Study of the rainfall-runoff process in the Andes region using a continuous distributed model

Abstract

The Precordillera area of the Andes Mountains (Mendoza, Argentina) is affected by severe flash floods, caused by heavy rainfall events of short duration and high intensities. A telemetric network, installed on a pilot zone since 1983, provided a set of about 150 rainfall events. In addition, about 50 flood events were recorded at the outlet of a 5.47 km² catchment. The distributed model Areal Non-point Source Watershed Environment Response Simulation was adapted to this catchment and applied continuously over the period 1983–1994. The year 1985 was used for calibration. The model was able to reproduce runoff volume with an efficiency of 0.6, and peak discharge with an efficiency of 0.46. The largest events were however underestimated, although the model was able to reproduce the sharp increases in streamflow registered by the sensor. Based on the assumption that the model had captured the catchment behaviour, it was used to determine the main processes involved in runoff generation. The combination of rainfall and soil variability, mainly associated with a quasi-impervious area in the middle of the catchment, was found to explain the rapid increases in streamflow. Vegetation, surface storage capacity, and initial soil moisture were also influential but with a much smaller magnitude than the combination of rainfall and soil variability.

Introduction

The Piemont and Precordillera areas of the Andes mountains situated in western Mendoza (33–33.5°S, 68.8–69.1°W) (Argentina) are subject to heavy summer rainfall events of short duration and high intensities, characterised by a large spatial and temporal variability over short distances (Fernandez et al., 1997). These events can generate severe flash floods, dangerous for the downstream city of Mendoza. Therefore, it was of particular interest to understand the mechanisms leading to the runoff generation in this region. A telemetric network of automatic rain gauges was set up in 1983 on a study area of about 600 km², westward of the city of Mendoza (Fernandez et al., 1988). A 5.47 km² catchment, the Divisadero Largo, was also equipped with a sensor recording streamflow (Fig. 1).

The collected data were already used in various rainfall-runoff and sediment delivery studies using mainly event based models (Mulders et al. 1990a, Mulders et al. 1990b; Hoefsloot et al., 1992; Ligtenberg et al., 1992; Sharma, 1993; Sharma et al., 1996a, Sharma et al., 1996b). The major drawback of these models is the initialisation of the moisture status of the catchment prior to each storm.

Therefore, the purpose of the study was: (i) to study the feasibility of using a continuous model on the Divisadero Largo catchment, and (ii) to determine the main processes involved in runoff generation in order to explain the sharp streamflow increases registered at the outlet of the catchment. The distributed and continuous model Areal Non-point Source Watershed Environment Response Simulation (ANSWERS) (Beasley et al., 1980), modified by Bouraoui and Dillaha (1996) was used for these purposes. Available data and model are presented in Section 2. Section 3 describes the calibration and validation of the model. Finally, Section 4 presents a sensitivity study conducted on the influence of the spatial variability of surface properties and rainfall on the hydrological response of the catchment.