FLOOD00113 - Laserfiche WebLink

Home Browse Search

288 J. l.AVABRE ET AL. HYDROl.OGICAl. RESPONSE OF A MEDITERRANEAN BASIN 289 TABLE 7 Nash Efficiency E 0.83 0.70 slightly to overestimate it. This bias, reflected in the E reduction, is evidenced both on the total annual runoff and in the daily runoff peaks. As for the AR, observed daily runoff exceeds the daily simulated annual water yield by 100 mm, which is a 220/0 increase over the estimated <normal' response (452 mm simulated instead of 552 mm observed). We should note that this estimate does not include the missed period (21 days including relevant storms), and should be taken as a lower limit for the estimate of the increment of the AR. This increase does not affect the peak flow in the same way as the recession flow. Table 7 shows different variables regarding the days where daily runoff is more than four times a,s great as the mean daily runoff (20Is.'), the high water days, and those where the daily runoff is less than one-fourth of the mean daily runoff, the low water days. These results point out that the increase of the runoff volume is practically concentrated on the high water day~ (a difference between the simulated runoff and Ihe observed of about 120 mm during these days). The low water days do not show any significant change after the fire (the model has fewer problems to reproduce correctly the number of these days. which is now reduced). The other index, particularly the ratios of the volume of floods and the number of high water days, confirms the relation between the increase of runoff production and the high water. On the other hand, if we compare the peak flow (see Fig. 7), it is confirmed that the new response of the basin can be characterised by sharper and increased peak flow (similar to thaI reported by Brown, 1972, and by Burch et a!., 1987). The zoom on the two most important daily peaks recorded shows this trend more clearly, which in the case of the higher peak leads the model to predict a peak flow of 4081 S-I instead of the 6631 S-I observed (62% of increase). Again, these results are in accordance with the previous estimates. They suggest that the consequences of the fire affect the hydrological response at all time steps, and lead to a 30% increase in runoff mainly concentrated in the high water days. (Note that this conclusion is not in contradiction with the conclusions reached concerning Fig. 4.) Performances of OR3 in the daily simulation Calibration period (1984-1989) 1990-1991 Annual ratio between observed and simulated volume of floods (%) 115 162 Days with flow greater than four times the mean daily runoff(>4 x 20Is-l) Number of observed days Number of simulated days Difference between observed and simulated runoff on these days (mm year-I) 105 112 43 29 18 121 Days with flow less than one fourth of the mean daily runoff ( < 20/41 s - I) Number of observed days Number of simulated days Difference between observed and simulated runolfon these days (mm year-I) 780 514 -9.6 113 104 -8.9 50% reduction in evapotranspiration leads to similar results to those observed as a consequence of the fire. . W~ should note that in fact we changed only the potential evapotranspira- tion (mstead of the real evapotranspiration). However, it was the only variable that could be changed in the simple model used. Thus the conclusion should be more precisely stated: that destruction of the vegetation cover induces a change in the loss power that in our case is hydrologically equivalent to a 50% reduction of the potential evapotranspiration. Daily analysis We carried out a similar analysis but using a daily time step. First the model was calibrated on the pre-fire years, and then the model simulation and the daily runoff observed for 1990 were compared. This time the model was calibrated using only a,5 year period which is cons~dered sufficient by the model (1984-1989). Table 7 summarises the performance of the daily slmulatlon. : The results of the post-fire year show again that the model systematically underestimates daily runoff. whereas for the reference period the model tends CHANGES IN THE FLOOD REGIME After a dry period in August and in September, the first floods of the post-fire year were observed on I and 3 October 1990. In both cases, rainfall quickly produced the peak response of the basin, which became zero a few minutes later (see Fig. 8). Although we do not have the flow records from 5 to 25 October, the field observations and the comparisons with other flow recording stations allow us