 1983; Andres,1989 ;M cGrego r &Green , 1978). Themechanis mo fterrac eformatio ni susuall ysough ti nthre e majorexterna lfactors ,climate ,tectonis man dbas elevel .A sth e exact terrace formation mechanism for the Allier terraces is unknown each factor isconsidered .

Climate Terracesedimen tcompositio nan dalteratio nindicat etha tmos t of theAllie r terrace deposits cumulated during and at theen d ofglacials .Th ealternation sbetwee ncumulatio nan dincisio nar e therefore ascribed primarily to climatic causes (Raynal,1984 ; Texier &Raynal ,1984 ;Bout ,1963 ;Kroonenber ge t al., 1988).

Tectonism The contribution of tectonism to terrace formation in the Allier basin is obvious from the tendency towards valley deepening throughoutth eQuaternary .Th etota l amounto fvalle y deepening at Randan since the Quaternary and consequently the probableamoun to fgradua luplif ti sabou t15 0m .Excep tgradua l neotectonicuplift ,whic hstil ltake splac e (Giote tal. ,1978) , some terrace sequences and longitudinal profiles suggest a tectonic faultcomponen t (Larue,1979 ;Gio t et al., 1978).

Base level Adirec to rindirec tinfluenc eo fse aleve lvariation so nth e erosionan dsedimentatio ni nth eAllie rbasi nappear sno tt ohav e been possible as there is an area of continuous Quaternary depositionbetwee nth estud yare aan dth ese ai nth ePari sbasi n and the lowerLoir ebasin . Ingenera l itca nb econclude dtha tclimat ean dtectonis mar e the main external factors which have determined Quaternary terrace formation inth eAllie rbasin .

11 2.2 THE LATEQUATERNAR Y TERRACE CHRONOLOGYO FTH EALLIER .

A.Veldkam p &S.B . Kroonenberg

Abstract The Late Quaternary terrace chronology of the Allier has been reconstructed by means of terrace litho-stratigraphy with Th/U disequilibrium and UC datlngs. The terrace level at 25 m above present riverbed (Wb) near Coudes has a Late Saallan age. The two Weichselian terrace levels Xa and Xb (20 and 10 m) have at least four different litho-stratigraphical units, a Middle Plenlglaclal, two Late Plenlglaclal and a Younger Dryas unit respectively. The oldest Holocene terrace sediments have Atlantlcum ages. The timing of Allier Incision and sedimentation during the Late Weichselian seems mainly climatic related. Major fluvloglaclal sediment fluxes from the melting glaciers on the Mt. Dore and Cantal at the end of the Late Plenlglaclal caused a strong rise of Allier riverbed level. This rise of approximately 20 m In the Limagne contributed to formation of lakes like Marals de Ravel and the Grand Marals.

2.2.1. Introduction Within lower Allier terraces major differences in sediment composition are known to occur (Kroonenberg et al., 1988; Larue, 1977). The major difference between the present riverbed and the Weichselian terrace (X) is the higher amount of basaltic fragments in the X terrace sands. This difference has partly a climatic origin as the Late Weichselian sediments are thought to have a fluvioglacial origin from melting glaciers in volcanic areas (Bout, 1963; Kroonenberg et al., 1988). A previous bulk geochemical study (Kroonenberg et al., 1988) of Allier sands showed that basaltic rock fragments determine the

Ti02, Fe203/ MgO, CaO and P205 content in the Allier sediments. Changes in bulk geochemical sand composition are well illustrated by CaO content, concentrated in the basaltic rock fragments, K20 content, concentrated in crystalline rock fragments, and Na20 concentrated in sodic plagioclase. CaO and K20 contents are virtually independent of grain size because their host minerals are found in almost all grain size fractions and Na20 is enriched in the finer sand fraction due to selective abrasion (Veldkamp, 1990).

In this chapter a new and more elaborated Late Quaternary

12 terrace (chemo)litho-stratigraphy ispresente d with new UC and U/Th datingsallowin g amor eprecis etim esettin g ofth eAllie r fluvial dynamics during theLat eQuaternary .

2.2.2. Materials andmethod s The various lower terrace levels have been thoroughly investigated by studying all available exposures along terrace scarpsan d ingravel/san d pits. Based onthes esedimentologica l fieldobservation san dmeasure dbul kgeochemica lsan dcompositio n a new litho-stratigraphy wasmade . Sampling methodology and laboratory treatments and measurements ofth esand swer edon eaccordin gt oKroonenber g et

al. (1988).Th ebul kelemen tconcentration so fSi0 2,Ti0 2, Fe203,

A1203,MnO ,MgO ,CaO ,Na 20,K 20an dP 205,wer emeasure dwit h X-ray fluorescencespectroscopy . Meanbul kgeochemica lcompositio no f each sampled sand layer is plotted for each described key section. Special attention was paid to lithological discontinuities associatedwit hpaleosol san dkryoturbate d sediments/paleosols. The litho-stratigraphy was elaborated at those sites where datable materials were found. Sediments were dated with 14C on organic rich samples, and by both UC and Th/U disequilibrium datings (Kroonenberg etal. ,i nprep )o ftravertines .Ther ear e threemai n groups of terraces, from old toyoung ,W , X and XY. Thedifferen tterrace slevel sar eindicate db ya na o rb (Wban d Xa) and the different litho-stratigraphical units by roman

figures (XIVan dWb :). Th estudie dW terrac ei sth eW bleve lwhic h is found at2 5m abov e theactua lAllier ,Th estudie d Xlevels , Xa and Xb,ar e found at 20an d 10m respectively, while the ZY level starts at 7m downt oth epresen tAllier .

2.2.3. Terrace litho-stratigraphy Wbterrac e sediments The ages of the W terrace levels (25 to 45 m above river level)ar estil luncertain .A stratigraphical ,palynologica lan d sedimentologicalstud ya tPont-du-Chatea uo fRayna le tal .(1981 ) madethe mconclud etha tth eW a (45m)terrac esediment swer emos t

13 probably deposited during the Cromerianglacial .The y observed reworked ash layers inth estudie dW asediments .Th emineralog y ofthes eas hlayer sstrongl yresemble sth echaracteristic so fa n ashlaye ruse da sa chronostratigraphi cmarke rb yDebar d& Pastr e (1988). Thi s green clinopyroxene rich ash was given an age of approximately 300,000year sB.P. ,thu sindicatin gtha ta younge r age ismor e likely. Near theAllier/Couz eChambo n confluence at Coudes the 25m terrace is well exposed along the RN 9. At this key section terracesediment sar efoun di na forme rgull yincise di ngranite . Thesediment shav ea maximu mthicknes so fabou t1 0m an dtw omai n unitsca nb edistinguishe d (Fig.2.2.1.) .Th elowe runit ,th eWb j sediments,ar epoore ri nvolcani cclast s (lessMg Oan dmor eK 20) than the overlying Wbll sediments. Wbx sediments are slightly alterated before they were buried with Wbn sediments. Both Wb litho-stratigraphicunit s are impregnated with travertine from nearby formersprings .

Figure2.2. 1 Keysectioni nW bsedimant sa tCoude s aMg O xK 20 •Na 20 14 The2 5m terrac eleve lwa soriginall y dated asLat eWeichselia n based on fossils and artefacts found on and in the travertine capping of this terrace (Daugas &Tixier , 1978). Although both terraceunit swer eoriginall ythough tt ohav ea Weichselia nage , radiometric dating showedotherwise . Thetravertin eimpregnatin gbot hW bunit sa tCoude swa sdate d withth enon-equilibriu m Th/Umetho dgivin g anag eo f 119,000± 13,000 (GrU-90420) forth eWb ^ and93,00 0+ 500 0GrU-90421 )fo r

Wbn travertine.

DMgO

xK 20 •Na 20

Figure2.2. 2 Keysectioni nW bsediment sa tInngue s 15 Asth etravertin eshow sn ostratigraphica ldiscontinuit ybetwee n the two units, it must have been deposited after the Wbn

sediments. The Wblr sediments are almost unaltered before travertine impregnation started indicatingtha tthes e sediments are only slightly older than the oldest travertines. The travertines give therefore only the minimum age for the Wbn sediments.A sth eW bsediment swer ealread ydeposite dan di ncas e ofth eWb xuni tals oaltere dbefor etravertin e impregnationtoo k place these units are thought to have at least Saalian ages. Travertines impregnating terrace sediments of the 50m terrace have Th/U ages up to 160.000 ± 10.000 (GrU-90418) years BP, indicating that travertine deposition took place also place duringth eSaalian .Thes eolde rtravertine ssugges tals otha tth e

Wb:l sediments are not much older than the Late Saalian travertines cementing them. AtLongue sanothe rterrac eleve la t± 25 m abov erive rterrac e level is found. This terrace level displays a similar litho-

stratigraphy (Fig. 2.2.2.) as the Coudes key section. The Wbx sedimentsar eals opoore r involcani cclast s (lessMg Oan dmor e

K20) than the overlying Wbll sediments. The underlying Wbt sediments arealtere d and locally kryoturbated beforethe ywer e buried with Wblr sediments. The kryoturbated paleosol clearly evidencesa majo rtim e lagbetwee nth edepositio n ofth etw oW b units.A tLongue sth eterrac esediment sar eals oimpregnate dan d covered with travertines. The travertines capping of the Wb terrace sediments at Longues were preliminarily dated by Kroonenberge tal . (1989),wit hth enon-equilibriu m Th/Umethod , giving Middle Weichselian ages as minimum ages for the Wb sediments. The distinguished Wb terrace litho-stratigraphy is schematically shown inFigur e2.2.3. , bytw okey-sections ,on e atCoude san don ea tLongues . Thegenera lcharacteristic so fth e terrace litho-stratigraphy aredescribe d below.

Wbz and WbXI: Based on the similarities in litho-stratigraphy, bulkgeochemistr yan dgeographica lposition ,th etw oW bsite sli e only7 kilometre sapart ,th eW bterrac esediment sar ethough tt o

16 belong to the same terrace units and to have Saalian ages. The

Wbn sediments areno tmuc h oldertha n 120,000year swhil e the Wbx sediments are thought to have an agebetwee n 160,000 and 120.000 years.

altitude(m ) alt tude (m) 30- 3\ Coudes ,0°\ 25- Long WB„ 20- WB„ 20- ,I *\A \ WBi •* • *^> \ WBi /': J 10- IU tntnntt" :::::::::iK^ y 1 Allier md * i* tii II * i Svl 0 500 100 200 300 m U0 basement rocks iH oligocene sediments I I fluviatile sediments |p°H colluvial sediments

Figure2.2. 3 Schematicallitho-stratigraph y

X terrace sediments For the younger X and Y terrace levels, fragmentary palaeontological, palynological, archaeological and UC datings evidence that they were formed during the Weichselian and Holocene (Rudel, 1953; Lambert et al., 1980; Raynal, 1984). An important key section of the X terraces is situated at Les Jarrauds (Fig 2.2.4.), where all four Weichselian litho- stratigraphical unitsar efoun dnearby .Th ehighes tterrac e level is found at approximately 20 m above present riverbed. This terrace level, Xa, consists of two litho-stratigraphical units, a gravelly topuni twhic h ischaracterize d byhig hbasalti c (high

MgO and lowK zOcontent ) sand and gravel content (Xm). Thi s unit is locally overlying a more sandy unit relatively poor in

basaltic fragments (Xn). At the transition to the lower Xb terrace level the Oligocene clays surface. The Xb terrace (10 m

17 abovepresen triver )ha sals otw omai nlitho-stratigraphi cunits .

On top again relatively basaltic rich sediments (XVI)overli ea basaltic poorer sediment (Xj). Sometime s the basaltic poor x„

sediments are foundbetwee nX :an dX IVsediment s indicating that the Xzsediment s areth erelativ e oldestX sediments .Th e lower

sediments (Xx)ar ealtere dan dlocall ya paleoso li sfoun da tth e contact with the overlying (Xn, Xm and XIV) sediments. An U organicric hcla ylaye ra tth ebas eo fth e (X:)uni twa s Cdate d at 29,560 ± 330 years B.P. (GrN-17242), revealing a Middle Pleniglacialage .A colde rclimati cenvironmen twa ssuggeste db y thepolle n content ofthi slaye r (Tuffery, 1986). Thefou rdistinguishe dunit sar eals ofoun da tothe rsites .Th e generalcharacteristic so fthes efou rLat eWeichselia nunit sar e schematically shown inFig .2.2.5 .an d described below.

•Mg O

xK 20 • Na-,0 Figure2.2.4 . Rsysectio na tLe sJarraud s

18 X:: Thislitho-stratigraphi cunit , XWaccordin gt oKroonenber g etal . (1988),consist so fgravell ysediment srelativel ypoo ri n volcanic components.Th e thickness changes from almost 20m to less then 3m . This unit is an altered and strongly dissected relict of a former terrace level at about ±20 m above present

riverlevel .Th eX :lithologica luni ti sfoun d inbot hX terrac e levels (Xaan dXb) .

upstream Allier-Dore confluence downstream Allier-Dore confluence

xm p. x„ K

XIV •m>w/'/ *-^0T ZY 1 i. 1 V//////// •777777777,vz #^Mw/M/////w/> 7W////7//////////ZZ

Figure2.2. 5 Schematicallitho-stratigraph ynorther nsectio n

South of Maringues and West of Le Bas Lachamp uneroded remnants of this 20 m terrace are found, the upper one to two

metres of the X: sediments are commonly disturbed by kryoturbation and frost wedges (Fig 2.2.6. and 2.2.7.). A kryoturbated clay layer in the upper two metres of this unit (Fig. 2.2.6.)Wes t of LaBa sLacham p hasa n UC ageo f 16,585±

250 years B.P. (GrN-17243) giving a minimum age for the Xx sediments and a maximum age of the last major kryoturbation

activity inth eX xsediments .

19 ITT 333JEE m ** ] ° & | clay colluvium

Xi gravel clay

V77A *>nd

1m 14C sample i GrN 17243

Figure2.2. 6 Kryaturbationo fth etc pX ,sediment sa tLa sba sIachanp s

wf/.Ox. 2 3

\x• m

paleosol • MgO xK 0 •Xi 2 • Na20

Figure2.2. 7 X,u overlyingkryoturbate dpaleoso li n X,sediment sa t culhat. X„: This lithological unit, the crystalline X terrace according to Larue (1977), predominantly consists of sandy sedimentswit ha lo wvolcani ccontent .Thes esediment soccu rver y

locally onto pth e XTan dbelo wth eX m andX IV sediments,an d havea limite d thicknesso fthre et oles s thanon emetre .Thi s unit is found in terrace level Xa as well as inXb . The sedimentsar eunaltere dan dn opolle no rothe rdatabl ematerial s were foundi nthes esediments .

Xnl: Thislitho-stratigraphi cuni tusuall y formsth eto playe r ofth eX aterrac eleve l (+2 0m abov erive r level),an dconsist s

ofgravell y sedimentsver y richi nvolcani c components.Th eX nl

sedimentsar eoverlyin g bothth eX xan dX „ units.Th evolcani c rich sediments, which lack anydatabl e material, areclosel y related todeglaciatio n inth euppe r Allier basin andhav ea

20 fluvio-glacialorigi n (Bout,1963 ;Kroonenber ge tal .,1988) .Th e

Xin sediments did locally bury some strongly kryoturbated

paleosols inth eX x sediments (Fig2.2.5. )indicatin g amaximu m

ageo fth eX m sedimentso fapproximatel y 16,500year sB.P . (age

ofburie d kryoturbated clay inpaleoso l inX x sediments). Correlation between the Late Weichselian chronology in the Artierebasi nan dth eAllie rsediment sshowe dtha tth eag eo fth e

Xin unit isbetwee n 11,500 and 41,000 years B.P. (Lenselink et

al., 1990). The pre-Allerod age of the Xm sediments was

confirmed byth eobservatio n nearLempde stha tX ni sediment are buriedb yth esam etrachy-andesiti c Allerod asha sdate d inth e Ravelan dArtier ebasi n (Kroonenberg etal. ,1987 ;Lenselin k et

al., 1990;Juvigne , personal communication).Th e xm sediments were thusdeposite d between 16,500 and 11,500year s B.P..

XIV: This litho-stratigraphicuni t isth e top layer of the Xb terraceleve l (±10m abov erive rlevel )an dconsist so fgravell y andsand ysediment sric hi nvolcani ccomponent s (Fig2.2.8.) .A n

organic clay block of aburie d paleosol atth ebasi s of theX IV sediment unit (Fig2.2.8. )a t St Yorre has an "c age of 11,380 ±10 0year sB.P. , (GrN-16793).Anothe rburie dpaleoso ldevelope d

inth eto p layer of the XIV sediments at Les Granvaux buried by a local sand fanha sa nag eo f731 0± 7 0year s B.P.(GrN-16795).

The XIV sediments were thus deposited between 11,380 ± 100 and 7310± 7 0year s B.P.. At Longues another terrace ±8 m above river level is found which resembles the Xlv unit as described in the northern section. It consists of gravelly and sandy sediments rich in volcanic components (much MgO and less K20), impregnated and capped with travertine. This level was dated by means of a correlation with a nearby dated archaeological site by Raynal (1984)a sYounge r Dryas.W edate d organicmatte r ofa palaeoso l inth eto psediment sburie db yhal fa metr epur etravertine .Th e organicmatte rha sa nU C ageo f9,63 0± 9 0B P (GrN 16912), this EarlyHolocen e ageo fpos tdepositiona l organicmatte r confirms theYounge r Dryas ageo fth eX IVsediments .Bu tth e CaC03 inth e samesampl ewa s14 Cdate da t2520 0± 90 0year sB.P . (GrN-16912).

21 St. Yorre

[TTT71 so;!

colluvtum

fluvial sand

|£| 9'avel

IE) oiigocene marl

C sample

1m • GrN IG793

O MgO

x K20

Figure2.2. 8 Keysectioni nX IVsediment sa tS tYarr e

As the CaC03 was precipitated in and around the dated organic matter, the CaC03 should be younger then the organic matter.

These contradicting ages made us conclude that the dated CaC03 contained 'fossil' carbonates probably originating from the nearby Oiigocenelimestones .

ZY: This unit consists of at least three different terrace levels with predominantly sandy sediments, relatively poor in volcanic components. The oldest dated Holocene sediments from 14 gulliesincise d inth eX IVsediment shav ea Cag eo f6,23 0± 10 0 yearsB.P. , (GrN-16794).A similar agewa s found forth eoldes t Holocene terrace level (Lambert et al., 1980).

22 2.2.4. A regional reconstruction of the Late Quaternary Allier dynamics Ourreconstructe dchronolog ystart si nth eSaalia n(Riss )whe n

Wbx and Wbn sedimentswer emos t probably deposited. As nomor e detailsar eknow no fthes esediments ,the ycanno tb e correlated 7withan yclimati ceven to renvironment .Ou rdate dchronolog yo f theAllie r sediments allows only a tentative reconstruction of theAllie rdynamic sdurin gth elas t30,00 0year s (Fig2.2.9. ). As standard the oceanic deep sea curve is used (Hays et al., 1976; Kominz et al., 1979) and matched with the continental chronologieso fLe sEchet s (DeBeaulie ue tal. ,1984 )an dGrand e Pile (Woillard, 1978; Woillard & Mook, 1982). More regional sedimentologicalan dpalynologica l studies (DeBeaulie u etal. , 1982; Juvigne et al., 1988;Rayna l et al., 1984;Reill e & De Beaulieu, 1988) served asextr achecks . Thelitho-stratigraph y(Fi g2.2.9 .fift hcolumn )start sdurin g aMiddl e Pleniglacial interstadialwhe nth eAllie rwa sprobabl y incised toth epresen t riverlevel .

Figure2.2. 9 BieLat eWeichselia nchronolog yo fth eAllie r

23 When the XT sediments were deposited during the Middle Pleniglacial the riverbed level rose up to 20 m above present

river level. These Xx sediments were locally at least 15 m dissected and altered during the subsequent Late Pleniglacial

before deposition of the Xn sediments. During the Late Pleniglacialknow na sth ecoldes tepisod edurin gWeichselia nth e glaciersha dthei rmaximu mextensio nan da dr ytundr aexiste di n the Limagne (Veyret,1980; De Beaulieu et al., 1982; Raynal, 1984). Under this severe climate the strong periglacial

deformation of the upper XT sediments took place. Within this

environment local deposition of thevolcani c poor XIX sediments took place. This cold period ended at the end of the Late Pleniglacialtriggerin ga larg efluvio-glacia lsedimen tflu xfro m thevolcani careas .Thi sclimat eimprovemen twhic htoo kplac ei n several thousands of years, resulted in the deposition of the

gravelly volcanic rich sediments of unit Xm (Bout, 1963; Kroonenberg etal. ,1988) . Theirdepositio n caused aris eo f2 0 m of the river bed level and must have happened during catastrophic floodscausin ga ninfil lo fth epalaeo-Auzo nvalle y (Lenselink et al., 1990) and lake formation in the lower

tributaries (Morge, Litroux, Buron etc) in the Limagne. Xln sedimentwer eprobabl y1 5m dissecte ddurin gBoilin gan dAllerod ,

after which the XIV sediments were deposited during the colder YoungerDryas .Thes eLat eGlacia lsediment swer edissecte ddurin g the Early Holocene. The oldest record of fluvial Holocene deposition inth eAllie r starts inth eAtlanticum .

TheX m sediments and lake formation inth e Limagne Ovalshape dsemi-close ddepression s(locall ycalle d'marais' ), severalkilometre s indiameter ,occu r inth e lowerpart so fth e Limagnerif tvalley .Withi nthes emarai slak esediment sar efoun d

(Gachon, 1963). Theto p altitudeo fth eX m unitcoincide s with the altitude of the upper lake sediments inth e Limagne. Their geographical relation isvisualize d inFi g 2.2.10. Asth elak e

sediments,lik eth eX m sediments,overli ea kryoturbati csurfac e their deposition seems related. The Xm sediments originating from large sediment fluxes are thought to have served as a

24 barrierfo rth eloca ltributarie scausin glak edevelopmen ti nth e Limagne. These minor Allier tributaries inth eLimagn e could never compensate a sudden rise inriverbe d level of2 0m b y sedimentationdu et othei rlimite dsedimen tsupply .A sth eoldes t lakesediment shav ea nag eo f12,37 0+ 23 0year sB.P .(GrN-12891 )

(Kroonenberge tal. , 1987)th eX ln sedimentsar ethough tt ob e deposited between 16,585± 25 0an d12,37 0± 23 0year sB.P. .

\i**^ Hercynian basement Kff&ffi Vokamcs [ 1 01igocene sediments [':-:':':::] Terraces Har a i s

77777y////.

Geographical relationship top All sediments and lake sediments near Culhat/Ravel

Figure 2.2.10 Schematical morphological relation between the top of the Xa terrace level and lake sediments.

25 Kroonenberg et al. (1987) explained lake development by thermokarst due to permafrost degradation a process which may havecontribute dt oth eactua llak ebasi nmorphology .Bu ti tar e thelarg esedimen tfluxe sdurin gth een do fth eLat ePleniglacia l which arethough tt ohav ecause d lake formation inth eLimagne .

2.2.5. Climatic and fluvial dynamics inth eAllie rbasi n Theregiona lreconstructio no fth eAllie rdynamic sdurin gth e LateWeichselia na s inFig .2.2.9. ,allow sth epostulatio n ofa simplified model for the relationship between climatic and fluvialdynamic s inth eAllie rbasin .

Climatic environment Fluvial activity

Glacial Deposition of sediments with a relative low volcanic content (<

MgO and >K 20). Th eolde r surface terrace sediments are kryoturbated.

Transition from Glacial to interglacial Deposition of coarse volcanic

rich (> MgO and < KzO) fluvio- glacial sediments. A strong rise in river bed level due to the large sediment flux from melting glaciers.

Inter-glacial Predominantly incision with temporary deposition of sandy sediment with a low volcanic

component (K 20).

Thissimplifie dmode lha ssom eresemblanc ewit ha mode lwher e the general textural characteristics of terrace sediments are related to climatic environments (Texier & Raynal, 1984). Our

26 terrace litho-stratigraphic reconstruction also shows that the InterglacialHolocen eterrac esediment shav ea fine rtextur etha n the coarse Glacial and Late Glacial sediments. But within our model the most important changes in the Allier system are sediment composition changes induced by glaciermeltin g on the Mt.Dor ean dCanta la tth een do fth eLat ePleniglacial ,a facto r notconsidere d inth esimple rmode lo fTexie ran dRayna l (1984).

27 2.3 TRACHYTIC PUMICE CLASTS IN MIDDLE PLEISTOCENE ALLIER TERRACE DEPOSITS:A CHRONOSTRATIGRAPHIC MARKER.

A.Veldkam p &A.G .Jongman s

Abstract Trachytic pumice clasts with similar characteristics as pumice of the Sancy volcano at Neschers have been found at different sites in the 65 m (Va) terrace of the Allier, implying a maximum age for this terrace level of 800,000 years. Micromorphology of an intercalated paleosol and kryoturbatic features indicate that Va sediments have known at least one glacial/interglacial cycle before dissection of the Va level took place. By using the pumice clasts as a stratigraphic marker a reconstruction of the longitudinal profile of the Va level was made. This reconstruction shows that the Randan region has been relatively uplifted, while the Lezoux section has been subsided after dissection of the Va level.

2.3.1. Introduction Fluvialterrace s arecommonl ycorrelate d onaccoun to fthei r relative altitude or sediment composition. Inth e Allier basin (Limagne, France., Fig. 2.3.1.), a first detailed terrace correlation was made for the geological map of Maringues (Jeambrun et al., 1980). The distinguished terrace levelswer e based on their relative altitude and general sediment composition. As inmos t other fluvial systems the stratigraphy ofth eAllie rterrace si sfa rfro muniform .I nth eAllie rsectio n South of Clermont Ferrand which was studied by Van Dorsser (1969), itwa s almost impossible todemonstrat e differences in heavymineralogica lsan dcompositio n forth edifferen tterraces . Pastre (1987)wh oextensivel y studiedth esan dmineralog yo fth e differentAllie rterrac e sediments adaptedth eexistin g terrace classification almost purely based on sand mineralogy. He also established the chronology and ages of the terrace sands by correlating theirheav yminera l compositionwit hth emineralog y ofknow nmajo rvolcani cevents . We studied the Va (Jeambrun et al., 1980) terrace of the Allier, which has an altitude of approximately 65 m above present river bed, along a stretch of 40 km, from Clermont Ferrand near the village of Beauregard l'Eveque toVich y (Fig. 2.3.1.).

28 •v » T"*V v-s—sr

;ChqinevI;.">

;Puys«;";»

EvT]Hercynia n basement [|111111| Volcanics | |Oligocen e sediments • samplesite s

Figure2.3.1 . Mapwit hsampl esite s

The average Va gravel composition, 50%basalts , 5% granites, 5% metamorphic rocks and 40% quartz, reflects the different lithologies within theAllie r basin (Larue, 1977;Bout , 1963;Va n Dorsser, 1969). Mineralogical sand compositions aregive n byVa n Dorsser (1969), Rudel (1963), Pelletier (1971), Pastre (1987)an d Tourenq (1986) and bulk geochemical compositions by Kroonenberg et al. (1988) and Veldkamp & Kroonenberg (1989). Theag eo fth e Va terrace was established at one million years based on sand mineralogy (Pastre, 1987).

It was Pastre (1987) who observed pumice clasts at one site in theV a terrace level (at Saint-Sylvestre-Pragoulin, Pastre, 1987, p. 500). W e found more pumice clasts inth esand y units of theV aterrac e level atfiv e widely separated sites.B y comparing the pumice characteristics we conclude that this pumice originates from asingl e trachytic source eruption from theSanc y volcano approximately 800.000 years ago, implying a considerable younger maximum agefo rth eV a terrace level as given by Pastre (1987).

29 >

Q. 01 TO 3

re 1= 0>

u1_0 — o^ N 01

QJ c »_OJ» QJ o a en

C71 o S2. o m

c o c

Figure 2.3.2. Terracestratigraph yo fsample dsite si nFv aterrac e

30 From five profiles in the Va terrace undisturbed samples were taken (Fig. 2.3.1., 2.3.2. and Table 2. 3 .1 ) for mineralogical and micromorphological examination. To enable a more detailed reconstruction of the Va terrace in terms of chronology and climatic environment, a kryoturbated paleosol was studied in more detail. Thin sections (10*10 cm) were made and described following the terminology of Bullock et al. (1985). Mineral countings (n=150) were performed in these thin sections.

site name X-coor. Ycoor. Ornon 528 8 5078.8 Bogros 527 9 5079.7 Drevoux 535 5 5085.7 Chez Faure 532 0 5085.9 Puel Chauvin 533.2 5096.4

Table 2.3.1. tfcp coordinates of sample sites.

2.3.2. Stratigraphic marker

Terrace stratigraphy Thestratigraph yo fth eV aterrac esediment si sbase do nfiel d observationsan dmicromorphologica lexaminations .Fou rdifferen t lithological unitsar edistinguished , fromto pt obottom .

Unit I. occurs inal l 5profile s and commonly consists of a4 m thick gravel unit in which basaltic clasts dominate. Unit I facies indicate abraide d gravel river.A planosol (FAO,1974 ) has been developed in the upper two metres of all profiles (Feijtel etal. ,1988) . Unit Up underlies unit Ian d predominantly consists of sands withman ytrachyti croc kfragment san dpumic efragments .Uni tI I facies suggesta sand ybraide d system.Thi suni tha sa naverag e thicknesso f1 metre .Th epumic efragment sar erounde dan dfoun d

31 in clayey and loamy units or as pure pumice layers of several centimetresi nmajo rsan dbodies .Th ebes tlocatio nt ostud yth e pumicemacromorphologicall yi si nth egrave lpi tnea rOrno nwher e pumice layersar eexpose d atth epi tfloo rnea rth egroundwate r table. UnitIII ,underlie suni tII ,an dconsist so fbasal tric hgravel . This unit of at least half a metre is only found in the Puel Chauvinprofile .Th efacie so fthi suni tsugges ta simila rrive r type aswit h theuni t I,a braide d gravelriver . Unit 0. consists of Oligocene clayey sediments which form the base for all theV a terrace sediments. This unit is only well exposed intw oexcavations .

2.3.3. Pumice Inthi nsection so fal lth efiv eprofile sth epumic efragment s are fine grained hypocrystalline to holohyaline, with varying contento fphenocryst so ffeldspa r (predominantlyK-feldspa ran d albite, 20-500 jim), biotite, green and brown clinopyroxenes, brownamphibole ,sphen e (20-200/Ltm ), an d opaquemineral s (20-40 ^m). Th epumic eshow sa vesicula rtextur ewit hnumerou sspherica l to ellipsoidal shaped holes, discontinuous wavy laminated oriented in a longitudinal section (Photos I and II in Fig. 2.3.3.). The pumice has a trachytic chemical composition and shows in-situ alteration and clay neoformation within the fragments (Veldkamp& Jongmans , 1990).

2.3.4. Correlation ofpumic ewit h dated Sancy eruptions Micromorphological and mineralogical (Table 2.3.2.) observations of the Va pumice show a uniform composition and texture of the pumice fragments from the different sample locations. Theoccurrenc eo fon etyp eo fpumic ei nsimila rstratigraphi c units atmor e than hunderd kilometres from the eruption centre (Sancyvolcano )suggest stha tth eV apumic eca nb ecorrelate dt o onemajo reruption ,an dca nb euse d asstratigraphi cmarker .

32 Q.

Vj^v-

Figure2.3.3 . Rimioefragmen to fNescher s(Hiot oI )an dpumic efragmen t fromuni tI Io fOma n (RiotoII) .

33 Only a few major trachytic volcanic phases are known as possiblesourc efo rth epumic eclasts ,th eignimbrit eo fRioube s (±900,000 years BP) and the ignimbrite of Neschers (±800,000 yearsBP ) (Bessone tal. ,1977 ;Chambaude t& Couthures ,1981 ;Ly , 1982;Pastre , 1987).

Fva Fva Neschers Rioubes pumice pumice pumice pumice Mineral Ornon Puel (Pastre, (Pastre, Chauvin 1987) 1987)

Zircon 1.0% 0.5% 0.5% 0.2% Sphene 18.5% 20.0% 23.0% 24.0% Browna n phibole 6.0% 8.0% 8.0% 2.0% Greenc l inopyroxene 65.0% 64.5% 63.5% 68.8% Brown cl inopyroxene 6.0% 5.0% 3.5% - Apatite 1.5% 0.5% 1.5% 2.8% Other 2.0% 1.5% - 2.2%

Table2.3.2 . Percentageo fheav ymineral si nFv apumic e(Orno nan dPue l Chauvin),Nescher spumic ean dRioube spumice .

Heavyminera lcompositio no fth epumic efragment si nth eOrno n and Puel Chauvin profiles are compared with the two potential source types in Table 2.3.2. Biotite content was excluded to facilitatecomparison .Th eoccurrenc e ofbrow nclinopyroxen ei n the Va pumice excludes the trachytic ignimbrite of Rioubes as sourceeruption .A sther ei sa fairl ygoo dsimilarit ybetwee nth e Va andNescher spumic ecomposition ,th eV apumic e fragmentsar e thought to belong toth e Neschers generation. Consequently the age of the pumice containing sediments (unit II) can be considereda tabou t800,00 0years .Thi smaximu mterrac eleve lag e ispartl y inaccordanc ewit hth echronolog yo fLaru e (1979),bu t ismuc h younger thanth eag eestimat e of Pastre (1987).

34 2.3.5. Paleoenvironment

Micromorphology ofpaleoso l The Puel Chauvin profile near Randan displays all stratigraphicunit sincludin ga kryoturbate dclaye yban di nuni t II. The less weathered and kryoturbated parts show that the clayey layerwa soriginall y afin elayere d sediment.Th eclaye y bandha sa distinc tangula rblock ystructur e ina greyis h (10Y R 5/1) clayey groundmass, indicating the formation of a well developed physicogenic structure. Channels, surrounded by porostriated b-fabrics and the occurrence of loose continuous excremental infillings suggest root and faunal activity. Walls of vughs and chambers demonstrate predominantly lenticular prints, indicating a former growing of euhedral masses of lenticularcrystallaria ,possibl egypsum , inthes evoid s (Porta & Herrero, 1990). Such gypsum formation is known from other paleosols inth eMassi f Central (Pierre, 1989). Occurrenceo fyello wno nlaminate d isotropican dgreenis hyello w anisotropic speckled oriented clay coatings (Feijtel et al., 1989; Jongmans et al., 1990) in the sandy parts of unit II indicatesweatherin go fth epumic efragment san dneoformatio no f clay.Sinc eth edescribe dweatherin g featuresar eno tnotice di n deoverlyin g deposits ofuni t Ii tca nb econclude d thatbefor e unit II was kryoturbated and buried by unit I, a paleosol was formed inth e sediments ofuni tII .

Themai nsoi lformin gprocesse si nthi spaleoso lwer ein-sit u weathering ofpumic ean dtrachyti c rock fragments,neoformatio n ofclay ,biologica lactivit yan dsolutio no fcrystallaria .Thes e processes suggest that after and probably during the pumice deposition,a temperat eclimat eprevailed .Thi senvironmen twit h aninterglacia l signaturewa sfollowe db ya col dclimati cperio d duringwhic hperi-glacia l conditionscause dth eobserve d strong kryoturbation ofth euppe rpart so funi tII .

Theothe rfou rprofile shav eals oweatherin gfeature si nthei r unitI Isediments .Bu tth eabsenc eo fdistinguishabl edifference s

35 between weathering features in unit I and unit II makes the recognition ofothe rpaleosol sdifficult .

2.3.6. Terrace chronology Va During deposition of the Weichselian terrace sediments the majority ofth ebasalti cgravel swer edeposite d atth een d ofa glaciala sa resul to fglacie rmeltin go nth ehighe rpart si nth e Allier basin (Bout,1963 ;Kroonenberg , et al., 1988). Assuming similardepositiona lmechanism sfo rth eV aterrac esediment sth e gravelly basaltic richuni t Ii sthough tt ob edeposite d atth e end of aglacia l period too.Afte r deposition,dissectio n took place under more temperate climatic conditions, causing the formation ofth eV a terracelevel .

On account of these observations and assumptions the following tentative chronology forth e events of theV a sediments can be reconstructed : Fluvialdepositio no fpumic eric hsand ysediment s (unitII ) took place around 800.000 years ago.Th e roundness ofth e trachytic fragments and pumice clasts indicate that these sediments have been fluvially transported before sedimentation. The deposition of these sandy sediments seemsno tdirectl y related toclimati c environment. After deposition, these sediments were weathered under temperate conditions (paleosoluni tII) . Next, climate shifted to periglacial conditions during which a strong kryoturbation of the existing soil took place (kryoturbated paleosol unit II).At other sites the unit IIsediment swer epartl y orcompletel y eroded. At the end of aglacia l period the remaining kryoturbated paleosol wascovere d bybasalt-ric h gravels (unitI) . These sediments were incised under more temperate conditions causing the formation ofth eV aterrac elevel .

As the Va terrace sediments were subjected to at least one glacial/interglacial cycle,dissectio no fth eV aleve lmus thav e takenplac e considerably latertha n 800.000years .

36 2.3.7. Tectonic implications Usingth epumic eclast sa sa chronostratigraphi cmarke ra mor e reliable reconstruction ofth e longitudinal Va terrace profile can be made then on the base of topography alone. In order to reconstructth elongitudina lV aprofile ,th esurfac ealtitud eo f the studied profiles from the 1:25.000topographica l maps were used (Fig.2.3.4. ). Th epumic elayer sar eindicate dindividually . The pumice site described by Pastre (1987) downstream of Puel Chauvin is shown as well. Based on:its similar stratigraphic position we include this apparently identical pumice marker horizon in our reconstructed longitudinal profile. The Oligocene/terrace sediment boundary is very irregular and thereforeno tconsidered .Th eprofil eo fth eactua lAllie rrive r bed isver yregula ri nth estudie dAllie rsection ,an dtherefor e a relative longitudinal Vaprofil e ispresente d (Fig. 2.3.4.).

relative longitudinal profile Fva (Fva altitude - Fz altitude)

Fva altitude - Fz altitude (m) 90 # pumice marker horizon J_ thickness Fva sediments 80 Ornon IDrevou x

70

60-

50 i 10 15 20 25 30 km kilometres downstream from Beauregard I'Eveque

Figure 2.3.4. Longitudinal profile of the relative altitude of the Fva terrace level. 37 Near Clermont Ferrand theV a terrace level runs parallel to the actual riverbed, but more downstream the relative altitude decreases steadily. Thisgradua l decrease in relative altitude in the longitudinal profile was also reported by Giot et al., (1978). North of the Allier/Dore confluence, the relative altitude of the Va terrace increases steadily. The scale and changes in the relative altitude clearly indicate that postdepositionaltectoni cdisturbanc eo fV asediment stoo kplace . TheRanda nregio nha sbee nuplifte dwhil eth eLezou xregio nha s been subsided after dissection of theV a level.Thes e vertical movements seem simultaneousan d related torif t faultmovement s inth eLimagn erif tvalley .Simila rtectoni cuplif twa spropose d byLaru e (1979),wh oexplaine dth ecours ediversio no fth eMorge , anAllie r tributary, by tectonism.Th eV a terrace of the Morge is found along the old course,whil e Fw is the oldest terrace levelalon gth eactua lMorge ,suggestin ga nuplif to fth eRanda n region afterth eV aan dbefor eth eW sedimen tdeposition .Alon g theAllie rirregula rlongitudina lprofile sar eonl yfoun dfo rth e V and older terraces,whil e theW and younger terraces show a very regular longitudinal profile. Our observations, confirm those of Larue (1979) and Giot et al., (1978) indicating a regionaltectoni cdisturbanc eafte rth eincisio no fth eV a (<80 0 ky) level andbefor edepositio n ofW terrac esediments .

2.3.8. Conclusions Pumice of the Neschers generation serves as a chronostratigraphic marker in the Allier Va terrace level. Va terrace sediments were subjected to at least one glacial/interglacialcycl ebefor edissectio no fth eV aleve ltoo k place.A geographica lreconstructio no fth eV aterrac eleve lwit h the stratigraphic marker confirms that the Randan region has known a relative uplift and the northern Lezoux section a subsidence,afte rdissectio n ofth eV a terracelevel .

38 chapter 3 SAND BULK GEOCHEMISTRY

As Quaternary research is more and more interested in the quantitative aspects ofmajo r environmental changes itwil l be necessaryt odevelo pne wquantitativ emethods .A wa yt odetermin e quantitativepaleohydrologica lchange si sb ystudyin gth echange s inbul kcompositio no ffluvia lsediment s (Kroonenberg, 1990).A s it is rather cumbersome to derive a bulk composition from the mineralogical composition of separate fractions and point- counting of thin sections istim e consuming, itwa s decided to measurebul k sand composition geochemically.

3.1 THE APPLICATION OF BULK SAND GEOCHEMISTRY IN QUATERNARY RESEARCH. A METHODOLOGICAL STUDY OF THE ALLIER AND DORE TERRACE SANDS.

A.Veldkam p &S.B . Kroonenberg

Abstract The bulk geochemistry of unconsolidated sands of river terraces in two drainage basins of contrasting geology in the Limagne rift valley, France, has been studied. Data processing and interpretation was done with multivariate statistics, notably factor analysis. In the Allier basin, underlain by volcanic rocks and crystalline basement, the abundance of basaltic rock fragments causes significantly higher bulk concentrations of Ti02, Fe203, MgO, CaO, P205, V, Cr, Ni, Zn, Sr, Zr and Nb. In the Dore basin, underlain by crystalline basement alone, sands have a significantly higher Si02, K20, Rb and Pb content, originating from quartz, K-feldspars and mica's. Comparison between different terrace levels shows the impact of weathering and changes in supplied sediment composition over time. Approximately 65% of the total variance in the basaltic element content in the Allier terrace sands can be explained by the combined effect of parent material and weathering. The effect of grain size on sediment composition is only significant for A1203, Na20 and Rb content in the Allier and for Ti02, Fe203 and Nb content in the Dore sands. Approximately 60% of the total variability in these elements is grain size related. These data show that bulk geochemical studies of Quaternary terrace sands can provide valuable data complementary to traditional sedimentary petrographic research. They also indicate that data obtained during mineral exploration are potentially applicable in Quaternary research.

3.1.1. Introduction Fluvial terraces result mainly from a complex interplay of changes in climate, tectonism and base level (Green and Mc

39 Gregor,1987 ;Bol le tal. , 1988; Veldkampan dVermeulen , 1989). These external factors are directly or indirectly reflected in thecompositio no fth eresultan tsediments .Sedimen tcompositio n isals ocontrolle db y local factors,suc ha ssortin gprocesses , postdepositiona lweatherin getc .Majo rchange swithi na fluvia l system do not only change sedimentgranulometr y and mineralogy but also the quantity of sediment delivered to the river. The total sample composition quantitatively reflects such changes. AsQuaternar y research ismor edirecte d towardth e quantitative aspects ofmajo r environmental changes itwil l be necessary to determine such changes by studying the bulk composition of fluvial sediments (Kroonenberg, 1990). It is rather cumbersome toderiv ea bul kcompositio n fromth emineralogica lcompositio n of separate fractions or from time consuming point-counting of thin sections. Therefore we decided to measure bulk sand composition geochemically. Although iti sa commo npractic e inexploratio n geochemistry tosampl eonl y onesiz e fraction fromstrea m sediments inorde r toobtai nmeasurabl ean dcomparabl eresult s (Plante tal ., 1988) , weus egeochemica lmeasurement sfo rbul ksan dsample so fal lsiz e fractions. As a case study the Allier and Dore basins in the Massif Central were selected as two basins having different basin geology and consequently different sandcompositions .

3.1.2. Materials andMethod s TheAllie rwa ssample d in1 4location s (totalo f 57samples ) by Kroonenberg et al. (1988), along a section between the confluenceswit hth eCouz eChambo n inth esout han dth eDor ei n thenorth .Th esampl e locationsar esituate d inth epresen tbe d (Z) and four different terrace levels (Y, X, W, V) (Figure 3.1.1). Along a stretch of 30k m along the Dore river, 43 sand sampleswer e taken at 12 sites from the same terrace levels as sampled along the Allier. The Dore was sampled from the confluence with the Allier up to its emergence from the rift valley scarp. Inorde rt oinclud eth eeffec to fsorting , 2-6 sampleso fdifferen tgrai nsiz edistributio nwer etake nfro meac h

40 I.'.'.'l Hercynian basement MM Volcanics * Isolated volcanic plug CD 01igocene sediments 1=1 Holocene basins (marais) EE) y and z terraces EMI x terraces w terraces

mv terrace s Bilal t and u terraces III s terraces • sample site

Figure3.1.1 . Geologicalsketc hma pan dsampl elocation si nth eAllie r andDar ebasins . sampling spot. Three physical parameters were considered during sampling: RELALT (altitude relative to the present river bed), UPSTR (upstream distance from confluence Allier/Dore) and MEDIAN (mediano fgrai n sizeo fanalyze d samples).Thi swa sdon e ina n attempt to evaluate the importance of lateral and downstream sorting and postdepositional weathering irrespective of the impacto fclimati cchang ean dtectonism .Car ewa stake nt osampl e sand from foreset laminae in small-scale cross-bedded sets, in ordert oavoi d concentrations ofheav ymineral s inhorizontall y laminated lag deposits. Sampleswer e taken as deep as possible toavoi d effectso f soil formation. In the laboratory, detailed granulometrical analyses were carried out for 13 fractions.Th e <16 fim, <0.053, 0.053-0.075, 0.075-0.106,0.106-0.15 ,0.15-0.212 ,0.212-0.3 ,0.3-0.425 ,0.425 - 0.6, 0.6-0.85, 0.85-1, 1-2, 2-4.8, >4.8 (mm) fractions were determined.Th emedia ngrai nsiz ewa sdetermine dfro mcumulativ e frequency plots.Media n grain sizes ranged from 210 nmt o 2000

41 3/im.Par t of the sample,wa s separated with a sample splitter. Organicmatter ,calciu mcarbonate ,fre eiro nan dal lmateria l<1 6 nm was removed by standard methods. This allowed direct comparison of chemical composition with optically determinable mineralogy and allowed us to avoid the effect of accumulated secondary products. Organic matter and manganese coatings were destroyed with H202 and 'free ' iron was extracted with Na- dithionite-EDTA at pH 4.5. Although the clay content was measured, the fraction < 16 urnwa s notgeochemicall y analyzed. Infe wcase sthi sfractio n (<1 6/xm )amounte dt omor ethe na fe w percentb yweigh t ofth etota lsample . Afterpretreatment , 0.6 go fth esampl ewa sseparate dwit ha micro-sample splitter, fused with 2,4 lithium tetraborate and analyzed. The major elements Si02, Ti02, A1203, Fe203, MgO, CaO,

Na20,K 20an d P205an dmino r elementsV , Cr,Ni ,Cu ,Zn ,Ga ,Rb , Sr, Zr,Nb ,La ,B aan d Pbwer emeasure dwit hX-ra y fluorescence spectroscopyo na Philip sXR Fassembly .Th esyste mwa scalibrate d using USGS geochemical standards as listed by Abbey (1980, p. 16). The mean, maximum and minimum bulk concentrations of the elements inDor ean dAllie r samplesar e listed inTabl e 3.1.1. Statistical treatments were performed with SPSSpc (Norusis, 1986).Th esignificanc eo fdifference si nmea nelemen tabundance s betweenth eAllie ran d Doregroup swer eteste db y aStudent st - test. Theinterpretatio no fthi scomple xmulti-variabl edat ase t wascarrie d outusin g factoranalysis .Facto ranalysi s examines the interrelationships among the variables (elements) in an effortt o finda ne wse to fvariable s (factors) fewer innumbe r thanth eorigina l seto fvariables ,whic hexpres stha twhic h is commonamon gth eorigina lvariable s (Dillonan dGoldstein ,1984) . A factor analysis is presented as a matrix giving the factor loadings of each variable. A factor loading indicates the relative contribution of a variable to the factors made; the larger the loading the more important the variable in the interpretation of the factor. The factors are rotated with a varimax rotation, to allow better group identification and interpretation.Multipl elinea rregressio nanalysi swa sdon ewit h the stepwisemethodology .

42 Dore Allier

Var. LLD Mean Minimum Maximum Mean Minimum Maximum value value value value

Si02 .007 81.09 69.42 89.00 74.95 63.01 85.02 Ti02 .01 0.14 0.02 0.53 0.78 0.27 1.78 A1203 .075 10.15 5.25 16.36 11.63 7.84 15.61 Fe203 .003 0.65 0.13 1.91 3.27 0.82 7.87 MgO .01 0.30 0.06 0.95 1.60 0.22 4.60 CaO .0014 0.26 0.01 0.98 2.07 0.49 5.38 Na20 .01 1.68 0.28 3.42 2.23 1.50 3.63 K20 .006 4.32 3.21 5.22 2.88 1.92 3.80 p2o5 .01 0.06 0.03 0.13 0.19 0.06 0.44 V 4 25 5 57 99 22 232 Cr 14 10 10 10 33 10 148 Ni 2 2 2 2 21 2 82 Zn 1 7 1 85 26 1 76 Ga 4 11 5 18 12 8 16 Rb 1 147 106 197 105 54 218 Sr 1 146 75 208 282 149 531 Zr 2 63 7 349 112 33 215 Nb 4 5 3 18 11 5 21 Ba 36 657 446 802 676 506 879 La 7 18 6 73 22 6 68 Pb 1 20 10 29 11 10 29 sandfractions : <4.8 0.1 11.0 0.5 35.5 2.0 0.1 13.1 2-4.8 0.1 18.2 0.5 38.5 3.6 0.1 23.0 1-2 0.1 15.9 1.0 28.0 11.9 0.1 47.9 0.85-1 0.1 14.8 1.0 31.0 5.6 0.1 16.2 0.6-0 85 0.1 10.5 1.0 32.0 17.3 0.5 41.8 0.425-0 6 0.1 7.3 1.0 28.0 18.1 1.5 51.9 0.3-0 425 0.1 5.5 0.8 17.0 16.8 2.6 43.2 0.212-0 3 0.1 2.8 0.5 19.0 12.4 0.8 47.0 0.15-0.212 0.1 1.8 0.5 29.0 7.1 0.1 35.5 0.106 0 15 0.1 0.9 0.5 6.0 3.0 0.1 22.9 0.075-0 1060. 1 0.8 0.5 4.5 1.0 0.1 12.4 0.053-0 0750. 1 0.8 0.5 4.5 0.4 0.1 5.4 <0.05: 0.1 0.9 0.5 5.5 0.6 0.1 7.6

Table3.1. 1 Means, maximum and minimim values of major and minor elementsan dgrai nsiz eclasse so fth eDor e (32samples ) andAllie r (49samples )sands .SiO zt oBa Oan di nweigh t percentages of oxides,R b to Nb in weight ppm of the elements,siz efraction s(mm )i n% .

43 3.1.3. Results Meanbul kgeochemica lcompositio no fth eAllie ran dDor esand s

isgive ni nTabl e3.1.1 .Th eaverag eamount so fTi0 2,A1 203,Fe 203,

MgO,CaO ,Na 20,P 205,V ,Cr ,Ni ,Zn ,Sr ,Z ran dN bar ehighe rfo r theAllie rsand swherea s Si02, K20,R ban d Pbar ehighe r forth e Dore sands.Th e differences inmea n element abundances between theAllie r and Doredat a aregive n inTabl e 3.1.2. In spite of high standard deviations,th edifference sbetwee nth eelementa l composition of sands frombot h rivers are significant with the exception ofGa ,B aan dLa . Asi ti sdifficul tt ointerpre tan dcompar eth era wAllie ran d Dorebul ksan dcompositions ,a facto ranalysi swit hcomponentia l extractionwa sdon efo reac hbasi n (Table.3.1.3 ). Th einpu tdat a set for each basin included not only the bulk geochemical composition but also the physical parameters RELALT,UPST R and MEDIAN.

Allier Dore — — T-test Var. Mean Standard Mean Standard 2-Tail Deviation Deviation Prob.

Si02 74.95 5.52 81.09 4.47 .000 Ti02 0.78 0.40 0.14 0.11 .000 A1203 11.63 1.61 10.15 2.49 .003 Fe203 3.27 1.75 0.65 0.45 .000 MgO 1.60 1.13 0.30 0.24 .000 CaO 2.07 1.26 0.26 0.26 .000 Na20 2.23 0.31 1.68 0.87 .001 K20 2.88 0.39 4.32 0.37 .000 P205 0.19 0.10 0.06 0.03 .000 V 99 49 25 11 .000 Cr 33 37 10 0 Ni 21 19 2 0 Zn 26 23 7 15 .000 Ga 12 2 11 3 .079 Rb 105 30 147 23 .000 Sr 282 92 146 35 .000 Zr 112 51 63 64 .000 Nb 11 4 5 3 .000 Ba 676 60 657 69 .187 La 22 15 18 15 .273 Pb 11 4 20 5 .000

=belo wdetectio nlimit ,st .dev .— 0 .

Table3.1. 2 Comparisono fth emea nelemen tconcentration so fAllie ran d Doreb yStudent st-test .SiO zt oP^O ji nweigh tpercentage s ofoxides ,V t oP bi nweigh tpp mo fth eelements . 44 Allier Dore var. Fact1 Fact2 Fact3 Com. Fact1 Fact2 Fact3 Com. 58.3% 13.8% 6.1% 51.6% 11.4% 8.3%

Rel.alt . 0.72 0.63 0.93 0.87 UPST 0.50 0.50 0.05 Median -0.79 0.63 -0.55 0.41 Si02 -0.87 0.98 -0.83 0.98 Ti02 0.96 0.96 0.79 0.90 A1203 0.76 0.93 0.87 0.95 Fe203 0.96 0.95 0.74 0.52 0.88 MgO 0.95 0.94 0.77 0.86 CaO 0.94 0.95 0.83 0.93 Na20 0.80 0.76 0.94 0.95 K20 -0.55 0.61 0.84 0.81 0.73 V 0.91 0.89 0.59 0.58 Cr 0.84 0.84 ------Ni 0.87 0.76 ------Ga 0.59 0.55 0.77 0.80 Rb 0.82 0.79 0.83 0.83 Sr 0.91 0.91 0.93 0.91 Zr 0.89 0.93 0.90 0.88

Table3.1. 3 Factor analysis including physical parameters, bulk geochemistryAllie ran dDor esands .

Interpretation ofAllie rdat a TheAllie rsan ddat ase tca nb edescribe db ythre esignifican t factors (Table 3.1.3), together explaining 78.2%o f total bulk geochemicalvarianc e (cf.Kroonenber g et al., 1988). Factor1 Factor 1explain s 58.3%o f the total variance, contains the

variables UPSTR, Ti02, Fe203, MgO, CaO,Ni ,Cr ,V , Sr, Zr,-K 20

and -Si02. The contributing elements areknow n tooccu r inman y differentmineral so fth eAllie rsands ,notabl yolivine ,augite , amphiboles, calcic plagioclase and Fe-Ti opaque minerals. The distribution of the elements between these minerals can be established ina napproximat ewa y fromgenera l crystal-chemical considerations andpublishe d mineral analyses fromothe rsites .

45 Olivine, augite, amphiboles and plagioclases are common constituents of alkali basaltic rocks such as abound in the CentralMassif .Well-rounde d fragmentso fdens e alkali-basaltic lavas,wit ha specificgravit y around 3g/cm 3,for mth ebul ko f theheav y fractiono fth eAllie rsands ,an dma yconstitut eu pt o 50% byweigh t of the total sediment (Kroonenberg et al, 1988). It can be expected that this alkali-basaltic fraction strongly determinesth evarianc e ingeochemica lpropertie so fth eAllie r sands. Therefore Factor 1 is interpreted as an alkali basaltic factor. The positive factor loading of UPSTR might indicate a downstream backlagging of alkali basalt grains due to density sorting asreporte d by Daviese tal . (1978). Factor2 Factor2 ,explainin g13.8 %o fth etota lvariance ,include sth e variables,MEDIAN , K20,Na 20, A1203, Rb and Ga.Th e contributing elementsar efractionate db yalkal ifeldspar san dmica' s (K,Na , Si, Al).R b substitutes forK especially inmica's ,an d Ga for Al. Therefore these micro-elements are logical constituents of this factor. Rb(ppm) 240

0.9 1.1 1.3 thousands median (mm)

Figure3.1.2 . Bivariateplo t ofK > (ppm)wit hMsdia no fgrai nsiz e distributiono fAllie rsands .Dat apoint sar eindicate db y terracesymbols .

46 As these minerals may be derived both from the crystalline basementa sfro maci dvolcani crocks ,facto r2 i sinterprete da s analkal ifeldspar-mic afacto rwithou tfurthe rspecification .Th e negativeMEDIA Nfacto rloadin gindicate stha tvariatio na tsingl e sample sites by lateral sorting is probably related to the specific hydraulic properties ofmica' s and feldspars.Thi s is well illustrated by Figure 3.1.2, inwhic hK ba sa typica lmic a element is plotted against MEDIAN. Mineralogical analysis of samples of different grain size confirm these relations (Kroonenberg et al., 1988). Factor3 Factor 3, which explains 6.1% of the total variance, only contains RELALT. This suggests that bulk geochemical variation between terrace levels is partly independent from other variables, and probably reflects long-term changes in sediment supply.

Summarizing, the factor analysis indicates that variance of bulkgeochemica lpropertie so fth eAllie rterrac esand si smainl y due to an alkali basaltic component, which shows some longitudinal sorting, and to a lesser extent to local lateral sorting effects,especiall y ofmica' s and feldspars, and long- term changes insedimen t composition.

Interpretation ofPor edat a TheDor edat aar eals ogroupe d intothre esignifican t factors explaining 71.3%o fth etota lvarianc e (Table 3.1.3). Factor1 Factor 1,explain s 51.6%o fth etota lvariance ,an d contains

RELALT, A1203, Fe203, MgO, CaO, Na20, Ga, Sr, and -Si02. This factor includes elements which are known to be found in plagioclases, amphiboles and augite as major rock-forming minerals.Th epositiv eRELAL Tloadin gindicate sa decreas eo fth e factor1 element s inth ehighe rDor eterrace s (Figure 3.1.3).A s thealtitud ecriteriu mca nb edirectl yuse da sa nag ecriterium , the RELALT factor loading indicates adecreas e of the factor 1 elementswit h increasingage .

47 relative altitude (m)

100-

80- O O

O O0 0 o 60-

40- oo o

20- o oo

0 °0( £> 08 ©§o 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 Na20(% )

Figure3.1.3 . Bivariateplo to fNa^ 3(wt% )wit hrelativ ealtitud eo fDar e sands.

Since most elements of factor 1 are known as mobile with respectt oweatherin gan dpredominantl yoccu ri neasil yweathere d minerals (Bear,1964 ,Wedepohl ,1970 )facto r1 ca nb econsidere d asa weatherin g factor. Factor2 Factor2 ,explainin g11.4 %o fth etota lvariance ,contain sth e variables MEDIAN, Ti02, Fe203, Zr and Nb. These elements are commoni nstabl eheav ymineral slik emagnetite ,ilmenite ,rutil e and zircon. The interpretation of Factor 2 as a stable heavy mineral factor issupporte d by the negative MEDIAN loading, as these minerals are commonly concentrated in the finer sand fractions (Figure 3.1.4). Factor3 Factor3 ,explainin g8.3 %o fth etota lvariance ,include sth e elementsK 20,Rb ,Ba ,an dPb .Thes eelement sar efractionate db y K-feldspars and mica's.Rb , Ba and Pb can substitute for K in these minerals. If these elements originate from mica'salone , astron gnegativ eMEDIA N factorloadin gan dR bdependenc ewoul d

48 Zr(pp m 350- p

1.2 1.4 1.6 1.8 2.0 thousandsmedia n(mm )

Figure3.1.4 . Bivariateplo t of Zr (ppm)wit hMedia n of grainsiz e distributiono fDar esands . havebee nexpected .However ,a sth emedia nan dR bfacto rloading s areles stha n0.5 ,th econtributio no fK-feldspa rt oFacto r3 i s probably different from the influence of the mica component. Factor3 i sinterprete da sa stabl eligh tminera lfactor ,relate d tocrystallin e rockfragment san d fragments oflarg eK-feldspa r megacrystsfro mcoars eporphyriti cbiotit egranite scommo ni nth e Doredrainag ebasin .

Insummary ,th eDor efactor syiel ddifferen tconclusion sthe n theAllie r factors.Th emai n part ofth e totalvarianc e of the geochemical properties reflects progressive weathering of the unstablemineral swit hincreasin gag eo fth esands .11.4 %o fth e totalvarianc ei scause db ysortin go fheav yminerals ,whil e8.3 % ofth etota lvarianc e isdetermine db yvaryin gamount so fmica' s andK-feldspars .

3.1.4. Regressionwit h grain sizedat a Grainsiz ean dweatherin gar eth emai nfactor sgovernin gloca l bulk composition for both Allier and Dore sands. In order to

49 better assess the impact of these factors their effects were quantified bymean s ofmultipl e linearregression . Both Allier and Dore sands show approximately 10%o f their total bulk geochemicalvarianc e isrelate d to themedia n grain size of the sand sample. Inthre e Doregrai n sizeclasse smos t measurementsar ebelo wdetectio nlimit ,an dconsequentl yhav en o normal distribution, these fractions, 0.053-0.075, 0.075-0.106 and 0.106-0.15 (mm),ar e excluded from further analysis and interpretation.

For each grain size related element a multiple linear regression model wasmade .Fro m thetwelv e grain sizeclasses , classeswer e selected with stepwise methodology as independent variables (Dillonan dGoldstein ,1984) . Both Allier and Dore sands resulted in three significant models (at the 0.05 level). The standardized regression coefficients (Betas) and coefficients of determination (R2)o f theregressio nmodel sar epresente dfo reac hmode l (Table3.1.4) .

Elements R2 fraction Betal fraction Beta2 fraction Beta3 (wtX) (mm) (-) (mm) (-) (mm) (-)

Allier: A1203 0.63 (0.212-0.15) 0.77 (0.85-0.6)-0.2 7 Na20 0.56 (0.212-0.15) 0.39 (0.85-0.6)-0.6 9 Rb 0.60 (0.212-0.15) 0.73

Dore Ti02 0.66 (0.15-0.106) 0.67 (0.6-0.425)0.3 8 (0.075-0.053)0.3 7 Fe203 0.55 (0.3-0.425) 0.40 (<0.05 3 ) Nb 0.78 (0.212-0.15) 0.79 (0.6-0.425)0.3 1 (<0.05 3 )0.4 4 Zr nosignifican tmode la tth e0.0 5leve l Table3.1. 4 Multiplelinea rregressio nmodel sfo rgrai nsiz erelate d elements, model fit (R2) and standardized regression coefficientsfo rth egrai nsiz efraction sselecte dwit h stepwis emethodology .

A multiple regression on extracted components from the grain size data of both Allier and Dore samples resulted is similar regressionmodel swit hcomparabl emode lfits ,indicatin gtha tth e roleo fcollinearit y isver y limited.

50 3.1.5. Regressionmodellin g ofweatherin g effects TheDor eterrac e sequencei sstrongl y affectedb yweatherin g asreflecte d inalmos t52 %o fth etota lgeochemica lvariability . Theweatherin geffec ti sno ts oapparen ti nth eAllie rsands .A s Allier and Dore arebot h situated inth eLimagn e rift valley, Allieran dDor eterrace sshoul dhav eexperience dsimila rclimati c conditions andhenc ea simila rweatherin g trend. The Allier data show that both clay content and contento f factor1 elementa l ("basaltic")oxide s (MgOfo rexample )o fth e sandshav ea rathe rerrati cbehaviou rwithi nan dbetwee nterrac e levels (Figure.3.1.5) .Thi simplie stha tneithe rbasal t (parent material) nor clay (weathering product) content are simply relatedt oterrac eage ,a smigh tb eexpecte di na chronosequence . Despitethei runpredictabl ebehaviour ,"basaltic "oxide san dcla y content have an inverse semi-logarithmic relation (Figure. 3.1.6), suggesting a relationship between parent material and weathering.

MgO(% ) 5-

9D 2^

200 400 600 800 age (x10

Figure3.1.5 . MeasuredM g(O xwt% )an dag erelationshi p (accordingt o PASTKE,1986 )i nyear si nth eAllie rterrac esands .

51 clay(wt-% )

ID a

14- a 12-

a 10- a a a a a «- a D D DO 6- a • a ODD • a a D 4- a a • a D • a a a a 3 l- •D D an a D a a • c a

U "1 i ' 1 i ' i •

Figure3.1.6 . MeasuredM g(O xwt% )an dCla y(wt% )relationshi pfo rth e subsoilfluvia lsand so fth eAllier .

in order to determine the effect of parent material in the Allierchronosequenc ea multipl eregressio nanalysi swa scarrie d out. Results showed actual Mg content with the independent contributions of (1) a clay formation factor, (2) a parent material factor and (3)thei r interaction factor. Thecla yconten t seems relatedt oth eactua lM gconten tb ya semi-logarithmic relationship and is transformed by a natural logarithm In(CLAY (wt%)) (Figure. 3.1.6). As independent parent material factor only the downstream compositional sorting variability of alkali basaltic rock fragmentsca nb e included (cf.Davie se tal. , 1978;Kroonenber g et al.,1988). The interaction term is In(CLAY) . UPSTREAM. A high contribution of this term will indicate a strong relationship betweenweatherin g andparen tmateria l composition.

In this way the following multiple regression model can be established topredic t actualM g content inMg Owt% .

52 Mg= Cons t+ BjUPSTREA M+ B 2ln(CLAY)+ B 12(UPSTREAM .I n(CLAY ))

Theregressio nmode li ssignifican ta tth e0.0 5 level.Mode lfi t expressed by the coefficient of determination (R2) was 0.71. Similar regression models were made for the other basaltic elementsTi ,Fe ,Mn ,C aan dP .Al lthes eregressio nmodel swer e significant atth e0.0 5 levelan dhav emode l fitso fabou t 0.65 (Table 3.1.5).

2 Elements R Beta^ Beta2 Beta^2

Ti02 0.59 0.19 0.70 -0.64 Fe203 0.60 0.23 0.74 -0.64 MnO 0.63 0.09 0.68 -0.55 MgO 0.71 0.26 0.78 -0.75 CaO 0.66 0.29 0.75 -0.80 p2o5 0.60 0.30 0.74 -0.76

Table3.1. 5 Multiple linear regressionmodel s farbasalti c elements, model fit (R^)an dstandardize d regression coefficients. wt% basaltic element = B„ + B1*ln(wt%CXAY) + B2*OTSTREBM(km)+ B ,2 *ln(wt%CXAX)*uTSrREAM(kin)

The good model fits and the high loading of the parent material/weathering interaction term show that theAllie r sand composition is strongly affected by parent material controlled weathering. A result also found for gravel weathering in the Allierterrace s (Veldkampe t al., 1990).

3.1.6. Discussion Theresult ssho wtha tsevera lregiona lan dloca lfactor shav e a significant contribution to the actual bulk sand composition inth eAllie ran dDor edrainag ebasins .Th emos timportan tloca l factors are sorting processes and postdepositional weathering. The factors provenance and changes of sediment composition in time have a more regional character and are of interest for Quaternary research. The origin of the variability in bulk geochemistry willb ediscusse d individually foreac h factoran d

53 finally the value of the bulk geochemical methodology will be discussed inth econclusions .

Provenance

Itwa s shown that Ti02, A1203, Fe203,MgO , CaO,Na 20, P203/ V, Cr,Ni ,Cu ,Zn ,Zr ,an dS rlevel sar esignificantl yhighe ri nth e

Allier sediments, whereas Si02, K20, Rb and Pb levels are significantly higher in the Dore sediments. These element abundances are evidently related to differences in sand provenance. The Dore sands derived from crystalline basement only,hav ehighe rlevel so felement sfro mquartz ,K-feldspar san d mica's (factor 3),whil e the Allier sands derived from both volcanic rocks and crystalline basement have higher amounts of elements ofvolcani c origin (factor 1). Th eoccurrenc e ofheav y mineralsderive d fromvolcani cas hi nth eDor ebasi nreporte db y various authors (Pelletier, 1971;Va n Wijck, 1985;Etliche r et al., 1987), is not demonstrably reflected in the bulk geochemistry ofth eDor esands .

relative altitude (m)

100-

80-

60- o o o o o

40- oo o o

20-

r-°0^f— ^OofO-Or -H—|—,—,— 3.0 3.4 3.8 4.2 4.6 5.0 5.4 5.8 K20 (%)

Figure3.1.7 . Bivariateplo to f ¥/>(wt% )wit hrelativ ealtitud eo fDor e sands.

54 Changes insedimen t compositionwit htim e The emergence of relative altitude as an independent factor inth eAllie rterrace s ismos tprobabl y causedb ymajo r changes in sediment supply or composition with time. Bout (1963) and Kroonenberge tal . (1988)propose da relationshi pbetwee nclimat e (glaciated Cantal and Mt Dore) and sediment type (basalt content), large amounts of basalt are supplied to the Allier terracesediment sdurin ga meltin gphas eo fth eglaciers .Pastr e (1986) suggested that not climate butvolcanis m is responsible forth esedimen tdifference sbetwee nth evariou sterrac elevels . Inth elargel yunglaciate d Dorebasi nwher evolcanis m isabsent , nosuc hchange shav ebee nobserved ,a sshow nb yth eK 20-Altitude diagram of Figure3.1.7 .

Sorting processes The heavy mineral factor of the Dore and the alkali feldspar/mica factor of the Allier are grain size related and dependupo nth eloca lpaleo-discharg edurin g sedimentation.Fo r bothfluvia lsystem sth eloca llatera lsortin gcomponen ti sabou t 10% of the total variance in bulk geochemical characteristics. The sediments ofAllie r and Dorediffe r mainly in the elements for which grain size dependent relations are found (Allier:

A1203, Na20, K20an d Rb; Dore: Ti02, Fe203, Nb and Zr). Th e main causes for these differences are probably the contrasting provenance.Th estandardize d regressioncoefficient s (Betas)o f the multiple regression models, show high positive Betas for finer fractions and negative Betas for the coarser fractions. This suggests that the grain size related elements are predominantly present inth e finer sand fractions. This observation can easily be explained by the following processes and causes. Rb, a typical mica element, tends to concentratei nth efin esan dfractions ,du et oth especifi cshap e related hydraulic properties of mica's. Na20 and A1203, originating from feldspars, are often reported to be more abundant inth e finersan d fractions,du et oselectiv e abrasion (Riezebos, 1971;Basu ,1976 ;Odo me tal. ,1976 ; Daviese tal. , 1978; Potter, 1978;Pettijoh n etal. ,1987) .

55 The enrichment ofTi ,Zr ,N ban dL a (stableheav y minerals)i n the finer Dore sands is related to the well known and often observed relationship of specific settling velocity with high densities andsmal lprimar y grain sizeso fthes eminerals .

The UPSTR factor loading of 0.5 inth emai n Allier factor, shows that there is a decrease of alkali-basaltic fragments downstream.Becaus en oimportan ttributarie soccu ri nth eregio n sampled, the downstream decrease can be explained by density sorting orb yphysica l breakdown during transportation (Davies etal .1978 ). A densit ysortin geffec ti ssupporte db yth elarg e average bulk density ofth ealkali-basalti c fragments whichi s about 3 (g/cm3). The susceptibility of those fragments to physical breakdown isillustrate d byth ewell-rounde d shapeo f most alkali-basaltic grains. It is not certain which effect predominates.

Postdepositional weathering Inth eDor e basin factor analysis indicates that theamoun t ofmobil e elements (Factor 1)decrease s with terrace altitude, indicatingprogressiv eweatherin go funstabl eminerals .Plot so f individualFacto r1 element sagains trelativ ealtitud e (e.g. Na; Figure. 3.1.3) corroborate this. Mica-feldspar (Factor 3) elements such asK plotted against relative altitude (Figure. 3.1.7)sho wn osignifican tchange .Thi sobservatio nsubstantiate s therelativ e immobility ofK i nth eDor esands . Inth eAllie rbasin ,however ,n oweatherin gtren di sdirectl y discernable fromfacto ranalysis ,i nspit eo fth ehig hamoun to f easilyweatherabl evolcani cmateria li nth esedimen tan dth eman y weathering phenomena (Jongmans et al., 1991). But multiple regressionmodellin go fth eweatherin geffect si nth eAllie rbul k sand composition showed that the variability of the present "basalt" content is strongly related to parent material controlled weathering. Theregressio nmodel so fweatherin gfo r MgOan dCa Ohav eth ebes tfi t(highes tR 2)probabl ybecaus ethe y have easy weatherable mineral sources (olivine and Ca- plagioclases),i ncontras tt oth eothe r"basaltic "element swhic h

56 110

too

90

80

70 - FM F MC 60 -

SO C « F C FT 40 C F

so H F F M M 20 F CfF M P FC F F FF M FM OK 10 -

0 2 CoO/TI02

Figure3.1.8 . Bivariate plot of decreasing CaO/Ti02 with relative altitudeo fAllie rsards .Dat alabel sindicat emedia no f grain size distribution, C = coarse, 900-2000 im; M = medium,600-90 0jmi ;F = fine ,210-60 0/on .

110 -1

FM 100 -

90 -

80 -

F CF 70 - B MC F 60 -

50 - CC c 40 -

JO -

20 - F FC C M * C 10 - F F C M

F M F c C F ^C FMM F c 1 1 0 2 4

Co0/T102

Figure3.1.9 . Bivariate plot of decreasing CaO/Ti02 with relative altitude of Dore sands.Dat a labels indicatemedia n of grain size distribution, C = coarse, 900-2000 im; M = medium,600-90 0 im; F= fine ,210-60 0 fan.

57 have more weathering resistant basaltic mineral sources, like heavy opaque minerals. The regression models have an average coefficient ofdeterminatio no f 0.65, indicatingtha tabou t65 % of the total geochemical variability in actual "basaltic" elementsconten tca nb eascribe dt oth ecombine deffec to fparen t material and clayneoformation .

Empiricalweatherin gindice sfo rAllie rsediment s (Ca/Ti;Figure . 3.1.8;Kroonenber ge tal. ,1988 )an dfo rDor esediment s (Figure. 3.1.9) show similar trends. Apparently the great fluctuations inth eamoun to fbasalti c rock fragmentssupplie d toth eAllie r strongly influenced the gradual loss of mobile elements by weathering.O nth econtrary ,th egeochemica l composition ofth e Dore sediments remained essentially constantwit htime .

Conclusions What knowledge was obtained from our approach to analysis sandsbul kgeochemically ?Th emai nadvantag ea sw ese ei ti sth e easy quantification of bulk sediment composition allowing statistical processing of the data, facilitating the identification and quantification of sediment composition determining factors. The local grain size and weathering factors were easily determined and quantified.Th egrai nsiz e factorha s apparently onlya limite dimpac to nth ebul ksan dcompositio ni nth estudie d areas, only a few specific elements and probably some minerals too are strongly grain sizedetermined . Post depositional weathering which strongly controls the present Doreterrac esedimen tcomposition , isals oprominen ti n the Allier sands where it is strongly related with sediment composition.Th eexterna l factorswhic har eo fmor einteres tfo r Quaternary researchsee mt opla yonl ya majo rrol ei nth eAllie r basin. A significant factorwa s found which was interpreted to display climatic or volcanic induced changes in sediment composition.

The added value of the bulk geochemical approach is

58 illustrated by the facttha t inbot h fluvial systemsmor e than 50%o fth evariabilit y inbul kgeochemistr y isdu et ocomponent s which are often neglected in routine sedimentary petrography, i.e.roc kfragment san dligh tminerals .Thi sals oindicate stha t bulksedimen tgeochemistr yca nresul ti nne wan dadditiona ldat a of interest inQuaternar y research.However ,th e interpretation ofgeochemica l factors can only be done ifth e sand mineralogy isknown .Therefore ,sedimen tbul kgeochemistr yi scomplementar y tosedimentar y petrology,bu tcanno treplac eit .

59 3.2 PARENT MATERIAL CONTROLLED SUBSOIL WEATHERING IN CHRONOSEQDENCE,TH EALLIE R TERRACES.

A.Veldkam p& T.C .Feijte l Abstract The chronosequence in the Allier terraces is subject to erratic changes in parent material composition with time. Approximately 65% of the total variance in the basaltic elements content in the Allier terrace sands can be explained by the combined effect of parent material and time related neoformation of clay in the subsoil. It is shown that parent material composition, as a function of transport distance, strongly influences weathering rate in the Allier terrace chronosequence. In order to evaluate parent material controlled weathering effects a process model simulation was made based on an exponential decrease of MgO and neoformation of clay. Long term simulations with this model suggest that parent material controlled subsoil weathering is significant for sediments with easy weatherable fragments. After prolonged weathering the role of parent material decreases and becomes untraceable in strongly weathered materials.

3.2.1 Introduction River terraces are often used as chronosequences to study rateso fweatherin gan dsoi lformation .Suc hchronosequence sar e selected in order to keep soil-forming factors other then time constant (Stevens &Walker , 1970). However, the assumption of constancy ofparen tmateria l forterrac esediment s isofte nno t very realistic (Kroonenberg et al., 1990). In many terrace sequences of major rivers, like the Rhine and Meuse, the composition of the supplied sediments changed with age as a result of uplift and incision in time (van Straaten, 1946; Brunnacker & Boenigk, 1983). Within such chronosequences the terracesedimen tmaturit ycommonl yincrease swit hag ea sa resul t of both postdepositional weathering and changes in parent material. When such chronosequences are studied it is always difficult to separate the role of time and parent material in chronosequence weathering. As it can be expected that parent material strongly determinesweatherin g withina chronosequenc e we have focussed our study on the role of parent material in chronosequenceweathering .

We studied the Allier terrace sequence (Massif Central, France), a chronosequencewit h irregularchange s insedimen t composition

60 in time (Kroonenberg et al., 1988). Such a chronosequence is expected to separate the effects of parent material and time related subsoilweathering .

The goal of this chapter ist o determine and quantify the effecto fth eorigina lsedimen tcompositio no nth eactua lterrac e sand composition in the Allier chronosequence. Further it is tried to determine the control of parent material on subsoil weatheringi na chronosequenc eb ymean so fproces smodellin gwit h a theoreticalmodel .

3.2.2 Materialsan dmethod s Allierterrace s (fig.3.2.1 )ar enumbere daccordin gt oFrenc h usage fromZ (youngest)t oS (oldest).Alon ga stretc ho f4 0k m alongth eAllie rrive rsan dsample swer etake n fromth epresen t bed (Z)an d from sand and gravel pits in 4 different terrace levels (fig.3.2.1 )a tY =5 ky ,X =1 5an d3 0ky ,W = 300k yan d V= 900k y (Pastre,1986 ;Raynal , 1984).

||%++*J Hercynian basement

* Isolated volcanic plug | j 01igocene sediments

| 1 Holocene basins (marais |:;;;; ;|y and z terraces |:::::::'::;] x terraces [jiij:::::^ Wterrace s

l::::::::-:-:-^ v terraces i;j:;;;;:j|::lt an d U terraces E::'?-:-iv$s terraces 0 sample site

Figure3.2. 1 Mapo fth e survey area, showing theAllie r andit s terraces. 61 At each sampling site 2-6 samples of different grain size distributionwer ecollected ,amountin gt oa tota lo f4 9samples . Carewa stake nt osampl ea smuc ha spossibl esan dfro mforese t laminae in small-scale cross-bedded sets, in order to avoid concentrations of heavy minerals inhorizontall y laminated lag deposits. Samples were taken as deep as possible (subsoil) to avoid effects of soil formation. In the laboratory, detailed granulometric analyses were done (15 fractions). To compare chemical compositionwit hopticall ydeterminabl emineralog y and to avoid the effect of secondary products, organic matter, calciumcarbonate ,fre eiro nan dal lmateria l<1 6ju mwer eremove d froma representativ esub-sample ,separate db ymean so fa sampl e splitter.Organi cmatte rwa sdestroye dwit hH 202an d 'free'iro n wasextracte dwit hNa-ditionite-EDT Aa tp H4.5 .Althoug hth ecla y andfin esil tconten twa smeasured ,th edecantate d fraction< 1 6 nmwa sno tanalyze d geochemically. After this pretreatment, 0.6 g of the sample was separated with a micro-sample splitter, fused with 2,4 g lithium tetraboratean danalyzed .Th emajo relement sSi ,Ti ,Al ,Fe ,Mg , Ca, Na, K and P were analyzed with X-ray fluorescence spectroscopyo na Philip sXR Fassembly .Th esyste mi scalibrate d using USGS geochemical standards as listed by Abbey (1980, p. 16).

In order to determine whether parent material has a significant contribution to the actual terrace sediment composition of the Allier chronoseguence amultipl e regression analysis is carried out (Dillon & Goldstein, 1984) with the SPSSpc package (Norisus, 1986). Thefollowin gste pi nou rapproac hi st oevaluat eth epossibl e effects ofparen t material controlled weathering by means ofa theoretical process response model. Modelling was done in accordancewit hchemica l kinetics.A simplifiedparen tmateria l controlled weathering model was made where basalt containing sedimenti sweatherin gunde rcertai nstric tassumptions .A sbasi s forthi smode l a spread sheetmode l of Feijtel &Meije r (1990) was used who used a similar modelling approach as Levine &

62 Ciolkosz (1986).

3.2.3 Results and discussion

Regression modelling onbul kqeochemica ldat a

Geochemicalvariabilit yo fth eAllie rsand si slarges ti nTi , Fe, Mn,Mg ,C aan dP ,a sa resul to fvaryin gamount so fbasalti c rock-fragments. Also weathering effects, on which this study focuses, are most conspicuous in the basaltic elements (Kroonenberg et al., 1988; Jongmans et al., 1991; Veldkamp & Kroonenberg,1989) .Th emean ,maximu man dminimu mconcentration s ofthes eelement s (Oxwt% )o fth efractio nbetwee n1 6ji man d200 0 Hmar eliste d inTabl e 3.2.1.

Var. Mean Minimum Maximum Standard value value Deviation

Ti02 0.81 0.24 1.78 0.42 Fe203 3.36 1.16 7.87 1.82 MnO 0.04 0.01 0.10 0.03 MgO 1.63 0.36 4.60 1.19 CaO 2.10 0.49 5.38 1.34 P2°5 0.19 0.07 0.44 0.10 Clay 4.6 0.1 15.5 3.3

Table3.2. 1 Means,minimum ,maximi man dstandar ddeviation so fbul k basaltic elementsan d claycontent so f 49Allie rsan d samples.Elemen tvalue sar ei noxid eweigh tpercentages , fraction< 1 6/j mi nweigh tpercentage .

Previousresearc hshowe dtha tth ebasalti celemen tvariabilit y inth eAllie r sands isindependen to fchange s insamplin g depth and grain size (Kroonenberg et al.,1988 ;Veldkamp , 1990). The lack of correlation between basalt content and median is explainedb yth efac ttha treworking ,concentratio nan ddepletio n of basaltic grains take place simultaneously in sands of all differentgranulometri c characteristics.Th einsignifican trol e of sampling depth for each terrace level indicates that all samples were taken deep enough to avoid direct soil formation effects.

63 Thefac ttha tth eloca lsit efactors ,grai nsiz ean dsamplin g depth, have no significant contribution to the variability in basaltic elements, makes these elements suitable to study regionalweatherin g effects inth eAllie rchronosequence .

Weathering of basaltic sand particles in Allier sediments results in an overall loss of Ti,Fe ,Mn ,Mg , Ca and P in the sand fraction, and an increase of the amount of clay in the terrace (Pastre, 1986;Feijte l et al., 1988;Jongman s et al., 1990). Change si nbasalti cfragment sconten tar ewel lillustrate d by Ca and Mg content, because they originate from the easier weatherable minerals in these fragments. We use Mg as representative element for basaltic particles in our study. Weathering ofbasalti c sandparticle s istherefor edescribe d in thispape r as adecreas e inM g content and an increase incla y content in the bulk sample, though strictly speaking clay formation isno tnecessaril yth eresul to fth eweatherin go fMg - bearingminerals .

MgO(% ) 5- B

3-

2-jl -•a7 a

1T!] a a 1 § - 1 1 " I 0• i • i • I u 200 400 600 800 age (x10

Figure3.2. 2 MeasuredM g(O xwt% )i nth eAllie rterrac esand swit hag e (ageaccordin gt oPastre ,1986 )i nkyears .

64 Our data show that both Mg and clay content of the Allier terracesand ssho wa rathe rerrati cbehaviou rwithi nan dfro mon e terraceleve lt oanothe r (fig.3.2.2) .Thi simplie stha tneithe r Mgno rcla yconten tar esimpl y related toterrac eage ,a smigh t be expected in a chronosequence. Despite their unpredictable behaviour,M gan dcla yconten thav eroughl ysee na ninvers esemi - logarithmic relation (fig. 3.2.3), suggesting a relationship betweenparen tmateria l andweathering .

clay (wt%)

ID" D

14- D

12-

D 10- D

D D 8- D • a a a DO 6- a D D DDE 0 D a 4- • a IB a D D a a a D a 2- D m o an D D D a D (J I I i ( () 1 2 3 4 5 MgO(wt% )

Figure3.2. 3 MeasuredM j(O xwt% )an dCla y(wt% )relationshi pfo rth e subsoilfluvia lsand so fth eAllier .

Inorde rt odetermin eth eparen tmateria leffec ti nth eAllie r chronosequence a multiple regression analysis is carried out, predicting the actual Mg content with the independent contributions of (1) a clay formation factor, (2) a parent material factor and (3)thei r interaction factor.

(1)Th e clay content seems related toth e actual Mg content by a semi-logarithmic relationship (fig.3.2.3 )an d istransforme d by anatura l logarithm ln(CLAY (wt%)).

65 (2) As independent parent material factor the downstream compositional sorting variability of alkali basaltic rock fragments is included (cf. Davies et al., 1978). Within the studiedAllie rsectio nth edownstrea mdecreas eo fbasalti c grains is almost linear (Kroonenberg et al.,1988). Thistren d is caused bydownstrea m density sorting andb yselectiv ewea ro fth e softer basaltic particles in comparison to harder quartz and feldspar grains. This is the only known variability in original parent material composition which is still traceable. Of course other factorslik eprovenanc ean dvolcanis mhav edetermine d variability in parent material but these factors cannot be reconstructed quantitatively any more. The sorting effect is expressed as UPSTREAM,th edistanc e inkilometre supstrea m froma chose n point in the Allier. (3) The interaction term is the product of the independent contributions of clay formation and parent material, In(CLAY) . UPSTREAM. This term gives an indication of the interaction between the two terms and is used as an independent variable. A high contribution of this term to the regression model will indicate a strong relationship between parent material and weathering. In this way the following multiple regression model can be established predicting actual Mg (in MgO wt%) content.

Mg= Const + BiUPSTREAM + B2ln(CLAY) + B12(UPSTREA M .I n(CLAY)) . (1)

Theregressio nmode lwa s significant atth e 0.05 level.Mode l fit expressed by the coefficient of determination (R2) was 0.71. Predicted Mg contents plotted with age (fig. 3.2.4) show a good correspondence with measured Mg content with age (fig. 3.2.2). Furtheranalysi s ofth eresidual s ofth eregressio nmode l did not reveal any new relationship. A plot of the residuals with age (fig. 3.2.5) shows that there are some differences for the different terrace levels. The X terrace has mainly positive residuals while the Z terrace has mainly negative residuals. These differences suggest different individual terrace

66 predicted MgO(wt% ) 5-

I DC -a at H

200 400 600 800 age(Pastre ,1986)(x10 )

Figure 3.2.4 Predicted Mg (Ox wt%) with age (according to Pastre, 1986) in kyears.

MgO wt% Residuals 3

-3 200 400 600 800 Age (Kycars) Figure 3.2.5 Residuals of regression model with age (according to Pastre, 1986) in kyears.

67 characteristics which aremos t probably caused by factors not considered in the regression model such as provenance and volcanism. Similarregressio nmodel swer emad efo rTi ,Fe ,Mn ,C aan dP . All these regression models were also significant at the 0.05 level andhav egoo dmode l fits (Tab. 3.2.2).

2 Elements R Beta^ Beta2 Beta^2 Ti 0.59 0.19 0.70 -0.64 Fe 0.60 0.23 0.74 -0.64 Mn 0.63 0.09 0.68 -0.55 Mg 0.71 0.26 0.78 -0.75 Ca 0.66 0.29 0.75 -0.80 P 0.60 0.30 0.74 -0.76

Table3.2. 2 Multipl elinea rre t jressionmodel sfo rbasa l modelfi t(R 2)an dstandardize dregressio ncoefficient sfo r thefunction : wt% basaltic element = B„ + B1.ln(wt%OA30 + B2.UFKEREaM(km)+B ,2 .in(wt%CXAY).uPSTREaM(km ) The regression models forM g and Ca have the best fit (highest R2) probably becausethe y haveeas yweatherabl eminera l sources (olivine,augit ean dCa-plagioclases ). Th eregressio nmodel shav e anaverag ecoefficien to fdeterminatio no f0.65 ,indicatin gtha t about65 %o fth etota lgeochemica lvariabilit yi nactua lbasalti c elementsconten tca nb eascribe dt oth ecombine deffec to fparen t material and clayneoformation .

As it is inappropriate to interpret the regression coefficients (B) as indicators of relative importance of variables,th edimensionles sstandardize dregressio ncoefficient s (Betas) are calculated (Tab. 3.2.2). They allow a comparison between the contribution of each independent variable. These standardized regression coefficients indicate a relative large contribution of the interaction term (high negative values) to the overall model.Thi s result indicates, as already suggested by fig. 3.2.3, thatweatherin g product and parent material are strongly interrelated. This means that parent material significantly determines weathering rate within the Allier chronosequence,a resul twhic h agreeswit hth ecommo n knowledge onweathering .

68 TheoreticallyModellin go fparen tmateria lcontrolle d weathering Inmos t chronosequences weathering products increase ina non linear manner with time, and primary minerals decrease correspondingly (Goldich, 1938; Hay, 1960; Ruxton, 1968; Lowe, 1986). The resulting exponential or semi-logarithmic equations are thought to be in accordance with chemical kinetics (Yaalon, 1975). We also tried to model in accordance with chemical kinetics. We made a parent material controlled weathering model where basalt containing sediment is weathering under certain strict assumptions. Each assumption is listed and itsvalidit y for the Allier chronosequence is pointed out. The weathering process model is based upon the following nine assumptions:

(1) Weathering is decomposition of primary minerals, leaching losses and clay neoformation. (2) Weathering is confined to basaltic rock fragments. This assumption is corroborated for the Allier sediment by Jongmans et al. (1990,1991), who reported the volcanic fragments as major source for coatings of secondary weathering products. Although differential weathering susceptibility ofth evariou sbasalti cmineral s is observed (Veldkamp et al., 1990), the process simulation model treats the basaltic fragments as a whole. (3) The model is only valid for subsoil sand deposits, i.e. interstratal or vadose zone weathering. This assumption is included to obviate the need for considering vertical profile development in the upper few metres within individual terraces. This assumption is justified for the Allier sediments to some extent as no significant effect of sampling depth has been observed (Kroonenberg et al.,1988). (4) The model is valid for one million years only. This is the approximate age of the oldest sampled Allier terrace (Pastre, 1986). (5) Constant weathering in time.

69 As it isto o speculative to include climatic changes within the simulation model, the effects of weathering are considered tob e constant and cumulative. Strictly speaking this assumption is not correct in the Allier basin. During the last million years the region has known many climatic changes which certainly affected water quality and temperature (permafrost) of the vadose zone (Bout, 1963; Veyret, 1978). (6) No net clay illuviation or eluviation. This assumption is supported by the fact that our samples originate from sediments (beyond 2-3 meters) where no distinct change in total clay content has been observed in the Allier terraces (Jongmans et al., 1991). (7) Primary clay content is constant (2w t %). Many terraces are stratigraphically complex, with primary variations incla y contentwhic h are not taken into account within the model. (8) Chemical composition of basalt and clay is constant. This assumption if of course also incorrect for a real system. The average compositions are based on measurements in the Allier sequence given by Jongmans et al. (1991). (9) Neoformation rate is not hampered by availability of other elements than Mg. Although most Mg is usually lost to the groundwater it is assumed that only Mg is limiting in respect to the neoformation rate.

Different simulations with changing initial compositions (basalt content) allow an evaluation of a real chronosequence like inth eAllie rbasin .Th emode l istune d by adapting the rate parameters ofbot hdissolutio n and neoformation processes (kxan d k2) to obtain comparable basalt and clay quantities as measured in the Allier terraces. Clay and MgO contents of the older terraces and the unweathered sands were used As standard reference values. This trial and error tuning was done because no comparable subsoil longterm weathering rates and rate parameters areknown .Mos tweatherin g ratesar ebase do n detailed studies under well defined conditions. Studies which may give

70 comparable weathering rates asuse d in ourmode l are catchment studies.Thes e studieshav eth edisadvantag etha tthe y register only actual weathering environmentswhic h areprobabl y not the samea slongter mweatherin g rates.Th euse ddissolutio n rateo f Mg (7x 10"5k gMgO/(h ayr) ) isno tdirectl y comparablewit han y known dissolution rate. When this dissolution rate is recalculated into a geochemical denudation rate a value is obtained (2x 10"3 kgMgO/(h a yr)) which isa factor 10*t o10 5 lessthe nmeasure dvalue s inth ecatchmen t studiesa spublishe d in Colman & Dethier (1986). This large difference is mainly causedb yth eassumptio n thatal lM g isbuil d incla yminerals . In reality most weathered MgO seems to dissolve in the ground water. Parentmateria l controlled weathering isincorporate d inth e modelb ymakin gth erat eparamete ro fneoformatio n (K2)dependen t onth eamoun t ofunweathere d basalt (M(b)). Thesimulatio nresult so fthi stheoretica lan dver ysimplifie d model have conceptual validity only because no realistic conditions are simulated. Only possible effects of parent material controlled weathering are simulated.

Thebasi c quantitative assumption isthat : Mg (basalt) insan d fraction (awt .%Mg )— > Clay (bwt .%MgO )

Theequation sused :

Dissolution(t) = Kx .M(b ) (2)

Kx= k i- a exp(t ) (3)

Where: M(b) =Mas savailabl e primary Mg (kgMgO/ha )

Kx =Rat eparamete r ink gMgO/(h a yr) 5 kx = 7x 10" k gMgO/(h a yr) forM g a = scale correction factor, 5 x 10"6k g MgO/ha forM g exp(t)= tim edependenc e ofdissolutio n in (yr)"1

71 Neoformation(t)= K2 . M(d) (4)

K2 = k2/M(b) (5)

Where M(d) Mass available dissolved Mg (kg Mgo/ha)

K2 Rate parameter in kg MgO/(ha yr) k, 2 (kg MgO)2/(ha2 yr) for Mg. kg and ha are

squared because K2 (equation 5) has kg MgO/ (ha yr) as units. M(b) Massavailabl eprimar yM g (kgMgO/ha )

In order to run themode l aspecifi c volume of sediment with a 'realistic'primar y compositionswer echosen .

Thickness of sediment layer (fixed) 10 cm Initialbul k density of sediment layer 1.5 g/cm3

Initial basalt content in sediment = 42 wt% (S0) , 35 wt%

30 wt% (S2) , 25 wt%

(S3) Mg content inbasal t 8 wt% MgO Mg content incla y 1.9 Wt% MgO

^ 3.4

600 time (thousand, yrs) Figure 3.2.6 Decreaseo fbul ksedimen tM g(O xw t% )conten twit htime , forfou rsimulation swit hth eproces smodel . 8-

6- j/' ^<^

k- _^V s——————

1 2- i i — 1 1 1 200 400 600 time (thousand, yrs)

Figure 3.2.7 Increase of neoformed day (wt%) in time, for four simulations with the process model.

3.0 3.U MgO(wf% ) Figure3.2. 8 Totalweathere dbasal t(M gi nO xwt% )wit htota lne wforme d clay (wt%),fo rfou rsimulation swit hth eproces smodel .

73 Thesimulatio n results,whic hshoul db eviewe d qualitatively only,ar epresente di nfigs .3.2. 6 and3.2.7 . Figure3.2. 6 shows the decrease of bulk Mg concentration, as a measure of the decreasing basalt content in time. Four simulations (S0 toS 3) arerepresente dwit heac hhavin ga differen tinitia lMg O (basalt) contento fth esediment . Thesedifferen t simulations showtha t sediments with a relative large content of basalt have faster weathering ratestha nsediment swit hlowe rbasal tconten tunde r similarconditions .I ti sals o illustrated thatafte rprolonge d weathering allterrace s remainwit hcomparabl ebasal tcontents . Figure3.2. 7 showingth eincreas ei ncla yconten tdurin gth efou r weathering simulations, illustrating the stronger increase in clay content of the sediments rich in easily weatherable components.Th ecurve so fFig . 3.2.7 alsosho wa larg e increase in clay content at the start of basalt rich sediment (S0) weathering,indicatin ga hig hrat eo fcla yformatio na tth estar t oftha t simulation.Afte rprolonge dweatherin g therat eo fcla y formation decreases and the difference in original sediment composition remains only traceable by the amount of neoformed clay, an unfortunately unmeasurable variable in a real chronosequence.Th eultimat eweatherin gstag ewil lb ea complet e alteration ofth ebasal t particles into clay. Such a stage is reported fromth eCaquet abasi n inth eColombia nAmazons ,wher e strongtropica lweatherin g erased alldifference s inprovenanc e leaving onlythic k clay soils (Kroonenberg etal. , 1990). Fig. 3.2.8, showing the Clay-Mg relations, displays qualitatively comparable relationship as the measured curve of fig. 3.2.3. Although the resulting Clay/Mg curves are not matching the measured effect, it illustrates that this type of relationship is most probably the result of parent material controlled weathering.

The simulation results suggest clearly that parent material composition has especially a strong weathering effect in sedimentsric hi nbasalti cparticles .Afte rprolonge dweathering , moretha na 0. 5millio nyear si nth emodel ,th ebasalti cfragmen t contentdecrease san ddifference si ninitia lparen tmateria l (S0 toS 3)ar ealmos terase dleavin ga chronosequenc ewit hver ysmal l

74 differencesi nMg Oconten tbetwee nth edifferen tterrac elevels .

3.2.4 Conclusions

Theanalysi so fth eAllie rterrac echronosequenc eshowe dtha t parent material can strongly determine weathering within a chronosequence.Th eeffect so fparen tmateria lcompositio nshoul d therefore not be underestimated in the study of other chronosequences. Long term simulations with a simplified theoretical process model of parent material controlled weathering indicate that parent material controlled weathering is only prominent in •young'sediment swit hman yeas yweatherabl efragments .Sediment s with many weatherable components will show faster weathering rates than sediments with less weatherable components. After several 100 thousands of years of simulated weathering the differences in weathering rate due to the parent material decrease and becomes finally untraceable in very old deeply weathered chronosequences. The differences in initial parent materialcompositio nremai ntheoreticall ytraceabl eb yth etota l amount of neoformed clay during weathering, but, unfortunately this is an impossible measurable variable in real chronosequences.

75 3.3 SPATIAL VARIABILITY IN FLUVIAL SAND COMPOSITION AT THE ALLIER/DORECONFLUENCE .

A.Veldkam p &I.G .Staritsk y

Abstract Spatial sand compositional variability in Allier (Limagne, France) terrace sediments is investigated and compared with spatial trends in the present river bed caused by sediment transport processes. In the actual Allier riverbed complete sediment mixing at the confluence with its tributary the Dore is delayed for approximately one kilometre, while the Younger Dryas terrace sands at this confluence display a delay of several kilometres. Geostatistics allow a discrimination of mixed and unmixed sediments within the same floodplain thus allowing better representative sampling for palaeohydrological research. The delay of complete sediment mixing at a confluence as displayed in sediment composition may be related to the amount of river channels and their sinuosity and floodplain width during deposition.

3.3.1 Introduction River junctions are points of significant changes within a fluvial system. The site at which two rivers meet is marked by highly complex three-dimensional patterns of flow and sediment transport and deposition. The riverbed morphology reflects the complex flow dynamics and distinctive patterns of sediment transport within the confluence, which are largely controlled by junction angle and the ratio of the discharges between the confluent channels (Mosley, 1976; Best, 1986; Bathurst, 1988). Experimental results of Best (1986, 1988) and Roy & Bergeron (1990) showed that sediments from the tributary channel are preferentially concentrated along a distinct pathway and deposited in separate bars, displaying delayed sediment mixing.

Within palaeohydrology it is generally known that terrace sediments near former confluences have a complex spatial composition, displaying both local and regional variability. The local variability is mainly caused by sediment mixing processes of the main and tributary sediment flows, and the regional variability predominantly caused by past environmental changes. Because the palaeohydrologist is commonly interested in fluvial dynamics related to paleoenvironmental changes, samples near confluence terraces are commonly avoided (Green & McGregor, 1986). It is difficult to discern mixing effects from more

76 regional effectsbecaus ether e isonl ylimite dknowledg e onth e quantitative effectso fsedimen tmixing .Thi slimitatio n isdu e inpar tt oth ediscrepanc ybetwee nth eshor tter mexperiment si n wellcontrolle dshoot so rbrooklet san dth elon gter mmacr oscal e studies inpalaeohydrology .

Inorde rt obette runderstan dth elarg escal esedimen tmixin g at a confluence, the spatial effects of such mixing are investigated. The goal of this chapter is to discern in the Allier riverbed and its associated terrace the spatial compositional variability componentcause db yth emixin g oftw o distinct sedimentflows .

This study will focus first on the spatial mixing characteristicsi nth epresen triverbed .Subsequentl yth espatia l compositional trends of terrace sands are investigated and compared with the riverbed sands of the same river allowing a more reliable interpretation ofth esedimen tmixin geffects .

ISlBar l~~ I riverbed • sample site [771 Late Glacial terra I I Holocene deposits —r** river

Figure 3.3.1 Study area with sample sites inAllie r riverbed and Younger Dryasterrace .

77 3.3.2 Study area The study area (Fig.3.3.1 ) includesth eactua l riverbedo f Allier river and its Younger Dryas terrace (Xb) (10 m above presentriverbed) ,a tth econfluenc ewit ha majo rtributary ,th e Dore.Previou sbul kgeochemica lstudie s (e.g.Kroonenber ge tal ., 1988) inchapte r 3.1 showed that basaltic rock fragments cause significantly higher levels ofTi0 2, Fe203, MgO, CaOan dP 205i n theAllie r sediments,wherea s theDor e sands are characterized by significantly higher Si02 and K20 levels, originating from quartz,K-feldspar san dmica's .Change si nbul kgeochemica lsan d compositionar ewel l illustratedb yCa Ocontent ,concentrate di n the basaltic rock fragments, and K20 content, concentrated in crystalline rock fragments. Ca and K contents are virtually independento ngrai nsiz ebecaus ethei rhos tmineral s arefoun d inalmos tal lgrai n size fractions (Veldkamp, 1990). A major difference between the present riverbed and the Younger Dryasterrac e sediment composition isth ehighe r amount ofCa O(basalti cfragments )i nth eterrac esands .Thi sdifferenc e has a climatic origin as the Younger Dryas sediments have a fluvioglacial origin from melting glaciers in volcanic areas (Bout, 1963;Kroonenber g et al., 1988). The actual riverbed sediments arerelativ e poor involcanic s asthe y have noextr a volcanic addition frommeltin gglaciers .

3.3.3 Materials andmethod s

Sampling procedures andmeasurement s Sand sampleswer e taken frombar s inon ekilometr e riverbed at the actual Allier/Dore confluence (Fig. 3.3.1) yielding a total of4 5 sampl esites . As the investigated river bars are predominantly gravelly, almost each sand bar in the investigated river section was sampled. TheYounge r DryasAllie r terrace nearth eAllier/Dor e confluence was sampled along a stretch of 15 km in sand and gravelpit samountin gt oa tota lo f6 7sample sa t1 6site s(Fig . 3.3.1). At every sample site 1-4 samples of different grain size

78 distributions were collected, to incorporate a possible grain sizeeffec tdu et oloca l sortingprocesse sdurin g sedimentation

(Veldkamp,1990) .San dbul kelemen tconcentration so fSi0 2,Ti0 2,

A1203,Fe 203,MnO ,MgO ,CaO ,Na 20,K 20an dP 205,wer emeasure dwit h X-ray fluorescence spectroscopy.Detaile d samplingmethodology , laboratory treatments and measurements were done according to Kroonenberg et al. (1988).

Spatialvariabilit y procedures Changesi nsan dcompositio na tth eAllier/Dor econfluenc ear e described by longitudinal profiles showing the average sand compositions of the sample sites onth e right and left side of theriver .Spatia lvariabilit yi squantifie dwit hbot hgeologica l clustering and geostatistics. In geological clustering it is assumed that a relationship exists between spatial sediment composition and geomorphological units,suc h as gravel bars in thepresen triverbe dan dterrac eunit sa smappe d (Fig3.3.1 )fo r the Younger Dryas terrace level. Within the geostatistical approach the spatial dependence is characterised by means of semi-variograms. In geological clustering the mean sand composition is determinedfo reac hindividua lsample dgeomorphologica lunit .Th e unitcomposition sar eclassifie d intothre ecompositio nclasses . The significance of differences in composition class is determined by means of multiple comparisons according to the Tukey method assuming independence between the distinguished units (Dillon &Goldstein , 1984). Geostatistics starts with calculation of semi-variograms, after which the nuggets and slope angles for the spatial dependenciesar eestimate d (Journel& Huijbregts ,1978 ;Stei ne t al., 1989).T oinvestigat ewhethe rvariable schang emor erapidl y in one direction than in another direction, directional semi- variograms were used. Kriging was used (Burgess et al.,1981 ) topredic tvalue sa ta regula rgri dwhic hi suse di ntur nfo rth e computerized construction ofmaps .

79 3.3.4 Results and discussion

Present riverbed Descriptivestatistic so fthe'presen triverbe dsand sar egive n in Table 3.3.1. Longitudinal profiles of mean CaO and K20 contents (Fig.3.3.2 )i nth epresen triverbe dsho wth edownstrea m changes in sand composition of the riverbar s onth e rightan d left side and in the centre of the river. After about one kilometre downstream there is no clear difference between the three distinguished bar types. These trends suggest delayed sedimentmixin go fth erigh tan dlef tban ksid eriverbars ,whil e thebar s inth erive rcentr edispla y amixe d sand composition. Cat3 (wt%)

4-

3-

2-

1- —*-— • —

0- 1 1 i 1 ' 1 ' I i 1 300 600 900 1200 1500 K20(wt% ) —• right side river - * centre river —D left side river

3-

0 -i 1 1 1 1 1 1 1 1 1 0 300 600 900 1200 1500 distance downstream confluence (m) Figure3.3. 2 Longitudinal trends of mean CaO and I^O contents of the riverbars in the middle (M) and on the left (L) and ricjit (R) river bank side of the AUier. 80 Element All bars | Allier bars| Dorebar s 1 Mixed bars (wt%) mean std | mean std | mean std | mean std

Si02 77.40 3.68 73.54 3.72 78.75 3.28 76.58 2.60 Ti02 .51 .31 .98 .40 .40 .23 .52 .12 A1203 10.86 1.21 11.11 1.23 10.66 1.10 11.17 1.34 Fe203 2.27 1.49 4.43 1.99 1.68 1.06 2.38 .59 MnO .03 .02 .06 .03 .02 .02 .03 .01 MgO .98 .49 1.69 .31 .74 .43 1.10 .19 CaO 1.26 .54 2.06 .28 .94 .43 1.50 .17 Na20 1.44 .20 1.49 .13 1.39 .20 1.54 .20 K20 3.68 .56 3.03 .28 3.99 .50 3.23 .16 p2o5 .14 .05 .21 .05 .11 .05 .16 .02

Table 3.3.1 Descriptive statistics present riverbed. Geological clustering

The mean CaO and K20 river bar concentrations were classified into three classes (Fig. 3.3.3). The unmixed Allier sands are

rich in CaO (>2 wt%) and poor in K20 (<3 wt%) , while sands with a Dore signature have opposite bulk geochemical characteristics, poor in CaO (<1 wt%) and rich in K20 (>4 wt%) .

K20 (wt%) in sand CaO (wt%) in sand pTTTTTTn >2 E33-4 ESD1"2 ^<1

Figure 3.3.3 Geological clustering maps of sand composition in river bars at Allier/Dore confluence.

81 Doresand sar efoun di nth eDor ean dalon gth eeaster n (right) riverban ko fth eAllier ,whil ealon gth eopposit ewester nban k sandswit h typical unmixed Allier characteristics are found.A river bar in the middle of the Allier has sand with a mixed origin (CaO 1-2 wt%;K 203- 4 wt%) illustrating theexistenc eo f sedimentmixin g atan d downstream theAllier/Dor econfluence .

Spatial variability Since semi-variograms were expected to differ in two directions, parallel to the river and perpendicular to it,tw o directional variogramswer e compiled for the variables studied (Staritsky,1989 ). N osignifican tanisotropi cspatia lvariabilit y was found, however. The two directional semi-variograms of the

CaO and KaO content are equal, indicating that there is no relationshipwit hth eflo wdirectio n ofth erive r (broken lines in Fig. 3.3.4 and 3.3.5). This rather surprising effect is probably due to the sharp bend of the river causing each directional semi-variogram to have a component parallel and perpendicular toth e flowdirection . 0.6

0.5

o (3 0.4

0.3

0.2

0.1 -

0.0 ' i ' i i i i i i i i i— 0.0 0.5 1.0 1.5 2.0 separation distance Figure3.3. 4 CaOSemi-variogram so fth espatia ldependenc eo fth esan d compositioni nth epresen triverbe d( ),an dsemi - variogramo fflo wdirectio nwit hadapter )gri dcell s( ), illustrating dependence of variability with flow direction. 82 0.7

0.6

0.5

c 0.4 - o • > E 0.3 -

0.2 -

0.1

0.0 i I I I I I I 1 1 1 1 1 r- 0.0 0.5 1.0 1.5 2.0 separation distance Figure3.3. 5 Kp Semi-^variogramso fth espatia ldependenc eo fth esan d ccnpositioni nth epresen triverbe d( ),an dsemi - variogramo fflo wdirectio nwit hadapte dgri dcell s( ), illustrating dependence of variability with flow direction. Toinvestigat ethi sstatemen ta semi-variogra malon gth eflo w direction was made using a transformed grid with grid lines parallel to the Allier (continuous lines in Fig. 3.3.4 and 3.3.5). These semi-variograms have high variability at short distanceswhic h iscause db yth eprojectio n ofbot h riverside s ona gri dparalle lt oth eflo wdirection .Th ene wsemi-variogram s havea muc hstronge rspatia ldependenc efo rth elarge rdistance s than the semi-variograms of the actual grid cells, indicating thatther e exists,a sexpected ,a directiona l dependence ofth e spatial sand compositional variability ina river . Withth esemi-variogram s ( )o fFig .3.3. 4 and3.3.5 , maps of the predicted CaO and K20 content are made with an isotropic kriging procedure. These maps show the spatial

variability ofCa Oan dK 20conten twithi nth egrave lbar s inth e riverbed (Fig. 3.3.6). Thestandar ddeviatio n ofth epredictio n errorsar eal laroun d0. 5wt% .Downstrea mfro mth econfluenc eth e bars situated in the middle of the Allier display a mixed composition,whil etypica lDor ean dAllie rsan dbar sremai nalon g theedge s ofth e riverbed.

83 Predicted CaOconten t Cu)f.3 [~~] < 1.0 1.0 - 1.5 11.5- a. 0 >2. 0

N

100m,

Predicted K20conten t Cut*] | 1 <3. 0 |3.0- 3. 5 |3.5 -4. 0 I > 4.0

Figure3.3. 6 Mapso fth egenera lsan dccmpositiona ltrend si npresen t riverbed.

84 This is in agreement with the maps of Fig. 3.3.3. The more downstream situated bars have a more homogeneous composition illustrating amor ethoroughl y mixed character.Th ekrige dmap s showmor edetail so fth echange si nsan d compositionwithi nth e sandbars .The yals od osho wth edelaye d sedimentmixin gi nmor e detail compared toFig .3.3.3 .

Origino fspatia l variability Spatialsan dcompositiona lvariabilit yi nth epresen triverbe d is clearly caused by the sediment mixing at the Allier/Dore confluence (Church & Kellerhals, 1978;Best , 1986;Bathurst , 1988). The observed compositional trends clearly show delayed sedimentmixin ga sobserve db yothe rconfluence s (Komura,1973 ; Best, 1986). Sands with amixe d composition arefoun d inth ebar s inth e centre of the Allier and at the first major river bend (Fig. 3.3.3 & 3.3.6), thus indicating that sediment mixing starts immediately buta complete mixing isdelaye d fora t leaston e kilometrei nth epresen tAllie rriverbed .Thi sdela yo fsedimen t mixingi smos tprominen tfo rth eba rsediment salon gth eriverbe d edge. This implies that terrace sediments are more likely to displaya maximu mdela yo fcomplet esedimen tmixin gdurin gthei r deposition.

Younger Dryasterrac e Downstream changes inmea n terrace sand composition onth e rightan dlef trive rban k (Fig.3.3.7 )sho wdifference s forth e two populations. Only many kilometres downstream the presumed former confluence,difference s between therigh t and left bank terraceunit sdiminish ,suggestin g adelaye d sedimentmixin go f atleas tte nkilometres .Spatia lvariabilit y withinth eYounge r Dryas terrace is determined only by means of geological clusteringdu et oth elimite damoun to fsampl esite s (n=16).Th e units used areth edifferen t terrace units from thegeologica l map. The sampled terrace units (Fig. 3.3.1) are classified according tothei rmea n sand geochemical composition.

85 CaO(wt%

T ' 1 ' 1 ' 1 ' I 4 8 12 16 20 distance downstream former point of confluence (km)

Figure3.3. 7 Longitudinaltrend so fmea nCa Oan dKg Ocontent so fterrac e unitso nth elef t (L)an drigh t (R)han dsid eo fth e Allier.

Themea n CaO andK 20conten t ofth eunit s isdivide d intothre e classesshow no nmap so fspatia lCa Oan dK 20trend s (Fig.3.3.8 ).

TheAllie rsediment sar eric hi nCa O (>3wt% )an dpoo ri nK 20(< 3 wt%) while Dore sands display opposite characteristics. Sands with Dore characteristics (CaO <2 wt%; K20 >4 wt%) are found along the eastern valley slope a fewkilometre s downstream the presumed former Allier/Dore confluence. The occurrence of volcanicfragment si nthes esediment sindicate stha tsom emixin g withAllie r sediments tookplac ebefor edeposition .

86 Figure3.3. 8 Geologicalclusterin gmap so fsan dcompositio ni n Younger Dryas terrace sediments at Allier/Dore confluence.

Sedimentswit h strongAllie rcharacteristic sar efoun dalon g the opposite western valley slope toabou t 10k mdownstrea m of thepresume d formerAllie r Dorejunction .

Comparison of spatial compositional trends ofpresen t riverbed and terracesands . Althoughth esedimen tcompositiona lpatter n(Fig .3.3.8 )foun d in the Younger Dryas terrace sands is only tentative, the resemblancewit hth epresen triverbe d (Fig.3.3. 3an d3.3.6 )i s

87 strong. In both Younger Dryas and present Allier sediments an elongated body ofDor elik esedimen t isfoun d alongth e eastern Alliervalle y slope,downstrea m theconfluence .I tseem slikel y that this has a similar origin, delayed sediment mixing at a confluence.

The most striking difference between the Younger Dryas and presentsediment si sth escal edifferenc ei nspatia lvariability . This difference may originate from the differences in fluvial systemcharacteristic sdurin gdeposition .Younge rDrya ssediment s were deposited by anAllie r with amor e braided character than thepresen tmor emeanderin grive r(Bout ,1963 ). Thes edifference s in fluvial system behaviour include differences in river sinuosity floodplain width and amount of channels. Sediment mixing in the actual Allier is maximally delayed to the first riverbend . A riverwit h ahighe r sinuosity can be expected to displaya shorte rdela y incomplet e sedimentmixing .Thi stren d becomes more complicated when different channels occur in the floodplain. Sediment mixing at a confluence with different channels will show a longer delay than in one channel. From aerialphotograph si tca nb esee ntha tth eDrya sAllie rha sknow n morechannels .Durin gth eYounge r Dryasth eDor emixe dwit hth e Allierchannel son eb yon econtributin gt osedimen tmixin gdela y ofsevera lkilometres .Th eactua lDor eflowin gsevera lkilometre s parallel to the Allier is still occupying a former Allier channel.

3.3.5 Conclusions The strong resemblance of the spatial compositional distribution within the present riverbed and the Younger Dryas terrace indicates that sediment mixing at the Allier Dore confluence strongly determined the spatial sand compositional variability in the Younger Dryas terrace.Thi s result supports thegenera l ideao fth e impacto f formerconfluence s onterrac e sedimentcomposition .Th escal edifferenc ei nspatia lvariabilit y betweenth eYounge r Dryasan dpresen tAllie r sediments suggests that changes in fluvial dynamics caused major changes in the

88 spatial variability of fluvial sediment composition at this confluence. The large regional differences between fluvial sediment variabilityo fYounge rDrya san dpresen tsediment sindicat etha t sedimentmixin g inlarg escal esystem s ismuc hmor e complicated than in small scale experiments. Therefore more macro scale research at confluences is needed to allow a more reliable palaeohydrological interpretations.

Quantitative determination of the spatial compositional variability in terrace sediments allows discrimination of the mixed and unmixed sediments within the same terrace level thus facilitating theselectio no fsuitabl esampl esite s forfurthe r regional palaeohydrological research. Geological clustering allowsa firs tan dquic kdiscriminatio no fth esedimen ttype sbu t a more accurate discrimination can be obtained from a geostatistical approachwhic h needsmor e samplesites .

89 3.4 CLIMATE CONTROLLED SEDIMENT FLUXES IN A FLUVIAL SYSTEM: A RECONSTRUCTION OF THE ALLIER BASIN DURING THE LATE QUATERNARY.

A. Veldkamp

Abstract Longitudinal trends in bulk composition of Holocene and Late Weichselian terrace sands of the Allier (Central France) show that sand composition is mainly determined by longitudinal sorting and tributary sediment fluxes. These fluxes are mainly related to basin lithology and environment. The past changes in bulk geochemical composition of Allier sands as found in terrace sediments at a confluence are used to reconstruct the past sediment flux dynamics. The good match between the reconstructed Allier sediment flux dynamics and known environmental changes in it basin during the last 30.000 years indicates that the methodology of calculating sediment mixing ratios is a promising quantitative tool for paleoenvironmental reconstructions.

3.4.1 Introduction Sediment composition within a fluvial system is the result of both internal factors such as transport mechanism and relative grain resistance during transport, as external factors such as source area, and tributary dilution (Cameron and Blatt, 1971; Basu, 1976; Davies et al., 1978). These external factors can be related to present and past relief, source rock and climate by studying the longitudinal variation in fluvial sand composition (Franzinelli and Potter,1982; Knighton, 1982; Ichim and Radoane, 1990). In palaeohydrological research it is often attempted to reconstruct past environments. Fluvial sediments are studied in order to establish the external factors determining sediment characteristics. Such studies do usually not rely on the longitudinal variation but on the interpretation of the stratigraphy of fluvial deposits (Starkel, 1983; Mitchell and Gerrard, 1987). There are many ways to relate sediment properties to hydrological regime and environmental conditions during deposition. A successful approach is estimating palaeodischarge peaks from maximum grain size (Maizels, 1986), thus allowing the registration of large scale changes within a fluvial system. In general the tendency in fluvial terrace research is to rely on sediment characteristics such as grain size and composition mostly combined with more or less fragmentary palynological and

90 paleontologicaldata .

Climaticchange ,uplif tan dothe rmajo revent s ina drainag e basind ono tonl ybrin galon gchange si nth egranulometrica lan d mineralogicalcomposition ,bu tals o inth equantit y ofsedimen t deliveredt oa stream .Thes eeffect sma ygreatl ydiffe rfro mon e tributary to another within the same drainage basin. One tributaryma ycarr y sediment fromdeglaciatin gmountai nranges , another may show admixture of contemporaneous volcanism. Such events can be detected qualitatively e.g. by heavy mineral analyses,bu tth eeffec to nth ebul kcompositio n ofth e fluvial sediments is rarely studied. Yet the study of bulk sediment compositioni sa wa yt oquantif yfluxe so fsedimen to fdifferen t composition ina drainag ebasin .

Whenchange si nbul ksedimen tcompositio na ta confluenc ear e measuredi ti spossibl et oreconstruc tth eforme rsedimen tmixin g process at this confluence by calculating the ratio of the two sedimentfluxe swhic hmixe da tthi sconfluenc edurin gdepositio n (mixing ratio). Environmental changes in a basin which cause changes insedimen t supply canpotentiall y beregistere d byth e mixing ratios of tributaries and the main river at their confluences. Itar ethes esedimen tmixin g ratioswhic hhav eth e potential to monitor environmental changes in the fluvial sedimentary record quantitatively.

In this study the proposed concept of measuring the bulk composition of fluvial sediments in order to reconstruct the sediment flux dynamics is applied on theAllie r river terraces in Central France. At first the effects of tributaries on the longitudinalsedimen tcompositio no fth emai nrive r (Allier)ar e investigated. The climatic dynamics are evaluated by comparing Late Weichselian and Holocene sand compositions. After this regionalstud ya mor edetaile dstud yo fth etim erelate dsedimen t mixing dynamics ismad eb y focusing onon econfluence .

Thesedimen tbul kmeasurement sar elimite dt oth esand sonly ,

91 because sands facilitate representative sampling and they are suitablesediment s forreconstructin g fluvialdynamic s (Potter, 1978; Pettijohne tal. , 1987).Bul ksan dcompositio ni smeasure d geochemicallybecaus emineralogica lan dpetrographica l analysis is rather cumbersome to derive the bulk composition. Previous bulk geochemical analysis of fluvial sands showed that bulk geochemistryi sa suitabl etoo lt ocharacteriz eth ecompositiona l variabilityo ffluvia lsediment s (Kroonenberge tal. ,1988 ;Mour a andKroonenberg , 1990;Veldkam pan dKroonenber g 1989). Only the younger terraces are sampled because the older terracesediment sar ealread ys ostrongl yweathere d (Veldkampan d Feijtel, 1990)tha tth eorigina l sediment compositioncanno tb e reconstructed.

The study area The study area (Fig. 3.4.1 and 3.4.2) includes the Late Weichselian terraces of theAllier .Thi s study startswit h the investigation of two relative young terrace deposits along a stretcho f 100kilometre s (Fig. 3.4.1).

IXO Hercynian basement fffffil Volcanics I I Oligocene sediments * sample site

Figure3.4. 1 Allierdrainag ebasi nan dsampl esite salon ga stretc ho f 100km .fo rth elongitudina lvariabilit yalon gAllier .

92 l-'-'-l Weichselian terraces I IHolocen e terraces ' river sample site

Figure3.4. 2 Studyare awit hsampl esite sa tth eAllier/Dor econfluenc e fara stud yo fsedimen tflu xchange sa tthi sconfluence . Thesample dsand sar edevide dint othre egroups ,th eAllie r sandsupstrea m(A )an dthos edownstrea m(B )th eAUier/Dor e confluencean dth eDar esand s (C).

(1)Th e Late Weichselian terrace level (Xm), approximatel y 15 m above present river level,characterize d by a large volcanic component (Xao ngeologica l map) (Cloziere t al., 1980). (2)Th eHolocen e level (ZY),0 t o 5m abovepresen t river,ha s sediments with a considerably smaller amount of volcanic components than the Late Weichselian deposits (Kroonenberg et al.,1988;Tourenq , 1986). After this regional study amor e detailed study of sediment fluxdynamic si scarrie dou ta tth eAllier/Dor econfluenc ewher e very contrasting sedimentsmi x (Fig. 3.4.2). TheLat eWeichselia nterrace so fth eAllie rar echaracterize d by two levels at 20 and 10 m above present river level respectively. They have at least four different litho- stratigraphical units (Fig. 3.4.3), AMiddl e Pleniglacial (Xt), twoLat ePleniglacia l (Xn andX ln)an da Lat eGlacia l (XIV)uni t respectively (Veldkampan dKroonenberg , 1991). Th eX IVunit swa s incised during the Pre-Boreal and Boreal, because the oldest known Holocene sediments have Atlanticum ages indicating that mostHolocen e sediments arerelativel y recent.

93 upstream Allier-Dore confluence, downstream Allier-Dore confluence

^rr^PWWMWsW ^„„,.,*MM'''''

Figure 3.4.3 Schematicstratigraph yo flat eQuaternar yAllie rterrace s atth eAllier/Dor econfluence .Th edifferen tterrac eunit s haveapproximatel yth efollowin gage sbas eo n Cdatings : 30,000- 29,00 0yB.P . X„ 25.000- 16.50 0yB.P . xm 16.500- 12.50 0yB.P . 11.000- 10.00 0yB.P . xIV <700 0 yB.P . ZY 3.4.2. Materials andmethods . To study the regional longitudinal variation of the Allier sand samples were taken from the Holocene (ZY) and Late

Weichselianterrac e (Xm) alonga stretc ho falmos t10 0k m (Fig. 3.4.1). Sampling amounted to a total of 38 and 46 samples for Holocene and Late Weichselian terraces respectively. At each sampling site 2-6 sampleso fdifferen tgrai n size distributions were collected (both fine and coarse sand samples), to incorporatea possibl egrai nsiz eeffect .Grai nsiz eeffect sar e caused by the depletion or enrichment of certainmineral s like heavy minerals, mica's and feldspars in specific fractions (Strakhov, 1969). Previously bulkgeochemica l investigationso f the Allier sands in the section between the Dore and Couze Chambon showed that this grain size effect determined approximately 10% of the total geochemical variability (Kroonenberg et al., 1988). The grain size related sand oxides inth estudie dAllie rsection ,mainl yNa 20an dA1 203,ar es owel l correlatedwit hgrai nsiz etha tthe yca nb ereasonabl ypredicte d fromgrai n sizedat a (Veldkamp, 1990).

94 Nearth eAllier/Dor econfluenc eextr asan dsample swer etake n along a stretch of 30 km of theAllie r river in the different LateWeichselia nan dHolocen elitho-stratigraphica lunit s (Fig. 3.4.2). The sampled sands aredivide d intothre e main sampling groups (Fig. 3.4.2), the Allier sands upstream (A) and those downstream (B) the Allier/Dore confluence, and the Dore group (C). Th eeffect so fspatia lmixin g (Best,1988 )an dlongitudina l sorting effects (Davies et al., 1978) were taken into account (Veldkamp and Staritsky, in press, section 3.3). Sampling amountedt oa tota lo f17 6sample s (X:=39,X XI=22,X m=50,X IV=65) from LateWeichselia n stratigraphic units and 54 from Holocene deposits.

Sampling methodology and laboratory treatments and measurementswer edon eaccordin gt oKroonenber ge tal . (1988)a s described inchapte r3.1. . Statistical treatments were performed with SPSSpc (Norusis, 1986). Mean bulk geochemical composition of the terrace sands upstream (A)an d downstream (B)th eAllier/Dor e confluence are listed in Table 3.4.1 and 3.4.2. The average Dore sand (C) composition (Tab. 3.4.3) is known from previous research (Veldkamp and Kroonenberg, 1989).

3.4.3. Results and discussion

Regionalvariatio ni nsan dbul kgeochemica llongitudina lprofile s Bulkgeochemica l longitudinalprofile so fAllie rsand sshowe d thatgeochemica lvariabilit yi sdetermine db ythre emajo relemen t groups. These groups were also found by the first exploratory investigation by multivariate statistics (Kroonenberg et al.,

1988). The first group is characterized by the oxides, Ti02,

Fe203, MnO, MgO, CaO and P205/ which are related to different abundanceso fbasalti croc kfragments .Th esecon dgroup ,Na 20an d

A1203 has a variability related to grain size, and the third group represents variability of the acid crystalline mineral group (K20 and Si02) which is mainly due to differences of K- feldspars,guart z and crystalline rockfragments .

95 terraces ZY XIV XIII XII XI Element

Si02 72.57 70.50 67.51 77.54 78.09 Ti02 0.97 1.08 1.35 0.60 0.53 A1203 10.84 11.53 12.13 10.49 11.02 Fe203 4.36 4.83 5.78 2.41 1.87 Na20 1.50 1.57 1.93 1.77 2.04 MgO 1.93 2.85 3.16 1.07 1.03 CaO 2.42 3.62 3.89 1.65 1.53 K20 2.82 2.68 2.56 2.98 2.57

Table 3.4.X Mean ood.de contents upstream Dore/Allier confluence.

terraces: ZY XIV XIII XII XI Element

Si02 76.25 74.31 68.87 78.86 76.61 Ti02 0.60 0.85 1.23 0.37 0.56 A1203 11.19 10.99 12.22 11.09 11.74 Fe203 2.68 3.66 5.56 1.45 2.38 Na20 1.53 1.40 1.56 1.60 1.48 MgO 1.16 1.74 2.48 0.59 0.69 CaO 1.50 2.32 3.52 0.94 0.98 K20 3.41 3.32 2.87 4.02 3.85

Table3.4. 2 Meancsd.d econtent sdownstrea m Dore/Allierconfluence .

terraces: ZY XIV XIII XII XI Element

S102 77.99 79.49 81.31 81.31 82.08 Ti02 0.18 0.15 0.15 0.15 0.25 A1203 11.83 11.27 10.19 10.19 10.04 Fe203 0.81 0.76 0.73 0.73 0.90 Na20 2.25 2.25 1.70 1.70 1.03 MgO 0.40 0.38 0.33 0.33 0.32 CaO 0.44 0.39 0.22 0.22 0.19 K20 4.50 4.11 4.23 4.34 4.22

Table3.4. 3 Meanoxid econten tDor eterrace .

96 K20 (wt%)

-10 10 30 50 70 90 KM from Dore/Allier confluence

Figure 3.4.4 K^ (wt%) longitudinal profile (km) from Bricude to Vichy late Weichselian terrace sand. Allier flew direction is from right to left. The main tributaries are indicated by vertical lines and symbols, A=Allagnon, P=Oouze Bavin, OOcuze Chambon, D=Dore.

K20 (wt%) 5-

-, 1 1 , r 30 50 70 90 KMfro mDore/Allie rconfluenc e

Figure3.4. 5 Kp (wt%)longitudina lprofil e (km)fro mBrioud et oVich y presentriverbe dsand .Allie rflo wdirectio ni sfro mrigh t toleft .Th emai ntributarie sar eindicate db yvertica l lines and symbols, A^Allagnon, P=Oouze Bavin, OOouze Chambon,D=Dore . 97 One oxide of each group is presented in two longitudinal profiles,on efo rth eHolocen ean don efo rth eLat eWeichselia n sands.K 20(Fig .3.4. 4an d3.4.5 )represent sth eaci dcrystallin e component, Na20 (Fig.3.4. 6 and 3.4.7) the grain size related component and MgO (Fig. 3.4.8 and 3.4.9) the basic volcanic componenti nth eAllie rsands .Th emai ntributarie sar eindicate d in each longitudinal profile as vertical lines and symbols, A=Allagnon, P=Couze Pavin, C=Couze Chambon, D=Dore.Th eAllie r flow direction is from south to north (right to left inth e profileso fFig .3.4. 4t o3.4.10) .Th echange si ngenera ltren d asindicate d byth ebroke n linear ebase d onaverag echange si n bulk concentrations forth e concerningoxide .

Longitudinal K20 contents (Fig. 3.4.4 and 3.4.5) indicate gradualchange si nsedimen tcompositio ni nbot hHolocen ean dLat e Weichselian sands. The Allier bulk sand composition shows a relative decrease in K20 (and Si02) content at each tributary drainingpredominantl y volcanic areassuc ha sth eAllagno n(A) , the Couze Pavin (P)an d the Couze Chambon (C). Th e Dore (D), draining ancrystallin e area,cause sa nincreas e ofK 20conten t

in the Allier sands. The differences of K20 concentrations betweenth eLat eWeichselia nan dHolocen etrend sar ever yslight .

It can be observed that the K20 concentration levels in the Holocenesand sar egenerall yhighe rtha ni nth eLat eWeichselia n sands, illustrating the higher acid crystalline component in thesesands .

Na-,0trend s

TheNa 20(an dA1 203)bul kconcentratio ntrend s (Fig.3.4. 6an d 3.4.7), predominantly representing changes in the albitic endmember of plagioclase (Kroonenberg et al., 1988)an d grain size effects,sho w a largevariance . This canb eattribute dt o thesamplin gstrateg yt oselec tsample swit hdifferen tgrai nsiz e distributions (Veldkamp, 1990). Nevertheless, the Na20 curve showsmajo rtendencie sa ttributarie swhic hsee mindependen tfro m grain size sucha sth estron g increaseo fNa 20 (andA1 203)

98 Na20 (wt%) fig 3.4.6

30 50 70 90 KM from Dore/AUier confluence

Figure 3.4.6 Na20 (wt%) longitudinal profile (km) from Rrioude to Vichy late Weichselian terrace sand. Allier flew direction is frcm right to left. The main tributaries are indicated by vertical lines and symbols, A=Allagnon, P=Oouze Pavin, OOouze Chambon, D=Dore.

Na20 (wf%)

2.6 a p D C P A 2.U- i l

2.2- ~\ ! ! I i / • \; ; : | / p \ i i pi : -/ - \> - !/• I | a I D 1.6-- !/ I ° a 1.4 I »" ] B a |

—p—,—— -1 ' 1 ' 1 ' -10 10 30 50 70 90 KM from Dore/AUier confluence

Figure 3.4.7 Na20 (wt%) longitudinal profile (km) from Brioude to Vichy present riverbed sand. Allier flew direction is frcm right to left. The main tributaries are indicated by vertical lines and symbols, A=Allagnon, P=Oouze Pavin, OCbuze Oiambon, D=Dore. 99 concentrationsa tth eAllier/Couz eChambo nconfluence .Th eNa 20 trends south of the Couze Chambon only show a slight general decrease downstream. Further downstream, north of the Couze

Chambon there is a stronger decrease in Na20 concentrations. These decreasing trends downstream seem independent from the tributariesan dar eprobabl yrelate dt oth elongitudina lsortin g effectcause db ybacklaggin go f 'heavy1grain san dth ebreakdow n of grains during transport as described by Cameron and Blatt (1971)an dDavie se tal . (1978). Theprofile sfo rth eHolocen ean dLat eWeichselia nsand ssho w a strong resemblance. They have approximately equal trendsan d similar bulk concentration levels.Th eresemblanc e ofth eNa 20

(andA1 203)trend sfo rth eHolocen ean dLat eWeichselia nsand si s probably only related toth especifi c origin ofNa 20.Th eCouz e Chambonbasi ncontain sth emajorit yo fth eaci dvolcani cdeposit s of the Mont Dore, like the trachytic ignimbrite of Neschers. These acid volcanic deposits arewel l known sourceso falbite .

This canexplai n thesudde n change inNa 20 content inbot hth e Holocenean dLat eWeichselia n longitudinal profiles.Apparentl y nomajo rchange si nprovenanc eoccurre d fromLat eWeichselia nt o recenttimes .

MqQtrend s The bulk concentrations ofbasi c volcanic elements likeMg O inth eLat eWeichselia n (Fig.3.4.8 )an dHolocen e (Fig.3.4.9 ) showdifferen ttrends .I nth eLat eWeichselia nsand s(Fig .3.4.8 ) fromBrioud edownstream , therei sa nincreas e inMg Oconten ta t confluences (Allagnon and the Couzes) draining the higher volcanic areas which were glaciated during Late Weichselian (Veyret,1980) .A stron gincreas ei nMg Oconcentratio nresemblin g theNa 20trend ,i sfoun da tth econfluenc e ofth eCouz e Chambon andth eAllier .Furthe rdownstream ,th ehig hMg Ocontent so fth e LateWeichselia nterrac ea tPon td uChatea u(k m3 5i nFig .3.4.8 ) clearlyreflec thig hconten to fvolcani ccomponent s (Bout,1963 ). This section showsa larg e spread inMg Ocontent .

100 MgO (wt%) b-l c D I D D | P A

4- • D | I I I I a I . D ' I 3- /7 • I B / a ^ i a 2- a 3 o ^ - D D en 3 a B 1- 3 3

I I ' I 1 1 °- -10 10 30 50 70 90 KM from Dore/Allier confluence Figure 3.4.8 MgO (wt%) longitudinal profile (km) frcm Brioude to Vichy late Weichselian terrace sand. Allier flow direction is frcm right to left. The main tributaries are indicated by vertical lines and symbols, A=ftllagnon, P=Couze Pavin, OOouze Chambon, D=Dore. MgO (wt%) b-

D c P A 4- [

3- I I •i • ! n 2- D • i P -D— I — • n • ID ? ° I a DD| \ C D • in a 1- O I a ' ° ° f I b , 0- 1 r- 1 i I I -10 10 30 50 70 90 KM from Dore/Allier confluence Figure 3.4.9 MgO (wt%) longitudinal profile (km) from Brioude to Vichy present riverbed sand. Allier flow direction is frcm right to left. The main tributaries are indicated by vertical lines and symbols, A=Allagnon, B=Couze Pavin, OCbuze Chambon, D=Dare. 101 Furtherdownstrea m downt oth eDor econfluenc eth e longitudinal sorting effect causes a decrease in MgO concentrations downstream, as reported by Kroonenberg et al. (1988). The Dore tributary causes a dilution of basic volcanic particles with abundant crystalline material, resulting in MgO concentration levelswhic h arealmos t equal tothos ea tBrioude . Inth eHolocen esand sMg Oconcentration s (Fig.3.4.9 )chang e with less sudden changes atth e confluences.Th e amount ofMg O inth eAllie r sands gradually increases at theAllagnon , Couze Pavin and Couze Chambon, anddecrease s atth e Doreconfluence . A longitudinal sorting effect seems also be detectable between theCouz eChambo nan dth eDore .Bot hlongitudina l profileshav e comparabletrend sbu tth eamplitud eo fth eLat eWeichselia ncurv e ismuc h largertha ntha t ofth eHolocen eone .

average difference (wf%)

30 50 70 90 KM from Dore/Allier confluence

K,0 MgO

Figure3.4.1 0 Longitudinalprofil e(km )o faverag edifference s(wt% )o f KjO,Na^ Oan dMg Oconcentration sbetwee nHolocen ean dlat e Weichseliansand s(Averag eHolocen e-Averag eWeichselian) . Allierflo wdirectio ni sfro mrigh tt oleft .Th emai n tributariesar eindicate db yvertica lline san dsymbols , A=Allagnon,P=Couz ePavin ,OCouz eChambon , D=Dore. 102 A comparison of the Late Weichselian and Holocene sand composition Both Late Weichselian and Holocene sands show the effect of sediment mixing at confluences. MgO mainly of volcanic origin increases inth eAllie r sands at theAllagnon , Couze Pavin and

Couze Chambon confluences at the expense of K20. K20 from crystallinebasemen trock sincrease sonl ya tth eDor econfluence .

Thedifference sbetwee noxid econtent si nth eLat eWeichselia n andHolocen edeposit sar evisualise db yplottin gth elongitudina l trendso fth ecalculate d averagedifferenc e of (K20h0l- K 20welch),

(Na2Ohol - Na20„eich) and (MgOhol - MgOKelch) (Fig. 3.4.10). The differencebetwee nth eaverag eHolocen ean dLat eWeichselia nNa 20 concentrations arealmos tzero .Holocen esand sar eriche r inK 20 and poorer in MgO than the Late Weichselian sands. The most strikingchang efro mLat eWeichselia nt oHolocen ei sth erelativ e largenegativ evalu e for (MgOhol- MgO weich)a tan ddownstrea m the Couze Chambon/Allier confluence (Fig. 3.4.10). Apparently much relatively basalt-rich material was supplied to the Allier in Late Weichselian times by the Couze Chambon (Fig. 3.4.8), but thissuppl ycease ddurin gth eHolocen e (Fig.3.4.9) .Thi schang e can obviously not be due to basin lithology or longitudinal sorting. Possible causes forthes e changes are either climatic changes or Late Weichselian basic volcanism (Bout, 1963; Kroonenberg et al., 1988).

Causes fordivergin g longitudinal trends Bout (1963) related the volcanic rich Late Weichselian sediments to a melting phase of the glaciers. His theory is supportedb yfiel devidenc e (Kroonenberge tal. ,1988 )an db yth e observationtha ttributarie s relatedt oglaciate d areascaus ea stronger increase inMg O content inth e LateWeichselia n sands compared toth eHolocen e sands (Fig. 3.4.10). Pastre (1986)an d Toureng (1986)bot h namevolcanis m as the major factor causing the relative highvolcani c content of the LateWeichselia n sediments.Th evolcani c history of the Allier basinindicate stha tdurin gth eLat eWeichselia nactiv evolcanis m

103 ofth eChain ede sPuy stoo kplac eindeed .Aroun d 27.000B.P .th e Tartaret volcano at the bottom of the Couze Chambon valley emitted a2 8k m longbasal t flowdownstrea m (Raynalan d Daugas, 1984; Raynal et al., 1985). However, no Late Weichselian volcanismi sknow ni nth eothe rmajo rtributaries ,th eAllagnon , CouzePavi nan dDor ebasins .Therefore ,onl yth erelativel ylarg e volcanic component atth eCouz eChambo n inth eLat eWeichselia n sands may be related to the volcanic activity of the Tartaret volcano. Onth eothe rhan dther ear eman yHolocen evolcanoe sknow ni n theAllier ,Couz eChambo n andCouz ePavi nbasin s (Camuse tal. , 1983).Th etrachyti cmaa ro fLa cPavi n (5990± 11 5B.P. ;Juvign e and Gewelt, 1987) is the most recent volcanic activity in the Allierbasin .Nevertheless ,Holocen evolcanis mdi dno tresul ti n any notable increase inth evolcani c component inth e Holocene Allier sands.Apparentl y active volcanism had no direct impact on sand composition. Therefore Bout's (1963)theor yo na paleoclimati corigi n for the Late Weichselian increase in volcanic components is more likely. Whereby the role of Late Weichselian volcanism in the CouzeChambo nbasi ncanno tb ecompletel yexcluded .At th een do f theLat ePleniglacia lmeltin go fglacier si nth ehighe rvolcani c areas must have resulted in the release of a considerable quantity of volcanic-rich fluvioglacial sediments. After the glaciers had molten away the fluvioglacial flux from the Allagnon, Couze Pavin and Couze Chambon stopped and the steady base flux of crystalline components became automatically more important during theHolocene .

Theprecedin g evaluation ofth e longitudinal changes inbul k geochemical sand composition from LateWeichselia n to Holocene indicatedtha tno tvolcanis mbu tclimati cchange splaye dth emos t important role in determining fluvial sand composition. This resultimplie stha tclimati cchange swithi nth eAllie rsyste mca n bestudie d fromchange san dtrend s influvia l sandcomposition . As bulk sand composition was measured it is also possible to reconstruct the past sediment flux dynamics which caused the

104 measuredchange si nsedimen tcomposition .No wpaleoenvironmenta l changes within the Allier basin can be measured both qualitatively as quantitatively. As a case study sediment mixing at one confluence the Allier/Doreconfluenc ewa sstudie dbecaus ea tthi sconfluenc etw o very contrasting sediment typesmi xan dmixed .

3.4.4. Sedimentmixin g atth eAllier/Dor e confluence Asargue di nth eintroductio nma ya mixin grati ogiv ea direc t indication ofth esedimen tmixin ghistory .Thi srati ouse d asa measureo fth erelativ econtribution so fbot hAllie ran dDor et o sediment mixing. The following procedure was used to determine the ratios of the two fluxes which mixed at the Allier/Dore confluence. Thesignificanc eo fdifference si nmea noxid econten tfo reac h litho-stratigraphicaluni t(Tabl e3.4. 1an d3.4.2 )wa sdetermine d bymean so fmultipl e comparisonsaccordin g toth eTuke ymethod . The significance of change inmea n Allier sediment composition atth eAllier/Dor econfluenc ewa sdetermine db ystudent st-test s foreac hstratigraphica lunit ,usin gth eSPSSp cpackag e(Norusis ,

1986). Mean Si02, Ti02, Fe203, MgO, CaO and K20 content are significantdifferen t (atth e0.0 5level )fo rth eZY ,X IV,X m and

Xn units,whil e theX juni tdiffer sonl y significantly fromth e otherterrac eunit s inMgO ,Ca O andK zOcontent .Therefor e only

MgO, CaO and K20content s are used further as trend oxides for sediment compositional changes of the Allier during the last 30,000 years. The compositional dynamics are plotted as bulk chemo-stratigraphy of the sediment groups upstream (A) and downstream (B)th eAllier/Dor e confluence inFigur e 3.4.11.Th e composition of the litho-stratigraphic units upstream (A) and downstream (B)th eAllier/Dor e confluence showdifferen t trends intime . Iti sobviou stha tth eoxide so fbasalti c originMg Oan dCa O have complementary trendswit h thecrystallin e K20 (Kroonenberg etal. , 1988).Th eoldes tLat eWeichselia nsand s (X1an dX n)ar e relatively poor involcani c oxidesMg Oan d CaOan d rich inaci d crystalline K20.

105 cold cone (vt%) cone (wt%) cone (wt%) upstream section (A] downstream section (B) upstream-downstrea m .J. J_ L_

10.000

20.000-

30.000 years B.P.

Figure3.4.1 1 Correlationo fchange si nbul ksedimen tcompositio nwit h climatea tth eAllier/Dor econfluence .

TheX m sandshav eopposit e characteristics,the y arever y rich inMg Oan dCa Oan dpoo r inK 20.I nth eX IV-unitsand sth eCa Oan d

MgO contents are slightly lower (K20 higher) than in the Xm sands. The basaltic oxide content decreases further in the Holocene (ZY)sands .

Withineac hAllie rterrac euni tmea nMgO ,Ca Oan dK 20content s alsosho wsignifican tchange sa tth eAllier/Dor econfluence .Th e

MgO and CaO content decrease at the confluence while the K20 content increases in all Allier terrace units. These relative changesar evisualize d inFigur e3.4.11 ,wher eth echange si n(O x wt%) CaO,Mg O and K20conten t (Upstream Oxwt % - Downstream Ox wt%) of the Allier sands are plotted against age. The general compositionaltrend so fFig .3.4.1 1confir mth epictur eo fa mai n

106 streamwit hvolcani c richsedimen t (Allier)whic h isdilute db y a volcanic poor sediment (Dore). This dilution effect is very smallfo rth eX ni sands (UpstreamO xwt %- Downstrea mO xwt %ar e small)an d large forth eX n sands.

Mixing ratio Themas srati oo fth eAllie ran dDor esand swhic hmixe da tth e confluence (sandmixin g ratio) can be calculated for different oxides (bulk Ox wt %) and time periods. This is done by the following formula:

Sandmixin g ratio = (Oxwt %Allie r downstream -O xwt %Dore )

(Ox wt% Allier upstream - Ox wt% Allier downstream)

Sandmixin g ratio'so fterrac eunit sar eonl ycalculate d for those oxideswhic h show a significant change atth e confluence (Table 3.4.4). A sandmixin g ratio> 1indicate sa large ran da ratio <1 a smaller Allier sand contribution toth eAllier/Dor e sedimentmixin gtha nth eDore .Difference s incalculate d ratios (Tab. 3.4.4)shoul d becarefull y reviewed.

terrace ZY XIV XIII Xll *I Element

Si02 0.47 1.36 9.15 1.86 ns Ti02 1.14 3.04 9.00 0.96 ns A1203 ns ns ns ns ns Fe203 1.11 2.48 21.95 0.75 ns Na20 ns ns ns ns ns MgO 0.99 1.23 3.16 0.54 1.09 CaO 1.15 1.48 8.9 1.01 1.44 K20 1.85 1.23 4.38 0.31 0.29

Average 1.12 1.80 9.42 0.91 0.94 ns =n o significant change at confluence

Table3.4. 4 Mixingratio sa tAllier/Dor econfluence .

107 A difference between a ratio of 2 and 10 equals the difference between a ratio of 0.5 and 0.1, and with a large or small mixing ratio the estimation error increases strongly. Both effects contribute toth e largevariabilit y inmixin g ratioswithi n each terrace unit.Therefor e only average sand mixing ratios are used in this research.

The average sand mixing ratios (Table 3.4.4) in time, illustrating general sandmixin g dynamics ofAllie r and Dore are shown in Figure 3.4.12.

sand mixing cold warm total sediment mixing ratio Allier / Dore ratio Allier / Dore 1-4-.-I 1 1 ' 1 L

10.000-

20.000

30.000 -1—i—i—i—i—i—r years B.P ^68 10 12 % 20 30 UQ 50

Figure3.4.1 2 Changesi nsan dan dtota lmixin gratio' si ntim ea tth e Allier/Doreconfluence .

108 Strikingi sth elarg emixin grati oo fth eX m sandssuggestin g a very large sediment flux inth eAllier . During deposition of

Xn sediments the Dorecontribute d 1.7time smor e sand thanth e Allier, whilst the Allier contributed about 6 times more sand thanth eDor ewhe n theX m sedimentswer e deposited.

Mixing ratios for Fe203 and Ti02 areusuall y higher than the mixing ratios of MgO and K20 for the same terrace units (Tab. 3.4.4), suggestin ga noxid edependenc eo fth emixin gratio .Thes e higherratio' spoin tt oa nenrichmen to fth emineral scontainin g Fean dTi ,th eheav yminerals .I ti sgenerall y knowntha theav y minerals tend to concentrate at confluences (Schumm, 1977). Further research isneede d to seewhethe r themixin g ratio has any value in detecting mineral enrichment (placers) at former confluences.

Asbot h Allier and Dorehav e known changes inth e amount of supplied sediment in time, their mixing ratio only records changes relative to each other. The actual compositional differencesi nth eDor eterrac esand sar emainl ycause db ypost - depositional weathering which does notpla y a dominant role in the younger sediments studied in this research (Veldkamp and Kroonenberg,1989 ). A sther ear en oindication stha tth eDor eha s knownan yconstrain s insedimen t supply intime ,Dor esediment s canb esee na sa relativ econstan tgeochemica lreferenc e forth e Alliersedimen tdynamics .A sth eDor ebasi nha dn omajo rchange s during the last 30.000 years it can be assumed that the Allier/Doremixin grati opredominantl yreflect schange si nAllie r sedimentfluxes .

Theproportio no fsan dwit hrespec tt oth etota lgravel ,san d and clay content inth eDor e ismuc hhighe rthe n inth eAllier . Fromth esample dgrave l and sandpit sestimate s ofth e relative sand proportions of Allier and Dore deposits have been made (Table 3.4.5). Assuming that theterrac e sediments give agoo d representation of the sediment fluxes that mixed during their deposition,th eestimate dproportion sca nb euse dt oconver tth e average sandmixin g ratio intoa tota l sedimentmixin gratio .

109 Terrace Estimatedsan d Totalsedimen tmixin g unit contribution Allieran dDor e Allier/Dore

ZY 0 5 2.3 XIV 0 33 5.4 XIII 0 2 47.1 XII 0 5 1.8 XI 0 2 4.7

Table3 4 .5 Totalmixin gratio so fsediment sAllie ran dDor e This mixing ration is the mass ratio of the bulk terrace sedimentswhic hactuall ymixe ddurin gdepositio no feac hterrac e units (Table 3.4.5).

Thetota lmixin g ratioso fth eX I# Xm andX IVsediment s (Tab. 3.4.5, Fig. 3.4.12) increase considerably compared to the sand mixing ratio. The Xm sediments show a sediment flux of the Allier almost 50 times the Dore contribution during sediment mixing.

3.4.5. Sediment fluxesan d climate Mixing characteristics ofAllie r and Dore sediments suggest that sediment fluxes of the Allier, compared to the Dore sediments, show cyclic changes in time. The Xx sediments depositeda tth een do fth eMiddl ePleniglacia l showa fiv etim e largercontributio n ofAllie rtha no fDor esediments .Th e local

Late Pleniglacial sediments (Xn)yiel d asedimen t mixing ratio which indicates an Allier contribution of only twice the Dore supply. This small mixing ratio may be explained by the periglacial environment and by the building up of extensive glaciers in theuppe r Allier basin storing large quantities of potential Allier sediments. Deglaciation during the end of the Late Pleniglacial released an enormous mass of fluvioglacial sediments intoth eAllie r (Xm),causin gth elarg esedimen tflu x representedb yth eX nlsedimen tmixin gratio .Thi slarg esedimen t flux (about fifty times the Dore contribution) was deposited duringcatastrophi cevent swithi na fe wthousan dyears .Evidenc e ofth ecatastrophi c character ofth eX m deposition isgive nb y Lenselink et al. (1990). Sediments deposited during the short

110 colderinterva lo fth eYounge rDrya s (XIV)hav ea mixin grati oo f about 5, similar as that of the Xx sediments. This isprobably - due to the fact that both cold intervals were too short for extensive glacier build up, but long enough for a major environmental change.

The reconstructed relative sediment fluxes of the Allier at theAllier/Dor econfluenc e confirm thegenera l ideaa s foundb y comparison of the Late Weichselian and Holocene longitudinal profiles,tha t sediment fluxes inth eAllie rbasi n aremainl ya function of climate. The observed trend is that during periglacial and interglacial environments the Allier supplies only twice asmuc h sediments as theDore ,onl y atth e end ofa glacial the relative Allier contribution increases strongly to fivetime sth eDor esupply .Afte ra prolonge dcol dperio ddurin g which extensive glaciers can built up in the headwaters of the Allier,ver y largesedimen tfluxe sca nb eexpecte d inth eAllie r as a result of glacier melting. These fluvioglacial sediment fluxesca n increaset o fifty timesth e Dorecontribution . The results suggest that there exists a correlation between the sediment flux magnitude and the duration and intensity of glacials. It would be interesting to investigate such quantitative relationships in areas with well known past environments.

3.4.6. Methodological evaluation It has been shown that changes in bulk geochemical sand composition in a fluvial system can be applied in Quaternary research as ametho d to reconstruct paleohydrology in relation with paleoenvironmental changes. It is obvious that this methodology isonl y applicable inwel l dated fluvialsystems . Within the described case study it remains difficult to validate the assumption that the calculated mixing ratio's displaymainl yAllie rdynamic sinstea do fDor edynamic so rboth . Thereconstructe d timerelate d sediment fluxesar edifficul t to validate,the y canb eonl ycompare d toa relativ e sediment load curve for the last 15.000 years, based on interpretations of

111 concepts for Central European river valleys (Starkel, 1983). Althoughbot hcurve sar econstructe db ydifferen tmethods ,i tca n be observed that they show a fairly match for the 15.000 to 10.000year sB P interval,bu tth emagnitude so fchange s inbot h curves differ strongly. Fig. 3.4.12 has of course only local validity but illustrates that it is also well possible to reconstruct sediment flux magnitudes more directly and quantitatively thanStarke l (1983)did . The results suggest a quantitative relationship between sediment flux magnitude and paleoenvironment. When more reconstructions are made it may be possible to construct a continental sediment flux record which gives quantitative indications of the past glacials. Such a record would be complementarywit hth eexistin gpalynologica lrecord swhic hgiv e mainly accurate descriptions ofth ewarme rpaleoenvironments .

3.4.7. Conclusions Longitudinal abrasion,sortin gan dtributar y sediment fluxes mainly determine longitudinal trends in geochemical bulk composition of the LateWeichselia n and Holocene Alliersands . Changes intributar y sand composition from Late Weichselian to Holocene seem predominantly caused by climatic changes, as volcanicactivit yo rothe rchange si nsourc eterrai ndi dno tsho w any direct measurable effect. Sediment flux dynamics at the Allier/Dore confluence are reconstructed for the last 30.000 years by calculating mixing ratios from changes in bulk geochemical sediment composition in time. These reconstructed sediment fluxescorrelat estrongl ywit hclimati cchange sdurin g the last 30.000 years.Allie r sediment supply seems to depend mainly onth etyp ean d duration ofclimati cepisodes . The surprising good correlation between the reconstructed Allier sediment fluxes and paleoenvironment suggests that the methodology applied is a promising technique in Quaternary research.

112 chapter 4 LONG TERM MODELLING OF RIVER TERRACE FORMATION

Within geomorphology the use of long term models has increasedstrongl yi nrecen tyears .A sgeomorphologica ltheorie s havemostl ybee nvalidate d forsmal lspatia lan dtempora lscale s only,the y areno tnecessaril y appropriate forth e larger range ofscale s (Howes& Anderson ,1988) .Constrain si nth edevelopmen t of longtermgeomorphologica l modelsar etherefor emostl ydu et o the interrelationships between conceptualization and scale (Anderson &Sambles , 1988). Especially longterm simulations are hamperedb yth elac ko f largescal equantifie d knowledge.Ther e aretw omai n reasons,fo rthi spoverty :

1)On eca nenvisag eshort-ter m (high-frequency)variation sneste d within long-term (lower-frequency) variations (Bradley, 1985). Asonl yshort-ter mprocesse sar emeasure di nexperimenta lstudie s we can only guess at the quantitative effects of the long,ter m processes. 2) Depending on the scale at which an environmental system is viewed, there are different sets of laws which operate. This scale effect causes that process or system variables which are system dependent at higher order scalesma y become independent ifth e order of scale isreduce d (Huggett, 1985). Duet o these scale effects, extrapolation of experimental results;t o large scales will certainly lead to considerable errors (3chumm, et al., 1987;Howe s &Anderson , 1988). These scale gaps ingeomorpholog y can only partly be filled bymor e research.Accordin g toHugget t (1985)ther e isno tmuc h thatca nb edon eabou tth escal eproblem ;th eimportar tthin gi s tob eawar eo fit sexistence .Th escal eproble m illustratestha t the description of the geomorphological system as .icomplex , inaccessible, scale dependent and inherently random natural system (Howes& Anderson ,1988) ,i swel lchosen .Accordin gt oth e sameauthors ,ther ei sa nuppe rlimi tt oth edegre eo fnumerica l modelling which is possible for such systems. They found it necessary to consider alternative modelling strategies and 113 techniquesfo rus ei nthes esituation s (Howes& Anderson ,1988) .

Theai mo fthi schapte r ist odiscus sa possibl e alternative modelling strategy forth egeomorphologica lscal eproblem stha t arise with long term modelling. Instead of using unreliable quantitative relations in numerical modelling, additional qualitative descriptivemodellin g isproposed .

Scaledependen t modelling Animportan tste pi nlon gter mmodellin gi sth edeterminatio n ofth escal edependen tsyste mvariabl ehierarchy .Th esyste man d processvariable sar eliste di nhierarch ywit hincreasin gdegree s ofdependence .Dependin g onth etim espa n involved,tim ema yb e eithera nextremel yimportan tindependen tvariabl eo ro frelativ e littlesignificanc e toa geomorphologi cstud y (Schumm& Lichty , 1965).

Anotherscal emodellin gaspec ti sth esourc escal eo fth euse d numerical relationships inth emodel .Thi s scale aspect canb e determinedb ya systemati cscal eanalysis .Suc ha scal eanalysi s isbase d on the assumption that relationships can only beuse d in amode l when they are applied onth e same scale aso nwhic h the original measurements were done.Th e scale analysis canb e donei na ver ysimila rwa ya sth ecommonl yapplie duni tanalysis , except that not only the units should match but also their magnitude. Such a model scale analysis can be applied successfully for relative short time spans. On such scales,a model scale analysis should be incorporated as a standard procedure ingeomorphologica l modelling.

Modelling over longtim e spans When a scale analysis is strictly applied on longer time spans, thousands of years or even longer, it can be concluded thatreliabl equantitativ emodellin gi sactuall yimpossible .Mos t knowledge on such timespans isdescriptiv e and interpretative.

114 But it is the large time scale which attracts many geomorphologists asmos t landformsar eth e result of long term processes.

The scale problem is sometimes 'solved1 by a number of assumptions. Insuc hcase s iti sassume dtha ta shor ttim espa n relationship can be extrapolated to a longer timespan. A very creativesolutio ni sgive nb yTetzlaf f& Harbaug h (1989)wh ouse d 'compute-and-drift' and 'compute-and-stop' schemes to overcome long time spans.Thes e schemes use short term calculations for longterm simulations. In case a direct application of such a relationshipdoe sno twor kou tproperly ,a scal e (tuning)facto r isinclude dt oobtai nmor erealisti cresults .I ti sobviou stha t this approach obscures the lack of knowledge and suggests a simple straight forward solution of the scale problem which certainly doesno texist . Anotherapproac ho flon gter mmodellin gi st oabando nth egoa l tomak ea ful lnumerica lmodel ,a sther ei st oles squantitativ e knowledge for that purpose. Consequently it is decided to use alsoth eknowledg ewhic h issufficientl y available,qualitativ e descriptive relations. Modelling with both quantitative and qualitative relations can be done with finite state modelling (Zeigler, 1976).

Finite statemodellin g A finite state model describes a system which can be in different states at different times. The basic principles of finite state modelling are to choose a finite sets of inputs, statesan doutputs ,an dt ospecif yfo reac hstat ecombinatio non e and only one transition to another state in case a change in statetake splace . Asyste mbehaviou rca nb erepresente da sa finit estat emode l in a scheme, flowchart ortable , showing the system states and statetransitions ,includin gth econdition swhe nchange so fstat e takeplace .Th estat edescription san dtransition sca nb ea swel l descriptive asquantitative . The most uncertain and difficult part of finite state

115 modelling is extracting the qualitative information from literature. A major problem is that descriptive knowledge contains many uncertainties commonly indicated by words as, 'probably1, 'may be', 'could be', 'is thought to', etc. When extracting the essential facts one has to generalize certain factsan d relations.A wa y tosuppor tth eextractio n ofthi s kind of knowledge is by keeping in mind that at the end of the modelling activities themode l has towritte n down in a computer language, forinstanc ePASCAL .Whe nth esimpl ean d limited syntax ofPASCA L istake n intoaccoun t oneca nconstruc t several 'rules' for knowledge extraction who facilitate finite state modelling. Two important rules are for example: 1) No uncertainties are allowed in the extracted relationships, a state or arelationshi p exists or does not exist (A state can only be true or false). 2) The reasoning and combining of information can only be done withus e OfAND , OR, THEN, UNTIL, CASE, FOR ... DO, IF ... ELSE, REPEAT ...UNTIL, WHILE ... DO etc. (Findlay & Watt, 1981). As it isth e task of linguists to develop such kind of rules, it isbeyon d the scope of this chapter and will therefore not be discussed indetail .Bu t itbecome s obvioustha tb y applying this kind of modelling the computer is not only used as a calculator but also as a reasoning machine.

Because model states are unique and they have only one possible transition to another state,decisio n rules canb e used to define these transitions. Itar e these decision rules who are directlyprogrammabl e andca nb elooke dupo na smode l thresholds. Although,Begi n &Schum m (1984)showe dtha t systemthreshold s are commonly gradual, they can be looked upon as abrupt on a large timescale, implying that finite state modelling is scale dependent.

Validity of finite state models By using qualitative knowledge in finite state modelling the

116 validity ofmodellin g isonl yo nth econceptua l level.Concept s and hypothesis can be tested and evaluated by simulations of certainscenario's .Whe nth etim escal ei sreduce dth enumerica l portion increases, but the validity remains at the conceptual level.

Conclusions Finite state modelling, allowing the application of qualitativedescriptiv e relations ina model ,i spropose d asa n alternative strategy for long term modelling ingeomorphology . By this kind of modelling the computer is not only used as a calculatorbu tals oa sa reasonin gmachine .Model smad eaccordin g toth efinit estat eprincipl ehav eonl yvalid yo nth econceptua l level and should display the actual state of knowledge on the subject.Eve ni fi ti spossibl et oconstruc ta complet enumerica l model, the strategy of finite state modelling can make the geomorphologistmor eawar eo fth elimite dvalidit y of longter m models.

117 4.1.RIVE R TERRACE FORMATION, MODELLING, AND 3-D GRAPHICAL SIMULATION.

VeldkampA . &Vermeule nS.E.J.W .

Abstract A model on river terrace formation is presented, written in PASCALand run on a VAX 8600. The model calculates the influence of a fluvial system on the relief of an area with macroscopical dimensions (10 km x 20 km x 0.5 km) over a period of 2.5 million years. Model input relies on uplift and alternations in discharge and sediment load as a function of climatic changes. The output of the model are 3-dimensional grid drawings which visualize the impact of uplift, discharge, and sediment load on a landscape. Model formulation is based on empirical information on fluvial systems, which was incorporated in the model by means of a slightly adapted way of finite state modelling in which decisions act as thresholds. The model is organized using two entities, 'River' and 'Landscape' with attributes that have values within a specific realistic domain. The model produces plausible (x,y,z) and (x,y,t) plots in the light of existing geomorphological theories. The described modelling procedure shows that it is possible to simulate river terrace formation three dimensionally with the use of empirical information.

4.1.1 Introduction The fluvial system in dynamic equilibrium is able to adjust itself to changes of external variables by changing its internal variables like channel depth and width, river roughness, mean velocity, channel form, and slope (Schumm, 1977; Dawson & Gardiner, 1987). River terrace formation can be looked upon as a result of changes in equilibrium, caused by variation in external variables like climate, tectonic and base level (Dury, 1970; Leger, 1983). The natural systems are so complex, and develop on such long time spans, that even laboratory experiments (e.g. Schumm et al., 1977) can only partly reveal part of their functioning. Computer simulation is increasingly recognized as a novel way to understand the way geomorphic systems work (e.g. Anderson, 1988) . Modelling and simulation comprise the activities involved in constructing a model of a real world system and simulating it on a computer (Elzas, 1978). To gain insight in river terrace formation, a simulation model was developed which graphically describes changes in a landscape due to changes of a fluvial system in time. This is an extension of the research on simulation of river terrace formation by Boll et al. (1988). Their model appeared to be unsuitable for three-dimensional

118 application,makin ga ne wapproac hnecessary .T omak ea noveral l model constructionpossibl e someoversimplification shav et ob e made inthos e cases inwhic h noaccurat e data or relations are available. The developed model consists of an algorithm of 575 lineswritte n in PASCAL and runs on aVA X 8600. This algorithm isavailabl eo nrequest .Th emode lconstruction ,it sproperties , and relationswil lb ediscusse d first,followe d bya simulation examplewit h input andoutput .

4.1.2. Materials andmethod s Model construction inthi s study started with the choice of a type of model which would best fit the objective to gain insight in terrace formation. Due to lack of accurate and detailed information, amacroscopica l empirical description of fluvial systemswa s selected as abasi s formode l formulation. Subsequently, a discrete time model was constructed, which simulatesgeneralitie s ofterrac eformation .

INFORMATION J— quantitative formula qua I ifafive descriptive like information information

modelling -

formulas state definitions decision rules

J_ model

Figure4.1. 1 Modellingwit htw ogroup so finformatio n

119 4.1.3. Model construction The information used in modelling can be divided into two major groups (Fig. 4.1.1): (1) quantitative, often empirical formulas,an d (2)descriptive ,qualitativ eknowledge .Sinc ebot h types of information are indispensable for the model, the qualitative knowledge was incorporated as well, using finite statemodellin gmethod s (Zeigler,1976) ,slightl yadapte dt oou r needsa sdescribe d inth eintroductio no fchapte r4 .Grammatica l analysis can be used to test the certainty and utility of descriptiveinformation .Sinc en osimplifie dgrammatica lanalysi s exists, a set ofprohibitio n rules on Dutch grammarwa s stated and used in analyzing texts.Thos e prohibition rules arebase d on the simple and limited syntax of PASCAL (Findlay & Watt, 1978),an dth esequentia lcalculation so fth ecentra lprocessin g unit. Since we are no linguists the grammar rules used can be lookedupo na sver yprimitiv eimplement sonly ,an dwil ltherefor e notb ediscusse d inthi spaper . States Erosion 111111111111000000000000 Channel form MMMMMMBBBBBBMMMMMMBBBBBB Uplift 111000111000111000111000

Resulting processes: Incision 111???111000000000000000 Bank-erosion 000???111111000000000000 Deposition 000000000000111111111111

M= Meandering channel B= Braided channels (note meandering excludes braided and vice versa) 1- true 0- false ?-Unknow nwhic herosio nproces swil lact . Table4.1. 1 Matchingo fstate san dth eresultin gdepositio nan derosio n processes. Inorde rt odefin eth etransitio n formon estat et oanother , decisionrule sar eused .Th estate so fth emodel ,whic har eonl y true or false, change at certain defined boundary conditions. Different states, are separated by thresholds and as such represent discontinuities is a system (Hugget, 1985). In the following example the state combinations of a fluvial system which result in a combination of incision and bank erosion or

120 deposition are listed in Table 4.1.1 (extracted from Schumm, 1977, 74-91). However,i tha st ob enote d thatothe rtext swil l yieldothe r states.Th equestio nmark s inTabl e4.1. 1 visualize gapsi nknowledg eabou terosio nunde rvariou suplif tan dchanne l form states.T o fillu p these gaps some assumptions weremade . In the model the question marks were replaced by 'true1. An advantage of using finite state modelling for the qualitative knowledge is that the relations of Table 4.1.1 are easily translated intodecisio n rules inPASCAL :

IFEROSIO NAN D UPLIFT ANDMEANDE R THEN INCISION UPLIFT AND BRAIDED THENINCISIO NAN DBANK_EROSIO N NOTUPLIF TAN DMEANDE R THENINCISIO NAN DBANK_EROSIO N NOTUPLIF TAN D BRAIDED THEN BANK_EROSION IFNO TEROSIO N THEN DEPOSITION (disregarding all other conditions) These rules canb e incorporated inth emode l algorithm.

Model organization When different textbooks are used, a well defined and formalized organization isneeded . A hierarchical organization structure,simila rt ocommo ndat abas eorganization swa schosen . Model organization, aspresente d inTabl e 4.1.2,wa s donewit h theus eo fentitie s and attributes.A nentit ywa sdefine d asa n independent, commonly compound, unit in the object system. An attributei sa property ,mar ko rcharacteristi cusuall yvariabl e ofa nentit y inth estudie dobjec tsystem .Th eentitie sdefined , RIVERan dLANDSCAP Ear edescribe db ythei rattributes ,eac hwit h values in a realistic domain.Th e interactions between the two entities are defined as PROCESSES. These processes are simplifications of the geological processes,an d were directly programmed in the model algorithm. The LANDSCAPE relief was defined asa n (x,y)matrix ,wit htw ostore dvalue s ineac hgri d cell (100mx 200m ): altitud ez (m )an dtim eo fdepositio nt (ky). Thismatri x isuse d asth emode l outputan dmake sGI S

121 Table 4.2.2

Time :1< Tim e<250 0(Ky )

B. Entities

1.Entity :RIVE R Attributes Domain 1.1Discharg e 20< Discharg e< 28 0(m 3s"1) 1.2Input_loa d 0< Input_loa d< 3. 0E" 3 (m3s_1) 1.3 Form meanderingo rbraide d 1.3.1Widt h 400< Flood_plain_widt h< 150 0(m ) 1.3.2Maxloa d 7.05E":1< Maxloa d< 3.5E" 3 ([nrV1) 1.4Erosio n 2.1E"J< Qerosio n< 3.5 E (n^ky-1)

2.Entity :LANDSCAP E Attributes Domain 2.1Quplif t :0< Quplif t< 0. 4 OAy) 2.2Relie fx,y, z :100< x < 1000 0(m ) 100< y < 1000 0(m ) 1< z < 500(m ) 2.2.1Valley_dept h :20< Valley_dept h< 50 0(m ) 2.2.2 Stratigraphy :0< stratigraph y< 250 0(ky ) 2.4Valley_widt h :400< Valley_widt h< 950 0(m )

Processes

1Uplif t :Upliftwhol erelie f 2Erosio n :Reliefdenudation ,a relief fvolum edecreas e 2.1 :Verticalerosion ,denudatio no fth evalle y bottom. 2.2 :Horizontalerosion ,denudatio no fth erive r banks. 3Sedimentatio n :builtu po fth erelief ,a nrelie fvolum e increase

Table4.1. 2 Modelorganizatio n

122 applications liketh ecreatio n ofsingl evalu emap spossible . The selected relief magnitude, 10k mx 20k m x 0.5 km (x,y,z), sizeo fmiddl escal erivers ,reflect sth emacroscopica lscal eo f the simulation. This also applies forth e selected timespan of 2.5millio nyear s (intim estep so f100 0years) , duet oth efac t thatmos tterrace s inmajo rrive rbasin sstarte d tofor ma tth e beginningo fth eQuaternary .Th edomai nlimits ,a suse d foreac h attributes (Tab. 4.1.2), were derived from published data on middlescal efluvia lsystem s (Leopolde tal. , 1964).Climat ean d tectonics areexterna l inputvariable s forth emodel . Discharge andsedimen tloa dar einpu tfunction swhic hwer eassume dt ovar y asa resul t ofclimati c changeonly .

4.1.4. Model Operation Ateac htim e step (of1 ky )a certai n scenario isconsidere d tocomput emode lbehaviour .Calculation so fth evolum et oerod e ordeposi twer efollowe db ystat edetermination san dcalculation s which determine the boundaries of relief changes and the processes which change the landscape. Discharge and input load areth einpu tfunction suse dt ocalculat eth erelie fvolum etha t changesdurin gon etim estep .MAXLOAD ,whic hi sth emaximu mloa d atth ecurren tdischarge ,i scalculate d accordingt oequatio n1 derived from Boll etal. , (1988).

MAXLOAD = LI x DISCHARGE2 + L2 X DISCHARGE + L3 X DISCHARGE (1) with: 3 MAXLOAD :Maximu m load (m s -1) LI :Suspende d load factor (snf ') L2 :Dissolve d load factor (-) L3 :Be d load factor (-) 3 DISCHARGE :Discharg e (m s -1)

LI, L2 and L3 are factors which represent respectively the dissolved,suspende d andbe d load component ofth emaximu m load ofa river .Thes efactor shav edifferen tvalue sfo ra meanderin g

123 and a braided river and are used to tune the model. For a meandering riverL I= 2.2 x 10"8(sm~ 3), L 2= 3x 10~8 (-) andL 3 = 1.15 x 10"6 (-),fo ra braide d riverL I= 1.9 x 10~8 (sm"3), L 2 = 3.2 x 10"6 (-) and L3 = 1.3'x 10"8 (-). One of the model assumptions is that at each timestep the river enters the landscape with input load and leaves itwit h themaximu m load. This condition makes it possible to calculate the erosion or deposition quantity (Qerosion)wit h use ofth emas sbalanc e of Equation 2.

QEROSION =MAXLOA D -INPUT_LOA D (2)

with: QEROSION :Erosio n orsedimentatio n quantity (irrV1) MAXLOAD :Maximu m load ofth esyste m (HlV1) INPUTJLOAD: Inputloa d (m3s"1)

discharge (I) input load (I) quptift (I) relief (I) (t=t-1)

channel form {*) I flood plain width (V) max load (V)

Figure4.1. 2 Flowcharto fth emode linput s(I) ,state s(* )an dvariable s (V).

124 The flowchart of Figure 4.1.2 shows how inputs, states and variablesar euse di nth emodel .[Uplift] ,[Erosion ]an d [Channel form]ar eth estate suse dt odetermin eth eprocess(es) . [Uplift] becomestru ewhe nth efluvia lsyste mcanno tcompensat eth euplif t by erosion during one time step. [Erosion] is true when the calculations ofth eerosio nquantit y indicatea positiv evolum e (QEROSION). The [Channel from] (meandering or braided) is determined bychange s indischarg e and inputload ,accordin g to thefollowin gdecisio nrules ,whic hwer eindirectl yderive d from qualitative relations formulated bySchum m (1977,135) .

IFQt- j< Q tAN D LOAD,.^> LOAD tTHE NMEANDERIN G ISTRU E

IFQ,.- !< Q tAN D LOADt-!< LOAD tTHE N FORM^ ISTRU E

IFQt,. !> Q tAN D LOADt-i> LOAD tTHE NFORM ^ ISTRU E

IFQ,.. !> Q tAN D LOADt.!< LOAD tTHE N BRAIDED ISTRU E with: Qt-! :Discharg e ofth eprecedin g time step (m3s_1) Qt :Discharg e (iti3s_1) LOADt-i :Inpu t load ofprecedin g time step (m3s_1) 3 1 LOADt :Inpu t load (m s" )

FORMt-i :Channe l formo fprecedin g time step (-)

Floodplain width and valley slope are two model variables whichac ta sboundarie sbetwee nwhic hrelie fchange stak eplace . Since the valley slope calculations did not yield any visible changes, the slope was assumed constant (0.005 m m"1). This insensitivity of valley slope is due to the relatively high minimumquantit yt ob eerode do rdeposited .Th efloodplai nwidt h for abraide d river isno t calculated, but comprises thewhol e width of the lowest level in a cross-section. Chorley et al., (1984,310)state dtha ti ti slikel ytha tther eexist sa relatio n between meanderwav e length (Lambda) and valley width. In this studyth erelatio n 'floodplainwidt ho fa meanderin grive ri s0. 5 Lambda' is used. Equation (3) is derived from the sinuosity formula ofa meanderin g river fromSchum m (1977,115).

FLOOD_PLAIN_WIDTH= 0.5 x3 0X (DISCHARGE/0.028)05 (3)

125 with: FLOOD PLAINWIDT H Floodplai nwidt h ofa meanderin g river (m) DISCHARGE Discharge

Themode lprocesses ,i.e. , (1)erosion ,subdivide dint oincisio n andban kerosion ,an d (2)deposition ,ar euse dt odistribut emas s transport over the x and y-axis (for instance increase of z- values of grid cells because of sedimentation process), as indicated in Figure 4.1.3. The incision process reduces the z value of the grid cells in the lowest part of a cross-section with oneunit ,an d bank erosion raisesth ewidt h inth e lowest parto fth ecros ssectio nwit h twounit s (Fig. 4.1.3).

new bank old bank 1--L bank erosion

old levve' l I newv level LJI incision JI | I

L old level _deposition J "**e w level

1m 100 m LJ erosion or deposition volume

Figure 4.1.3 Model and their inpact on a cross-section

126 4.1.5. Results A simulation yielding illustrative results was selected to demonstrate thesimulatio nprocedure .

Input Althoughth erelationshi pbetwee nfluvia ldynamic san dclimat e behaviour depends on the nature of climatic change and the effects of such changes on discharge and sediment load (Lowe& Walker,1984 ), a simpl erelatio nwa sassume dbetwee ntemperature , dischargean dsedimen tloa dchange .A climati ccycle ,consistin g ofa glacia l anda ninterglacia l period,wa sdivide db yBol le t al. (1988) into four phases, in which each have different discharge and sediment load characteristics. Direct sedimentological and paleoclimatical evidence in many natural terrace systems suggest that inglacia l periods,sedimen t load increases as a result of reduced vegetation cover and hence increased lateral sediment supplyt oth eriver .I n interglacial periods slope stabilization by vegetation and higher precipitation lead to increasing discharge. This more or less complementary relationship between discharge and sediment load is translated in the model as sinus and cosinus-functions , respectively. Inorde rt oobtai na mor e refined climaticcurve , theinpu tfunction so fdischarg ean dinpu tloa duse dar edirectl y based onth ewidel y acceptedMilankovi c curve (Imbrie &Palmer , 1979), which describes climatic changes in the Quaternary asa result of astronomical parameters. Consequently, discharge, Equation 4, is simulated asth e sum of three SIN functions and inputload ,Equatio n5 ,a sth esu mo fthre eCO Sfunctions ,o fth e earth orbit parameters periods of 23.000, 42.000 and approximately 100.000 years,a s found ind 180curve s byHay s et al. (1976).

DISCHARGE =DAV E+ DAMPx (SIN(27Tx TIME/100 ) (4) + SIN(27TX TIME/42 )+ SIN(27 Tx TIME/23) )

INPUT_LOAD= INLAV E +INLAM P X (C0S(2TT XTIME/100 ) (5) + COS(27Tx TIME/42 ) +COS(27 TX TIME/23) )

127 with: DISCHARGE :Discharg e (nrV1) DAVE :Averag edischarg e (150m 3s_1) INPUT_LOAD :Inpu t load (m3s_1) INLAVE :Averag e input load (0.005m 3s"1) DAMP :Discharg e amplitude (130m 3s_1) INLAMP :Inpu t load amplitude (0.0015m 3s_1) TIME :Tim e (ky)

Theinpu tfunctio no ftectonic si nthi ssimulatio n isa constan t uplift of 0.1 m each 1000years .

Output Each 0.5 million years pictures are made of the landscape (x,y,z,t)data .Three-dimensiona lpicture sar edraw nwit hus eo f thecompute rpackag eUNIRAS .Th e(x,y,t )picture so fFigur e 4.1.5 were drawn manually. The (x,y,z) relief diagrams show the development anddestructio n ofrive rterrace s (Fig.4.1.4) . The uplifto fth elandscap ea swel la sth eincisio no fth evalle yar e visible.Th e (x,y,t)diagram sindicat eth etim eo fdepositio no f thesurfac edeposits ,a tth een do fth esimulatio n (2.5millio n years)ther ear estil ldeposit swhic hwer e laiddow ndurin gth e first 0.5 million years (Fig. 4.1.5). Those remnants are only preserved at small high spots. Another visualization of the dynamic properties of the simulated example is shown in Fig. 4.1.6,whic hportray sth e formation,reduction ,an ddestructio n of several terraces in one cross-section aty=18km , during the simulation process.Onl y some small terrace remnants remain in the cross-section. Besides graphical output,th e number ofth e model states and processes which occur during simulation, is recorded (Tab. 4.1.3).. Theperiod sdurin gwhic hth emode l statesremai nconstant ,hav e anaverag eduratio no fabou t7 ky ,wit h aminimu m of 1k yan da maximum of 19 ky. Figure 4.1.7 gives a detailed insight in terraceformatio nbetwee n1.4 2 and1.4 5millio nyears ,cause db y thesequenc e ofmode l statesan dprocesses .

128 Figure4.1. 4 x,y,zrelie fpicture so fth elandscap edevelopmen teac h0. 5 millionyear sdurin gth esimulation .

129 130 131 t = 0 t=0.5 t =1. 0

-10 km-

t= 1.5 t=2.0 t=2.5

2

0

Figure4.1. 5 x,y,t pictures, showing the time of deposition of the surface deposits each 0.5 million years during the simulation.

Uplift Erosion Meandering Incision Bank-erosion n^

1 1 1 1 0 71 1 1 0 1 1 2 0 1 1 1 1 1044 0 1 0 0 1 674 1 0 1 0 0 43 1 0 0 0 0 70 0 0 1 0 0 104 0 0 0 0 0 492

1720 •>2: 186 1791 1262 1117

1- true 0= fals e n^= Frequenc y oftim e steps thata mode l state occurs (sumi s 2500). n2= Frequenc y oftim e steps thata stat e variable orproces s istrue .

Table4.1. 3 Frequencieso fth emode linputs ,states ,an doutputs .

132 Figure 4.1.6 x,90,z picture, showing the development of one cross- section (y=90) during simulation.

time in 1000year s 1 I 1300 1400 1500

uplift j i

•H k -i-L _W |_ ± 1 PM j—±—LJ 1

meander

-*-i I*I 1*1 i*l

bank _rt i- erosion -*—tru e false

Figure4.1. 7 Model states and processes occuring between 1,3-1,5 million yearsdurin g simulations.

133 A period of erosion (incision and bank erosion) is the two channel types interchanges with a period of sedimentation. Due to this sedimentation the fluvial system cannot compensate the uplift by incision, and the [Uplift] state becomes true. When sedimentation hasstoppe d ameanderin g river immediately starts incisingunti lth euplif ti scompensated .Thi sincisio n (without bank erosion) causes a new terrace to develop. During the following time steps the new terrace decreases inwidt h due to bankerosion . Observingth esimulatio nresult si tca nb econclude dtha tbot h intrinsican dextrinsi cchange so fstat etak eplace .Whe ninsid e the system thresholds are crossed without changes in input variables,lik euplift ,a nintrinsi cchang eo fstat etake splace . Whenever the system responds directly to external influences, extrinsicthreshold sar ecrosse d (Huggett,1985 ;Schumm , 1977). Due to the simple model organization the input strongly determines the output causing the majority of the threshold crossings to be of extrinsic character. By changing model parameters other simulations were carried out. Simulations without uplift orwit h a constant uplift ofmor e than 0.5 m/ky resulted in terraces which disappeared quickly after their formation.

4.1.6. Discussion and Conclusions Theobserve ddifference si nth evariou ssimulation shav emad e it possible to formulate some conclusions concerning terrace formation inth e described model.Withi n the simulated fluvial system model, terraces are formed by changes in discharge and sediment load. Commonly terraces are quickly eroded. At some specific inputvalues, however , such as a constant uplift rate of 0.1 m/ky, some terrace remnants are preserved from being eroded.Bot hextrinsi c and intrinsicchange so fstat eca nbrin g aboutterrac eformatio ni nth emodel .Althoug hthi smode lha sno t and cannot be validated, it produces plausible (x,y,z) and (x,y,t)plot si nth eligh to fexistin ggeomorphologica ltheories . The described model and simulation show that it ispossibl e to simulate river terrace formation and to graphically display

134 the results with use of both quantitative and qualitative information. The decision ruleswhic h areth e result of finite state modelling, allowing the incorporation of qualitative relations. They act as thresholds and strongly determine the model output. This indicates that further model improvements could be found in extension of the used decision rules or refinements instate san dstat etransitions .Whe nth epresente d modeli sapplie dt oa rea lsyste mi tca nb euse dt oformulat ean d analyzehypothese so ffluvia lterrace,formation .Th emethodolog y used is also applicable on many other geomorphic subjects, especially whenno tal l information iso rca nb e quantified.

135 4.2 A 3-D MODEL OF FLUVIAL TERRACE DEVELOPMENT IN THE ALLIER BASIN.

A.Veldkam p

Abstract The combined effects of climate and tectonism on general terrace stratigraphy and valley asymmetry during the last half million years in the Allier system (France) are simulated by a 3-D conceptual model (LIMTER). This model allows the formulation and evaluation of long term terrace formation scenarios for the Allier system. Simulation results suggest that terrace stratigraphy in the study area is mainly the result of the internal Allier dynamics and climatic change. Local tectonism caused the development of unpaired terraces while the general regional uplift played a dominant role in terrace formation and preservation in general.

4.2.1. Introduction Riverterrace sar ea fundamenta lpar to f fluviallandscapes . The existence and formation of terraces has been subject of research for many years because they provide a relative chronology to which other geological, geomorphological or palaeohydrological events can be related (Dawson & Gardiner, 1987). It has always been tempting to link Quaternary terrace chronologies with established chronologies of major Quaternary environmental changes, such as climatic and sea level oscillations, especially when other dating evidence is not available. A first andwel l known generalization about the response of fluvialsystem si nsemiari dt osubhumi dregion st osligh tshift s in climate was set forth by Huntington (1907, p. 358) and is called"Huntington' sprinciple "(Fairbridge ,1968 ,p.1125 ;Bloom , 1978, p. 248).Huntington' s principle, was established as a general model which many followed in constructing terrace chronologies,whereb y glacials are thought to be characterized by aggrading braided rivers and the interglacials by incising meandering rivers. This generalization is now recognized as being exceedingly simplistic. The other major generalization of a driving force behind terraceformatio nwhic hi so rwa softe napplie dconcern schange s inbas e level. Such base level changesma y be due to sea level

136 changeso repirogeneti cmovements .I nBritai nbas eleve lchange s received much attention and the relative changes in sea-level were seen as a major external driving force for river-level changes (Green & Mcgregor, 1987). In continental Europe a climatic school dominated which applied base level changes due to tectonism as second important external variable (Brunnacker & Boeningk, 1983;Buch , 1987;Texie r &Raynal , 1984). However, aperfec tassociatio no fterrac edepositiona lphase san dincisio n episodes with climatic phases of certain base-levels is rarely possible, or even likely. Terraces are in fact very complex features,resultin gpartl yfro mexterna lfactor san dpartl yfro m thenatur eo frespons e inth e fluvial system itself.River sma y maketerrace swithou tther enecessaril ybein ga nexterna lchang e (Schumm, 1977). Ofte n the external model breaks down on a lack ofdetaile d understanding ofth eprocesse soperatin gwithi nth e fluvial system.

Stratigraphy offersa direc twa yt ogai nmor e insight inth e internaldynamic so fa fluvia l system.Th e sedimentproperties , deposition conditions and their sequence can give detailed insight in formercondition swithi nth e fluvial system studied. Breakso rfacie stransition swithi nth esedimentar yrecor dofte n mark major environmental changes, indicating possible external forcescontrollin g fluvialdynamic s andterrac e formation.

In general it has long been recognized that alluvial sedimentary successions arecompose d oftw omai n faciesgroups . Onegrou pcomprise ssandstone san dconglomerate sth eothe rgrou p consists of,mudstone s and siltstones (Bridge & Leeder, 1979). A similar alluvial sediment discrimination wasmad e in fluvial terracesediment sb yTexie r& Rayna l (1984)wh ocorrelat ecoars e sediments to glacial environments and fine sediments to interglacialenvironments .The ybelon gt oth eclimati cschoo la s they linked the reviewed terraces with the major Quaternary climaticcycles .

A simulation of alluvial stratigraphy within one floodplain

137 understric tassumption s (Bridge& Leeder ,1979 )showe dtha tth e internal dynamics can strongly determine the stratigraphical record. Tectonic movements had only a significant influence on the simulated successions if a preferred direction of tilting maintained. Fromthi ssimulatio n andth ecorrelation s ofTexie r & Raynal (1984) itbecome s obvious that if onewant s to create a more realistic model of fluvial terrace formation the stratigraphical record hast ob e incorporated.

More effects of changes in fluvial dynamics can be viewed, whenconsiderin gbot hgeographica lpositio no fterrace san dthei r stratigraphical record.Th erol eo ftectonis m seemed onlymino r inth estratigraphica lmode l ofBridg e& Leede r (1979)bu twhe n similar movements are considered in a model including terrace formationmor edistinc tresult sca nb eexpected .Tectoni cuplif t is likely to cause vertical erosion of a river and thus abandoningo fth efloodplain ,a neffec tno tincorporate d inthei r stratigraphical model. Inrespec tt otectonis mrive rterrace sar emor evaluabl etha n stratigraphybecaus ethe ygiv edirec tindication so fforme rrive r gradients, and successive terraces show successive river positions. In areas of tectonic activity, valleys and their terracesma yb easymmetrical ,tilte d orwarpe d invariou sways , givingvaluabl eclue s forinterpretatio n oftectoni cmovements .

Modelswhic hdescrib efluvia lsyste mdynamic sfo rlonge rtim e spans are sparse because most knowledge on fluvial systems is mostlybase do nshor tterm ,wel lcontrolle dexperiment s (Gregory, 1983; Dawson& Gardiner , 1987).The yar etherefor eusuall ybase d on the correlation of terraces with known major changes during the Quaternary. Most of these models are used for the interpretation of largeterrac esequences . Recentlya quantitativ ebu tstil lconceptual ,lon gter m (2000 ky) macro scale (100 km2) fluvial system model was constructed whichsimulate sth edevelopmen to ffluvia lterrace sa sth eresul t ofexterna lan dinterna lchange si nth esimulate d fluvialsyste m (Veldkamp &Vermeulen , 1989). This so-called coarsescal emode l

138 (Thornes, 1987) did not consider the effects on the fluvial stratigraphical record.

In this paper the combined effects of climatic change, tectonisman dinterna lfluvia ldynamic so nalluvia l stratigraphy andvalle ymorpholog yar esimulate dusin ga comprehensiv eversio n ofth eexistin g 3-D fluvialterrac e formationsimulatio nmodel . As an exercise a plausible terrace development scenario is simulated for theRanda n section inth eAllie rvalle y with the 3-DLIMTE R (LIMagneTERraces )model .Th eprogra mlistin g canb e found asappendi x II inthi sthesis .

4.2.2. Model characteristics Modelconstruction ,organizatio n andoperatio nar edescribe d in a previous paper (Veldkamp & Vermeulen, 1989). Model organizationwa sdon ewit hentitie san dattributes .A nentit yi s an independent unit inth e object system, the entities inthi s model areLANDSCAP E andRIVER .A nattribut e isa property ,mar k orcharacteristi c ofa nentity ,suc ha sdischarg e forRIVE Ran d Valley_depth forLANDSCAPE .Attribute shav ebee nconstraine d to lie in realistic domains for the Allier system. The LIMTER entities and theirattribute s aredescribe d inTabl e1 . Interaction betweenth etw oentitie sar edefine d aserosio nan d sedimentationprocesse s acting intimestep so f 1000years . Thesemode lprocesse sar elon gter man dlarg escal eanalogie s of real erosion and sedimentation processes. They are artificially constructed processes reacting to changes in the 1000year sdischarge/sedimen tloa dequilibriu mwhic har edefine d as a function of climate.Whe n the sediment input load exceeds the sediment transport capacity the difference is deposited (DEPOSITION = true), and incas eth etranspor t capacity exceeds the input load thedifferenc e iserode d inth esimulate d system (EROSION= true). Thequantit yt oerod eo rdeposi ti nLIMTE R iscalculate d from a sediment budget,bu twhic h processes act inth e LANDSCAPE is determined by decision rules extracted from general literature onhydrolog y (Schumm, 1977).

139 Time :1< Tim e<80 0(Ky )

B Entities

1 Entity:RIVE R Attributes Domain 1 1Discharg e 1.58xl012< Discharg e< 9.47xl0 12 (m3ky~ 1 2Inpu tloa d 0< Input_loa d< 9.46xl0 7 (m3ky"1) 1 3 Form meanderingo rbraide d 1.3.1Widt h 0.15

2 Entity:LANDSCAP E Attributes Domain 2 1Quplif t 0< Quplif t< 0. 4 (m/ky) 2 2Quplif tunequa l 0< Quplift_unequa l< 0.0 5 (mAy) 2 3Relie fx,y, z 150< x < 1500 0(m ) 150< y < 1500 0(m ) 1< z < 500(m ) 2.3.1Valley_dept h 20< Valley_dept h< 50 0(m ) 2.3.2Stratigraph y 0< stratigraph y< 100 0(ky ) 2.3.3 sediment_composition 1< sediment_compositio n< =4 (-) 2.4Valley_widt h 0.15< Valley_widt h< 14.3 0(km )

Table4.2. 1 Modelorganizatio n

Thesedecisio n rulesare : IFEROSIO NAN DUPLIF TAN DMEANDE R THEN INCISION IFEROSIO N ANDUPLIF TAN D BRAIDED THEN INCISION AND BANK_EROSION IFEROSIO N ANDNO TUPLIF TAN DMEANDE R THEN INCISION AND BANK_EROSION IFEROSIO N ANDNO TUPLIF TAN D BRAIDED THENBANK_EROSIO N IFNO TEROSIO N THEN DEPOSITION (disregardingal lothe r conditions) WherebyUPLIF Tan dNO TUPLIF Tar esyste mstate sindicatin gth e impact of tectonic uplift.UPLIF T istru e (NOTUPLIF T = false)

140 whenth efluvia lerosio ni sno tabl et ocompensat eth eLANDSCAP E uplift.MEANDE R istru ewhe nth eAllie r ismeanderin gwhil eth e BRAIDEDstat ei stru ewhe nth eAllie ri sbraiding .Th eimpac to f INCISION, BANK_EROSION AND DEPOSITION in a cross-section are shown in Fig. 4.1.3. Erosion processes are headward migrating alongth elongitudina lprofile ,whil esedimentatio nmigrate s in downwarddirection .Irregula rheadwar derosio ndurin gerosio no f a meandering RIVER is caused by changes in the effective floodplainwidt hdurin gth esimulation .Th eresultin g irregular bankerosio ncause sa simila reffec ta sa meanderin grive rdoes .

During a simulation the volume to erode or deposit by the RIVERi nth eLANDSCAP Ei scalculate dfo reac htim este po f1 ky . These calculations are followed by state determinations and boundary calculations within which LANDSCAPE changes must take place. The LANDSCAPE is changed by modifying grid cells elevations in response to the acting process. The changed LANDSCAPE isstore d ina geographica l information system (GIS).

Terrace stratigraphy Terrace stratigraphy was also incorporated within LIMTER. During the simulations the stratigraphy (sediment age and composition class) of the upper ten metres (grid cells) are stored. The sediment composition classes arederive d from bulk geochemical sediment research ofth eAllie rterraces .Ther ear e four sediment composition classes with the following characteristics: Class 0, isth eunderlyin g bedrock; Class l, are sandy and clayey sediments with a low volcanic content which are deposited during interglacial periods ina meanderin g system; Class 2, arepredominantl ysand ysediment swit ha nintermediat e volcanic content which are deposited in a braided system,durin g glacialperiods ; Class 3 are gravelly sediments with ahig h volcanic content, whichar edeposite d ina braide d system atth een do f

141 a prolonged glacial (>1 0ky) .

4.2.3. Model Input The inputs and outputs of theAllie r system simulations are described after which simulation results are presented and discussed.

Climate The Quaternary has known many astronomically controlled mondial changes in climate, which can be very satisfactory described by Milankovics curve (Berger, 1978). These climatic changesar edirectl y registeredi ndee pse acores ,ic ecore san d lossprofile sgivin gindication so fth erelativ eamount so fwate r stored in ice masses, and changes in mondial circulations. Although the relationship between fluvial dynamics and climate behaviour depends on the nature of climatic change and the effects of such changes ondischarg e and sediment load (Lowe& Walker, 1984), a simple linear relation was assumed between caloric insolation (Berger, 1978), mean discharge and sediment load each 1000 years. During Glacials much water is stored in glaciers and drier continental climates prevail causing lower mean100 0year sdischarge si nrivers .Du et oth edrie ran dcolde r glacial environment the vegetation cover decreases causing an increase in sediment supply to fluvial systems. Interglacials yield the opposite picture,a n increase inth emea n 1000year s dischargean da complementar ydecreas ei nth e100 0year ssedimen t supply.Thi ssimplifie d relationship betweenmea ndischarg ean d meansedimen tloa di spartl ysupporte db yth echange si nmagneti c susceptibility as found in the long loss sections in China, indicatingdr yglacial san dwe tinterglacial sdurin gth elas t2. 4 millionyear s inthi scontinenta l area (Heller &Wang , 1991).A similarconditio n canb eexpecte d inth eAllie rbasi nwher eth e glaciers caused a permanent high pressure area during the glacials. Asmode lsimulation srequir ea climati c inputwit ha constan t reliability during the simulated time span a simplified

142 Milankovics curve was used as basic climatic input. Mean discharge each 1000year s issimulate d asth e sum of three SIN functions with the periodicities of the precession (23,000 years),obliquit y (41,000years )an deccentricit y (96,000years) . As the sediment load is assumed to have a similar behaviour complementaryt oth e100 0year sdischarg ei twa ssimulate da sth e sum of the COS functions of the same three periodicities. Although these curves do not exactly match the climatic curves derivedfro mdee pse acore san dlos ssediments ,the ysufficientl y describe climatic changes during the Quaternary for our conceptual longter mmodellin gpurposes .Th egenera l assumption thatther eexist sa simpl estraigh tforwar drelationshi pbetwee n Milankovicscurv ean dbot hth e100 0year sdischarg ean dsedimen t loadcurve si so fcours ea to ostron goversimplificatio n tohav e any realisticvalidity . Iti sobviou stha ta refinemen t ofthi s assumptionwoul d improveth egenera lvalidit y ofth emodel . Exceptth edifferenc e inbehaviou ro fth efluvia ldynamic si n aglacia l and interglacial stadiath etransitio n froma glacia l toa ninterglacia lenvironmen twa sals otake nint oaccount .Bul k geochemical research showed also that changes in sediment flux magnitude and sedimentcompositio n arerelate d toth e intensity and duration of a climatic episode (Veldkamp, 1991). During prolonged glacials, largeglacier s werebuil t up on the higher partsi nth eAllie rbasin ,th eCanta lan dth eMon tDor evolcanoe s (Veyret, 1978).Whe nclimati c conditions improved relative fast glaciermeltin gcause dth ereleas eo flarg equantitie so fcoars e volcanic rich sediments into the Allier. During the Late Quaternary this extra sediment flux ranged from 2t o almost 50 timesth e Interglacial fluxquantities . The fluvioglacial sediment fluxes are incorporated within LIMTER as follows: when a glacial lasts longer than 10 ky an extra large (4 x normal flux) sediment flux with composition class 3,i srelease d intoth esyste m atth etransitio nt oa ne w interglacial period. The fluxmagnitud e of 4 isbase d onmode l tuning.Th eresultin g alternating erosionan ddepositio n states during thesimulatio n areclearl y illustrated byth e cumulative erosion graph (Fig. 4.2.1).

143 cum erosion " 3 10 M 0.3

Figure4.2. 1 Cumulative erosicn curve, derived frcm changes in discharge/sedimentloa dequilibriu mdurin gsimulatio no f UMTER.Thi scurv eillustrate sth ealternatin gerosio nan d sedimentationstage sdurin gsimulations . Tectonism Two different components of tectonism (Fig. 4.2.2) are incorporated in LIMTER. A component of gradual uplift of the whole simulated landscape (QUPLIFT), and an uplift component describing the difference in uplift rate of the landscapes on both sideso fth eAllie r (QUPLIFT_UNEQUAL).Th elatte rtectoni c component assumes an active fault in the middle of the Allier valley,dividin gth eentit yLANDSCAP Ei ntw opart swit hdifferen t uplift rates. There is some field evidence which suggests the presence of sucha faultingzone . Because there isn oproo f that tectonism has known changing ratesi nth eAllie rbasi ni ntime ,onl ysimulation swit hconstan t gradualuplif tar emade .A simulatio nwit ha QUPLIF To f0. 1m/k y and aQUPLIFT_UNEQUA L of O.Olm/kywil lb eshown .

144 1 1 initial relief \ QUPLIFT + \ QUPLIFT QUPLIFT-UNEQUALL // / \ /

t / rn_i_ _ J ^ I- •aUPLIFT -• -J 1 I

Figure 4.2.2 Twotectoni c components as input of the TJITTE R simulation. Quplift is the general uplift of the modelled area, Quplift-unequal is the difference in uplift rate between the two valley sides, assuming an active fault in the middle of the valley parallel to the flow direction.

Initial relief The initial relief (Fig. 4.2.3.a) is also a model input. The simulated LANDSCAPE has a surface of 225 km2 (15 x 15 km), a maximum altitude of 270 m and a minimum altitude of 240 m. The initial relief consists of a broad (valley_width is 7800 m) , shallow ( valley_depth is 30 m) valley. Terrace stratigraphy displays only age 0 and composition class 0 indicating that no fluvial sediments occur in the initial LANDSCAPE.

4.2.4. Model Output Theoutpu to fLIMTE R isa raste rGI S filewit hth e LANDSCAPE altitudes and stratigraphy for each timestep. With thesedata , cross sections,maps , 3-D relief graphs,o r 3-D graphs of one cross section development intim e canb edrawn .

4.2.5. Simulation results A simulation with plausible realistic simulation results is presented inmor e detail.Thi s simulation had the climatic and tectonic inputs as described by themode l input.Th e LANDSCAPE developmenti ntim edurin gthi ssimulatio ni sillustrate dthree - dimensionally inFig . 4.2.3 foreac h 100ky .

145 initial relief

relief version D2 after 100 timesteps

Figure4.2. 3 3-D valley relief development during simulation, in timesteps of 100 ky. Ihe effects of changing fluvial dynamicsar eillustrate db yincisio nan dterrac eformation . Irregularities inth evalle y slopes are also caused by changesi nfluvia ldynamic sdurin gth eheadwar derosion .

146 relief version D2 after 200 timesteps

relief version D2 after 300 timesteps

147 relief version D2 after 400 timesteps

Relief version D2 after 500 timesteps

148 relief version D2 after 600 timesteps

350 cross section in time version d2

Figure4.2. 4 3-Dgrap ho fth edevelopmen to fon ecros ssectio ni ntime . Anasymmetrica lterrac esequenc edevelop sa sa resul to f distorteddowngradin go fth eAllie rsystem . 149 Valley morphology Simulations without tectonic uplift result in temporary terraces only, while large constant uplift rates (>0.3 m/ky) cause steepcanyon swithou tterraces ,whil euplif trate saroun d 0.1 m/kydispla yman y terraces.Simulation swit hgradua luplif t ofth eLANDSCAP Ea sa whol eresul ti npaire dterrace sonly .Whe n a difference inuplif trat e forbot hhalve so fth eLANDSCAP E is introduced (QUPLIFTJDNEQUAL> 0 ), asymmetrica lterrac esequence s develop with both paired and unpaired terraces. Most terraces remain on the valley side with the highest uplift rate. Simulations with both general and unequal uplift components result in landscapes with comparable characteristics as the presentRanda nterrac esequence .T o illustrateth e incisionan d sedimentation dynamics inmor e detail a cross-section sequence intim ei sdraw ni nFig .4.2.4 ,showin grelie fchange sdurin gth e first 350 ky of the simulation. Both figures clearly show the alternating incision and sedimentation of the Allier in time causingsevera ldifferen tterrac elevels .Th eeffect so funequa l uplift also become prominent during the simulation. An asymmetrical terrace sequencedevelop sa sa resul to fdistorte d downgrading ofth eAllie r system.

Duringth esimulation sunpaire d terracesdevelo p onlydu et o changes in unequal uplift. This result contrasts with the traditionalvie wtha tunpaire dterrace sar egenerate db ychange s ininterna l factorssuc ha sa widel yswingin gmeande rbel twhic h produces unpaired terraces during the slow lowering of abroa d floodplain.

Terrace stratigraphy Terrace stratigraphy displays the total effect of both sedimentation and erosion processes related to the sediment fluxes dynamics during simulation. Simulations without major changesi nsedimen t fluxdu et oclimati cchang eresulte d inthi n sediment layers with a very simple stratigraphy. When the sediment composition and fluxmagnitud e arerelate d toclimati c environment aver y complexterrac e stratigraphydevelops .

150 t = 420 t=440

ITITHITI1 ffff]

iffi^w

t =46 0 t = 480

MM m

t =50 0 t= 520

nn Inn

10 m -i 1 ikm

Figure 4.2.5 A cross-section and terrace stratigraphy development during the last 100 ky of the simulation shown each 20 ky. 151 It isno tsurprisin gtha tth elarges tsedimen t fluxes (Class 3), i.e. those related toglacie rmelting , are relatively well preserved in the terrace stratigraphy. During these large fluvioglacial sediment fluxes the floodplain and the lowest terraces areburie db ythes esediments .Thi sburia l causesth e formation of the 'standard1 Allier terrace stratigraphy with relatively volcanic poor sandy units buried by volcanic rich gravelly units. This development can be seen in the cross- sections inFig . 4.2.5., displaying the changes inbot h relief and stratigraphy (sediment age and composition). Six cross- sections illustrateth e fluvial systemdynamic sdurin g thelas t simulated 100 ky. The last cross-section shows the LANDSCAPE during an interglacial period like the present Allier and can directly becompare dwit hth eactua lRanda n cross-section. Although the river has exhibited a meandering state many times, iti ssurprisin gtha tth etypica l interglacial sediments (Class 1)ar e rarely found inth e stratigraphical record. This limitedoccurrenc e ofmeanderin g sediments istru e forbot hth e described simulation and the actual Randan terrace sequence. During LIMTER simulations interglacial sediments are deposited duringeac hinterglacia lbu tthe yar ealmos talway serode ddurin g the same interglacial or the subsequent glacial. The limited occurrence of interglacial sediments inth esedimentar y records seems to be due to the mainly eroding characteristics of a meandering Allier,th emos t common interglacial Allierstate .

4.2.6. Evaluation of the simulated Randan terrace sequence InFig .4.2.6 . thecross-sectio n after 520timestep s (ky)i s plotted together with the cross-section near Randan, allowing comparison of the LIMTER simulation results with the actual Allier system atRandan . Itha s tob e realized that the Randan crosssectio n inFig .4.2.6 . isonl ya stron gschematizatio no f reality. The effects of mass movements on slopes and the dissection ofterrace sb ymino rtributarie s areno t included in theschemati ccross-section .Becaus eLIMTE Rha sonl y conceptual valueth ecorrespondenc e betweenth emode l andth eRanda n

152 relative altitude(m) randan Jfflr 800

ffl! -300

15- 3lfl-ff^0 ^ P-; 1km

volcanic rich sediments (class 3 ) intermediate volcanic rich sediments (class2 ) • volcanic poor sediments (class 1 ) 300 age in kyear s

Figure4.2. 6 Comparisono fth eRanda nterrac esequenc ean dth emode l output.Not eth egoo dcorrespondenc ebetwee nth eyounge r terracesfo rmode lan dreality . sequenceshoul db ereviewe dqualitativel yonly .Thi simplie stha t the rates for discharge, sediment load and uplift, have no validity outsideth emodel ,bu tthe ymigh tgiv esom e indication ofth eorde ro fmagnitud ethes evariable smigh thav eha di nth e Allier system. Themajo rpoint so fcorrespondenc e betweenLIMTE Routpu tan d the Randan sequence are the number of terraces, the general terracestratigraphy ,an dth erelativ ealtitude san ddistributio n of theterrace s onth evalle y slopes. Also clear differences existbetwee nth emode lan dreality .I nLIMTE Ra relativ eag eo f 420k yi sfoun d forth e6 5m terrac e (Va)whil ei nrealit y this terraceleve li sthough tt ohav ea nag eo fapproximatel y80 0ky . This large agedifferenc e suggests that the incision rate in LIMTER wasprobabl y toofas t compared with what actually took placei nth eAllie r system.Whe nth eLIMTE Ri sru nfo r80 0k ya much deeper and steeper valley exists as the actual Randan sequence. On the other hand, the age difference between the youngerterrace si nth emode lan di nth eAllie ri smuc hsmaller , suggestinga muc hbette rsimulatio no fth eAllie r system during

153 the last few hundred thousands of years. This change in age correspondence between simulation results and reality suggests thatth euplif trate snea rRanda nchange ddurin gth eQuaternary . Probably analmos t stand still inuplif ttoo kplac ebetwee n80 0 and 400 kyears BP. This interpretation implies that the model assumptiono fconstan tuplif trate si sincorrec t forth eRanda n terracesequence .A chang e intectoni c activity issupporte db y somedistorte dlongitudina lterrac eprofile sindicatin gregiona l tectonic activity between the formation of theV and W terrace levels (Larue,1979 ;Gio t etal. , 1978). Another striking difference between simulation results and reality isth e complete lack ofterrac e remnants of theV a and Vb terraces on the easternvalle y side inth eRanda n sequence. Thisdiscrepanc y isdifficul tt o interpretbu t itmigh tals ob e relatedt oth esam etectoni cevent swhic hcause dth edifference s interrac e sedimentage .

In general it is striking that relative simplemode l inputs canaccomplis ha complicate dterrac esequence ,includin gunpaire d terracelevel san da relativ ecomplicate d terrace stratigraphy. Despite the differences between LIMTER output and the Allier terracesequenc eth esimulatio ni sthough tt odispla ya possibl e generalscenari ofo rth edevelopmen to fth eactua lRanda nterrac e sequence.

General applications ofLIMTE R LIMTER has of course no validity outside the Allier basin becausei twa sadapte dt othi suniqu esyste monly .Th eimpac to f climatewa ssimulate d relatively straight forwardb ychange si n the100 0year ssedimen tan ddischarg ebecaus eth eAllie r system isstrongl ycontrolle db yfluvio-glacia lfluxe sfro mglacier si n itsheadwater s (Veldkamp,1991) . Itremain stherefor et ob esee n whether this input assumption has any validity in a system without such a fluvio-glacial controldurin gth eQuaternary . The hydraulically-based clastic sedimentation model of Tetzlaff and Harbaugh (1989) seems to allow a more straight forwardinterpretatio no fsimulatio nresult stha nLIMTER .LIMTE R

154 calculates only an average impact each time step (1000 years) whilethe ycope dwit hlonge rtimespan sb yth e'compute-and-drift ' and the 'compute-and-stop' strategies. Although Tetzlaff & Harbaugh's basic assumptions seem realistic their complete numerical model has so many calculation operations that no reliable results can be expected. During each operation the calculation errors increase resulting in unreliable outputs. LIMTER has less realistic basic assumptions but has much less calculation operations reducing the calculation errors and limiting the computing power demands. It is obvious that both approaches have different advantages and disadvantages and are more or less complementary. Iti sbeyon d anydoub ttha ta nelaborate dmode lwit ha soli d hydraulically-based foundation such as constructed by Tetzlaff & Harbaugh (1989), suggests amor e general validity. But their numeric model is mainly focussed on quantified transport and deposition processes neglecting other complex dynamics in the fluvialsystem .Th eremainin gproble mi stha tsuc ha mode lshoul d be tuned to both sedimentary sequences as valley morphology (river terraces). The latter is the main goal of LIMTER while long term hydraulogically based models can only be tuned to sedimentary sequences.Anothe r problem ist o obtain a reliable long term input for such a hydraulogical model, because our oversimplified inputassumptio n isunsuitabl e forsuc ha model .

4.2.7. Conclusions Thelon gter msimulation so fa nAllie rlik esyste mwit hLIMTE R suggesttha tth eAllie rterrace sa tRanda nar emainl yth eresul t ofth einterna ldynamic srespondin gt obot hclimati can dtectoni c factors. The conceptual model (LIMTER) illustrates that inth e Allier system tectonism may have played a dominant role in determining the terrace formation and preservation, while climatic dynamics seem to have strongly determined terrace stratigraphy bycausin g changes incompositio n andmagnitud eo f sedimentfluxes . The simulated valley morphology is the result of the interaction oftectoni cmovement san d fluvialdynamics .Unequa l

X55 upliftresult si nunpaire dterrace san dasymmetrica ldowngradin g ofth e simulated river.

156 Chapter5 SYNTHESIS

Themai nai mo fthi sresearc hwa st oestablis ha quantitativ e largescal ereconstructio no fth elongter mAllie rdynamic sa sa result of global environmental changes. Therefore a model was made simulating these long term dynamics. Model dynamics were based on the most commonly applied concept in fluvial palaeohydrology,th econcep to fdynami cequilibriu ma spostulate d by Schumm (1977). Within a complex natural system such as a fluvial system a single event isthough t tob e able to trigger a complexmorphologica l and/or sedimentologicalrespons ea sth e various components of the system react to the change.Withi n a fluvial system indynami cequilibrium ,terrac e formationma yb e concentrated into relatively short time periods of dynamic equilibriumassociate dwit hinterglacial ,interstadia lo rglacia l conditions (Green &McGregor , 1987). A key problem which has to be solved is to determine under which conditionsthreshold sar ereache d ina fluvia l system. In the Allier system we found a rather straight foreward relationshipwhereb y theAllie rca nb eenvisage d asoscillatin g betweena cold-climat ebraided ,terrac ebuildin gconditio ndurin g and at the end of a glacial and a more meandering, incision conditiondurin gth einterglacials .Th eAllie rthreshol dbetwee n deposition and incision isfoun d atth etransitio n fromglacia l to interglacial. For NW Europe most investigators agree on a general cold depositiono fmos tterrac esediment sbu tthei rinterpretatio no f the exact system dynamics involved in creating such deposits differstrongl y (Starkel,1983 ;Gibbard ,1988 ). Par to fth eMeus e (M.W.va n den Berg,pers . comm.)an d Thamesbasin s (Dawson and Gardiner,1987 )ar ethough tt ohav esimila rdynamic sa sfoun dfo r the Allier. Starkel (1983) reports that the main phases of erosion in the Vistula and other European rivers were in late glacial and early interglacial times in as well meandering as braidedsystems .Thi sinsigh ti sno tshare db yGibbar d (1988)wh o statestha tincisio noccur swhe nrive rru nof fi shighl yseasona l but when limited supplies of detritus are available, i.e.,

157 predominantly undercol dclimates . Aver yplausibl egenera lstatement ,whic hw esupport ,i smad e by Green and McGregor (1987) who state that different river systemsreac tdifferentl yt osimila renvironmenta lchanges .The y givetw omai nreason sfo rsuc hdifferences :th esiz eo fth erive r and the position of itsbasi n in relation to the geographical patterno fth eenvironmenta lvariable stha tgover nth ebehaviou r of fluvialprocesses .Accordin gt othe mth edifference sbetwee n basins of similar order arecause d by the facttha t the amount ofenvironmenta l changeneede d tobrin ga fluvial systemt oan d across a critical geomorphological threshold for terrace formation isno tth esam e inal lenvironments .

Byapplyin gth econcep to fdynami cequilibriu m alarg escal e model LIMTER was made simulating Quaternary Allier dynamics. Starkel (1983),Gree n& McGrego r (1987)an dBul l (1990)envisag e at each phase a leading factor which pushes the whole system towardsdow ncuttin go raggradatio nleadin gt oa ne wequilibriu m between mean discharge and mean sediment load. A similar behaviour isfoun dfo rLIMTER .A chang ei n100 0year sdischarge , 1000year s sediment load oruplif t ratewil l usually triggera system response oferosio n ordeposition . LIMTERsimulation ssho wtha tterrac eformatio ntake splac ei n a model in dynamic equilibrium thus confirming the general validity ofth eSchumm' sconcep to fdynami cequilibrium . LIMTER also demonstrates that both climate and tectonism play a significant role interrac e formation. But themai n conclusion istha t it is well possible to model such a complex and large scale system.

Methodology The following methodologies were applied inthi s thesis:1 ) bulk geochemical characterization of terrace sediments, 2) sediment flux reconstruction,

158 1 Bulksan d geochemistry This research tool facilitates the discrimination and quantification ofth eimpac to fdifferen t factorsan dprocesse s on terrace sand composition. The effects of provenance, weathering, grain size and fluvial transport could be satisfactory discerned and statistically modelled. Within the Allier and Dore sands,th e role ofgrai n size distribution was very limiteddu et oth elarg eamount so frock-fragment si nbot h sediment types.I nsand swit hmainl yminera lgrain s like inTh e Netherlands (Moura& Kroonenberg ,1990) ,grai nsiz edistributio n playsa mor eimportan trol ei nbul kgeochemica lvariability .Th e effects of weathering are very dominant in the Allier terrace sands. Itturne d outtha tmos to fth eorigina l sand composition of terraces older than Weichselian was altered by weathering processes. Although it is possible to simulate this parent material controlled weathering, it remains impossible to reconstructthei rorigina lsedimen tcompositio na characteristi c necessary fora paleoenvironmenta l reconstruction.

2 Sediment flux reconstruction Byextensivel ymeasurin gth echange si nsan dbul kgeochemistr y ata majo rconfluence ,pas tsedimen tmixin gca nb echaracterize d and quantified. This quantificationca n be used to reconstruct thepas t relative sediment fluxeswithi nth estudie d system.T o permit a reliable reconstruction of the past sediment mixing, spatialvariabilit y withinth eterrac e sandswa sstudie d first. The sediment flux reconstruction could only be done with Weichselian and younger sediments because the effects of weathering processes are limited in these sediments. The good matcho fth ereconstructe d fluxesan dth eknow npas tenvironmen t indicates that this methodology is a promising new tool in Quaternary research.

3 Finite statemodellin g Althoughbul kgeochemica lstudie so fth eAllie rterrac esand s yielded much new information of the large scale and long term Allier dynamics there still remains a considerable shortage in

159 largescal equantitativ edata .I norde rt obridg eth ega pbetwee n the limited quantitative and abundant descriptive knowledge on Quaternaryterrac eformation ,finit estat emodellin gwa sapplie d toconstruc ta 3- Dmode lsimulatin grive rterrac eformation .Th e developedmode l (LIMTER)ha stherefor eonl yconceptua lvalidit y becauseman yuse dmode lrelationship san d functionsar ederive d generalities from theoretical concepts and empirical measurements. The aim to establish a quantitative large scale modelwa stherefor e notcompletel ymet . Amai ndisadvantag eo fth euse dmodellin gmethodolog y istha t the model can never be validated because it is impossible to measureth e longter mdynamic s of sucha larg e scale system.

We saw that there exists agenera l agreement on the concept of dynamic equilibrium (Schumm, 1977). This concept is typical for Quaternary geology and geomorphology because it isdynami c and applicable on a long time span.Withi n such a concept the qualitative aspects of changes dominate. A fluvial system is described interm so fbraide dversu smeanderin g and aggradation versus incision divided by thresholds, while the real fluvial systemha sbot hmeanderin gan dbraide dcharacteristics ,an dwhil e deposition and incisionusuall y takecontemporaneousl y place. Stratigraphic models (Bridge & Leeder, 1979; Tetzlaff & Harbaugh,1989 )simulatin g fluvialsedimentatio noriginat e from thefiel do fsedimentology .Thes emodel sar emor efocusse do nth e variability within sedimentary bodies and their hydraulic parameters. There are a few essential differences between the geomorphological and the sedimentological approach. The geomorphologist is mainly interested in where, when, how and underwhic hconditions ,sedimentatio nan derosio ntak eplace .Th e sedimentologisthowever ,i smainl y interested inth edepositio n processitsel fan dles si nwhe nan dwhere .Tha ti swh yth emodel s of Bridge & Leeder (1979) and Tetzlaff &Harbaug h (1989) have only limited tectonic or climate related inputs or processes. Theirmode l relationsar efocusse d on (sub)processeswhic hhav e a direct effect on the sediment characteristics like discharge determininggrai nsiz edistribution .Bu tth emor egenera laspect s

160 of climatic and tectonic settings are often neglected in such models. These limitations find their origin in the scale difference between stratigraphic models and geomorphological models. The existing fluvial deposition models result in detailed stratigraphical output causing an overestimation of general model reliability. The geomorphological models like LIMTERresul ti nmor egenera lan dgloba loutput sgivin ga bette r visualisation of the complex real world system. It are the geomorphological modelswit hthei rmor eholisti c approachwhic h cangiv ea bette rrepresentatio no fth ecomple xoveral lprocesse s involved inshapin g fluvialdeposits .Althoug h there islimite d overlapi nth einterest so fgeomorphologist san dsedimentologis t itwoul d enrich both disciplines when theirmodellin g attempts would bemor e related.

Regional conclusions Theapplicatio no fne wcombination so fresearc hmethodologie s in the Allier basin resulted in some new regional insights. Becauseth eAllie rsediment sno whav ebee nradiometricall ydate d with1A Can dTh/ Umethod sa considerabl erevisio no fth eexistin g Allier terrace chronology (Fig. 5.1) is necessary. This new chronology hassom eresemblanc ewit hLarue' s (1979)chronology , whostudie da muc hlarge rarea .Pastre' s (1987)chronolog ygive s systematically much older ages for the terrace sediments, indicating that his methodology of linking the mineralogy of dated volcanics with sand mineralogy is not as accurate as he assumes. The new chronology shows a large time gap between the deposition of the V and W terrace sediments. The LIMTER simulations clearly indicate that this must have a tectonic origin. Itseem stha tth euplif to fth eRanda nare aan dprobabl y the Limagnegrabe na sa whole ,cam et oa nalmos tstandstil l for a fewhundre d thousandso fyear sdurin g theMiddl ePleistocene . LIMTER simulations also suggest differences in uplift rates of both sideso fth eAllie rvalley .

161 ACE in Larue (19791 Pastr e(1986 ) thisThesi s

Kypar^ BP |Fx I10-15m) |Fx (10-20ml 100 . |Fx(0-Sm) |Fwb(25m)

200 . I lFwd(10-15m) |Fw(25-35m> 300 . •FwaCSm) * |Fwc(25m) 400 .

|Fwb(25-30m) 500 .

|Fwa(30-35m) 600 . Fv(60-70m)

700 _

800 . |Fval65m)

900 . |Fvb(60-70m]

1000 _ |Fva(60-70m) 1100 .

Figure5. 1 The reconstructed terracechronolog y Our paleo-environmental reconstruction shows that sediment fluxdynamic si nth eAllie rbasi ndurin gth eLat eQuaternar yan d most probably the whole Pleistocene is strongly related to glacier dynamics on the Cantal and Mont Dore,an d thus climate controlled. An indirect result of the strong climatical relationship of thepas tAllie r dynamics isth edevelopmen t of Holocenelak ebasin s (Marais)i nGrand eLimagne .Thes elake sar e theresul to fa stron g riseo fth eAllie rriverbe d levelcause d by enormous sediment supply duet oglacie rmelting .

Further Research Simulations of future scenarios with LIMTER suggest more incisionan d erosion ofth ecurren tY an dX terrace sdurin gth e next fewthousand s years. It is important to keep inmin d that suchsimulation sgiv eonl yver ygenera lprediction si ntim estep s of 1000 years. Because the future climate scenarios are still

162 underdiscussio nan dsinc eLIMTE Ri sstrongl yclimat econtrolle d and does not include human factors,i t isno t realistic to use LIMTER to simulate future scenarios. Only a more refined and improved version will be suitable for such predictive simulations.

Thebul kgeochemica lmethodolog yshoul db emor eofte nuse da s a quantitative large scale tool to gain insights in long term fluvialsyste mdynamics .Th etechniqu eo freconstructin gsedimen t fluxes within a fluvial system can be more refined when it is applied onothe r fluvial systems.A wa yt omak ea mor e reliable pastsedimen tflu xreconstructio n ist oappl yth esam etechniqu e on several confluences within the same basin. In case of the Alliera stud ya tth eSioule/Allie ran dLoire/Allie rconfluence s combinedwit hth eDore/Allie rstud ywoul d certainly result ina more accurate basin reconstruction. It is possible to combine such confluence studies with a sediment budget study in the delta.Th e reconstructed relative sediment fluxeswoul d allow amor e quantitative reconstruction ofth e sediment sourceareas . Moreresearc hi sneede dt ose ewhethe rfinit estat emodellin g can indeed contribute to better longterm modelling in geomorphology. Iti stherefor epropose d toappl ythi stechniqu e to other macro-scale objects in geomorphology. Another rather successfulattemp ti salread ymad eb ymodellin gcoasta lterrace s asfunctio no fchangin gse aleve lan duplif t (Veldkamp,i nprep) . The 3-D terrace model may gain more value by adapting it to different fluvial systems. This has been done for the Meuse system,a fluvia lsyste mi nwhic hth eimpac to fglacier s isver y limitedcompare dwit hth eAllie rsyste m (Veldkamp& va nde nBerg , subm). Another way to improve LIMTER is to combine finite state modellingwit hth emethodolog y of "fuzzyknowledge" .Th eresul t would be a model with less firm decision rules allowing more realistic systemstates . Although it is generally known that the current state of knowledge in geomorphology obstructs pure numerical modelling,

163 thisshoul dno thol du pth eeffort so ftryin g it.O fcours ene w methodologies have to be developed to measure large scale processeso nth eappropriat e scale.Suc heffort swil lhopefull y revealknowledg egaps ,indicatin gwher eadditiona ltheorizin gi s necessary. Finite state modelling will allow to test/simulate newlydevelope d concepts andtheories . It is a pity that as a result of increasing modelling activitiesth eattentio no fgeomorphologist sshifte dmor et oth e short term processes. The fact that many global theories have their roots in the 19th century indicates that theorizing on largescale si ngeomorpholog yha scom et oa nalmos tstan dstill . Betterbalance d research activitieswil lcontribut emor et oth e field asa whole ,an dwil l soonerlea dt oworkabl e solutionso f scaleproblem s ingeomorphology .

164 REFERENCES

Abbey, S., 1980, Studies in 'Standard samples' for use in the general analysis of silicate rocks and minerals. Part 6: 1979 edition of 'usable'values ,Geol . Surv. Canada Pap 80-14:30 pp Anderson,M.G. , 1988,Modellin g geomorphological systems.Joh nWile y & Sons Ltd, London. Anderson, M.G. & Sambles, K.M., 1988, A review of the bases of geomorphological modelling. In: Anderson, M.G., (Ed), Modelling geomorphological systems.Joh nWie y & Sons Ltd,London . Andres, W., 1989, The Central German upland, Catena supplement 15, p. 25- 44. : Autran,A .& Peterlongo,J.M. , 1980,L eMassi fCentra l In:Geologi c des Pays Europeens, France, Belgique, Luxembourg. Dunod (Paris): 3-133. Basu, A., 1976, Petrology of Holocene fluvial sands derived from plutonic sourcerocks :implication s topaleoclimati c interpretation.Jour .Sed . Petr. vol 46,no. 3 p.694-709 . Bathurst, J.C., 1988, Flow processes and data provision for channel flow models,in :Anderso nM.C . (Ed)Modellin ggeomorphologica lsystems .Joh n Wiley & Sons Ltd. Bear,F.E. , (Ed),1964 ,Chemistr yo fth e soil,2 n d ed.,Reinhol d publishing corporation,Ne w York Begin, Z.B. & Schumm S.A., 1984, Gradational thresholds and landform singularity: significance for Quaternary studies.Quaternar y research 21,267-274 . Berger,A.L. ,1978 ,Long-ter mvariation so fcalori c insolationresultin g from the Earth's Orbital elements.Quaternar y Research 9,139-167 . BessonJ.C .& Ly ,M.H. ,Cantagrel ,J.M. , DeGoe rD eHerv eA. ,e tVincen t P.M., 1977, Une coulee de ponces post-Villafranchienne sur le versant oriental duMont-Dore . C.R. Acad. Sc. Paris 284 Ser. D.P. 1875. Best,J.L. , 1986,Th emorpholog y ofrive rchanne lconfluences .Progres s Phys. Geogr., 10(2), 157-174. Best, J.L., 1988, Sediment transport and bed morphology at river channel confluences. Sedimentology, 35,p 481-498 . Bloom, A.L., 1978, Geomorphology, a systematic analysis of Late Cenozoic landforms, Prentice-Hall, Inc NewYork . Boll, J., Thewissen, T.J.W., Meijer, E.L., and Kroonenberg, S.B., 1988, A simulation of the development of river terraces, Zeitschrift fur Geomorphologie, 32,31-4 5 Bout, P., 1963, Observations sur la basse-terrasse de l'Allier a Pont du Chateau.Acte s du quatre-vingt-huitieme congres national des societes savantes. Clermont Ferrand. Bradley, R.S., 1985, Quaternary paleoclimatology, methods of paleoclimatic reconstruction.Alle n& Unwin,Boston . Bridge, J.S. & Leeder, M.R., 1979, A simulation model of alluvial stratigraphy. Sedimentology, 26:617-644 . Brunnacker,K .& Boenigk ,W. ,1983 ,Th eRhin evalle ybetwee nth eNeuwie dbasi n and the Lower Rhenish Embayment. In: K. Fuchs et al., (Eds). Plateau Uplift, Springer Verlag, Berlin: 62-72. Buch M.W., 1987, Spatpleistozane und holozane fluviale Geomorphodynamik im Donautal ostlich von Regenburg - ein Sonderfall unter den mitteleuropaischenFlusssystemen? ,Z .Geomorphologi eN.F .Suppl.-bd66 , pp.95-111, Berlin. Bull,W.B. ,1990 ,Stream-terrac e genesis: implications for soildevelopment . Geomorphology, 3,p.351-367 .

165 Bullock, P., Fedoroff,N ., Jongerlus ,A. , Stoops,G. ,an d Tursina, T., 1985, Handbook for soil thin section description. Waine research publications,England , 150pp . Burgess,T.M. ,Webster ,R .,& McBratney ,A.B. ,1981 ,Optima linterpolatio nan d isarithmicmappin go fsoi lproperties .I Vsamplin g strategy.Journa lo f Soil Science, 32,643-659 . Cameron, K.L. & Blatt H., 1971, Durabilities of sand size schist and "volcanic" rock fragments during fluvial transport, Elk Creek, Black hills, South Dakota,Jour , of Sed. Petr.Vol .41 ,no . 2,p . 565-576. Camus,G. , De Goer,A. , Kieffer, G., Mergoil,J. , Vincent P.-M.,1983 ,Par e naturel regional desvolcan s d'Auvergne,Volcanologi e del aChain ede s Puys, A.R.P.E.G.E. imprinterieModern e 15000Auriliac . Chambaudet, A., & Couthures, J., 1981, Datations pas traces de fission de gisements Plio-Pleistocene de la peripheric des Monts-Dore (Massif Central, France), C.R. Acad. Sc. Paris,t 293,se r II,pp67-71 . Chaput, E., 1917,Recherche s sur les terrasses alluviales de la Loire et de ses principaux affluents. TheseAnnale s de l'Universlte de Lyon nouv. serie I,Fac e 41. (from J.F. Pastre, 1987) Church,M . &Kellerhals ,R. , 1978,O nth e Statistics of grain size variation along a gravel river. Can.J . Earth Sci.,15 ,p .1151-1160 . Chorley,R.J. , Schumm, S.A., and Sugden,D.F. , 1984,Geomorphology , Methuen and CoLt d New York. Clozier, L., Fleury, R., Giot D., 1980, in: BRGM, Carte geologique de la France a 1/50 000,Maringue sXXVI-3 0Not e explicative,Orleans ,Franc e Colman,S.M . &Dethier ,D.P. , 1986,Rate so fchemica lweatherin g ofrock san d minerals. Academic Press Inc. Daugas, J.-P. & Tixier, L., 1978, Stratigraphie du Quaternaire recent et niveau archeologique protohistorique a Coudes (Puy-de-D6me). L'Antropologie, T. 82no . 3pp.439-450 . Davies, D.K., Vessell, R.K., Miles, R.C., Foley M.G., Bonis S.B., 1978, Fluvial transport and downstream sediment modifications in an active volcanic region. In:Mial l A.D., (Ed)Fluvia l Sedimentology: Canadian Soc. of Petroleum petrologists,memoir e 5,p.61-83 . Dawson, M.R. & Gardiner, V., 1987, River terraces, the general model and palaeohydrological and sedimentological interpretationo f the terraces of the Lower Severn, in: Gregory, K.J., Lewin, J. & Thornes, J.B. (Eds), Palaeohydrology inpractise ,Joh nWile y & Sons Ltd,London . Debard, E. & Pastre, J.-F., 1988, Un marqueur chronostratigraphique du Pleistocene moyen a la peripheric du Massif Central: la retombee a clinopyroxene vert du Sancy dans le site acheuleen d'Orgnac III (Bas- Vivarais, SE France), C.R. Acad. Sci. Paris,T.306 , Serie II,p.1515 - 1520. DeBeaulieu ,J.-L . &Reille ,M. , 1984,A lon gUppe r Pleistocenepolle nrecor d from Les Echets,nea r Lyon,France . Boreas,Vol . 13,pp .111-132 . De Beaulieu, J.-L., Pons, A., et Reille, M., 1982, Recherches pollenanalytiques surl'histoir e del avegetatio nd el abordur enor dd u Massif duCanta l (Massif Central,France) , Pollen etSpores ,vo lXXIV , no 2p 252-300 . Dillon, W.R. & Goldstein, M., 1984, Multivariate Analysis, methods and applications. John Wiley & Sons, NewYork . Drees,L.R .& Wilding ,L.P. ,1978 ,Elementa ldistributio ni nth eligh tminera l isolate of soil separates. Soil. Sci.Am . J., vol.42 . Dury, G.H., 1970,River s and river terraces,Macmillan ,London . Elzas,M.S. ,1978 ,Systeme ne nmodellen ,Landbouwkundi g tijdschrift 90-8a (in Dutch).

166 Etlicher,B .,Janssen ,C.R. ,Juvigne ,E. ,Leeuwen ,J.F.N. ,1987 ,L ehau tFore z (MassifCentral ,France )apre sl epleniglaclalr eWtirmien :envlronnemen t et tephra duvolca n de la Nugere. Bulletin de l'associatlon francaise pour l'etuded e Quaternaire, 1987-4229-239 . Fairbridge, R.W., 1968, Terraces, fluvial-environment controls, in: Fairbridge, R.W., (Ed),Encyclopedi a of geomorphology: Reinhold Book Corp. New York, p1124-1138 . F.A.O.,1974 Soi lma po fth eworld ,1:5000,000 ,Vo lI Legen dFA OUNESC OParis , 59pp . Feijtel,T.C. ,Jongmans ,A.G. ,Miedema ,R. ,Va nBreemen ,N. ,1988 ,Genesi so f twoplanosol s in theMassi f Central France.Geoderma , 43:249-269 . Feijtel,T.C. ,Jongmans ,A.G. ,Va nDoesburg ,J. , 1989,Identificatio no fcla y coatings ina nolde rQuaternar y terrace,o fth eAllier ,Limagne ,France . Soil Sci. Soc. Am. J. 53:876-882 . Feijtel, T.C. & Meijer, E.L., 1990, Simulation of soil forming processes. Syllabus of course with spreadsheets and exercises. Internal report Department of Soil Science and Geology, Agricultural University Wageningen, TheNetherlands . Findlay, W. & Watt, D.A., 1981, PASCAL, an introduction to methodical programming. Pitman publishing limited, London, second edition. Franzinelli, E.,& Potter, P.E., 1982,Petrology , chemistry, and texture of modern river sands,Amazo n river system, Journal of Geology, vol.91, p.23-39. Gibbard,P.L. ,1988 ,Th ehistor yo fth egrea tnorthwes tEuropea nriver sdurin g thepas tthre emillio nyears .Phil .Trans .R . Soc.Lond .B318 ,559-602 . Giot,D. ,Clozier ,L. ,Fleury ,R .,1978 ,Manifestion stectonique squaternaire s en Limagne d'Allier. Bull, du BRGM, section I,n o 2,p . 150-155. Goldich, S.S., 1938,A study in rock-weathering, Jour, of Geology, 46 .17-58. Green,C.P .& McGregor ,D.F.M. ,1986 ,Th eutilit yo f intercomponentratio si n the interpretation of stone count data, In: Bridgeland, D.R., (Ed), Clastlithologica lanalysis ,Quaternar yResearc hAssociation ,Technica l guide no. 3,p .83-93 . Green, C.P. &McGregor ,D.F.M. ,1987 ,Rive r terraces:a stratigraphi c record of environmental change. In: Gardiner, V., (Ed),1986 , International geomorphology 1986 Part I. John Wiley & Sons Ltd. Gregory K.J., (Ed) 1983, Background to palaeohydrology. John Wiley & Sons Ltd, London. Gregory, K.J., Lewin, J. & Thornes, J.B. (Eds), 1987, Palaeohydrology in practise, John Wiley & Sons Ltd, London. Hay,R.L. ,1960 ,Rat eo fcla yformatio nan dminera lalteratio ni na 400 0year - oldvolcani c ashsoi lo nS tVincent ,B.W.I .,Am .Jour ,o fScience ,vol . 258, p.354-368. Hays, J.D., Imbrie, J., Shackleton, N.J., 1976, Variations in the earth's orbit, pacemaker of the iceages. Science 194,p 1121-1132 Heller,F .& Wang ,J. , 1991,Magnetis mo fQuaternar y sediments:Los s inChina . Special Proceedings, Review reports, for symposia of the XIII International congress INQUA Beijing, 1991,p 88-98 Howes, S. & Anderson,M.G. , 1988,Compute r simulation ingeomorphology . In: Anderson,M.G. , (Ed),Modellin g geomorphological systems.Joh nWile y & Sons Ltd, London. Hugget, R.J., 1985, Earth Surface systems, Springer-Verlag, Berlin, Heidelberg. 262p . Huntington, E., 1907, Some characteristics of the glacial period in nonglaciated regions. Geol. Soc.Amer . Bull.Vo l 18,p p 351-388.

167 Ichim, I.,& RadoaneM. , 1990, Channel sedimentvariabilit y along ariver :a case study of the Siret River (Romania). Earth Surface Processes and Landforms,vol . 15, 211-225. Imbrie,J . & Palmer-Imbrie,K. , 1979,Ic e ages, solving themystery . Enslow Publishers Hill Side,Ne w Jersey. Jeambrun, M., Dadet, P., Clozier., L., Fleury, R., et al., 1980, Carte Geologique de laFranc ea 1:50.00 0Maringue sXXVI-30 ,Not e explicative BRGM, Orleans 54p. Jongmans,A.G. ,Va nDoesburg ,J. , andVa nBreemen ,N. , 1990,Micromorpholog y andmineralog yo fweatherin gan dneoformatio nphenomen a ina Quaternar y terrace sequence of the Allier, Limagne,France . Chemical Geology vol 84, no 1/4 p. 83-85. Jongmans, A.G., Feijtel, T.C., Miedema, R., Va n Breemen, N.. Veldkamp , A., 1991. Soil formation in a Quaternary terrace sequence of the Allier, Limagne, France. Macro and Micro morphology, particle size distribution, chemistry. Geoderma 49.p.215-239 . Journel, A.G. & Huijbregts, Chr. J., 1978, Mining Geostatistics. Academic press. New York,N.Y . Jung J., Symposiu m 1971, Geologie, geomorphologie et structure profonde du Massif Central francais; Plein Air Service ed., Clermont-Ferrand, 1 vol., 610p. , 154fig . Juvigne, E.& Gewelt ,M . 1987,L anars e d'ampoixcomm e tephrostratotype dans la Chaine des Puys Meridionale (France). Bulletin de l'Association francaise pour l'etude de Quaternaire, 1987-1, p.37-49. Juvigne, E., Milcamps, V., Delibrias, G. & Evin, J., 1988,Age s de traits polliniques et chronozonation du Tardiglaciare final et de l'Holocene dans leMassi f Central (France),Mededelinge n rijksgeologisch e dienst vol. 42p . 33-50. Kleffer, G., 1971, Apercu, sur la morphologie des regions volcaniques du massif central. Symposium J. Jung 1971 dlermont-Ferrand,p479-510 . Knighton,A.D. ,1982 ,Longitudina l changes inth esiz ean dshap eo f streambed material:evidenc eo fvariabl e transportconditions .Caten avol .9 ,25 - 34. Kominz, M.A..Heath, G.R., Ku,T.L. , Pisias,N.G. , 1979,Brunhe s time scales and the interpretation of climatic change.Eart h Planet Sci.Lett .45 , p 394. Komura, S., 1973, River-bed variations at confluences. Proceedings International symposium on river mechanics. International Association forhydrauli c research. Kroonenberg,S.B. ,va nde nBer gva nSaparoea ,R.M. ,Jonke rA.T.J. , 1987,Lat e glacialan dHolocen edevelopmen to fsemi-close dbasin s (Thawlakes? )i n theLimagn eRif tValley ,Frenc hCentra lMassif ,Geol .Mijnbou w 66:297- 311. Kroonenberg, S.B., Moura,M.L. ,Jonker ,A.T.J. , 1988,Geochemistr y of the sands of the Allier river terraces, France. Geologie en Mijnbouw 67: 75-89. Kroonenberg,S.B. ,va nde rPlicht ,J. ,Vlaanderen ,B .& Wassink ,W. ,1989 ,Th - Udisequilibriu mdatin go ftravertine-impregnate d PleistoceneAlluvia l terraces in the Limagne Rift Valley, France. Busche, D., (Ed), Abstracts of paper and poster of second international conference on geomorphology., Geookoplu s 1,p .162 . Kroonenberg, S.B., 1990, Geochemistry of Quaternary fluvial sands from different tectonic regimes.Volum e 84,n o 1/4 p 88-91. Kroonenberg,S.B. ,Hoorn ,M.C. ,Moura .M.L. ,& Veldkamp ,A. ,1990 ,Variabilit y in bulk geochemistry of fluvial terrace sands, consequences for the study of weathering chronosequences.'Pedologi e XL-1p 19-31.

168 Lambert. J., C. Orcel., J., Daugas & J. P. Raynal, 1980,Premier s resultats dendrochronologiques pour le Massif Central francais obtenus sur un bois fossiled e labass e terrasse de l'Allier aJoz e (Puy-de-D6me). C. R. Acad. Sc. Paris,t.29 0seri e Dp263-266 . Larue, J.P. 1977, etude du materiel differentes nappes alluviales de l'Allier., Extraitd el arevu ed'Auvergn e tome91-numer o3 ,Institu td e Geographie LIV, Clermont-Reproduction,Aubiere . LarueJ.P. ,1979 ,Le snappe s alluviales del aLoir e et de ses affluents dans le Massif Central et dans le Sud de bassin Parisien: etude geomorphologique. These geographie, Clermont II,multigraphe e 543p. , plus annezes: 30 cartes,12 6planches . Leger, M., 1983, Signification dynamique et climatique des formations et terrasses fluviatiles quaternaires, Colloque de 1'Associations francaise pour 1'etude de Quaternaire. Paris,Deuxiem e serie 21 annee no 17-18-19 /l /2/3 . Lenselink G., Kroonenberg, S.B. & Loison, G. 1990, Periglacial to Holocene paleo-environments in the Artiere basin in the Western Limagne rift valley,Massi f Central,France .Quaternair e 2,p.131-148 . Leopold, L.B., Wolman, M.G. & Miller, J.P., 1964, Fluvial processes in geomorphology.W.H . Freeman & Company, San Francisco. Levine, E.R. & Ciolkosz, J., 1986, A computer simulation model for soil genesis applications. Soil Sci. Soc.Am . J. 50:661-667. Lowe,D.J. , 1986,Control so nth erate so fweatherin gan dcla yminera lgenesi s in airfall tephras: a review and New Zealand case study. In: Colman S.M., Dethier,D.P. , (Eds), Rates of chemical weathering of rocks and mineral, Academic Press,In cNe w York.P.265-330. Lowe, J.J. & Walker, M.J.C., 1984,Reconstructin g Quaternary environments. Longman Group,Ne wYork . Ly, M. H., 1982, Le plateau de Perrier et la Limagne du Sud: Etudes volcanologiques et chronologiques des produits Montdoriens (Massif Central francais). These III™"*cycle ,Clermon t Fd. Maizels,J.K. ,1986 ,Modellin go fpalaeohydrologi cchang edurin gdeglaciation , geographie etphysiqu e Quaternaire,vo l XL no3,p.263-277 . McGregor,D.F . &Green ,C.P. , 1978,Gravel so f theRive rThame s asa guid e to Pleistocene catchment changes. Boreas,Vo l 7.,p p 197-203. Mitchell, D.J., & Gerrard, A.J. 1987,Morphologica l responses and sediment patterns. in: Gregory, K.J., Lewin, J., and Thornes, J.B., Palaeohydrology in Practice. John Wiley & Sons Ltd,London . Mosley,M . P. 1976,A n experimental studyo fchanne l confluences,Journa l of Geology vol 84,p.535-562 . Moura, M.L. & Kroonenberg, S.B. 1990, Geochemistry of Quaternary fluvial sediment in the southeastern Netherlands, Geologie en Mijnbouw, 69: 359-373. Norusis,M.J. , 1986,Advance d statistics SPSS/PC+ for the IBM/XT/AT SPSSinc, Chicago, USA. Odom, I.E., Doe,T.W. , Dott,R.H. , 1976,Natur e of feldspar-grain relations in some quartz-rich sandstones. Jour. Sed. Petr. vol.46 ,no. 4 p.862- 870 Pastre, J.F., 1986, Alteration et paleoalteration des mineraux lourds des alluvions Pliocenes et Pleistocenes du bassin de l'Allier (Massief Central, France)Assoc . Rf. Etude Quat. Bull. 3/4:257- 269. Pastre,J.F. , 1987,Le sformation sPlio-Quaternaire sd ubassi nd el'Allie r et le volcanisme regional (Massif Central, France). These doctorat de l'Universite Paris IV. 706p .

169 Felletier,H. , 1971, Sur lesminerau x lourds transparentes des alluvions anclennes et recentes de laLimagne , d'Auvergne. These Fac. Sci. Clermont-Ferrand, 79p . Pettijohn, F.J., Potter, P.E., Siever,R. , 1987,San d and sandstone, second edition. Springer Verlag,Ne w York, Berlin,Heidelberg . Pierre, 6., 1989,Le s alterites fossilisees par des coulees de lave: valeur paleoclimatique et implications geomorphologiques; l'exemple de l'Auvergne, de l'Aubrac et du Velay. These a l'universite de Paris I Pantheon Sorbonne. Plant, J.A., Hale, M., Ridgeway, J., 1988, Developments in regional geochemistry forminera l exploration.Trans .Inst .Min .Hetall . (sect. B: Appl. Earth Sci) 97p B116-B130. Porta, J. & Herrero J., 1990,Micromorpholog y and genesis of soils enriched withgypsum . In:L.A. , Douglas (Ed),Soi lMicromorphology : abasi c and appliedscience :proceeding so fth eVIII thInternationa lworkin gmeetin g of soilmicromorphology , SanAntonio ,Texa sJul y 1988,Elsevie r p.321- 339. PotterP.E .,1978 ,Petrolog yan dchemistr yo fmoder nbi grive rsands .,Journa l of Geology,vol .8 6p .423-449 . Raynal, J.-P., Paquereau, M.-M. & Daugas, J.-P., 1981, Stratigraphie - La formationfluviatil ed eSainte-Martin e (Pont-du-Chateau,Puy-de-Ddme) , nouvelle sequenced ePleistocen eMoye nd'Auvergne .C.R .Acad .Sc .Pari s t. 292,Seri e II -p .841-846 . Raynal, J.-P., 1984,Chronologi e des basses terrasses de l'Allier en grande Limagne (Puy-de-D6me,France) .Bulleti nd el'Associatio n francaisepou r l'etude duQuaternaire , 1.2.3., p.79-84. Raynal,J.P. ,Daugas ,J.P. ,Paquereau ,M.M ., Guadelli ,J.L. ,Marchianti ,D. , Miallier,D. , Fain,J. , Sanzelle,S. , 1984,L emaa r de Saint-Hippolyte (Puy-de-D6me, France), Datation pas thermoluminescence, flores et faunes fossiles,presenc e humaine, climatochronologie et dynamique du systeme paleo-lacustre. Rev. Sc.d'Auvergne , Vol.50 , Raynal, J.P., Paquereau, M.M., Daugas, J.P., Miallier, D., Fain, J. & Sanzelle, S., 1985, Contribution a la datation du volcanisme Quaternaire du Massif Central fran;ais par thermoluminescence des inclusions de quartz et comparasion avec d'autres approches: implications chronostratigraphiques etpaleoenvironementales .Bulleti n de l'Association frangaise pour l'etude de Quaternaire 1985/4 p.183- 207. Reille,M . & De Beaulieu,J.-L. , 1988,L a find e l'Eemiene tle s interstades duPrewur mmi spou r lapremier e foise nevidenc e dans leMassi f Central francais pas l'analyse pollinique. C. R. Acad. Sci. Paris, t. 306, serie IIp .1205-1210 . Riezebos, P.A., 1971, A contribution to the sedimentary petrological descriptiono fth eMaa sdeposit s insouther nLimbur g (TheNetherlands) , Geologie enMijnbouw ,vol . 50 (3),p.506-514 . Roy,A.G. ,& Bergeron,N. , 1990,Flo w and particle paths at anatura l river confluence with coarse bed material.Geomorpholog y 3,p . 99-112. Rudel, A., 1953, La faune Quaternaire des terrasses de l'Allier a Pont du Chateau. Rev. des Sc.Nat . d'Auv., pp 43-47. Rudel, A., 1963, Les mineraux lourds des terrasses Quaternaires de Limagne d'Auvergne et les eruptions montdoriennes Soc. Geol. Fr. Bull. (7):468-469. Ruxton, B.P., 1968,Rate s ofweatherin g of Quaternary volcanic ash innort h east Papua. Trans. IX Int.Congr . Soil Sci,Adelaide , 4:367-376. Schumm, S.A., 1977,Th e fluvial system,Wile y New York.

170 Schumm,S.A . &Lichty ,R.W. ,1965 ,Time ,spac ean dcasualit yi ngeomorphology . American Journal of Science,vo l 263,p 110-119. Schumm, S.A., Mosley, M.P. & Weaver, W.E., 1987, Experimental fluvial geomorphology. John Wiley,Ne w York. 413p . Staritsky,1.6. ,1989 ,Manua lfo rth egeostatistica lprogram sSPATANAL ,CROS S and MAPIT. Agricultural University Wageningen, Department of Soil Science and Geology, internal report. Starkel, L., 1983, The reflection of hydrologic changes in the fluvial environment of the temperate zone during the last 15,000 years. In: Gregory,K.J. , (Ed),Backgroun d toPalaeohydrology , JohnWile y& Sons, London. Stein,A. ,Bouma ,J. , Mulders,M.A., ,& Wetering sM.H.W ., 1989 ,Usin g spatial variability studies to estimate physical land qualities of a level river terrace. Soil Technology, 2,385-402 . Stevens, P.R. & Walker, T.W., 1970, The chronosequence concept and soil formation. The quarterly review ofbiology ,vo l 45.p .333-350 . Strakhov,N.M. , 1969,Principle s of lithogenesis,vo l 2.Consultant s Bureau, New York and Oliver & Boyd, Edinburgh. Tetzlaff, D.M. & Harbaugh, J.W., 1989, Simulating clastic sedimentation. Computer methods in the geosciences.Va n Nostrand Reinhold. Texier, J-P., & Raynal, J-P., 1984, Les dep6ts et terrasses fluviatiles d'aquitaine et du bassin de l'Allier. Bulletin de l'Association francaise pour l'etude duQuaternaire , 1.2.3., p. 67-71. Thornes, J.B., 1987, Models for palaeohydrology in practice, in: Gregory, K.J., Lewin, J. & Thornes, J.B. (Eds), Palaeohydrology in practise, JohnWile y & Sons Ltd, London. Tourenq, J., 1986, Etude sedimentologique des alluvions de la Loire et de l'Allier des roches des bassins versants Doc. BRGM 108:108 pp. Tuffery, C, 1986, Les formations alluviales recentes de l'Allier dans la region deVichy : chronostratigraphie etevolutio nmorphologique . these Institut de Geographie,Universit e de Clermont-FdII . Van den Berg, M.W., 1989, Toelichting op kaartblad, 59,60,61,62, Geomorfologische kaart van Nederland 1:50.000, Staring Centrum, Wageningen,Rijksgeologisch e Dienst,Haarlem . 33p . Van Dorsser,H.J. , 1969,Etud e geomorphologique dansun eparti e de lavalle e de l'Allier dans la Grande Limagne Publ. ITC, Delft, Pays Bas, Ser.B 50:66 pp. Van Straaten, L.M.J.U., 1946, Grindonderzoek in Zuid-Limburg. Meded. Geol. Sticht. Serie C-VI-2. Veldkamp,A .& S.E.J.W .Vermeulen , 1989,Rive r terrace formation,modelling , and 3-D graphical simulation. Earth Surface Processes and Landforms, vol 14,641-654 . Veldkamp,A .& Kroonenberg ,S.B. , 1989,A compariso no fth esan d geochemistry of the Allier and Dore terraces, Limagne Rift Valley, France, in: Busche D., (Ed),Abstract s ofpaper s andposters ,secon d international conference on geomorpholoy, Frankfurt, Geooko plus 1,vo l 1,p 304 . Veldkamp A., 1990,Predictio n of bulk chemical composition of fluvial sands from grain size data, Allier and Dore terrace sands Limagne Rift Valley, France. Chemical Geology vol 84n o 1/4 p208-209 . Veldkamp. A. & Jongmans, A.G., 1990, Trachytic pumice weathering, Massif Central,France :geochemistr yan dmicromorphology .Chemica lGeolog yvo l 84, no 1/4,p .145-147 . Veldkamp,A .& Feijte lT.C .,1990 ,Regiona lweatherin gwit hbul kgeochemistry : a case study for the Allier terrace sands, Limagne, France. Chemical Geology vol 84,nol/ 4p 142-144.

171 Veldkamp, A. & Kroonenberg, S.B., 1991, The effects of a perlglacial environment on the fluvial dynamics of the Allier during the Late Veichselian, Limagne, France. Abstract, Symposium perlglacial environments in relation to climatic change Maastricht/Amsterdam 3-6 may. p.13-14. Veldkamp, A., 1991, Reconstructing past sediment fluxes within a fluvial system: the Allier basin during the Late Quaternary. Abstract XIII INQUA, Beijing, China,p 368. Veldkamp,E. ,Jongmans ,A.G. ,Feijtel ,T.C. ,Veldkamp ,A .& Va n Breemen,N ., 1990, Alkali basalt gravel weathering in Quaternary Allier river terraces,Limagne ,France .Soi lScienc e Soc. ofAm . Journal. 54:1043 - 1048. Veyret,Y. ,1980 ,Quelque scaractere sd'un emoyenn emontagn eenglacee ,exempl e des hautes terres cristallines et volcaniques du massif central francais. Rev. de Geomorphologie dyn. 29,p.49-65 . Veyret, Y., 1978, Mode16 et formation d'origine glaciare dans le Massif Central francais, problemes de distribution et de limites dans un milieu de moyenne montagne. These de doctorat d'etat, Universite de Paris I. 2vol ,78 3p . Wedepohl,K.H . (Executive Editor), 1970,Handboo ko fgeochemistry ,Springer - Verlag Berlin,Heidelberg ,Ne wYork . Wljck van, H., 1985, Zware mineralen onderzoek Limagne Frankrijk, Internal report,Departmen to fSoi lscienc ean dGeology ,Agricultura l University Wageningen. (In Dutch). Woillard ,G.M. , 1978,Grand e Pile peat bog: a continuous pollen record for the last 140,00years .Quaternar y research, 9. 1-21. Woillard,G.M . &Mook ,W.G. ,1982 ,Carbon-1 4date sa tGrand ePil e Correlation oflan dan d seachronologies .Repor t in:Science ,Vo l215 ,Jan .p .159 - 161. Yaalon, D.H., 1975,Conceptua l models inpedogenesis : Can soil-forming functions be solved? Geoderma 14,189-205 . Ziegler, B.P., 1976, Theory of modelling and simulation: An introductory expositiono fconcepts ,Departmen to fapplie dmathematics ,Th eWeizman n institute Tehovot, Israel,Wiley ,Ne wYork .

172 Curriculum vitae

AntonieVeldkam pwer dgebore no p2 2me i196 3t eNieuw ePekela . In 1981 behaalde hij het Atheneum-B diploma aan de scholengemeenschapS tMichie lt eGeleen .Hi jstudeerd evana f198 1 Bodemkunde en Bemestingsleer (N33)aa n de Landbouwuniversiteit waarhi ji n198 5he tkandidaatsexame naflegde .Va nmaar t198 5to t april198 6wa shi jwerkzaa mbi jhe tTPI Pprojec t inEmb u (Kenya) waar hij zijn praktijktijd en veldwerk voor twee hoofdvakken volbracht. Deze twee hoofdvakken, Geologie en Tropische Bodemkunde, werden vervolgens in Wageningen afgerond. De doctoraalstudiebeston dverde rno gui tee nhoofdva kinformatica . In September 1987verkree g hijhe t ingenieursdiplomame t lof. Vanaf januari 1988 tot januari 1991wa s hij als Onderzoeker inOpleidin g (010)werkzaa mbi jd estichtin gAardwetenschappelij k Onderzoek Nederland (AWON) en gestationeerd bij de vakgroep bodemkunde engeologi eva nd e Landbouwuniversiteit. APPENDIXI

XRF analysis X-ray fluorescence analysis (XRFS) is a standard analytical tool to obtain quantitative geochemical data. At the Department of Soil Science andGeolog y ametho d fordeter ­ mination of trace elements in glass beads of fused soil and sediment samples was developed (Kuijper & Meijer, 1987). A Sc tube is used to determine both macro and trace elements. In 1989 this method was replaced by another method wherby trace elements aremeasure d with aRhodiu m X-ray tube in combination with the measuring of the Compton scattered radiation (inte­ rnal ratio method). Themacr o elements are still measured with a Sctube . The X-ray fluorescence intrumentation include a Philips PW1410 wavelength-dispersive spectrometer, LiF200 and LiF220 analyzing crystal, scintillation and gas flow-proportional counters and an automatic sample changer. More detailed instrumental information of the used method is given by Moura & Feijtel (1989).

A major disadvantage of the current methodology is that it is impossible to calculate the Lower Limit of Detection (LLD) for the trace elements. It is possible to estimate the LLD based on the standard reference sample which are used to calibrate the system. As LLD's estimated in this way still tend to give an under estimation of detection limits a more practical way to estimate the detection limits was used (Thompson & Howarth, 1978; Reimann & Wurzer, 1987; Reimann, 1988). The practical detection limit instead of the rather meaningless theoretical detection limit was estimated from 50 duplo'so fth eLimagn eprojec t sand samples (Tab1. 1& 1.2). Table 1.1 Determination of practical detection limits for macro elements of theAllie r and Dore sands. (50 duplos)

Element Mean diff. Std diff Est. LD line fit

Si02 0.64 0.71 "1.00 poor Ti02 0.07 0.06 <0.01 good A1203 0.22 0.22 "0.01 average Fe203 0.02 0.02 <0.01 good MnO 0.00 0.00 >0.01 no relationship MgO 0.03 0.03 <0.01 poor CaO 0.01 0.01 "0.01 poor Na20 0.42 0.17 "0.01 average K20 0.13 0.03 <0.01 good P2°* 0.01 0.01 "0.01 poor BaO 0.01 0.01 "0.01 no relationship

not very reliable element difficult to estimate as only high concentrations occur indat a set.

Table 1.2Determinatio n of practical detection limits for trace elements of the Allier and Dore sands. (50 duplos)

Element Mean difference criterion(%) LD line fit

Ba* 639 20 3 >800 nvt Co* 29 7 29 >50 nvt Cu* <6 - - >6 nvt Cr 51 2 5 4 poor Ga* 14 2 11 >25 nvt La 37 11 35 20 poor Nb 8 2 35 6 poor Ni 46 5 13 8 average Pb 36 3 7 1 average Rb 136 24 18 3 good Sr 280 35 12 6 good V 117 38 39 31 good Zn 37 2 7 3 average Zr 194 17 9 3 good

= unreliable element criterion p= x 2x 100% x + x' References

Kuijper,A.J. ,Meijer ,E.L. , 1987,He tmete nva n sporenelemen- ten in boraatparaels met de Philips PW1404 rontgenspectrometer. Intern rapport, Vakgroep Bodemkunden en Geologie. Moura M.L., Feijtel T.C., 1989, Analysis of trace elements using x-ray fluorescence system: application of theRh-Compto n correction method. Internal report Department of Soil Science andGeology . Reimann, C., 1988, Reliability of geochemical analysis: recent experiences. In:MacDonald , D.R., &Mills ,K.A. , (Eds), Prospecting in areas of glaciated terrain 1988; The Canadian Institute ofMinin g andMetallurgy ,p .485-499 . Reimann, C., Wurzer F., 1987, Monitoring accuracy and precision, improvements by introducing robust and resistant statistics.Mikrochim .Act a 1986I I31-42 . Thompson, M.-, & Howarth, R.J. , 1978, A new approach to the estimation of analytical precision. Journal of geochemical Exploration, 9.p 23-30. Bulk geochemicalmeasurement s ofAllie r and Dore sands used inthi s thesis are listed below. For each sample the following data are listed: sample number, location name,ma p coordinates (From 1:25.000topographica l maps), terrace code,absolut e and relative altitudes (m), samplin g depth (m), Distanc e upstream from actual Allier/Dore confluence (Upstr) in km, Median of grain sizedistribution , Si02, Ti02, A1203, Fe203,MnO ,MgO ,CaO ,

Na20, K20, P205,BaO ,V , Cr,Co ,Ni ,Cu ,Zn ,Ga ,Rb ,Sr ,Zr ,Nb , Ba, La, Pb. O HH0^^1(rl^lnHoo**nolnoD*Hnco^oalf,^^^^o\^^(J\n^oco^^coo^f^coH(Jlf^O(Ncofl|NlN^^o^ou^'a'il^O'^,Nln

(NtN(NOJ(NH(NHHHHMr ( rH H r-1 H H .H r < i-H H H r-t H < < rA O H H H T-I H .

O co^fo^oJcoo^H^(^o^l^coHMo^H^MM*^^^MM^^^oa3H(^(NHlnol^^H(NHcocoo^cowol^cooHffl^fllfllOf^lOH

HHHrlHHNNINrtlNHHrlHrtrlNHIMNINHHHHHHrlHMHrlHHHOlHOOHOHOOOOHHHOHHHOrlOHOtNIM

O l^llNnol^N^HH^rJ^lnHln^a>N(^lH^^Ho^^^Jlno3^r^*fONHnolltlOlflo^(»^^f^^ocorl^>)^o^rtlfl^ln^^lfl0^o<5lH O^ lno^H^H^MHl\Nlnoo^ol^^gc^co*HlrtONn(Jl(N^(Tl^^o(Nno^lJco^^^o^^H^^Mlnoco^f^l(Tl«^^(^p^(^l(^lHH (IN(NfN(NHHrHHHOrHrH^HHCN]^H-Hr I O O -H >H -H r 1 H O O H O r OOOOrHOiHOOHHOOO-HOiNfM

C OOOOOOOOOOOOOOOOOOOOrHOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO a ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo n ln^o^ooHHlnHlnxoa)M^nf^)mo^H^^*^co^^co(J^^^oa^o^^^Mn^lXllno^O'f*W(N«3(3\(^lrl<'0^,lnlrtH^o O ocoNnoo1^Hnco^HtNf^lon^colTlM^^o^lOO^no^nlnooa^^c^^HH*H^oD^^oalm^ocoo*»ln^^^^lfl^^^O(^lo iHN^Hfvi(NojHnHi\nNn^nHH(no(NOOOOiNH')iOHNnHrjONO^* ro n^^ln^Hn^^oHno^l/)^^HlJ\o^H*^^^^^lnMOMfS(^)^^^^M^lnlIlOl^f0^lncoo^coH^,(N^^o^o^(^l^olfloo^ O HOlOlnNH^coffiM^o^(^l^^ln(^l(Jl(^lHH^MOln^no(N^HONll^*o>Ol3^ococoHO^^^^Na\o3(^lO^H(^l^^o*l^l^^^ IN ....#.....#. , , , iH HOHO>(NH

»< .Hr-li-l.-lHHHHHi-IHr-lr-t.-1rH.-t.HiHHH.-t H H H H fHrH HHHHHHHHHHHHH HHrH Hi-l HHH HHH HH

CN o»lfl^coonHlDN^^^o\^3nHn^HH^MM^^lconHHO)^rocom(^lTl03««NOO^M't'fO'Jto^1HHa)'l,rl

H OOOOOOOHOOOOOOOOOHHHHOOOOOOOOOOOOOOOOOOOOOOOOOOOOHOOOOOOOOOOO CN »^^o^!^^HO^OM^a)*^\o^NHlOln^'lnco«o5nNlrl^KllJllnalO^Hno3lOH(Nco^no^H^^'Hoo*HOln^«)^*o O H^M^lfl(No^^nHlnoHco^HClHcolnodr^lfl'(Jlc^J^^olnn^rtlI3^(N^Ho^ln(N^^^'lN\D^o^u1^^^JlH(>^Hln^£lcol0^o u] *(Jo3H(»^lflo^c>^^^^H^HMcor^l*^ko^^H^^nnocoollI)ln^^ou^o\conoIJlH^olH(DOH(^\DlnNOlON*no^(^l

C oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooioo ro co^^fvDintN'^ooooooioojooinooooooLninoooooooooooomooooooooooooofNmoooooinH -H ™in[^Hn^vo^^^ocoiorv]^oc^H^a*^coifiLntNifl^CNincoHni^on(»

nnnooooooocooococooococooo •oooooooooooooooooooooooooooooooooooooooo 1 i ) J t 1 t t 1 » 1 1 I 1 1 I 1 1 1 1111111

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

a*a>a*HHHHt^r^t^[^[^r^r^^i©^\ovooooooooooooooooooooooooooooooooooooooooow5

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo oooooooNNNlf^lnlrllI1ooo^o^o^^l^llN'l'^^nnnHHHlJmolOlOOO^o^l)^^^^^^^HHHOooco«)coN^l(Nnn HHHa\^o^CT>^CT>a^HHHHfsicNicNri{^iriininininin^v£^^^o^o^^^^\o\o^ ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ifiinirii^ininininininininioininininininiriini^LnLninirtLnioioiniriioin inminioinininoooooooooooooooooooooooooooooooooooooooooooooooooooooo h-t^r^t^r^c^r^oooooooooooooooooooooooooooooooooooooooooooooooooooooo f^lMc^ioooo^^^<^M^^ooo^^w^nlnLO^fl^na)o)coooolflu^lf)ln(J^(^o^cococococo^^^^o^^(h^^^^^^ooo(oa) iriinincocococoncnroLniflininLnininifiLfi^o^vo^^vo^vovo^r^r^vD li^iriu^ininininininininininuiinLninLfiintninLnLnLnLntniniriiriininininin^ ^iDiOvOC-c^r^cocotNngojnmn HHHHHHHHHHHHHHH t-lrHHHi-(HHrHr-lrHHHHHH a) a) a) -a- ^r •* 1 H CO 00 00 A A ^ in ^ INNN d) m m H H H •H -rt -H 0 C C C -P -P +J +J +> +J +J +J +J 4J +J4J4J4J+JJJ-P-P-P4J4J+J-P 0) <0 •rt CCC-t-l-P+J-PtflflJfl © io to to sj (d to CS(tJ(0(d(Qfl3(Ot->t-t»-lk-lk-lk-lt-*h-l*-1»-lfc-lh->h4Mh-l»CCCCCCCCCCCCC 4 -H-"^ ^H1 Id ID MHHHH h ^ ^ t. t. t. 1. (. (. I. I. (. t. (I <> tl 1*1 1*1 fl0 W W WHHHrHd IB JJ <8l0lO(8(0(Oo^oHOco^lf)(N^(n^ln^o^o^Nco^(^vo^nn^co«N^(Na»H(*)kOf^(Nrl^ocDlfl^OHco^^^

^ co CM c* •* r*« 00 [^ 00 ^ CO HrHHNi-tHH *H HfNHHfM H .-t *H H '^ffiin^oounon

HnolNlOHm*owo«nNWH(NlnnHOOl0^oooNOl0^f^la)oo\H(Jl(,)^nHM^O(J^oo*0(^l^^NOH^o*ol^*0(NO(,JlO m(OOl^»^(Nw*n(^lHNttHOl«HHHHNr^ocooo^q|lnofll<]lPlfl,nHO^>OlnnN^f^lo^oN^**co^OHrtl^lal(^^(N^(DO H O) HHOJHHHHc-tHfH HcnNN^HHCNHi-l H H H r-l rHOJHH HHHCN H HH H ri HHfH H CM MM

N^HN**^OH^*o^OlnHlilon(^rl^rlNOl*l^o*^ l»^

OM^MO^ClO(Ncoo(Jl^*H^N0(\HD«»0«HNHnnOn*fl^'t«NN H i-i r4 r-i H HHHHrHHHH H iH H rS HrHHHHHHHHHHHHHtHHHi-IHHHHr-tHHHHHrHHHHHH

r-4

r^ r •- C- O) OJ H H H IOVO\O^DVO\0 \oi£v£iCsOi©V0\D^0^0 1(^(0 »© VO *D VO VO vO VO *C ^0 ^O V0 vvvvvvvvvvvvvvv V V V V V V vvvvvvvvvv V V V V V V V V V V V V V V

)OO0^0(N«o*0NW*i"i(0oain ^o^lr)oowHo^al^lO^(JlOr,)(NH

0 H^eNcn^fCioo^rvoCrtOror*^^wo>u^l^r1N*^no^n«o^lnHNalo^r^M^^^oco(^ltflnoMln^^^^^»olOln«(^r^lMHO^ocoH^l(J- l

nH^^*o\^^gf^ool•*(Jl'fONfflOr>nn^O(N^'t^^D^OlonMH(^l^no(NlIl^n•Tln't*»IllnQONOH[,J

Hr-IHc-IHrgHHi-l rH H rH n] -H rH

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

iNnnoiWNNNNnrinionrinnNMiNnnnnnnnnnnnnnnnn^^n^fi^^fl^nnnnrifirtrtn^nvnN O ^«^IwnH^rtHon^^^lONW«^^Mrt^^lXlOO^o*lno^^«^o^lflnln^N^^lJln«co^offloo^O(S^'OH(,lNcoolHCO CN ^^-if>v0\o\ocoaot^co^»ificocovoaico^coioeirNir^coo^ifiLnco^r^eiifiifiocoHco^ei^io I i-t H l-t H r (HHHtNNNrlMt

O ^colnnlnt^l0^^o^r^^NNrJln^^a)^J|nn^oHON^HOH^^rt«eoH^^|ooHcoop^coolnlO^*(MO)^HlnclmH^Nn to (^^HW*^^o^o^*^o>Hco^^(»^lfl^OH^H^o^Mrl(,^lJla)^on*^^lncon(N^HO(^lOOOln^coooHOHlJl^lD H^(NfHHHrHHrHtHOOHfHrHMHOOOrHr^nnn{NHHHOOHOOOvD^H

O conHHOov^o^n^*(^colnnmc^nOHChOlI^oo^co^(^lli1*HHOl^M^^rr^^^lnaHM^flHWco^^t^l^OH^Ml,^co(JlO U» ^lnNco(N^mln^^^Hnolco^ln(N«^^oDlflpllll(JlHHlnln*H^^JmolHln^ln^^o^mfflco^mHn^nlrlln^lfl(Nmco^ X HnCN)^HHfHi^i-trHOiHHHO(NHOOO'-lHfH011NHr-tHH0001HrHO^r'*H(NOHH(NHHH01{N(NPgC>l(N01WHHHHOHW

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

l l ^J^nnlNN(Nrt(sn(NN)<^lnHH^l•lNHHln^f^*H^n*(N^t^•lr)lnlnnf l^oln( n n n PI n •

'ai^Hcuooiw^uwco^HnotNfgifKMMs^wHrxnwMncofnnHOHOioirtHinrtriH^Hin

Or-HOOOOOOOOOOOOOOOOOOHr HOHHOOOOOOOOOCNHOHOHHHOOH^r-fHHHHHOOOOOOHO

^oco^Olcon(NO(NHH^lnnHln^lnlnlflNHHlfl^^o^colsnfg(^o^a^^na)o«f1Pl^'tMco^Of^ln^o^n««(Nln^^^*

oooooooooooooooooooooinooooooinooooooooooooooooooooooooooooooo OOOOOOHOOOHOHinOOlOrHOOOinHOOO^IOininOHOOOOOHOOOOHOOOOOOOOOOOOOOOLnO ^oolnoo^^^ln^o^fMO^DlN^^^on^olno^m^coH(N^Do^*o^ooo^^oow^^oo'*(N^1'^^o^^o^Ol (Nr-lrHlN (N fH (S H H r-i cHHrHrHCNrHH1-(HHHrHH(N(NHHrHHi-tHH

einr^r^cooococNjoinintniriiriinon^^^^ooinuiiiioininoocNrgcNf^ lnlncoa)cococo^oa)fflco^Mn^lnoooo^^fnno^nooo^OlOlC«^*Ol(J>^srnnn^^lJ^(^lTlO^Olf^flnn**^'*^tf mtn^j'^'iD^'^cococococor^r^r-r^ <* ** ic vo ivc^ococococococor^[^'*'*corv>iiiiiiiiii iliiil

OOOOOOOOOOOOOOOOOOOOOOOOOOOrHOHfNOOOO (HHHNHNiNoiHiNnnnncinflnnnMMNNN

OOOOOOOOOOOOOOOOOOOOOOOOOfNC

, 000101 (MM fgmn^0^^ooo^lJ>f^n(^n^^NNln^lnlnoo^<'*fllnlfi^fllnoOHHrtrlOlnll1lfllnlnHHHH0lfl l(Jl0\0l(Jl in in ** -^ t^ r* ^ool00o(^o^(Jl0^ln>o^Ol0^ococolfllfl^^lO^DOOrlHHH^tn>i)^HO^^^^^^o>o^OlD^o^^^^^OliJl0^o^O'D ei o n ei ei n

U U ft ft N N iti:>i>^>^:s:*>i:*>i>.:>iXXXXXXXXXXXX>

ininininininif)ininininuiinininir>ininininininifiintnLninini^

www n ecu wi««w_____ -H -H -H en w .nai-H-Hoifl) w cccc'O'O'a'O'0'0 3 3 S pj a « a) a> o o w ^ i4(/)WWDicO T3 -H -H -H -H 3 3 3 3 3 3 > > a a c c c •P -P +J 0 0 0 r -P -P ©•H33C-H 3 TJ'O'O'OtOWWUlOTW aj (0 -H -H -H 03 m ai a) o o rH(0 CO(0 n)9l$OUO^^ O O W (fl N N N ^ ^ M tO JIBOJ'O T3 T3 -3 0 0 -a -d > X •HH«l/mHBlBlWBl^>>>>>^^i<|(l)flUUiq»« j) j) H J) 1) x (o a> a ft 3 0> 4> 0> 0) 3 CO 0) 01 tT tfl X3 X3 X3 'C M -H -H -H -H -H CQ £D ffl CQ M U U U U U W W 3 3 *-l ^ fc d) 3 N N N JJ +J 4J « W •H -H d) (0 (d a a-aa-p 33 333310^^^ MUMMMM a«J < < < -n * +» 0 c C C G X! £ -o ^ - a a a a -Ptr&ca-PcaftJaitf-r-Kuaaiaiv-p-P-p-Poooooo a : 3 >« o OOOJJ uu o 0 0 s (0(d(a(d(d«ccaiu)^Mi4M 4J+J4J-P4J-H-^-H-H>4>>*&H>«>« o o o 2 S •p (OS p ) (0 ft ft CU ft w w 3 3j3 .CCC* 1 3 3 3 3 3 -P +> -P-P 0 o £3 ft tfl [0 ^ b u u u u i(0«j(tf[0(0a)nrinH,*nrinfOnrOrintM H

nj iD^ovD^f-ifioroncricoCTiffiLntN^oi^c^tNCTiairHiric^tNaiLDCNO •* H {Niot^>aiHinri]mco''tfcN J ^oooino)rnrir,,«.rn.H^ou,iuitvir^ir>ia"0(*i^'^i'>-(iOtnr^cO'3,r)rN-cor-] co i-t ^*tfifl[^(Nif)\£'

«J I^l^^^to^^^oO*conOlOH^^(Dn(^l*l^co^co(^Hl^o^(^HHO^^nHn^HO^^^OON*^^OlOn^^r)^COlJllfl^(Dy3^Jl^'co CQ ^ONtJl^o^^a^aJ^n^o^^M^(Jlo:^^D^lJ^^^oHli^^^^Olnn^lflalnlrlo:^(^f^lMr^lnl^o*Hl^l^o*co^^OlflW'^H<,rl (^o^^coJll^^D^a)coolco(^c^^o^Mo^lJ^ooo^ocococoa'^oco^^^(^

.0 ^l|H(^^rtMnMnolno(Nnll3^a^H^•JmlnHHlnH^^l^ln)^olnn(NooHln c- ^or* iif >ir >m 25 HNHrlrlHHHHHHHHHHHOIrlH (n(N(NHf>)HHHH oir-»HH*#CNHHHCNHC>JHOlHHHNHrMH«H rg V

U ^HN»onla^nHsr(?l0^cOlf)li^^^D^lrl^oolco^^vocovo^^olo*c^^co^^^oo^(NNcoo^^^^^o^HlOcooMOlflO^^^ N lDcoOHW(N^oo^^o^oco^conH^ln^lHH(J\^HalsrlD^^Oln^oD^nH^oco^lr)^a3n^rtcococoaMa^nNr^lHo^H'J^ i-fOJHH HtNrjHH HH H rl H \T fl fl H rl (N H CN m H n (N H N

U ifi^ifi^^^Hoi^o^or^covovo^ooocia^r^otfiu^^inLno^^cocovonvoor^c^c^^r^r^rnr}^ W ^o^r^\onr^v£>cocjr^a^co^if>c\uivo^Ha*rorncor^H[Nino^if)v0cocot^>ifiOHO^co(nHcofM M*NrNJ(NOlfNJHf^HHHOJ(NCNgr0

^OlO^**HOHH^o^(Po^ocoHo^M^o^r^lnoJ5l^J3^o^rlcolI^*^Ort(^lHO^l£lO(sr,lotN'ilH^*(0^ald^^no5^*oco S HIIHHHH^OWI/lflNHOHainvOiCO^^NOMOH^^N^itnHrlCOiO^Htf^rlNririHOOOiHIJintHHOOrlHffiir H HHHHHH HHHHHH HHHHHHHHHHHHHrHHHHHH HHHHHHHHHHHH H HHHHHHHH

3 MCN]OO)ivlO[^(T>r»«C0OOOOOffiOO OO (^ CTiO Q ^^HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHClOJHHHfNHHHHHHHHHHHH HH HH

V V V

•H OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO^OHOOJMr^

0 (N*o^oo(J>«^o^lnNln^HHlJl^^OlNHocooOlnln*^(Nal•T'J,^"t*^^rlOlCr1^or1HNl^^H«'J'M(>lO^*'^'^OlO(^i o ^vfl^[vinro^inn^iniriifi^\Din^^^H(n^

H

^4 f«Jl©n)CNMfNnjr«J(NrJMOJC>J(N(N(N01C>jr>ir^CNr«lfNjHCNH(NCNMCNr^CN U ooocococooooocooocococococooococooocococococoocooicooooococococoraoo^noooicocooocooocococooooocoooOT H H H H > (Tl^^^nnnHlnl»cocOl^N^llD^OHCOn^^ClNCOlDco\o^^OOl^H^OOO^O»Cr^r1^^cor^cOH*M^O HNH HHH H (NH rid M H ri N HH -* (N H (N tNHH H N rl fl «t (N H C

O iriLfiifiin^r^^^r^r^c^r^^iOt^^cocot^c^^Dcotri^Dic^^r^r^NHoit^^ui^^^rncoriio^o^ «J OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOHOOOOOOOOOOOOOOOOOOOOOOOOOOOOOC CQ ' ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooc

in l^mlJl0^r1^0NlnlnoM^coNno^*o\coa3^^^o^o^M^ow*oNlJlco^^^^«^nNO^o^M^^ONN^^f^lO^^^N^ O HmHHHH^HHHHHHHHCNHOOOHrNj(NOJlCOO (N a< ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooc

O olnolfl0^nMlncoN^^nlno^ONonHMOM^co(NNco^a^H^r1No^lnH^NHHOWlno^HN^coco^^^»^^o\

vor*iniO[^coa>o co en v , v i£ COO iO H ^*in»0t^C^C0(^OfN, < ~~ O0>CTi(Ti0%criO[^C0HHHCv(TiOicvcrior)M'rtininii)M)cor)M t >Lninii3MJC0r », »t^r>t^o^a>criooc —'nnnnrO^-COCOHHCOVOCNnHHHHHHHfNHiHHHHHHtNOJHHHHHHCNOj.p - - - — ' • i i i i M l—i i I _1 _l I IM AI n <<<«<<<<<<<<

O co*^^fNl00^^^ln^l^oo«oocoofOH^U1*tNnoH^gln^oo^«)^co^,^OlO^*OHln(N^'J•(^)^colOlfl^^O(D(^o50l*^OH 10 ^lnHHM^coH*co^(^r^oooooocoln(Nn^^ln^n(Xl^^al'f(»^Dto^(nlflln^N^^r1H^ool*lOlnln^NOrl^o(^r^Ol O HHPJINlNniNnNHNHnOOOOOtfN^^nMNOKNtSHrlHHNHHHNnnniNdiniflNnNNNiOHHV^HXJnHHNO

O ^o^^Oln^Hln^^^^ia^o^^n^n5rwHn^g|Jlln(^lo^N^*ocoln^H^^^H^(^l^lfl^lncoco^^ool

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

nMcifNjrovo^^nn^rninoooooiO^ulifi^^c^in^MNCNHri^ojCNjfn^^L^

OOOOOHHr-IOOHOHOOOOOHOHHHOf-tHHOOOOOOOOOOOHHOOHWHHOOOHOOHOOHHOOOO

- -1 v - 1 - i"» i^> i"» i^ i> *i n I* n i^ i" i> MI w w i"> w w * i " w « IM* i^ L^ i> i^ i* i^ i i> i^ i^ i i^ i * i" w i> i> [^ w 41 \u i> i - i> i^ in r* r* w 7 ooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooo o OOOOOOOOOOOOOOO^OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOCNOOOOooooooo o tN^inoioiocoovOfHr^oi^oon^rH^(N^Hcot^coin(N^ro^rvor)cop,''*co ^•o>\oin HHH.-trHHi-IHr-tHH.H H V H i-f r»l rHHi-l rHHHr-fi-IHHHH HHrHHHH ,_tHHr-t HHIN

inintniriifiiriirjifiifiinininininininininooooooococoQOcocoooooooococococooocoinirjininw

oooooioinomoooooooooooooooooooooooooooooooooooooooooooooooooooo oinooirir^^ot^inn^iiiininvD^ot^inuiiriLnoooooiriooooooooooointnooinooooooooocoooooocNiifi

HrHrHHrHHHHHHHHiHnfnnrif^OOOOOOH

NNiNNiNNiNNN(NNNNNNNNNNNN(NNNNMnMNNNMN(NiNiNN(N(NNiNwnnnnnnnnnnnnnrtPinnn

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ^CTia^o^ai^o\a\oi(TiinLninvovp^^D\Dojn]MrgMN[^(^(yiaia\(rii^ininininu^cftm

ini^LninifiLninLfiiriinifiLnLnLflinintnLnLnLnifiininLnininuiini^ & ^^^tno^lno(^^o^^allrlo^co^^^o^^o^f^lal(Noo^^^coNHf^l^^HconNHH^^N^ocDrto^ln(^]co«^N^oco(^lO»o rH H «H .-( r-t H fl

(0 n-sft^MHrHnNOHco-^Hoo •«3,r>jincNj\Dr*-t^OkDc>j^fcoco w^roc^inriM(NHoi(N]conro\oincoincor*-r** HO«(N«rinoiir)ai

OHifin^ooc^rntNHair^mHfNn^o^i-tocooeNrooo oj^CTirHMNvoor^cocou,i(*ia>inoo^,<,cyi«j'«a,'*oocots)H«)»*>fi^,inco(N,-i^)otNco"*[^oooNoco(ocoH\Dii' iiiinsovo^^in\Dvoir)iovoiriir)t^-^D'*^r^j'ioioiflkoini«ot^ifiin*oiriixiirivDirio

Nf^o-tfrvpjcot^nco ^orv^vuninni-r-t^^-jcN'^'ir) ifiLnoors-ininr^co^D'*vO[v[^vi)conj>ocyir>fftt^^'U^^,n^Dir

i^ ^

U 'to^NH^^l^^n^^r^ln^coH«OlH^oH^n^^^^lJlcoa3\OHf^^g^^(^lBn^^^^lfl^^^^[snolflOffllO^'alncooi^o ai N^HHH^N*n^u^^^go^HOcoM^^ln^MlnmHco^^^^^colJ1\o^oco^lnco^NcoMlnlnal^(^]co*no^flO^"*^^lr (NiNcinnnnNpiiNMNnH TH ^n'*^nnmN^(^l(^ItNn(Nc^l(NO^Nn^n^^nlfllil(N(1NNCN^(^lf^lln^(N^o^l(^^N(^l^

co^coMMHO(N^«^«^^f^N(nv(^HncoH^(Nco^(^(^lr^coH^NOo^o^^^(^nNcoHt^

'0«ilNrl(Nr IHHfHHrHHM-HrHHfH^HHHHHrHHrHr-tCNHHrHr-IHHrHr-lfHi-H-HHHHH! HrHHHrHr-IHHHHHiHi-li IO jH H H H T

c^^cTNO^^o^co^[^c^c^ric^ioiocoiH^irtvovoocooj,3'r*',*"*^roo

\C \£> lO ^ \0v 0 kO k£v Ov O kDk 0k Ok O* £ k0k 0v D^ O^ O^ D VV V V V V VV V V V V VV V V V

•H •*mMCTirMr^>ii)o,Dj ^ [f, O) ^.y H O in

O r)OM[^<,cNr)Hoo^>-cocoai,>*m'

•rnrHOvD<'iriinco(NrHOf-

, , H cn COHC air*ooc^cococN(NfHcot^co[^ ^tNmvD^n^DO^D'ocoorHiNCNcooininin^'cot^t^rHr^OTvoojin-HHiniriCTic <-l rH rH H <-' H rH H rHrHHrHrH

t^cocococovot^cor^co^co^cococor^r^t^cot^r^co^cor^c^cor^cocftcoc^oicQ^c^ ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooc ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooc

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO)

nnnnnNMMMncnnN^iri^«vD(NnN(NiNfSMNNnnnnrtnnnnnNN«nNNHMNnMnMnMNNnN O ococoolco^w^con^H^oH^o^lOHO^co<7^oo^^l^o^^^(Jl^nvo^coroH^Dl*l^fou1^wf»lO(,^^olf)^Do)WOOlH(Nf1^oo*o^o^lP

I H H H H -H r INNINNMNNNNNNNNINNHHHrlHHHHHHHrldrlHrlHrl

H^n^^^NlnolOWO(NcoNMln«^^(Blann^OH(Nna)^H^ln1NOcolnHCOln^!^Honncof^rl*o\nHcoHOlNN«coN ^^lfl'J,^OlHOcoHco«ftf)|J)^^*^^0(OM^(Doo'r^l,^^H^oa3

C OOOOOrHOOOOOHOOO'HO-HOr-lOHOOOOOOOOOOOOOOOO'HOOOOHHOOOOOOOOOOOOOOOO ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

^ror^Ln^r^uiinin^^c^M^^ooint^int^inroiDinio^inLn^^iri^rt^kOvD^^r^^inLn^r^

nc^l^cO(^(^l*n^^^M^lflc3^f^l^^Il^^H^co^^HHN^^HlJlo^lnolHO^H*Ol,^^H^cocJ^^o^l^^O(,^c^coH^O{^lo»o o^HcoMO^«^o^^^lnH^^«n^o^al{oco^o^^^no)^MN^^oco^rlO(^NHOoolr)lna)^,)O^HO^^^OlOr10lflc^^ inNNnNHNNWNINi IWOWOOlHNINN JHHHHHHHH

HncoN(NcoconHO\o^Ho^coli1N^^co^ln^^co^rt^o^Olfl^f^^ooco^^co^^lflMo^l,^Hn^Of^(^lOln'a'^l/lH\D^^n *l^l^lHn^HNNOHmMconNn*o)^(^l(^nnnoHH^(^ln(^(JlOln^CHrl^li)nnoo>ffl^o*ONO^B*Nn^lflN(N(N I O O -H (N H W -H r IrlrlHH. I O O -H .-t H H t JHHHHOOOOOOOOOO

l£lnCl^rl^lrt^tDOHO^^l^lo^M*N*no:H^o^ln(a^l^^^o^Hno^no^c\^lncoco^OH^o^o^lOO"J'»ol^^OHO^ln^^

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooomoooooooooooooooooooooooocoooh-ojLocoin-d'ioincN'^'vocooooooooooooooooooo co^*c^l^^^l^N(N^^col^^D\Dconlnal^lJ1o^^^ol»o^n^Da3n^DOfsal^^n^^OHln^o»ooi()r10(^conco^o^^'o^ •HH H .-H H <-H H H

cocooonfflcow^ow^oowcocooocooococotocococooocococoooo^y3vD(^o\^(TiWCii^W(^^ifiinifiinifiuioocow

rv]c>]r*ir^i>-:^^^^^^^r>JCNicNOJrjr«jHCNr^rNi<>jr^

inioinHHHiniriinin[^[^r^ojM(NM(Nirgnr)rorn(^Mrnrnrir^r^o)cooooooococoio^^ u)\o«HrtHHHHHHHcococoa)a30)coolIJl(^(^l^o}alracoooodHH!N(NNn(,^(^n(nn[N(N(NNNa)(flHH^^^^^^co«lcD

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx U-iU-,b-,U->b-

oooooooooooooooooooooooooooointnioooomininininmirtinmoooooooooooooooooo oooooooooooooooooooooooooooor*-r^f*.ininificNo]oiojNtNo»(No)oooooooooooooooooo (NM(Ncococo*«^^oo^^ln^nlntnlr)^^^^^oooo^o^^olfll(llntoMco(,1M(*lMrtn^^^^^^^oo(1^^,)^nn^^^

ooo^^^^^^^^^ooooooocococo(Oo)Oll^c^Ol^^^^^^^^^l0^o^o«^o^oa)cocococoala)^^Ol»(MJ\(MJ^cococo HHHOOOOOOOOOi-li-tHfHHHHOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO ininiriinintninioinioininininirtininininoifiininLninuiiriifiininininini^ ooooooooooooooooooooooooooooinminooooooininLninininoooooooooooooooooo OOOOOOOOOOOOOOOOOOOOOOOOOOOOt^t^t^OOOOOOOjrgtNMMNOOOOOOOOOOOOOOOOOO , ^^^^^^^^^^ininconaicoacDcoooacococococococorgoiNooo^^^^^^^^^HHHHHHHinirtnoococonooH^l ^^ooooooooo^1Clnl*ln^n^^^^^^^^Mr)^^HrlHO>o\ffl^^^^^^^o^D^OlO H nnnMMNMMMMMNnnMnflnnNOIO)(NMMMrin«MMPJNMHHHHHHHHHnPl(,llyl iriiiiininifiinininLn^LfiiniriinLriifiifiLfiinLnLntfiiriiniriinin inLnininininifitninininmintn uCO uC O idCO uC O CO 01 CO k hhlihhX to CO ^ u 91 0 CD CD CD CD 41 u u 0) (0 •H •H -H -H -H -H -H 01 CD Q) QJ (J) 0) (U+J+J+J+J-H-H C c c 1c c c e ,•H •H -OT3T3 CCCfflfl(0«JH^«)a)ai(Dll)4J4J.p4J+J.p4J4J4J-mJ V 41 01 0) 0 0 0 0) -P +J -P HHHh^ll3 333CCr(HHHHHH CQ CQ £fi X X X c c c c c c c> E> 4 J -U 4J +J JJ 4J C C C UUUUUUK{%P6r%Q> •H •H •H C C C B S C 0 O 0 0 0 0 o 41 -P -P 4J +J 4-> 4-> 0 O O OJ oooOfl3n3n3to m M IB fffftrtrtrtisiiijiii ,1! O O O SEK>>>>-IHHrH +J+J4J+J+>+J+J «J (0 «J ffJ (0- H -H -H -H .G .C XJ £ ,C« j =« J J JJ JJ •H -H -H -H -H -H J J H H k M U ij Ij u u u u CO CO CO CO CO CO CO CO CO 3333*** (t)[0!0[Draitlffltl(tl(!)(DlllOOOOOOOfflnJi«(flffl«fliBiD3333 33000010101010)0)0) CD 4>« « 333335 JJJJ 3333333

Hr-IHOJCNOJrHHr-IHHHHHH

Nn^oHwr^*lny5^cooH^go*ln^^lJ1^o^eooHc^lnTJ'ln^li^vo^HNrl^ru,)^o^ooa^^alfflOH ffion-^tnvo^.cotNro^' >>>>>>>>>>>>>>>>>>>>>>>>>>>> >>>>>>>>>>>>>>>>>> XI ^inocnt^ix>fHcoMJiriroirir«)[^

v v v

(0 (JlO^M^0MC0^0^nNr^l^na^OlXl^(Snn^Hln^o^lTlO^^lnH^H^*l/l0^lnrl^^HNf^lMC0^DOONO>-l^«'l'WH(^)(N»l^ oQ o^onco^^cOHOu^lf)r^l^o*co^MO^t^Jno^o^^^^HtNO^^o*^^co^cococo^OlBcoa)^nooeoH^^ON(N'^'S'co^M()l(,^ -H rH r-l H t-H Hr-tr-

^•coffiifir'icor^^'CTi woin ^ o eo

Hco^^lO^^OHNNOvnolonN*o^(N^NOlO'^^H^rlf^lflf>^o^^l^l0^cof^llo^H^corl^olnlf)^o^oOHlflO(Nl^lO(N(^ co^Olnlr)^oococo^^o^^^^^o5^^^Dr^^Mr^lOlcoN^Jl^(TlNHH^^•o^lr)o^oo^^ron<,HNOWln^^lnnno^oo^^^^ HH(M(MNNNHHHHNHHHNH(NHINrl(NHrlNHHH(N(N(NHHr(NOI(M01(MNNNNnnNH(NHddHHH HHH

i-i Hr^a>r^f-iininovDvoo(N(Nt^iri'i)Oo>c^nt,^{Nirtif)tvr--Ln w ^•rHinnoooo^inoa^oj^f^kDnoonmcocofNcoor^fnco

Of^lJlP^\ool^^^MoD(Ol0^o^^o^rl'J'O^OMln^^nlrlooH^n(N^^j|ocooofl^*lOln^oHOMcoo(,^^,10(0^^()l^ooolO^ S o>o^NlTl0^^cooolco^ooco*ls^o*^co^(^coHlflN(NWl^(J\HO^cooa5^*o^o\^o^wN(^H(^ncooH«nlNHr>Hoo

HH H HH HHHH H H r4 H r4 (N HHHHHHrHr-tHHrHHOJNCN

O HHHfNH^tMHrHHHHi-irHHr-lHrHi-tHHHi-tHHHHf-lH^rHCNrHHHHrHHHHHHHrH^IHHHrHHrHHHH HHHHHH

C *Non^^(DO^^o^nH^o^oolnln^nlnMHoolNa5(o^oco^o^^HnlJllflw^*H^lflln(Nn*nr)c^co^^Moo

(N t-- o o ojocricor^r^o^eNco

-H *NHOlT)H^0\C0\0lrtr,10101U1*Ln^^in[^^^coHvoco^r^r^r^co^r^c^^\ocoin^fo

0 lr)^f^^lno^oco^coH^oconH^^colfloolHlJlMNln<'Olgolno^^Nlno3*^noco*^o*o^coa)^or^*,*^lflOH^lIlH

h ^^a300c^r^^ooolfl*fl^ncl^o^MNHlNlflln^^o^H^cocoo)co^^1^c^l*H«*Hoa)ooJHM(^lNH^fl^lO>o*^ O HioocooHocor^rvr^Huivi)H(^cornco^a)irjoc^^vDc^oocoo>co^rs)^ff\^[^a\Horn{N

> (^ll3^^ln^oco^co^^lnolno*HHHwnnMa>co^coM^^nlNt^**o(NHlnNOl»oo3H*Hlfl«^^n^l»M^^go^^^(ON^£ H (NHHCMrHHrHr-liH

(0 ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooc 03 ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooc in oln*r^wo^o^^^co^coM^nM^D^^ofl^coooco^u^^oNO^^'JH^(,^n^o^(Noo^lfl^H^Ho^NlflH^HOHOlco(T r>j ft< ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooc

O Lno^r^«^oo^Hv0Mtf}^OLn^t^^lflf^oolnul*H^H«oln^m«(NHf^llfllnla^a)HN^^HOlNNOONcoMcoHOlnol^n^ocon(0'Jln^l >c; '

0'a'^^^cooHf,^M*lfllOlrl^o^Hf,iCl'J^l0^co(Tl^cowOrl o\oci,*w\or»cor»jc,i'3 r-t lJ>o^al^^^*Ovo**^*oDffloococococO'f*^*^^^^^^Ila^o^^OlO^HHHOOOOOO'*^''*lfllflr^^vo^OI^^>l(^l(^l(^]^*^ S >>>>>>>>>>>>>>>>>>>>>>>>>>>> >>>>>>>>>>>>>>>>>> 5 %%%%%%*:<<<<<<<<<<<<<<<<<<<<< << < < < < < < < < < < < < < < < < O o^lrlo^t^lMN^^*^ln^o^HN^nIJlmo^^(^HMnf^l^f(^lnal^OlCl^o^HHOooo^lfl^5l«3^coo)lflN^^^nHrln(n^o <0 2 ^H1HCNHHHHrHrHr-4rHrHHHHHHH^HrHHHHHHr-tHrHHHrHOr4(NCNJfN](NrHH£Nrg(NO]HHrHHHHHrgHHHHHHOIHH

H i-l O O H (N fN . IHHOOrHHHrHOlHHHOOOOOOrHOtN. IrHi-lHOOrHOOHHHf HOOHHHHor>innt^

O ^o^^Mt^^^iHinno<^tNH

O (NNHNlrl^lrlN^l^n(Nn^nHHNNNnfl01HnnO(S(^l(N(N^Hrl>Jl^NI^lHOIH^lNHN(^lISIN**(N^l^!^N(N(^H(^l^N C OOOOOOOOOOOOOOOOOOOOOOOOOOHOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

Qj HrHOHro^^r-iNrgMHrgntNOrHcNrgHNc^HHrorNi^OrHHHHnHrg^mNH n Mno\o5l^n^nd^«c^alHOlO^^HHMOH^^H^^o^OHflna\n^*co^o^*^^lP*lncoo^al^OlN^^<^l(^lO|OH(^^*Nli1 O «N«H^^Hcow*!OlO«Ollrto^HnMo^o^H^Nv^^o^^alOli^Cl^Nn^lrll500^Hoo^eo^H^^nnNlflo^iO(D^l

CN o^t^lco^l^^a\n(^Ma)BlnHOlOH(Nn(^onlflnno^^co^noN^no\HHt^icoll^l,10lOHlfllncon(^eo

o o o o O r OOOOOOOOOOOOOOOOOOOOrHOOOOOOOOOOOOOOOOOOOOOHOOOOOOOOrOHHO

CN ^0(Nrl^^(^w^HHmco^onlil^(^)lono^lfl

C OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOlOOOOOOOOOO (0 OOOT-4OOOOOOOOHOOOOOOOOOtNOOOOOOOOOOOfyimin'*(NC0O,^,OHOOOOOinor*«OOOOtNOOOO •O H H i-H fJH(NO)rl H

(NCNNrgin^^^^^^^^inu^in^EN^^^^^^^rn^rnrnnco^^ifl^iJ^CNJCNNHMH

cocococoifiinLninininLnLfii^intNcNcNininin^ininiALnr^nnnrinHrHCNMOJtyi^ cocol«cow«l0^o>c^ClO^D^o\o^^^oooooooocococcl»coco^^ooooooorlHHH^^^^^^^^^^HHHalO>o>c

NNIN(MNMMO)N01INMNMHHiHriHHHHHH'HrlHHHHrlHrlHHHrlHHHrlrHHrlHHHHrlHHHHHrlrlHilJ!l(Q£ xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxssss

iriininininintriinLfiininininiriiriiniriininLnininuii/ii^uiiniriiniriinin 01 0 01 01 01 O 41 V V V u u u U Hi 111 01 31 -H •H -H lOioiiiioidiuiOidxi^A^AA wuiwwwwwwwwwwww c c c C tji 0* CTi C i/i 1/1 in OOUUUOOO UHM^3333333333 III Id Id id 01 ai (11 (U (U III ill 01 111 0) •p -p -p p p p .p 0 0 o diaiBiUfliciSfflccccccojiii vaiaiaiiQtoidmiDaaiRjmto c c C c >H >• ;H> • >H - >H • >H >• >H >• >H c c c ^ *o "6 "6 I) o 0 M -M* M U M >-l u tl kj •H •H -H 3 3 3 3 £< )E S X 01 C) 01 IV 111 01 01 ^01 ai a. o< o. (0{0(0(fl(DflJ(flflJ^^MM^^a>(uo>a> («(Da)a)'^-H-H-Ha)a)Q)aja>a)a>a;o

\0>OiOHHH(NHHHH HHHNNNNN(N(N(N Xi ^oo^o^w^ovcocoa^^*^^^'^^oo(^weoHW(^lcocoo^OHH(^Ho^^l»^Hco^(^lncooNnH^^MlniNn(>in(N(Nn(MnriPiMfnncoHHH HHHHHHHroMn^r^'^'n^f^^fNnnHHHrr

3 inocHionino\oco(NNcor-or>-Ln \o co o coHOMMsin^sf o^ mno nin,OMfln'*HnNMrnoH\ono\c r-t H H H H r** r* [^ r*. r- HHH vvvvv vvvv v

BJ OHHH^mocoH^eocof^lul^^o(N^ol^^lnn(^^^^ln^lnHO^lncO(^o^cnHo^flc^lc\^oonol^fl^rt^oon(()^(^]<^^nnwr •3 ^o»^N^^oo*ONWcoaJl^lOH^Noco^lc^llnno^^^^H^oco^^^^roH'J^^'y)^•o^'0^ln^lnf^^D^Dt^mv^,o^'lOr)OUlO' woo^OlOH^lalHHOOlIJl(^^^'fllc^o^D'I>l0^o^'^lln^^olBlOlo^l^ln^^[s^^^ln^^^^l3^(Jlooooo^f,i(^o^oy)'Oco^la3lf «H H i-t rH c-l tH

oomcor-»ONoiHro

HO*oiftrnr«jco^co\ot^inr^[^rHmiflo^uicN^ooa\ ofll'fif,lo^HrlH*eo^HWH^oa^oo^(3lOOOHo>oo^oolJ>eol^Nnn«o«l(^^^«l^(*lwowo\MlIllf)ONo^OlOoolOl^^ N N M HrlH HHtN H H HHHHHN rH tH ,_i (>) ,_| ,-( H

fflNln^ooo^**r10'JH«HOlnco^rlHli)nr^\o^o^HcoH^opl*i)lOHH^ocoooln'J,Hnolnf^oo*^H^colJ ^^NlnolflO^ONlOnHl^O(OOlHol^'*^nr>ln^tfl^^^o^cfl^^lf^H>fla)^^lflr^H,sfH^^sl,*^ln^Oln1,ln^**HcolOO^c

NWNOinnVNnnNMnfgNHH(\0)NN(NNN(NHHHHHHHHnnnNNNI\NNP)M(NNfV(\INr«NMNN(>INNn^tflf

HWNlCO^lfl^nOf^lC0C0^^O^MC0Hln^C001MOO^^V^nO00lflC0^OC0OO0^lnH(NW^^NIBC0fflO*lflC0y3H^^'

N nfMrHrl\0OM0(M'J^rHHncMIN CN (N rH H .H H rH rH kO H IN rl H r-1 rH TH m(*> H

vO \D ^ ^0 ^0 • ^* *tf "tf ^f •* ** *tf < ^r <• <* * *r *o ID VVVVV vvvvvvvvvvvvvvvvvvvv

1 •H ininr-t^rcyv^c^o^f^omminvDrHincorvjocrirHr* ^ \D m o CJ CN 55 oooooooooooooomm^,mmio^rifiin'*^,,^,mmsfmif)m,*oji-im vvvvvvvvvvvvvv vvvvvvvvvv

0 ^^(^Nlno^o^HnHO^o3N*^^coo^«IJ^roo^M^ln^nsl,co^cocon**o^MH*^>lfgn^onoco^^HH*o^o^W!JlHr, U r^M^Hinu^^voir>^M^cNm(Nmmm^^m^^mcNmm(NOJri^rHm^^m^ir)^mi/iOi^Dm^mmm^tNvDr^m^oimrHmm^,r-

COCO0OCO0O0OCOCOC000C0COCOCO vvvvvvvvvvvvvv vvvvvvvvv

> rt^coroinHs*HOmtt»inc^ifiiflHifirHa^ >^\o^^Mvosococo^voc^omm^mif)^^vDm{Nin\oomojr^cscyi^r^vO[^r^HHOcoHH(^\o^oooorvmooo^o)^m^\ommr OINNHHH H IN rHHHHHHHHH HOJOJHHHH <«• N H

O t^c^cncoior^c^r^Wfflr^cot^t^r^c*o\cor^erico»cococoa>LnMC\ffl (0 oOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOr- CQ

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooc

LO n^oon^nlnlnHm^N»o^*eoo^^f*^o^mlnN^^rnMcoffl^^HH^Nlnn(Nlflcoo*NHn^^lnr1n^o*^o^o*o^O'fOC O HrHO«H(N(NOJi-((NHiHHrHrHf-IOOHi--tHHmHrHHOHHOOOrHHO)H(NHHiHrHOiHHHHHrHHCNOJHr-4HHHHHmm(Nf* (N O m^o)o^i^ONOOio^rvLom^rHr^NHO^rH^HHmo\r^in^co»oHmHOO^omr^H\orooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooc'

, i m •* ^c ^" m m t )mmmmmm^'^ ^'mtNmm(Mm^'mmmmmmmmincMm(Nmm(M

, s Bjj2^^op*-coo>orHNmo<-iMm^, , min>i)oooo>w, o ifivot -inin (M(MN(NN*0)NNNN(MONOI(NlNOJNNfg010J(M(NMn mmmmmmHHHH^HHHHHfH«*'*^ a\<^o\'! 5 55 S 5 5 5 2; 5 5 * >>>*•*•>>>>>>>>>>>>>>>>> 1^ <<<<< <<<<<< <<<< « | , 0 Hf^ln^o^l•H^H(^<^«ooo^covDlN^HN^l lnlno5^^(N'^ ^(^l'*o<1OH^noocoHH'J(nrl^^Jl(flln*lnnn^lfl^(NNrlMO<^ 01 Hn^^^Lnoj^^oroco^^^infHMrtooo^ro^rNioioojncorocN^inror^vDtfi^ IrHPJOr-trHr-lrHr-IHt i H H O O H H f

O OHM^wmlf1^H^ft^c\Ho^lnnHO^*o^oH*(^lOll/l^N^o^^Nnr^o^^^^NnM(]llJ^l«^o^Nn^'J•^Noo(,J^^(NH(^JO^^

Hrgnri(NHO,*

O NOl^lflH^(^r^nH^^^ln

(HOOHHOHOO(NHOOr OHHOPHP-IOOO* OOOOOOOOOOOOOJHHr OOOOOOHOOOOOOOOHO

n^^^*nHlnMN^(Nnnc^f^)^^n^n(N^co^ln^NHcoH^^MHon(o^Hn«co'flnMN(,l^c^ln^lNOHHH(N(NHfllN OOOfHrHOOOOOOOO-HOOOOOHHOOOOOOrHOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

(N^<'OHn(^l^^ln^o^o^*HflnrtHco«^n«^^^^Hl^Hn(^l^^oo^olnn^^rlH(>JHrl(N(NOHHHHNHnri

\oolO^NHlrlnHO^Hnlno^no^^M(^c^or1^^lnr^^^(Ncoln^gHco*^0'Nln^r^lo^N^^^^lnlfilf)HMr1^^l,^n(^l^on OOOfNOJOOHHOHOOOJHOHOOtNHO- HOOOfNOHOOOOOOOOOOHHfHOOOOOOOOOOOOOOOOOO

u^^^HO^(\l/^M^oco^DHlrt^NOM^Olflln'JHcolnu1lTlNcol^l^^lJlnon^o^oo^l»*nH^Noo(^(N^*lil^Olfl^o^^ (^HOH^o*^^^u)^WM*Hlr10ln^N^^co»HH^(^co^MncolOlC(N(*1^(,1n^^'^,^>fl^o^y)'0^^^(^l^'000^|*ln

ooooooooooo o1 oooooooooooooooooooooooinoooinooooooooooooooooooooo oooooooooo-"OvOOOOOOOOOOOOOOOOOOHOOrHOlDiflO>-IHLnOOOOOOOOOOOOOOOOOOOf>JO meNoait^^THfOfoo l/lHO^ooolr)(n^Df,10J^owo(N^^'l'^i)(N^'*o^«)^lno^^n^(JlOf^(J^(^lcoHHlrlrnlO*(^llfllnfflln(,^*o i-iVCNr-lrHHHpH ,_| H H H H H rH •HH HH H r-f H H H H .H

WMCoajococococoa)cococonfiioooooiriininin^inioooooooooMfl****nMttcocoa5cocoiriininiiicoiocooooooo

oa)^vfN^D^cocoocooooN(^H(NN(N*^l^]Tfln«^«fg^^cco^Olnlfl^^^^(N(^Nr^l'I)«\D(^^*^o^*^*^l''*l,>^^nn

n(^^nnnrlnn(^rtr,lr,lno^ll^^^^fflO\o^mol(^(3lOO^^'f^<'^nrlOOOlr)cotooooooooooolOlnlnlnlOlO>o^D

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

HHHHHHHHHd^HHOlnN(^I^IHH^^^^^^^HH^DlOalil«^OHHHHHINOO^I)^OO^D^OWlO\e^fi«()^(J^01Jln(,)l*ln OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOHHHHHHOOOOOOOOOOOOOOOOOOOO

oooooooooooooooooooooooooooootoinminmLnoooooooooooooooooooooooooo ooooooooooooooooooooooooooooor^t^r-c^r^r^oooooooooooooooooooooooooo

HHH.HrHrHHH.HHHHHHHtNr\inrO<^rHHHHHHHrHiH^rnMrnr0r0{nnnnn

OJ a; Q) 0) a) 0) a> (/) +J+J4->-P+J4J-P0)1 ><>1 >1 X X JJ JJ .p -P -P P 01 01 ZZ222ZZZZ2Z22aiCiJiJ!i]Ull0roron3flfl3flrtl3a 3 3 c c c c c C -P -P 4J .p -P -P SiSgggg > c a> a -H -H -H ,c .c .a x: x3 .a ,a to tf CO 10 01 01 01 01 0) 0) CCCCUh u u to (0 fl (d uiuiuuitaiooiuiuiiAiiiiaiAiQVCCtdidto .a .a 4J JJ p 4J JJ 4J 01 01 l< u u u h M 0 0 0 0 01 01 fl) jjooou IDIDIIIIDJIOOOJIIIIIDIDIIIO ^--4-HrHHH W B M W W W « WWWWWUl'O'O •P 3 3 3 3 3 3 'OTJTj'O'a'OTj'a'aTj'd'O'a mjsja OJ H W W c IS lit It) 10 r-io i-ia r4fl r-iBfl r-i r-tU rH rH 0> 0> D* 0> 3333333333333!fl>fl(IJ«fl5-<'0,OT3l3T3'D,dOO •H -H -H -H -H -H 0 s0 ID m 10 il 01 01 J J J l-l 3 3 3 3-H-H •H-rl C C C C 00000000000000 jiJ nj(0(0(onj(0«,a'a h u u u u uao; c c C 01 c c » 3 3 ««««UO UUUUOUOUOOOUOrJtfl iflifl WHHHHHHH33 Oj & Q4 0* Ck & HI 01 01 Q 0 0 01 01 01 V) W u> £ X X £ oo«3 « « a> -rH-H-Hiofltoiototoiooo 01 iOHHH e B-H •H •H •H •H 0) 10 •P JJ 4J 4J -J ooowwuii/iwwnuo JJ JJ JJ 4J 4J 4J 11 01-H •H •H 01 H •H O O 0 0 aaaao> o> CQ CD tfl «tnwwnoijH3>>>iJ>JiJpoiBia i a £ 33 J J J J *o r* c\ 01 fncO'*CT^f*irHCT\(N^,\DcO'*ommio^,in^,r-i'*D m>vornot^c^a\^r^rHrHr^oco[^co*0r^Minvoc>ioO[^o^c?»NC*cnro0*Hc^Hr,«-*o com^lnalln^coco(^»**HHHHnnnolOC^o^NOcon^HNN(\r^HNHHH^HnnnnnnNflr1N^1rtf1^(^r1nnnM H (1 ON HPIH

<0 ^oc\Nr^^ir)Hrgcoui*nsf<^(NinHrN)rHnrvco^rginuinooi^rH^^^(no)cn^vo^NNinininin^inrnincoo^coo>t^co^cocooocft^v0^t^v^

J3 \ooir>mojo**HOvoojr*-\o<-i*tfNr- in a\ co t-« r>i co in ^•[vrHcococoin^fcnr^^rHcoovoHf-iincominn H-tfnroei^-om rH H OJ r-l H CN H H H m i-t i-t

M i-tinr^ocoi0o\co^^cnvON^^inin^o^{nr^[^^c^coH\o^^NvGoiinuir^^t^r^o^a\LnNr^ N (^lNOHO^a)lr1lnHHOolNH^DNHHco^»ocoo*o^cl)lfl«^,NlO^OlnooloncoHHlOn(^ooH(JlO\H(,l^(J^f10lnooo!N(P HnNI/lfl'H nndlMciHIflPl C) H »H N CI ClHr-tHH-'S'HH i-t (NHi-l fHtNnjHf-1 HHH MHH «-H HW

w lJl0^eocoo^l^^^Jl0^^coo^^^^lJlOOco^flH^^cococoo^co^lNo^rl^al0^lfl^^*H^^^,o^lo^co<*lOI*>coco(J^o^*co HHnnrlNHNHMrl^HrHNHHMHNMH nHHHnndHINNMHNOlrlHtN'JNNNHHnMNHNnHHHHHINrlCJH

*wNlflHln^H^^*ln<\lH^^o^^lncoco^^cooco^^colDlnol3>^HlnlCOOlO^H«)(J^co

•"s-oooojoooooomnjt^r^o ^^vOvfl^O^OvO \0^D'Ol^'«0 vO vO \0 \D kO \0 V0 v£ v£ <£ i£ V V V V V V V V V V V V vvvvvvvvv V V V v V V V V V V V V V

Hn<^cocnoric^^r^co<^^ooooo<,r-N^o>Lncovooinn^oooooooooooo^oronHvDrNsj't^[^ifiiriniO'tifi^fiftoroini H

CO

0 0\<-^MHooo^co*<^vooLninrncor«i(N^r^r^cTio^Ln\D^Hvo^ooH^rd^nr>)invD^in

M o^oJroHnj\OH^OC^r^(NinrNjfSf^ai^(Norn^oovOcoo^r^c^o]f^f^fN)rvirgr«jr«)oJf^ O ^coniniriintNOco^riH^^ooajconnn^rit^convo^inrnaDcococooococooococoooo^ i-t H i-t <-t H

> o^^o^^^D\D(^(^^^^^^cof^lHonH^(^lno^^coH^c^coHcocolXlcoco^ocooo^|^l't^O'*cO'a,o^H^^o

HHrlfO .H rH H ^- CN r") rg fH H CM H r-t CM r-1 H

O cor^ovfl^cocor*\Dr^vD(NC^iHiHco^oocofn^cot»vD^>r^^r^^t^\or^t^cOTHcor^cocoir)cococoow © OOiHOOOOOOOOHOOOOOOOOOOOOOOOOOOOOOOOrHOOOOOOOOHO>HOOf-tOOOOOOOOOOO en • ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

in *con^ooNo^co^lnlOu^c^i^fflNn^oNlNtoa^^^*^o^IPHN^^oco*^^H^o*n^^l,1HOl(OHMHO^<^lI^o>NH^rl O HHnnrOHHC^r^rHnCOHNNOfHNHfOCNiHHOJHHHrHinoOHrHOOHOOOHfOi^riHOOHHOHNHOOOOOiHHi-tH fM Oi ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

O ^lnol0^^lnH^eo*NH(J^lflH^^^n^lnHt^ncoHcoNa>^nHH^Dln^o^coo^^^*o^ulOMlno^al(^^lr)^J^o^OH*^^«(*)^, ^(N ^ooco^HON^ou1co3lnooolO^ON^(OH(Jl^^^^n^o*^l^l,^nlfllnco^o^o^^lflHHOlco^l'^llf,^l-^coH^o(,^o^lf)H^o* ojrincsj(N(^<^rinNN(n(^o]N^r)rnrojnj(NO]nHnMr>p^ O lncnHrtln^«ln^^co^^n(^co^^DO(^l»HoolHnlIllnl»lnN^(^cooo(,1y3H^^coo^o^^oo^o^u>(Noo^ooln«mn^Oln CM n'^rl0^r^lnoovoalcoH'J'^o(N*nooooJnw)^nlnlilW'*ow^(N^o^^HH(N(N(3^a)r^^OHOlO»I>HOO^o**nnlno*^**

rtNHHrt(NN(NMrlHMNnNMt>iI>l«*HHriMHHHHHHrt(MHHOJlN(NnnOi.

OOOOHHHOOOO< OTHOrHrHHOJC^OOOHOOOOi-tOrgtNHOrHHHHHHr^OOHOHHHHOHOOOOOOHtNHO

OOOOOOOOOOOOOOOOOOOi-IOOOOOOOOHOi-lfHOOi oooooooooooooooooooooooooo

O N(NHHnnOJHnHNMHNHnNNO^HNHNNMINNC0in*mNMnn*in^nnH(NnHN*'tiDnyJHO)^HrlrlC0inOIIN C OOOOOOOOOOOOOOOOOOHOOOOOOOOOOOOOOOOOOOOOOOOOOi-HOOOHOOOOOOOOOOO

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

HtNHHNfMi IH(NO(NHfMnriHfMHf J(NrgNHO)rlHr*(NfNMrv^'rO^CJHr-(HinOl(NH

.H O^^OllOINHH^OOirlr

OOOOOOOOOOOOOOOOOOOOOOOOOOOOTHOHHOOOOOOOOOOOOOT-

o^nn^nNvo»(Tlf|*r^^o k[^i^cocor^r^t^^r^cocoi^k^kVcoi^c^t^voiooOk^coc^cot^r - k - %** ww k - k - » - i'*' i*w *v t - k - k - *w i - » - 1 - ^w v w k - M* i - tv k-»-l- k - k "* *fc^ f \fc/ MS *«**rffc-t--%v%fc/k^L-»k-*l-*w W **/ l^ i^ l^ V l^ L^ l^ l^ IV l^ bV ^ l^ l^

oooooooooooooooooooooooiooooooooooooooooooooooooooooooooooooo ooinoHvoooroiN^'ojricyiooooooHooointHoo^oofMinorooooooooooi-ioooooooooooooooo

H H rHrHHiHtNi-l ,-t<-l

Ot>ir'lCNjr^^.^.^HHHC^k^l^[^C^[^k^OOv0\0OOOOOOOOHH^HCN

Mi^^rivpvDvo^r^cNfMr^M(^^nfO(^n^irtkD(Nr«]^Ln*^inH\DcotN(NfN]rt

«>oa*ooooooo^^*^^^*a)*iriin^i^inLnin^i/iJiooooNMtNHriH^inLniniriinifiiriinir)irioooooooim(jim Mr>jMiNrorncn^^^^kNr^r>)kN(N(No]^D^k^r^r*]tN]^^

(0 (fl (0 (0 3 3 3 3 3333S33>>>>>>>UO>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> fcj fe CM CM [uhfabbbbli*liilibhli

ooooooooiOLninooooooooooooooooooo^LnLnLnininiooooooooooooooooooooooooo OOOOm^lDLOt^r^t^OOOOOOOOOOOOOOOOOOOcNlMCNOJCVk^t^OOOOOOOOOOOOOOOOOOOOOOOO coooonnnnHHrHLnininLnininincccoNN^^'t^^^^^cooscoffl^^sroi^aiHHHHHHHHHHH^tnwm

cocococoi^r>-[^[^k^r^k^CT\CTi^o\c^c^a>«r^inir>iT>^c^CT\aioicri(^ ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo inLni^uiLn^ifiiriinirii/iinLnininiriinifiinifiiotninLn

oininLnooooinLninooooooooooooooooooooooomifiinoooooooooooooooooooooooo OI^t^r^OOOOPJOJfNOOOOOOOOOOOOOOOOOOOOOOOCNCNfNOOOOOOOOOOOOOOOOOOOOOOOO ^nnno\o^ol!^co«)a)oooooooc^lN^o^oN^^rgt^^(^lt^lNN^^^^HHHoooooooooooooolnLfllfilnlnlnal*(^(^

inLni/iLni/iinvOinirtirtiniiitninLnLOLnLnioininininiflin

c c c -H -H -H >, a) ai ai e > > > cccccccc (fl tfl (fl CO j-( i i i 01 01 -^ -H -H -H -H -H -H -H ai ai >>>>>>> a) a) .c ja JC £ -C43.C X! J5 J3 fi O rH lH H -P+J-P 0)01(0(033333333 N N N U ^ n u o U O 0 U 0 I I I uuuu(tio>o)(Daci)£.c<--irH<0(O(S(O(diyiTiTcr'CriTa)a)CQpa cccc 'O *0 T3 'O 3 3 <0 (0 (0 (0 3 3 3 3 3 3 3 3 3 3 3 0 0 0 0 0 0 0 0 o o £ & & & -H-H-H-H-H-H-H35•H -H -H -H -H -H -H 25 22 HHHHHHHH O O O O •P > > > N N nj CO (0 (0 (8 (fl (0 ooo » > > > > > > M l< u it I k k U U U U U m m (i» (» ci) CD 0) 0) C C C C 3 3 3 0) 1 1 d) d) CD o © 0) D> B> 0CT 0 > C 0* 0* IJI t7» tjt Q> 0) V O 0 O 0>0* 33333333hHMH - k U U 333W0>(U0)V(D(U ppppuuuu 0 0 0 43 £ £J -H -H •H -H -H -H -H •H -H -H -H M J^ hUhhili ftiCUO-CU&jfXOjC^OOOO J u o o bbbU U U tf Ptt os a o! « « « « BS « Q Q a a a a m n

cococor^t^t^r^onfi rHrHrHHHrHHOItMO)

n^cocyiOfH^ciMn^f^^io^ nsr ocoH'*ifio>OHr«i(*)^,rHO)(nHoifn^,mvor,«coa>OHOH(Ncn^,tn\ot^cooi

nnnnflrtnnnnnnnnn>HH^OtHH>>>>rtrtH*flonnnnno)(N(N(NnMrtO)MMnrtNNp)nnffln(nnnni,l > <>> 5> >>><<<<>>>> > >>>>>>>>>>>>>>>>>>>>>>>> Xi ^^o^H^wnwcofln^^Ol0^cooc^o^^*o^^^^M^MoonooHH(N^^f^^rH^oo\'J'^o^a^coffleo!,1fl(^l(Ofl,f|1^olfloo^ (I4 CI H rH r-f H HrHHrHr-tr-IHr-l H (N H rHHHHHrHeifir'inr'i HHHNH(NO)CNnfnrinriCM(N(NCNN(N(nc*lnNrinn(MrnN

(0 colXlO^OlrtHmf^J(NOconalw*H^^Ho^Hcoo^^^(^(^lrlnlnvo^Jl0^o^N^n(^o^o^^no^^r(^Iln^H^^oo^ocOln(,^^on OQ co^o^*^alcoMo^ooN^^lN^lno*^n«CTlHMM*HH^\o^^ol0^nco^coHrl(N(^lHOlOo\H(N\D(Nlno^mlncc^TlO ^^•co^eo^^ol^^oco^Ol0^coMMHvIl^•cooH^oo^vI)^]•^Do^^^^^co^oy)co^^^kD^>l^l^^^o^ou1^rs^^^•^^oo^)co r-tHrHrHHr-tt-HHrHrH r-t

,Q (NHco^(^lnNHNH(N^no*alHOMa30H^*^^^^M^lnnHMH(»HINO)Mcor^(Pcon^o^^1nr1lf)^(J\o* ^HVO^O 41 2 rH 0JHHi-Hi-IHHHr-l

V

M ^^^ln^^^^^^o(^^o^lnN«^MM^^colnl^ln^fOMooaDf^]^^H^*ln^QO^^oooo^^Nnn^o^p^^^oo^o^co^D^c^l N o^Hln^fflno^coM^co^ala\Hlrt^Nco(Jl^coo^o^rn^^f^on^»Hl^lHO^H^^Hno\DN^o^^ln(NHrl0^cl^nHHO H NH rH H HHHHH^fl rH <-H H H f N N N (\ (N H M 1*1 CM (M H JH in (S H H

t/i o^lfi^a)*^^^n*HN^eo*lnln^o^ONO^o*la^nn^^*Nlnrtlnoo^nlflln^NnH(0^ln*NOO^N(OH(^H^f^nnrl

H^inc0O[^r0»^C0*Cftinc0iNrNlrHr«J^>^inr«ia*rHrHt^<^OLnOC0OTrHrHOl^OC^^<^**[^ 5 f^lcoOOlflHH^gnHO^n(N(^nfl^^ln^^0^^^^f^^Hngo^CT^OHMO(NINHCo^ccoJ^^n^nt^l(^lHCD^nNn^r^H(NH HHHHHHrHrHlNH rHrHrHrHrHrHrHHrHH

3 \D ^0 in OOOOMOVNINNNONOHVD^ [v *3- O (^ CO O HrHHrH.-lrHi-HH.-HHrH,-HrHrH.-H.-| H (N vvvvvvvv vvvvvvv vv

-H CN m cocftin mnr^ooooooooooooooo^ocyic^HcnH co r^oo^Hincocor^^oincoi-tc^H^r^noinvoonin 5; in H "fiifliij^'iD^^^m^^^yJ'ii'O^H^NnH rH ^inrN«'*(vicyirn<,oj^'ryiri'*rn^f

^ M n in (Nf>jog^^cN^^HH HHHHrHHrHH HHr-lrH VVVVVVVV VVVV

> Tfoc^\ocNin^<^oi^^ocorHa3coawt^Lna3coracocomLn^i^rni^

HfHrHHH rH rH HrHOlrHrHrHrHCN IN NH rH

O ^^D^^co^^ool^^^OlO}aJoOlJlOcoN^^^OH^ooOl3llrlco(N^^D^^^^r,^n(^lC1C^ocoo^lIl^oo^^^o^Olo^wo^ol{^lHM(Jl (0 OOOOOOOOOOOOOOOOrHOrHHOOHHOOHOOOHOOOOOOHHrHOOrHOOOOHOOOOOOOOOOHOO aa ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo in H^a)o^(Nmm<^ooHHcoo^^*^M^o^oa)o^c^co^^^\o^^co^llflnlnnNHlno^Ncoo)co^(^l^^^oo^(^HOlcoo5*(aoO'* O HOOOHOOOrHrHHHOHOHHrHrjnjOHOOOOrHHHOf^HOHHHrHHrHHi-HrHrHfHOfSHrHHHHOOrHOOOHHOH CM Oi ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

O ONHnwHHoco^ln^^owMN^o*H^clr^lMa)o^^^n^^^^o^NOlnlnolc^llnw^HlnMWHco^w*cooow(^o^lnro K

d) nMBOiOHNniNn^n-f in* r> ^r ocoH^inaiod(N(n*H(NnHtMrTj,ininM»o\oHOHNr)^infiMOO> CL nnnnnnnnflonnnnn>HHOi>>>HHH**nrtnnnW(\N(NNNNNN(\i(,i(>iPiN(Nnr,irinnnrir)rin g > <>>>>>><<<<>>>>> >>>>>>>>>>>>>>>>>>>>>>>> 5 < << << < << << << < < < < < <<<<< < < < <<<;.< < <<<<<< 4 < << < (N nt^min^COOCOOlU1^^0^0^^rn0^tNO)(Nlf>jfnroOOOH,OiNOrHf-((Ntr)(T10vDvO(^(N Hr-iHCNOJ-HMcHf CNH^HOrHOrHHfN^r-^HHHr-tOOOO' OOOOOOOHOOOO

Q ^ricocNfNrnieroooc^ifi^omrifNjinr^rHnjfNioiN^ooooooooo^HtNinfN^os , , o offi a\rat*»[ »if)00ooir>ooooHOH^ fNojoo HOtNH(N(N0JrH01HHHOfNOOO(Tirgr>O0'Hi-lOOOOOOOOOOr-

HO^OH^HHOi-IOrHOi-tOOOOrHOOOOOHOOOOOOOOOOOOOOOOO

ooooooooooooooooooooooooooooooooooooooooo

•*njcor)-t'j,^ioj(j>incsn^,f>]nmooooHooooorHOi-i

O O H O O r HOHOOOWHOOOHOOOOHHOOOOOOOOOOOOOOOOO

OOOOOOOOOOOOOOOOOOOOOOUIOOOOOOOOOOOOOOOOOO OinOOOOOOOOOOOOOOf-tOOHOr-linOHOOinvPOOOOOOOCOOfHOH i^^^^ininotJiOo)(NH\o\oc^o^vDfOOHOt^^'r^

, > , cococococoooooooooinincocoirtifiif>if»ininoif)iriinif)r -f «r «.fs.r»1vo^i^)ooooo

ooooo^iNMOiMMMO(N(,Jooooomiincini*"i**oooooininin^'i,floo r--tN[^[^^[^c^^r^i^r^tvvor^r^oo'fi^vDiOcooorHrHrHHr-(ooooOHr-jTHrj(Mr>)nn r-\ r-i H HHHHr-tHr^H(NfMfMHH

cncricn^o^rHHHHHi-frHtfiHHOornnnoootnoooooooooc^r^r^ooOfn inininir>int^^cvr^t^c^t^^a>(^c^ooor«if^(Na^

333U-P-P+JWWUJI

ooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooinminininoooooooo

ooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooinmininirtoooooooo

r^^c^r^c^nn(^nrnrorirnr^t^cococoa)aooo(N(Ncy\cocococoHHHfHHOoooooif>'0 (N(NIN(NOlINN(NNNN0it0HHHHHHHHN(NNN(NN(,Jnnnni,inn(nNNNPgOl inu^ininininininiiiii^inininininininiriLninininLniriininuiioininifii^

01 11 utotototoxxx 01 0) +J JJ mmmmmsas 01 c c 3 3 3 3 3 m o o 0003mm rHrHrHr-i[flU il o o • OOOOI

OH(vi^rif>vor,«-cowojrH(N HM no ^ ^invOHtNfo^m tninir>^^iriinininininLn[^HrHo\ooooo<^OMcoCTiOrHrHrHrHcooococ)Ooori^rifiin\o >>>>>>>>>>>>>>>>>>>>>>>>>>>ncnrnfHrHrnnfor^(^rornf^^H00O^rHfHHH^inHif)invOvDr^rnrirHi-tfHfHHt> >>>>>>>><,^r,v[,^><> nj *cmr-.mH'#cor>t-i.-ir*-o f^H-**crnor^vooot-ir^rH(MooomcriOfHO

*-i OH^Hnn*ornorni^)(TiM«>^DO^DootNco^,\D\ocooo^,u,ivo N r^c>)HOnc0^ir>\0mc0rt**,v0ifi

h ^^vrnrv^tHcoinrot^in\o^ininou,>iTicO'i*'(Tio4rga\fiLn

^nnc^o^no\Hcomoi>ooata>\oin*org'*ot^c^a^[^coojHn[^cot^HvDr-oH(^tn o^o^MHOO)o)Mf^'j)nHt^'X>o)inr)ooo(^HvO[voococoH^r'*f^'*oor,v^j,t^\Doo HHHHHHHH HHH HOJHHHHHH H HH HHHHHnjHHfOrg

co r*« -* o\ •fO CMM

OH(NiDC0(MOH(Nr)y)H "ercOCOHHCOHCOcon^COCOCOCOCOCOCOCOCOCOCOCOCOCOCOCOCO H^inoOOifNfnr^OvOO^O^^^OC^Or^^r^vOlTlcNiDvDvDiXi^^^D^D^^O^OiCVO^Ov^^OkO HCS OlfMrOHOCM CNH H(N

O coot^cNo^t^r^o^ocrih'COHfsjcri^^DvDr^cocor^^ofO'^^'int^r^H'^r^in^^rt^r^'^'^n (0 OHOHOOOHOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO CQ OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO in o)o^coCT>o)con^c7iHH^^HOt^or^r^coa\aicoinrnoj<^inr^tNH^^inrir^nvo^f>)M O OJOfMH01NN(N(MHO)H(N(NHOOJ(NHHHOOHOOOOOOOOOOHOHOOOO fN Cu ooooooooooooooooooooooooooooooooooooooooo

O (ncnc0r^M^0^^^^*C0OV0t^H^^^CftC0vOf,,)lilC0H0V'*C0HriC0Cftr0vDvD'*Oin'*li)

oooooooooooooooooooooooooooooooooooooooooooooooooooooo

oooooooooooooooooooooooooooooooooooooooooooooooooooooo

oooooooooooooooooooooooooooooooooooooooooooooooooooooo

Oi-tOOOOOOH

o H oooooooooooooooooooooooooooooooooooooooooooooooooooooo

ooooooooooooooooooooooootnooooooooooooooooooooooooooooo OOOOOOOOOOOtNfNOOOOOinoOOrHOinoOOrHOOOrsJOOJOOOOOOOOCNOOOtNsO^rOiNOOrH H f-l

inmmoominmooocooofonm ooonnonnnooooooom^ininminiflojininininmininoo I o o o o o o o o

OOOOOOOOOOOOOOOOOOOOOOOOOO^HOOOrHrHr-tHrHHHi-(HHHOrgr-IHn^,^,^'\Oininoirt

oooooooooooooooooooooooooooooooooomminooooooooooooooooo oooooooooooooooooooooooooonooJCNr-tHrHrHrHHHfMOJNr-'r^r^h.inooooiriininoin HHHr-fHHHN.-t

^^HHlfllnHHHffl^^m^'0^^lnlnl/llnlnlr)^lno^a^lf)lnlfll^y)nfnr1lnlnln^nno»t^Dvo*o^J»(^f

aoaQa a aaa a aaaaa aooaaaaaQQaN NNaoaQaQOQaQaQQQ 00000000

^(|t^,*^cococofllnlnlnlnlnlnHWHHHHHHNN^tNo^l^o>Neo cooocococo(yv<^crcTiff*<^o^a\^^c^o^c^cnoio>rvt^r^(X>comMnrn[^cooooo«^o\ooooocor*-r^^ 000000000000000000000000000000000000000000000000000000 LnLnu^u^^iniriin^ininini^tnininu^iriLniniriiriu^inLniriLntn^ini^

lnlnln^^^^o^o^^^^^^^^«)^o^fl««3cocoeolflln^^^^co^**cM^NN^^^oooo'0^ovo«o^DH^^H^n^ i^ininirjinininu^ininintniniriuiioinininuiinifluiuiinininuiwiniiiinininin

c c ai ai a> 01 10 «« E EE c c a a s 3 3 -H -w (0 CO 10 10 (0 *0 *0 •PrH £ J3 -H r-l r-l U IJ 01-H UUHHH v 0 n « •P (0 0c •rt-d-HTiafll! •P 01 0> 01 3 0> •H 4J*J*J • 113 3 3 H M (0 -H •* -H 0 D» < 4-1 10 (0«00 0 1JO<00< Mlj Ml(03 10 hkh O C C O H 0 ^MMMlfl>i>.>i (0(0 < 0 a>a)a>coioss3iobL4C0 -H -i-< *H O 1-1 hhfaionoiaA>ji]U>J SJJJ LLI [N.cNfo^,o)M^,m>i>mcoo>o«H H N n ^ in *© [N ^ (\nnndrlHOKNinHHNN OJMMfJHrHH rr c inHiHHH-HHNOl HHHH (N ^ in von <• in rH t^coo^cNcivoc^cor^coox^(Xl^Nn^^co^^ow(^lIllfl*^^^lrl^o^^IN(^lOHHlNn'^'^n(o l^(,^HHHlnlnrtHH^lH^lnlnlfll^)^^HrlO>lfl nro(ninin^vo\o(NCNgr>]rnoooorN[^rs>r^r--'*^'**MMininirttf> ***'*'t^ncnN^^^"d,mnnrtfn^^*riin a nnnmifiioioyjWpgM, , , , n , , v e >>>>>>>>>>>>>>>>>>>>finn^ir*^^***nnn^ ^j'^ '!i''*< ^ 'n <<<<<<>>>>> > <> <<>> ^o^oirif^^rH^in^Hrg^£isro[^Hoico

n o m O •* CO *J- o o o f M)OMJi CM p- CO !H oj f\] ci MINiOrlH m ro in •* CM <*> H CI •£>

v r^inovo^inr>^t^sorvt^r>jir)i^oojr^H[^oHcj\inif)ir>cnri(7ir)r--vfi^ot--*£><£>vor*-r*- lJ^u)^oy)^lO^(0\ooo

r-rH H^^H^coiDfiin^oooiNifinajcD r^iD<,mcor--cNnco^'r^r^cocTicoinro^o lOrHcoin^ococom^'tD cr»o j HrHHr-lrHrHrHrHrHrHrHOJ rH rH (NH (N H rH rH rH rH C\J V V V V ^l»NHMW^onH^OHH'*o^ln(3^fvlv^coHOfn|^lo^^l"^'coHHH'fy:^HOlrlln^lna>lXlf^^l|ol^o^mo^)HW N (NN NHNNINMNniNnN NN -i H C* rH rH (N

coco*^^^co*cocol»cocon)^1ln^^,lD^^{l0^^^ocoH(,1HoooH^(>lHH^to^oH^r^Cl^oo^crl^o^(,n(^H HHHHHHHrHHHHHHHHHrHHHrHHHtNrHMriNMrvltNMr^OJrHr^MMrHHHMMMHrHMMHnirHHHHrn

^o^cocQincoo\M^rHocftincof^voa\ov©Hin\0rncoLnt^oci>\oc>r^v0roncft^[^coin^^H H(NHrHHHrHHHrHrgrHHM(NtNJ

HHHHrHHrHrHHrHHrHrHrHrHrHHrHHHHrHOJHrHrHrHHrHHHrHrHrHrHrHHrHHrHHOJHHHHHHHHHHHH

in r-- en \o ^ in *r rH HO) H CM m rH M

, 0 r)lflO*M^l*HOCOHIJlO)0^roo^^O^H^N'j*ifi^NnNinrif,in^NtM<,rHnNniii«NNHN*'*

U «oH(N^oOlfl«lnnoo^n^on(^n^NNNI^^NMN(NP^NINf^o]!srl(N(^l(,J(>llNNlNPJf^llN^l(^lr^^|(^l^l(>l CJ ^HMNHNHnNrO^tNcnril^CNHrOnOlCOCOCOCOCOCOCOCOCOCOCOCO

O oocooc^coocncoc^CTicncocnoogcftcocftocor^cnv0OOt^ococrico\O[^\ocot^cnr^ tO HrHOHOOHOOOOOOOrHrHOOOHOOOOiHHOHOOOOOOOOOOOOOOOOOOOOOOOOOi-1 CQ • oooooooooooooooooooooooooooooooooooooooooooooooooooooo in lflco^o^H^Dtn^oHO(PHaHa>HO^^OHW^r1^^lnu1Mco«^lfl^lnHO^o^Of^'flCHcolil^no^lnlnln'O^OlflTf^I) O OOOOrHOOOHHOHOrHOrHOOHHOOHOOOOrHOOOOOOHOiHOOOOHOOOOOOOOOOOO N CU oooooooooooooooooooooooooooooooooooooooooooooooooooooo

X ^y^^^^rl^ro^^^^^lnln^^^^ln^n^^^^^tfOO^^^roln^^^^^^^^^^^^(n(Nr^^•^^*^*v^^^,^•

OHCN I^OHfMOH(>iMOaH£IMOO> n <• UHO fl V IT) r^coo>(Nnvot^cors-cocftoinin*o[^n,fl,Lnvi)csr>Nr>)NOHrHfMn^c^nrinrHrHHininrHHHHHr>.ininininHHHCTvi- n ^^^^r^lnln^o^D^D^^o^(N(noooo^•^»^'^N^r-^f^^cNojininin• m t'f'jsr^'trtnN'isf'a'^^nndn^vii'HW m >>>>>> > >> >>>>>>>>>>>>>>>>>>>>> >>>>>> < < to Z CNHOOOHOOOHOOOOOOOOO

O OO^CO"*(nOOO*H Q O r*H * H rt CO CO(Ny5HHOJOOOHOOOOOOOOO u • • ooooooooooooooooooo vvv vvvvvvvvv

O C0MOOJC0«C0C0h>Os0Ottr-tt«tf»©C0f,» (T r-imHONrlOOOfnOHOOOHOOO ac * OOOOOOOOOOOOOOOOOOO O HHOfNHfNOOOOHOOOHHHHH C OOOOOOOOOOOOOOOOOOO OOOOOOOOOOOOOOOOOOO VVV V vvvvvvv rt •4iHO«^««(Nin«Nnin^a«ainin O in00

OOOOOOOOOOOOOOOOOOO OOOOOOOOONONOOOfNOOO HHHHHO) i-i H H H

oacooooifltrtii^^^iflintninininiAO

)

OOOOOOOOOOOOOOOOOOO inoiflininmNNfsiNfNoooooooo H^nnnrn^flvc^sc^rN^tvOOOOH H H H H H oooocofflooooONiNNnnnnm tnfNr-toooLflininmiDMt^voooooH

QQQQQQQQQQQQQOQQQQ

03C0OCN

r-vD^inLninininminLncococor^r^c^r^trt lOinininininLrtiDinininiDiAioirii/iifiintD

en x TJ 0) »3k 13 H ID (0HHH )g rHoirfQiajoi x x x x « •H>rHCCC 3 3 3 3 0 c ««HOOO 0000 0 « « «J S E S CCCHHHH< ty uoooflioioco.-j 4J 310 £££CCC C Cltlllt HHH-H HaiOlOOOHrtrt-IHlHM^O.0.0.0.3 ffl0 m j ai aaaao.333 o 0 j id(0(SiO(QoooQ)Q>a)a)A 1-1 ^1 co

E >>>>>> (0 i< i<; < < (< i< (0 ["-• ^ ^ H c\ n n r- OOJ n ^ n co • j in IN mm •**!•<"•> t H HOIH r r^ r» r» t*- V V V V

(0 ^*inmHvDO<^>Ht^mt^vo^OrHcooo(*ir«-r,«- •3 r-ojm^frncoinr^cNcTt^'ojcoc^rioojrHtyi OHOO^O^OCI3^0^*^OMO^O^OMI3vO^Jl H H r-t H

XI ^onr^r^oojfMxj'vDt^'S'CNr^m vom^fco

^i rNjr^nmoocot^miniNH^,tNr>'*c>]ot^ en e»t^m^,ri[^Mr-i(N'

morNtv\ofO'*o\ocooco'5rrH^oot/i'!i'H3 OHrs^srco^ocooh-oJnnmoonnH (NtNHrHHWiH(NH(NHHHHrHHiHrHH

O otno^oocodc^vot^iovor^iflr^vocovoia^' (0 -HOOOOOOOOOOOOOOOOOO 03 ooooooooooooooooooo m nm^*^j,'3'mimcNm^i'n'*^,nm'!rnnm O OOOOOOOOOOOOOOOOOOO CM ft ooooooooooooooooooo

O ffNH(n^rorococNm,^'OCT>^i,^,(NH[^mMOO*OO^HnM»CO(M K

0) vor»»oocj>t^m<,invDr^cooirH s >>>>>> IQ i< i< i< (< < (<; * * * * ***** ****** ******** ******** s-ss

•d TJ V Hi a Tj e 4-) O H •H rn •H c n t0 •3H V)a) J C *J W -H -H m -M o ii) o c IM (i) (t )) 4-1 • , t-l CM w 4J > C rH Q) O crt 0) rto 4) o V o ^! 4J Tl O 0 o c> o •a (1) •H 11 0 3 u t) H t-l n i 4J 3 W w •>H 0 n) •H ffla) JJ JJ 4J (0 4-) -H o m •l-l •H •r-l n Tj Tl Tl Tl Tj TJ cfl o rH rH rH 4J c c •H •H -H •H 0> D. * o o .- o * o o •- 4J * o o o • o C * o o o o (d * rH rH m 1 r-l rH 4J *

1 1 1 s*i 7.96E - 2.2E- 8 3E-6 ; 5.23E - 1.9E- 8 3.2E- 6 S3 W w o w w w 3 3 d u a to m a c * www w w W U J I o * H H si pj W H <3 c U WW •H * S£3 eye

* * * *

******* ***** * * * * •**TX-X-X*-**-J<**:-}<-f< * * * * * *

0) m 0) R)

•H •H JJ

cr> C •* Tl C rH rH •-, < * rfflH a) td 4-J * •H ^-- u-i bi J2 * s U u tiO* t/1 -H * •H -0 I-H a> w 1-1 * 3 -H •H iH k CM c s^* 4-( B 4J w 0) M « « w « - a * o o * crj CLOv O * >i O 0) 41 Q X t>< >< «_C M a a* * ^ tu >H 0O3<<< J3TJ0 Ml 5j rH O * OJJ S £ £ £ « C-H o a ' o* * C a) 'O'O'HXFN "iJ ,J Tj U"l CT« * •HOrHHwHX >H O P.rH OH* S3 HHOilll 0 W U> w c w B tJ H HI 0 • H M VlH 01 4JbCO I OOO I I I c JJ

•OtdrH OOOrHrHM 0) fH fl) aiw Hi (UUH 5 8 3 3P S H <• 4J WOW *** ****** *** HO* I S3 rH * rH OOO o •- OOO o ,0 (0 0. -H v-t •^OrH rHrH-OiHrHiH iTl tl W * i C C (C w H in * X5= XM X 2 2Z O i W W H n i-i W *

SI

************ ********** •S~J ISa ,3 1si ! H I i 41 £.5 5 a ^ s, 8 IS w M <* c y *-' MS£t •3 H * 8 0 « ft V "ft £-' U 41 A) J _ . „ 0 « o — Sir S" 'S * S m g e •« tt°»1 ^:? -sit 0 >H | U I M jj M 8 _ « JS rt • •H M G O.JJ to no pi >- w >: £ p-Oj- nj y > t/1 *-" W --I **********

C C H •HUM e N as U O 1 •B• * H " W O M z M O H QUUH 3*£ " w £ J S w iJ 3 H O w o ax c w W Q ********** •o op 0) G

41

3 a WW — UP 8)+<+«i« «OH aY u 3a a a as s •a'B H.i.HUBBayeu _j»

**********

a a

********** ***** ******** ******* ******

•H U) &.S, ur

^ "1,0 (1 (I o f» «fl c y II V u u 9 13 Tj flj 4J O 3 3 ;> p u

********** ***** *******

sisii .a: ******* Jut) (J 0)

O M

M rH O OJ 3 •l-l iJ w a o X M cfl o a (0 (w J 5 9 «4-4 (fl •-c o a a c ** ^c< * ' 5 i 2 M U O w w 09 • 0 a< •r* * CO • Q 4) 3 C- 3 rC OS l-l M CM 05 a at z ?> : X « 0) •o a* \o * -C o 0 « CM o w * > I c U * a. e 4J n > 0) c* -o O o > c + o cu c fl a c •H (4-1 O*S CO *H V i> •r4 (J rH 1 W sg M 1-1 o0 iS u -t N u W QJ O •r4a H + > OS w o > c •O rH co a XI 10 o in c \o (0 cfl •H W tt) > M *0 c u a cfl •H m a •P 5 > O OtS 3 0 4 a 0) 0 * a) CO B + W * 0 0 M DJ + * <-t M a co o o U-i O 0 4J C -C 0 X ai (4 •u 4J m CM CO > * 3 * 1-1 -0 0) M CO e H > cfl H •0 *w a 0 a 3 MM * (0 tJ O « •• "X •0 o •o 53 O.Oi o 3 W Q t/i o o- O tti in U H Sis OS r-l aM H £ Q (W — £i < wia s O (4 **-H ! •3* w • g ^ *? w oS oj oJ oJ «s VO > \o > VO M3 > yD l-l >J5 > M3 ai I W I W W W JH M * >M * • * * w * * * s * * * * ' CtJ *^v *^ > " > > > — M £ M l-l M r-i oS « oS oJ «! rH O > O CU > U ft-— > M [d l-l td OS 03 Di cq

•x * * -x

n a i 53, V A •- rH tj cfl A H H -. W Ol I 0 I oS * H o K H O •• H (1) oi « 55 =5 - > I CO | — > Ul 4-i w <; •o O O o o -H q Si ,1 •'• k S a) * > z OS OS Q 2 (H Z. , X - 4J o m q > > a i C rH O O rJ -1 os -2 os r-t M P Z c^ U H U (d S3 OS os o M •H 3 ^ 3 MOM o c C a a Cn fa b fa] Z rJ 1 Cfl n 1 > > < > S3 3 a W U i u oS MDd OJ OMS O 3* 5 z 2 H -K 0) cfl x u aI) b0

a) a .-t u r-t at n

T3 cd bO V

— A

M H n Cd CT J Q fti f) J a w UJ *• CM til U, I* ES Q id w Cd <: w UJ CM H u. (') cd w

o o cd w * * ** CM CO

*****

C J H cd CM O U Ptfl\ W O Cd £ P 9 HOOPild « bJ H Q A V

u cd

U M o o cd w ******** CM CQ 1 * * * * J I * t < !

: • I : -1

! ' t ; ' c

E < C

E ' I I ' I 1 < 1 • ' E 1 ' 1 1 < 1

1 *-• '1 E W 'I

• 'Hi :! ! o1 : i *J, •• i

* i -1 Q § § § - i * * * * *

*****

(A 4) >> 0> rJ J §•1 flU *W -H J3 41 W H •i-4 r-l *fi £1 J! r-t iH O y u 4-> 4) 4) « fl u k U H •M (0 H a r-l 4-» 4> fl fl U 41 'H 41 U Q 2 iJ -fl S 1-1 U U 01 gV] •8 0 MC w -0 4» « §5 ft. r-l (0 0 41 W 41 fl XI •C fl •i-4 OJ S U O P • U U S x" •-< a an >£ Ul fr» H 4) 10 «H H •• i • • 1 r>»H a«) W H 4u) H w >H 1 H 4! iH Ed 55 $£$ >. w) a JO u N P w J3 M > £> fl 4) ^T i-* 3 § • 4) 3 a- td O w A. »•< 2^ H U H m o t3- Ul O 4> K N G -H IH Ml O (f0 l O 4^ ^ tU -rj a IT> l-l o J oi u u g.H* (fl •3 g J O^ •W^ 4^1 0) z 9 UJ O H H g 3s3 sX J3 -H> 4) d.u- z (*i ^o •H 01 0-, a! u. Orf * 41 S °§ u u w w -a- CN Mfl 4) Q z fi Q o *_• io Q U * JS 4-1 F H o +J JC - ># >H fl o -? 1 m 1 in rs w LTI M * * * * UH * * * * * * tH SiJi £l£l ttd 6* w N M s(4 s wa s tafZ tku I *~' « Cd W • i H ~~g H M M M ) v 3

rt 4) a I 3 a a> u J= a 4) « •35 •f H .23 a M 4t 'H 4J ,.5 5 $!§* e. n a E 2U 4) W> 8 . « V C w W » ^ -H •3 u -o ft M O •» 3 pi U ^N W U 6 I 35S

:i

*****

4) cfl J3

"0 X w « « fl> 41 I-I U 0 O H O U tl c u H 3 1-) 41 a w r

w

*****

03 •5 S,

— «H t-l A o fa C/l P> - u o Ed l/l in 5 w W < O 3 H 4) Ed Ed Ed fa W Crf Ed CM Hi sfa w Crf HH CM O n M O M Ed H -g r-l W Ed U> to H S M Ed O i/i Id z fafa •• P W P Ed U-) Z erf < oi cd 11 erf a: u s~ * A H Eod w — •—- C/l I as fa C/l M (0 at U •• > z * erf o Q <-< erf o O cd P o cd fa CQ

•d ^ a) 3 •d ^ C QJ Q M m O m a) .. n ij •- w > .H -CO •r^ «j en • 4J -J • i — rH 3 V J-.SC- 3 Ed u c o- o m Ed t-i u Ed o CM A fd Ed CO i) 3 o< H 1 < a H j •H U) X - •'•3 § w O O U a* O 0) ••3 O 4 o\ a O (U z s .-—i E zd rH Cd §£ CN Ed Ed fa erf •• Ed Pn H fa — J Crf , M O r*i Ed c/l H *. A; O W H Z •-O.I3 M crf O I erf O P O W Z to 1 O Cd J ft. CO Ed >H * 1 fa M NS 5 5 £ § =i » a t-4 -H o o o & ~ §' 1 QHSU . 8225 H W H-| O •I•I Q MH O 33 H X o H 33 Q o <; o U O -tf o o a O *£ O U O o HrJCdrlO fO fcd H W Q fa H fa faH H HP2H faO CM N O W O O N O W O O N O W O O • H • H r^ . H • H • • H • H • o — •—- r^ H }H r-l >« iH >iH>4 ?H r-l >« r-( JM O o 55 Xj" II x" II X*" . H X 1 X 1 X H xf 1 x" i xf 5} fa fa fa o HUbl H OS CO

* * -x * * *

oi a - —

OS faOS CfaM O Si w W l-H s •J ,-J o U CM •X U bJ ti fcd O s o •K fa fa fa O O H M fa * O Cd W fa W w EJJ Czl u fa 03 W M >H H OS H P H o faH o c •• fa * 00 *J a> td * 43 -o * o OS O OS o I *•X * # -X * ** * — * * *,£. ft i ' §

i W Z H M H £S >>.«s JG § a§ ! w E as*4

£> w u (X (U U) IflU 1M «

H « U-l 3 W CM •-g s -ga •'•gSx'is : Mai M « 1 H ;i g| gH g§ j£ ge ge : g Of o o w (J. CQ

*****

a) t> rt C § S j: a) o •H 4J -H 0 4-* a £, w • 4) H u W W W)|-H

•rt -H « « M C (fl bfl 0

Q O ^ B U E3 S W ul •-) N N N N H u .. _; 8-; M.J3 W •H 0 _•x H in >* W POT) •• O - H U 0 n) X t/i X - w 8S. §£ ^ S. 8H. x a rt c ^ >—• w (d ^ [* [K O 0) o Cb (H io"o fn M tn j3 p- S g £ m [d a M H H o W •• fH •^ 4J (0 teo •H 0 O (N M MM Qi Li | Q N N EHS >-i •Hs . - o J K 01 ^> B0 o w —S w — —. in H Id u a u -H ai X -rt 0 H « OJ 06 H M Bi O •H 0 4J o O v m -bO -fa a w 4-> U tfl j u H CM b] Q X xi a "^ 1 o l-i M Qj a) m ' —• X 1 >H 11 U o C tn o r 2 6 O (n U •• W l-< ^ M t-i -H W H XH I XM I LIE F • H H w o • rH O si REL I LIE F REL I CM M M W CM 3S co w a) fO l-l tH H 5= J O -J ig a W ^ w Sg VO (3 •-! r-t «» 3 ° 1 ••Ba-'Sa- oS w H Z • Pi I-- CM . o m •• X w w x w w X H ro o !» + H O MW N i H H A H H ^ I fH Cd H H aj H H OS i oi o 1 M Cd C ***** 1 I O W i W * * * CM P * * * J : U-> PQ • N§h §g w w w w !" 2 rt sggsggs QB..*! U S£

M h W p M « -p a - + • .---MO •- •- M « O CJ .—. -J U1 ^.^H M «^h H a G w * •"§ * Cub-SwwBSS^ bS5l * 3- ^ a * ^EdWHffiOpHEW •*-* 0 MMWUoQfcUiUP 4-1 0 kJJHOOHEHOZ Cl •.-< r-4 + WWMWQMMbJi-i O Q -K Q) "-) Q aiSpoiJpsow w a M T) CdWfcdWWWWWW •H H M wwwwwc/iwww S33S33333 U * ouoououuu

Q n H