On this poster, we present the results of a study realized in the PIREN-Seine research program, where we try to represent mixed watershed hydrological signal using hydrological signal of pure catchments. We work on the Orgeval experimental catchment, a basin followed by Cemagref since 1962. On this catchment, three small different sub-basins were added recently. One of them is completely agricultural, the second has only forest cover, and the third is a mixed one with almost 20% of forest cover and 80% of agricultural land. We measure fluxes of water and elements like nitrate and chlorure and characterize the three types of hydrological behaviours. The hypothesis is that the hydrological signal of the mixed catchment is a combination of the two pure signals from the two others basins. We present the results of linear and non-linear regressions trying to explain the mixed behaviour with the two pure hydrochemical informations and with only one of them. We compare the results using determination coefficient (r²) and Nash & Sutcliffe criterion. It clearly appears, at the elementary catchment scale, that we need (as expected) the pure informations to explain the mixed hydrochemical signal. With linear regression the results are quite good for either low flow or flood. The next step is modelling hydrological signal for mixed catchment at order 3 (Strahler classification), using pure and/or mixed catchments at order 1. It appears that the change of scale changes the relations. To obtain the mixed signal, we need more-linear relation to have a good modelling. We have now four years of hydrochemical information on these three elementary catchments. In the framework of the ACI Cat-Nat research program, in order to have better understanding of hydrological behaviours for flood prevention, we are planning to monitor an additional catchment at the second order to identify where the non-linear behaviour appears during the scale change.