Functioning of a Small Experimental Watershed in theSerra do Mar, Brazil The need of a Multidisciplinary Approach

L.Barbiero', P.Curmi2, S.Furian,,3 F.C.S.Arcova4

, V.CicC03,4

1 - IRD-CEFIRSE, Indian Institute of Science, Dept. of Metallurgy,Bangalore 560 012, India.

2 - French Institute of Pondicherry, 11, St Louis Street, PB 33,Pondicherry 605 001, India.

3 - USP-FFLCH, Indian Institute of Science, Dept. of Metallurgy,Bangalore 560 012, India.

4 - Instituto Florestal de Sao Paulo, Rua do Horto, 931 - 02377-00 Sao SP - Brazil

Abstract

In the Serra do Mar region, in southeastern Brazil, few is known about the flows of water into the soi Icover. A good understanding of these flows is however essential for better management of (i) thewater that supplies the highly 'populated Paraiba valley and the city of Rio de Janeiro, and (ii) thewidespread landslide processes that are likely related to the slope feature. The analysis of a databaseof seven years of rainfall-discharge makes it possible to highlight the temporary storage of water intothe soil cover of an elementary watershed. A structural analysis of the slope morphology revealed thatthe soil mantle is mainly characterized by (i) a gibbsitic saprolite, (ii) various kalinitic horizons withinthe ibbsitic material, (iii) kaolinito-gibbsitic topsoil horizons. The morphology and dynamics of thissoil cover are discussed according to the thermodynamic stability of gibbsite and kaoliniteacompnying the solution percolation through soil profiles. The physical features of the soil indicatethat water is retained briefly within a microaggreated horizon during intensive rainfall. Because of theinclination of kaoJinite compact horizon, any excess water within it flows laterally downslope andaccumulates in the lowest part of the slope. This leads to landslipping, the main process of landformdevelopment in the region. The organization of the soil cover along the slope allows to identiy twomain water reservoir, one superficial, which chemical composition is likely in agreement with thekaolinitic and organic environment and the other one deeper, and likely at the equilibrium with thegibbsite. The next research should be focused on the Identification of the geochernical signature ofeach reservoir and their contribution to the waterflow at the outlet of the watershed at the scale of theyear and of the rainy event.

Introduction

In the eastern part of the Sao Paulo State inBrazil, the Serra do Mar region, covered withAtlantic forest, supplies in water both theParaiba valley and the city of Rio de Janeiro. Agood management of water at different scales iscrucial in this area with high density ofpopulation. However, data about thehydrological functioning of watersheds in thisregion are surprisingly scarce compared to thatof Amazonian watersheds. An hydrologicalbalance has been carried out on two small

experimental watersheds by Fujieda et al.

(1997) using ten years of field measurements

and hydrograph analysis. The results indicate

that 15 % of the annual rainfall is interceptedby the forest and return directly to theatmosphere while 85 % of the rainfall reachesthe forest floor where it infiltrates and remainsin the soil cover. The total volume of thestormflow is about 11 % and theevapotranspiration has been estimated as 15 %of the annual rainfall. According to this study,about 59 % of the rainfall is stored in the soilcover and contributes to the stream baseflow.Another distinctive feature of this region is thewidespread landslides that are reported almost

every year (Furian et al., 1999). Many studieshave involved statistical estimates of the

location' of landslides and the conditions of

climate, bedrock geology, slope steepness,

vegetation cover and human occupation leadingto their occurrence. This has shown thatlandslides occur principally in the Serra do Marcomplex, at the end of the rainy season whenrainfall is most intense. Guidicini and Iwasa(1976) showed that the probability oflandsliding is increased when rainfall exceeds250 mm in 24 h, regardless of antecedentrainfall. Cruz (1974) emphasized the role ofslope steepness: slopes exceeding 40 %encourage landslides, regardless of vegetationcover and human occupation. Landslideprocesses occur mainly on the lowest third ofthe slope. Small landslides occur in the middleslope and readjust the slope topography afterthe main landslides have occurred downs lope.These landslides contribute significantly tolandscape evolution in this region.

Although De Ploey and Cruz (1979), Furian(1999), and Fujieda et a1. (1997) concluded thatthe hydrology of the slopes should be studiedin more detail, the flows of water into the soilcover remains unknown. The objectives of thiswork are to identify the organization of the soilcover in this region, to understand water flowsinto it to relate these to landslide processes andto propose a route for future investigations onthe hydrology of this' area at the watershedscale.

Site

The studied area is located in the Serra do MarState Park at nos between the Serra do Marand the Atlantic Plateau in the eastern part ofSao Paulo State, Brazil (Fig. I). Uplifted sincethe Oligocene, these plateaux of 1000-2000 maltitude belong, with the Serra da Mantiqueira,to the uppermost surface of the blockmountains of southeastern Brazil's Atlanticmargin. They correspond to the headwaters ofthe Paraiba basin. From 800 to 1500 m altitude,the region is covered with a mountainrainforest of the "Atlantic forest" domaincalled Mata Atlantica. The tree layer of theforest reaches 20-30 m in height. The wettropical climate belongs to type "Cwa" ofKoppen (Furian, 1999) with altitude andorographic influences, and with rainfalls in allthe seasons. Annual rainfall ranges from 2000to 2500 mm, with a rainy season fromSeptember to March giving 71% of the annualrainfall. The number of rain days ranges from127 to 179, with a mean of 153 days between1983 and 1992 (Fujieda et aI., 1997). Thechemical weathering of the crystallinePrecambrian basement has resulted in sandyloam regoliths with a maximum thickness of 15m (Bigarella et al., 1965). Recent evolution ofslopes in the Serra do Mar has been mainlycontrolled by mass movements, includingslumping and planar slides in the regolith,rockfalls and rockslides.

23°S

Sao Paulo State

AUantlc Ocean

sokm

Fig. 1: Location of the studied site

"287

/N

~5m• Ralngauge

• Gauging station

~ Riparian area

• Detailed 5011 study

- Soli toposequences

200 m

Fig. 2 : Morphology of the watershed and location of the soil study, raingauges and gaugingstation

The fieldwork was concentrated on a small (56ha) watershed of the River Paraibuna in thedistrict of Cunha, known as the "D" watershedin the Serra do Mar State Park, and managedby the Instituto Florestal de Sao Paulo (Fig. 2).Cunha is part of the granite-gneiss highplateaux of the Serra do Mar. The watershed ischaracterized by steep hills lopes (46% onaverage), a large riparian area and a Northwestaspect. The mean channel 'slope is 19.9 %. Thewatershed is covered with the rain forest.

Methods

Rainfall and discharge: This watershed isequipped to acquire hydrological data (rainfalland discharge) for more than 20 years, Rainfallwithin the watershed was measured with 4tiping-bucket raingauges. The actual rainfallwithin the watershed was calculated by anarithmetic mean method (Shimornichi et al.,1987). Streamflow at the outlet of the

watershed was measured with a trapezoidalflume with a stiIIing pond equipped with acontinuous stage recorder. The rating curve forthe flume was determined by current meteringat different stages at the gauging station (Ciccoet al., 1987). Seven years of daily rainfall­discharge measurements on the D watershed,ranging from November 1986 to October 1993,were selected for this study. Both rainfall anddischarge events were shared out into IIclasses «4; 4-6; 6-8; 8-10; 10-15; 15-20; 20­30; 30-40; 40-50; 50-100; >100 mm). For eachclass, the time of return was defined as:Time of return for class C = (number of days ofthe collect)/number of days with event in theclass C),

The times of return of both rainfall or dischargeclasses were compared.

In a second step, considering that about 20 mmof rainfall could be stored into the soil coverwithout significant contribution to the

288

stormflow, daily rainfalls were corrected asfollow:

Rainfallcorr. = Rainfall - 20 mm.

The times of return of both rainfallcorr. ordischarge classes were compared.

Soil cover: The soil mantle organization wasreconstructed along four sequences located onthe crest line and on the slope. The mineralogywas studied by X-Ray diffraction analysis(XRD) on randomly oriented powder andorientated deposit of the <Zurn fraction, whichwas studied in 5 conditions: K saturation, Mgsaturation, Mg saturation and treatment withglycol ethylene, heating at 450°C and 550°C.Undisturbed blocks were collected ar.dimpregnated with an acetone-diluted polyesterresin (Scott-Bader Crystic) after dehydration byacetone exchange, and vertical thin sections (70x 110 mm) were made. Some samples werecollected for SEM (XL 20 Philips at 15 kV)observations of the morphology of weatheringfeatures at the crystal level and punctualmicroanalyses (LINK Analytical eXL energydispersive X-Ray system).

In order to quantify the macroporosity, afluorescent dye was added to the resin (CibaGeigy Uvitex OB). Images from thin sectionswere captured by reflected DV light (Hallaireand Curmi, 1994), which caused pore space toappear as bright area on the dark background.Images were digitalised using the Visilogsystem with a spectral resolution of 256 greylevels and a pixel size of 10 urn. The poreswere identified by simple threshold partition.Overall macropore connectivity in each thinsection was classified visually as low, mediumor high.

The pore size distribution, between 0.0037 and100Jlm equivalent radii, was studied with aCARLO ELBA 2000 mercury porosimeter usingair-dried centimetric aggregates outgassed atroom temperature. Water retention curves weredetermined on 5-10 ern' samples by sorptionthrough membrane filters according to Tessierand Berrier (1979) for the low suction range(pF 1-3) and by pressure membrane equipment

for greater suctions. The shrinkage of the soilsamples was slight, so the shrinkage curveswere estimated from the apparent volumes atpF 2 and 6, using the kerosene methoddeveloped by Monnier et al. (1973).

Total analyses have been performed afteralkaline fusion using lithium tetraborate. Fe, Aland Ti were quantified by colorimetry, Ca, Mgand Mn by atomic adsorption, Na and K byemission. Si was quantified by gravimetry.Total chemical analyses of the secondaryminerals produced by weathering have beenperformed using a three acid digestion method.Semi quantitative estimations of the abundanceof secondary mineral phases of gibbsite andkaolinite were calculated from the Ki indice(Ki = Si02/ Ah03) according to Pedro (1966).

Results

Hydrology: The relationship between the timesof return for rainfall or discharge for each classin presented in Fig. 3. The probability ofrainfall in the class <4, 4-6, 6-8, 8-10, 10-15,and 15-20 mm is almost similar. Therefore, thetime of return for' rainfall events of theseclasses is constant. For classes higher than 15­20 mm, the time of return increases. On theother side, the time of return for dischargeincreases continually with the rate of the class.For the higher classes (10-15 mm to > 100mm), the time of return for rainfall increases 10times more rapidly than the time of return fordischarge.

After correction of the data, and in order totake in consideration that about 20 mm ofrainfall could be stored into the soil cover, therelationship between both rainfall.g., anddischarge times of return changes and increasesproportionally to the unit slope with the rate ofthe class.

Catena organization: Figure 4 shows thedistribution of soil horizons in sequence 2,which can be divided into two zones, theups lope and downslope domains.The upslopedomain consists of a 12m thick ferrallitic soil.It is mainly composed of a sandy gibbsiticweathered material derived from the parent

289

rock (G Horizon). The original mineralstructures and that of the rock itself arepreserved in gibbsite pseudomorphs, whichresults in a gibbsitic horizon with a box-workmicro-structure. The polysynthetic twinning ofplagiuclases and the cleavage of amphibolesare also preserved in the gibbsite, indicatingthat the formation of gibbsite was the first stagein weathering of the parent rock. Only quartzand muscovite were unaltered at this stage. Thestructure of the parent rock is also preserved inmillimetric gibbsitic veins that probablyinfilled fissures in the parent rock. The totalporosity of this horizon is about 40%,exhibiting a bimodal distribution (Fig. 5). On

thin sections, the macroporosity of this horizonis about 25 % with medium connectivity. Thisgibbsitic weathering level is overlain by apebbly and blocky' horizon composed offerruginised gibbsitic material. The structure ofthe parent rock is still preserved in the blocks.Although the soil fauna has burrowed into thesuperficial horizons, it is still possible toidentify relics of the millimetric gibbsitic veins.The proportion of kaolinite increases towardsthe top of the profiles and reaches about 60 %of the secondary minerals in the superficialhorizon.

10000 ..,-----------------------,

1000

100

10

<4

10

8-10

100 1000 10000

Time of return for discharge (days)

10000..>.III

~

01000

u

:!e 4-6e <4

100 • •...2E~~ 10.... Unit slope0

Gl

EI-

10 100 1000 10000

Time of return for discharge (days)

Fig. 3 : Relationship between the time of return for rainfall and discharge (a) and rainfall correctedand discharge (b) for each class.

290

o

10m

10m

. " u erficial horizons;e 2 1 - Kaolmltlc. s ~bbsitic horizon. n toposequeDc... rizons within gi .. horizons alo g K olinitic ho I e domam.. . fthe mam mh2); 3 - a . _Downs op. 4 Distribution 0 . Ons (mhI and _Parent rock, 6 . . that

FIt MiCrOaggregate~.~o~ibbsitic Horizons; 5 f the G material, in~l:':f and

Parent roe , . truncates that 0 d first from the . e terial.

domam, a r develope . kaolinitic maf the upslope . (here the latte hanged into a . solution of

I

n the lower part 0 kaolinitic hOflZohn lZ'on subsequentlYtacct evidence of dld

s

by SEM.I loam . the G or. this con, . bserve .

compact c ay n) occurs within t 85 % of the At and gibbsite IS 0 f gibbsite mtocalled K ho~lZoofkaolinite is abhou . on has a quartz~ re resilication 0 tulated in thisThe proportion rals This K onz Fi. 5), There 0 , has been pos 2002). In thinsecondary nune ti~n of pore sizes ( 9g% of Kaohmte t (Furian et al., to have arisenunimodal distribu . ely microporosity ( h the envtroDmthen mhl horizon see~b itic box-work' 1m st entir ) Althoug . ns e th gib SI .which IS a. 0 ded by mercury" the image sectto; destruction of e mh2 horizon IS

the pores mtru eaches 51 Y" rosily of from th whereas the .s burrowed by thetotal porosity r

ld that the macropo ith low structure'd f biofabrics and I f kaolinite reach. evea e 93 % Wl mpose 0 rtions 0 dprocessing r . only abou'.. d brinkage co The propo I, f the secon ary

the K ho?"on IS water retention an s almost sod fauna. 40 and 55 Yo 0 ated horizonsconnectivity. Th~hat this K horizon '\ongue- respectlvel\hese microaggr~fslribution(FIg.

curves ind~capt~3 (87 %, Fig. 6~~t :~perfiCial ":~i~ls~ bimodal pore :~:nd inter-aggregaltesaturated a. cated by mhl and e ding to mtr . (Cumn et a .,shaped and I:gg':;:ated horizo;(the slope 5), corre:~o~ther ferrallitics

O:::'ut

36 % with

porous nucro

are

conformable Wl kaolinitic pore4s)

aThse

macroporosity ISCtiOD (pF3), water

)

These lay loam 2 199. " At low su f themh2. by A second c downslope at - hi h connectivity. f the pore space 0~,;;r::P(K' hOrizon)t::et'::: sandy G hor:~~ oc~upies only ~Oh:riz':,ns (Fig. 6).depth, also WI h . on. Detatled ~ microaggregate2.5 m h the first K onz d G honzons

but beneat t between the K an K horizonf the contac tructure of theo d that the sshowe

291

0.04

0.02

0.03

o

0.' OIl--r'l0.7 E

u

0.6 E.:0.5 0~

0.4 f·0

0.3 Cl.Gl~

;;0.2 ~

=0.1 E

=u

0.06

0.01

Gib bslllc h0 rlzo n

0.07

OIl--E n.osU

Gl

E=e~

~oI:l.

0.07Kaolinlllc horizon

0.06

~E 0.05

uGl 0.04E.:e 0.03~

Gl; 0.02

I:l.0.01

o -I----.....eIIlt

0.' OIl

--r'l0.7 E

u

0.6 E.:0.5 0~

0.4 ~

e0.3 ~

~

;;0.2 ~

=0.1 E=

o U

0+-----.-...

0.04

0.05

0.' ~r'lE

0.7 uGl

0.6 E.:

0.5 0~

0.4 fe

0.3 Cl.Gl~

0.2 ;;~

0.1 Eo U

Mlcroaggregated horizon

0.03

OIl 0.06

--r'lEuGl

E=e~

f 0.02e

I:l. 0.01

0.07

-3 -2.5 -2 .1.5 -I -0.5 0 0.5 1.5

log (Equivalent radius, flm)

Fig. S. Mercury porosimetry curves ofgibbsitic (G), kaolinitic (K), and microaggregated (M)horizons: differential and cumulative pore volumes shown as a function ofcalculated pore radius

On the middle part of the slope, the upslopesoil cover is abruptly truncated over its wholethickness by the material of the downslopecover, which clearly differs from that of theupslope one. The downs lope domain (notdetailed in Fig. 4) is composed of superficialhorizons overlying a clay horizon with many

randomly oriented blocks. The clay horizonrests directly and abruptly on the hard bedrock.Remnants of box-work gibbsite occur, but arefrequently associated with unweatheredminerals such as hornblende, other amphiboles,microcline and biotite.

292

K borlzon M horizon

6532

o +---r-----.------.--r---......------.

o

1.4

1.2

0.6

0.8

0.4

0.2

65432

O+---r----r-----.------.----,.--;'

o

1.2

1.4

0.8

0.6

0.4

0.2

~........~..~'I:l ~C ....~~. .i. ;.. I.. >i'"'

Soli water tension, pF Soil water tension, pF

Fig. 6. Water retention and shrinkage curves of kaolinitic K and microaggregated M horizons.

One of the main features of the whole. soilsystem is the presence of kaolinitic horizons (Khorizon) into the gibbsitic weathering material(G horizon). According to the morphologicalstudy of the 3 other toposequences, the limits

(--:~~

~ NORTH

01_2IT j 3

0 4

of the K horizons and of the colluvium on theslope are drawn of Fig. 7. The distribution ofthe K horizons is complex; they occupy a largepart of the soil mantle, but are truncateddownslope by the colluvium.

080

Fig. 7 :. Distribution of the K and G horizons in the upslope domain, and areas of the downslopedomain. Map based on sequences 1,2,3 and 5.

1 - G horizons; 2 - K horizons within G horizon; 3 - Downslope domain; 4 - Toposequence.

293

Discussion

Soil cover organisation and mineralogy: Theupslope soil cover is derived from the in situgibbsitic weathering of the bedrock, thestructure of which is preserved in the soil, evenup to the surface horizons where gibbsitic veinsoccur. Therefore, three characteristics of thissoil cover do not match with the normaldistribution of gibbsite and kaolinite in lateriticprofiles. They are:

I. A bauxitic weathering of the parentrock, without kaolinite intermediateformation;

2. the presence of kaolinite in the topsoilhorizons;

3. the presence of kaolinitic compactclay loam horizons into the gibbsiticsaprolite.

Various observations in the region aroundCunha indicate that this type of soil cover iswidespread in the Serra do Mar. Therefore, therelative stability of gibbsite and kaolinite isdiscussed below.

The gibbsitic saprolite: Although the parentrock is rich in quartz, a gibbsitic weathering isobserved where we could expect kaolinite. Thegibbsite-quartz association is however possibleif the kaolinite stability shifts from K I to K2 asindicated on Fig. 8. KI value corresponds to arelatively well-crystallised kaolinite andwithout impurities. In the kaolinitic clay loamhorizon, the low discremination of thesecondary picks of kaolinite by XRD analysisindicates a low crista 11ini ty. Therefore, thegibbsite-quartz association (0 on Fig. 8) ispossible.

::cc,

M

+,..-.., 8+

<.-.--'

~"-"OI)0

--'

4

-7

Gibbsite ,E ".. ,~__ J

A' ---------,

log (H4 SiO4 )

-4

Fig. 8. : Advance of the soil solution in a gibbsite-quartz-kaolinite equilibrium diagram. A, B, C:The soil solution leaves the system before the equilibrium with respect to kaolinite being reached:

E, F: Advance considering the role of the complexing organic matter(modified from Lucas et al., 1996)

294

The gibbsite-quartz associanon can also beobtained by considering a fast renewal of thesoil solution, under conditions of humidclimate and good drainage. This is particularlypossible, as the kinetics of dissolution of thequartz is very slow. The solutions reach theequilibrium with respect to the gibbsite (B onFig. 8) but leave the system before the reactionof dissolution being finished, and before thedouble point of the equilibrium with respect togibbsitelkaolinite being reached. In this case,the profiles are, to some extent, truncated bythe bottom, the depth horizons classicallyexpected in a water percolating model, are notformed (Fritz, 1975), and gibbsite-quartzassociation is observed in the soil.

The kaolinitic topsoil horizons: The proportionof kaolinite is higher in the topsoil horizon thanat depth, in the sandy gibbsitic material. Thisphenomenon, which today could appearclassical, has been particularly studied inAmazonian forest environment of Brazil(Lucas et al., 1993). These authors have shownthat the forest recycles significant quantities ofchemical elements, particularly Si and AI,maintaining a dynamic equilibrium and causingthe stability of kaolinite in the topsoil horizons.A similar recycling is probable under thecoastal Atlantic rain forest, which still cover alarge area of the Serra do Mar.

The compact kaolinitic clay loam horizons:The presence of the compact kaolinitichorizons within a gibbsitic weathering materialsets more problems. It cannot indeed beattributed to a recycling of solution enriched inSi and Al by the top of the profile because ofthe geometry of this horizon, being intercalatedbetween two gibbsitic horizons. Theexamination of the contact inform us that thedevelopment of kaolinitic material at theexpense of the gibbsitic one is today in process.

The transition from the gibbsitic saprolite to thekaolinitic clay loam horizon corresponds to ahigh increase of the volume of fine pores, with

295

an equivalent diameter around 0.3 urn (Fig. 5).The gibbsitic horizon exhibits a strongconnected macroporosity, permitting a quickpercolation of fresh solutions, undersaturatedwith respect to quartz and gibbsite and able todissolve these minerals. In the compactkaolinitic horizon, the solutions are sloweddown and/or concentrated under the effect ofevaporation by the forest, and kaolinite canprecipitate. This assumption can explain theself-ups lope development of kaolinitic clayloam horizon. The phenomenon can be initiatedby the numerous micaceous layers, which wereobserved in the gneiss saprolite, the micasbeing transformed into kaolinite, contrary tothe feldspars, which are transformed directlyinto gibbsite. On the other hand, thisassumption does not make it possible to explainwhy the dissolution patterns of quartz andgibbsite is observed just ups lope of thekaolinitic clay loam horizon.

The role of organic matter is also determinantin the solution-kaolinite equilibrium conditions.The complexing organic matter formed by thebiological activity in the topsoil horizonssubstracts a large proportion of Al from thesolution as it has been observed in AmazonianOxisols (Cornu, 1995, Eyrolle et al., 1996).These complexes are retained onto material ofhigh surface exchange and cannot migrate outof the system. Between the rainy events and atdepth, due to the mineralisation of the organicmatter and the increase in Al contents' and pHvalue, the solution reached the equilibrium withrespect to kaolinite (E, F on Fig. 8) (Lucas etal., 1996). This process could also explain theintermediate kaolinitic clay loam horizonwithin the gibbsitic saprolite.

Identification of the landslide risk: Theobservations and measurements emphasize thecontribution of weathering and pedogenesis tolandscape evolution. Fast circulation of water ispossible into the G saprolite according to theconnectivity of the macroporosity. Thedissolution of gibbsite and quartz and the

formation of kaolinite generate tonguesarranged along the slope like tiles on a roof.During the fieldwork, waterlogging afterrainfall was observed in the M horizonsoverlying the K horizons, what is consistentwith the severe decrease in the vertical

infiltration rate at this contact, and influencingthe flow of water on the slope. A three-stagemodel of increasing development ofK horizons(Fig. 9) is proposed. In the first stage (a), theK horizons are thin and separated and vertical

(0

• 1

• 2 -.- 4

~ 3~

4Ilt 5...Fig. 9. : Model proposed for the contribution of weathering and pedogenesis to landslide

processes. a) Vertical drainage predominates - low risk oflandslide on the slope; b) lateraldrainage increases - medium risk oflandslide on the slope; c) lateral drainage predominates - high

risk of landslide on the slope. 1 - K horizons; 2 - M horizons; 3 - Water flows; 4 - Lateralevolution of the K horizons; 5 - Landslide.

296

drainage in the sandy gibbsitic weatheringmaterial predominates. Little lateral flow isobserved and the water infiltrateshomogeneously along the slope. The risk oflandslide is low. In the second stage (b), the Khorizons are more strongly developed and theproportion of lateral subsurface drainageincreases. The risk of landslide then alsoincreases. In the third stage (c), the K horizonsare strongly developed and the subsurfacelateral drainage predominates, canalising thewater downslope during intense rainfall. Thewater in the macroporosity of the M horizoncan create a consequent overload in the lowerthird of the slope. The mass of retained water,within a 0.3m thick microaggregated horizon,increases from 100 kg/m' at pF3 to 200 kg/m'at pF 1, increasing the slope instability. The riskof landslide is then very high. As the Khorizons evolve laterally and upslope at theexpense of the G horizon, the mass of waterana the risk of landslide increase with thedevelopment of this singular pedologicalsystem.

Hydrogeochemical model and proposal for thenext research: The non linearity of therelationship between the times of return forboth rainfall and discharge indicates a structureof the watershed and a probable storage ofrainwater into the soil cover. For rainfall eventsbelow 20 mm the stormflow remains low ormoderate. The increase of the time of return forrainfall.g., and discharge proportional to theunit slope. confirm the growth estimate of thewater stored in the soil before the contributionto the stormf1ow. The identification of the soilorganisation makes it possible to propose amodel for hydro geochemical functioning ofthe watershed. The soil cover is a substantialwater reservoir according to the thickness ofthe weathering saprolite. Two argumentssupports that the time of persistence of thewater is low. The first one is that the solutionsleaves the system before reaching thethermodynamic equilibrium with respect tokaolinite. The second argument is of physicalnature, the bimodal pore size distribution andthe connectivity between macropores of thegibbsitic saprolite allows a fast circulation ofthe water. This deep water, which likely exhibit

a chemical signature of the gibbsiticenvironment, provide the baseflow all along theyear. A second reservoir occurs temporarilyinto the superficial microaggregated horizonswhere the compact kaolinitic horizon preventthe vertical flows. Indeed, the water retentionand shrinkage curves indicate that almost 50 %of the porosity of the M horizons could bebriefly occupied during intense rainfall. Thehypodermic flows contribute to the stormf1owas a supplement to the runoff over the saturatedareas and likely exhibit geochemical features ofthe kaolinitic and/or organic environment.

This hydro-geochemical model has to bevalidated by a coupled hydrological andgeochernical monitoring of the slope (Currni etal., 1998 ; Durand and Juan Torres, 1996). Ourpresent knowledge of the soil coverorganization will guide for the location of tl.­piezometry and tensiometry measurementstations along the slope. One station will b.located upslope for the monitoring of thevertical transfert and the quality of the deepwater reservoir. Another one will be located inthe middle of the. slope, where kaoliniticcompact horizon occurs, in order to check thesuperficial water flow and its chemicalcomposition.

Conclusion

The management of the Cunha watershed in theSerra do Mar region is based on amultidisciplinary approach. The hydrologicalstudy indicates that the soil clearly contributeto the water storage during rainfall and alongthe year. The pedological approach based onthe structural analysis of the soil cover makes itpossible to identify the organisation of the soilmantle, and based on the mineralogicalfeatures, to understand the genetic relationshipsbetween the different horizons according to thequality of the percolating waters. The study ofthe physical features of the soil horizons allowsto identify the potential reservoirs and theprobable ways of water flow. The next researcheffort should focus on the identification of thechemical signature of both deep and superficialreservoirs and on their contribution to thedischarge using hydrograph deconvolution.

297

This work was supported by a CAPES­COFECUB Cooperation No 35/87.

References

Bigarella, J.1., M.R. Mousinho, and J.X.Silva.I965. Consideracoes a respeito daevolucao das vertentes. Bol. Parana.Geogr.I6117,85-116.

Cicco V., Emmerich W. and Fujieda M., 1987.Determinacao da curva chave do vertedouro dabacia hidrografica experimental D, no ParqueEstadual da Serra do Mar-Nucleo Cunha, SP.Bol. Tecn. I.F. 41, 79-96.

Cornu, S., 1995. Cycles biogeochirniques dusilicium, du fer et de l'aluminium en foretamazonienne. Ph.D. Thesis, University Aix­Marseille Ill, 411 p.

Cruz, 0., 1974. A Serra do Mar e 0 litoral naarea de Caraguatatuba. Tes. Monogr. 11, IG­USP.

Currni P., Durand P., Gascuel-Odoux c., MerotP., Waiter C. and Taha A., 1998.Hydromorphic soils, hydrology and waterquality: spatial distribution and functionalmodelling at different scales. Nutrient Cyclingin Agro-ecosystems 50, 127-142.

Curmi, P., Kertzmann, F.F. and Queiroz Neto,l.P., 1994. Degradation of structure andhydraulic properties in an oxisol undercultivation (Brazil). In: Ringrose-Voase A.l.and Humphreys G.S. (Eds), SoilMicromorphology: Studies in management andgenesis. Proc. IX Int. Working Meeting on SoilMicromorphology, Townsville, Australia. Dev.in Soil Sci. 22, Elsevier, Amsterdam, pp. 569­579.

De Ploey, J. and Cruz, 0., 1979. Landslides inthe Serra do Mar, Brazil. Catena 6, 111-122.

Durand, P. and Juan Torres, J.L., 1996 ­Solute transfer in agricultural catchments: theinterest and limits of mixing models. Journal ofHydrology 181, 1-22

Eyrolle, F., Benedetti, M., Fevrier, D. andBenaim, J., 1996. The distribution of colloidaland dissolved organic carbon, major elementsand trace elements in small tropicalcatchments. Geochimica et CosmochimicaActa 60 19,3643-3656.

Fritz, B., 1975. Etude therrnodynamique etsimulation des reactions entre mineraux etsolutions. Application a la geochimie desalterations et des eaux continentales. Mem, Sci.oeei 41, 152 pp

Fujieda M, Kudoh T., Cicco V. and CarvalhoJ.L., 1997. Hydrological processes at twosubtropical forest catchments: The Serra doMar, Sao Paulo, Brazil. Journal of Hydrology196,26-46.

Furian S., Barbiero L., Boulet R., Curmi P.Grimaldi M. and Grimaldi C. 2002Distribution and dynamic of gibbsite andkaolinite in an oxisol of Serra do Mar,southeastern Brazil. Geodenna, 106 (1/2), 83­100.

Furian, S., Barbiero, L. and Boulet, R., 1999.Organisation of a soil 'mantle in tropical south­eastern Brazil (Serra do Mar) in relation tolandslide processes. Catena 38 (1), 65-83.

Guidicini, G. and Iwasa, O.Y., 1976. Ensaio decorrelacao entre pluviosidade eescorregamentos em meio tropical umido.Publicacao IPT 1080, Sao Paulo, 48 pp.

Hallaire, V. and Curmi, P., 1994. Imageanalysis of pore space morphology in soilsections in relation to water movement. In:Ringrose-Voase A.l., Humphreys G:S. (Eds),Soil Micromorphology: Studies in managementand genesis. Proc. IX Int. Working Meeting onSoil Micromorphology, Townsville, Australia,Developments in Soil Science 22, Elsevier,Amsterdam, pp. 559-567.

Lucas, Y., Luizao, FJ., Chauvel, A., Rouiller,1. and Nahon, D., 1993. The relation betweenbiological activity of the rain forest and mineralcomposition of soils. Science 260, 521-523.

298

Lucas, Y., Nahon, D., Cornu, S. and Eyrolle F.,1996. Genese et fonctionnement des sols enmilieu equatorial. CR. Acad. Sci. Paris 322,Iia,I-16.

Monnier, G., Stengel, P. and Fies, r.c., 1973.Une methode de mesure de petits agglomeratsterreux. Application a I 'analyse des systemesde porosite du sol. Ann. Agron. 24, 533-545.

Pedro, G., 1966. Interet geochimique etsignification mineralogique du pararnetremoleculaire Ki = Si02/AI203 dans l'etude des

299

laterites et des bauxites. Bull. Gr. Fr. Argiles18,N.S.13,19-31.

Shimomichi, P.Y., Cicco, V., Arcova, F.C.S.and Faria, A.J., 1987. Correlacao entre metodosde calculo de precipitacao media mensal embacia hidrografica experimental. Bol. Teen. I.F.41 (I), 1-26.

Tessier, D. and Berrier, r., 1979. Utilisation dela microscopie electronique a balayage dansI'etude des sols. Observations de sols humides,soumis adifferents pF. Sci. du Sol 1,67-82.