Formation des crues «cévenoles» dans des bassins élémentaires du Mont Lozère Flood generation conditions in the Cevennes (southern part of french «Massif Central»)

Résumé

Les théories classiques concernant les processus générateurs des écoulements rapides permettent mal d'expliquer la soudaineté de la montée de crue dans des petits bassins versants élémentaires, situés sur le versant sud du Mont Lozère (sud du Massif Central, France).

Des études portant sur les conditions de genèse de ces crues « cévenoles » ont montré les conditions particulières d'apparition de tels événements: dans ces milieux de moyenne montagne granitique méditerranéenne, les formations superficielles sont peu épaisses et très filtrantes et les abats d'eau considérables; les surfaces saturées, seules capables de générer un ruissellement important, progressent, contrairement à ce qui s'observe habituellement, d'amont en aval des versants; ce type de fonctionnement empêche dans un premier temps toute arrivée d'eau jusqu'au drain principal et ce n'est que lorsque, de proche en proche, l'ensemble du bassin est quasi saturé que la crue se déclenche, entraînant immédiatement des débits importants.

Ce modèle de fonctionnement hydrologique se vérifie bien lors de la crue du 22 septembre 1992, responsable de nombreux dégâts et de pertes en vies humaines dans le sud de la France.

Mots-clés:

• Processus de genèse des crues,
• aires contributives saturées,
• crues « cévenoles »

Abstract

The southern and western edges of the French « Massif Central » are frequently subject to very high floods, known as « crues cevenoles ». They mainly occur during Autumn, and they are characterized by particularly sudden and high flows. Genesis conditions for these types of floods have been studied in a small experimental basin located on Mont Lozère.

The Latte River basin (20 ha) is located between 1200 and 1400 m above sea level. Bedrock is granite, covered with sandy, weakly weathered soil, ranging in depth from 0 to 1 m, generally thinner on the upper part of the slopes. Infiltration capacity is very high, generally over 70 mm .h^-1 (and locally more than 135 mm • h^-1), as indicate by rainfall simulation (COSANDEY et al., 1990). The basin was previously covered with a 70 year old spruce forest but the forest was cut down between 1987 and 1990, without evident effects on flood genesis conditions (COSANDEY and BERNARD-ALLEE, 1992; COSANDEY, 1993).

Rainfall is recorded by a network of three pluviographs. Facing the rainy winds coming from the Mediterranean sea, the basin receives about 1900 mrn of annual rainfall; large storm events, particularly in Autumn, can produce more than 400 mm of rain (e.g., in November 1986) with very high intensities (about 60 to 90 mm . h^-1 during half an hour one time every year; an intensity of 170 mm . h^-1 was observed during half an hour during an exceptional rainfall event of 22 September 1992). A V-notch weir was built at the outlet of the catchment, and the discharge measured using a hydrograph. During « cévenoles floods » discharge can reach 2 m³ • s^-1 • km^-2.

Initial observations showed that the rising of the flood waters is always sudden and rapid, suggesting the existence of a threshold above which the flood can occur. Previous studies have shown that this threshold is not simply related tothe amount of rain. For example' in November 1986 discharge remained very slow, after more than 110 mm rain (fig.3). In another case, in October 1987, 36 mm was sufficient to initiate flooding (COSANDEY, 1993a). In fact this threshold depends on the total water storage within the basin; it can be calculated as the sum of potential maximum values of both the soil and ground water content and is equivalent to 270 mm (COSANDEY and DIDON-LESCOT, 1990). Flood generation processes include the following:

The high infiltration capacity of soils does not allow the formation of Hortonian runoff, despite high intensities of rainfall. Rapid flow can occur only in conformity with the « saturated contributing areas » theory (DUNNE and BLACK, 1970). But here saturaton begins on the upper part of the slopes where the soils are thinner; runoff which occurs does not reach the stream because it can infiltrate downslope, where the soils are thicker. Saturated surfaces progressively extend downslope; from the moment where the saturated zones join the stream, runoff can contribute to rapid flow. At this moment, flood waters arise suddenly. Figure 4 shows the different steps of the extension of the saturated areas within the basin, and the model of streamflow generation:

- When it does not rain, the water table is drained by the stream.

- A small amount of rain can be sufficient to make the water table arise in the bottom of the valley; this saturated area becomes a source of rapid flooding, but the volumes remain very small.

- If the rain continues, new saturated areas occur where the soil is thinner, mainly upslope. But runoff generated on these areas infiltrates downslope; this runoff may cause a rise in water level and an extension of saturated surfaces, but it cannot join directly to the stream, and it does not contribute to rapid flow. Discharge remains very low.

- If the rain persists, saturated areas extend downslope, and finally join the saturated areas in the bottom of the valley. From this moment infiltration is impossible, and runoff generated by the whole slope contributes to the rapid flow directly or as piston-flow. The flood starts to form and increases rapidly; discharge becomes very high.

This model allows us to understand why, when floods begin, they are immediately important if rainfall continues.

This hydrological pattern has produced about 10 « crues cévenoles » during the studied period (1981-1992), with flood peaks around 300-500 l• s^-1 (1300 to 1500 l • s^-1 • km²). During the particularly high intensity event of 22 September l992, the peak flood reached 1500l • s^-1 (7.5 m³ • s^-1 • km²). The question is to assess if hydrological processes able to produce such a discharge are the same as those described here, and if the hypothesis of a threshold can be well verified.

When the rain tregan on 2l September at 6 pm, the basin was under very dry conditions. The second part of August and September had been dry. Soil water content was quite low (about 9 mm) and the base flow discharge, 5 l • s^-1, indicates, according to Maillet'law (ROCHE, 1967) a groundwater storage of about 5 mm (COSANDEY. 1993a): 256 mm of rain would be needed to reach the threshold of 270 mm. From Monday, 6 pm to Tuesday, 4:20 am the amount of rain was about 140 mm with medium intensities. Discharge remained very low (fig. 5; COSANDEY 1993b). Total soil water content within the basin was about 162 mm. Intensities then became very important, with 65 mm rain in half an hour. A flood occurred, with only a small peak (170 l • s^-1) despite the high value of rainfall intensities. The water storage was about 227 mm, and the threshold had not been reached. The discharge decreased from the moment that rainfall intensities became lower.

About 40 mm of rain were needed to reach the threshold, and this was exceeded at 5:52 am. A rainfall event with similar characteristics (65 mm in half an hour) then produced a peak flood estimated at 1500 l • s^-1 • km², seven times higher than previously reported. Rainfall stopped quickly, and the flow decrease was rapid. Nevertheless a small rainfall event of 10 mm over 15 min was sufficient to produce a peak of 334 l • s^-1, This particularly high rainfall event demonstrated that, despite very high intensities, no important flood can occur when the basin is not completely saturated and the total water content is less than 270 mm.

It is important to note that this hydrological pattern concerns a very small (20 ha) first-order basin. Flood genesis conditions are very different downstream, mainly due to the combination of tributary floods.

Keywords:

• Flood genesis processes,
• high floods,
• Massif Central,
• Cévennes,
• saturareted contributing areas