ABSTRACT: Precipitation-runoff modeling in cold and snowy climate: An alternative approach with artificial neural networks.

CGRG Bibliography of Canadian Geomorphology
Search Results

Author :

Date :

Title :

Publication :

Issue :

Page(s) :

Abstract

Watershed runoff in areas with seasonal snow cover is usually estimated using physically based conceptual hydrologic models. Such simulation models normally require a snowmelt algorithm to be part of the modeling system. Most snowmelt algorithms consist of a surface energy balance and some accounting of internal snow pack processes. More detailed surface energy balance models require inputs such as air temperature, relative humidity, wind speed and either cloud cover or radiation data. On the other hand, Artificial Neural Networks (ANNs) are flexible mathematical structures that are capable of identifying complex non-linear relationships between input and output data sets of observed historical records without trying to explicitly represent the different components of the hydrological processes. While the precipitations are the main driving factor for a hydrological model, the temperature data provides information on the state of the precipitation (rain or snow) and the available potential for evapotranspiration. Logical information regarding the season is also provided to the network by including the number corresponding to the month (1 to 12) in the input pattern. Moreover, since the runoff to be estimated with hydrological models is a function of not only the current but also the antecedent values of the input variables (such as precipitation and temperature), dynamic neural networks with time delay lines are found to be more appropriate. One such type of neural networks which brings a memory structure into the network, namely, the Time Lag Feed-forward Network (TLFN) is used in this study. The size of the memory layer depends on the number of past data that are needed to describe the input characteristics in time and is determined on a case-by-case basis. The ANN simulation results are then compared with that of HBV-96, a conceptual hydrologic model which has a routine for snow-accumulation and snowmelt based on a degree-day relation with an altitude correction of temperature. Simulation experiments are conducted for both the reservoir inflow and the streamflow simulation in the Chute-du-Diable watershed in northern Quebec. The study demonstrates that ANNs are capable of modeling the precipitation-runoff process in cold and snowy climate by implicitly recognizing the input-output patterns in the historical data rather than trying to represent the snow accumulation and snowmelt processes explicitly in the model. The validation results show that the ANN based models performed quite well, almost comparable to that of HBV-96, in simulating the precipitation-runoff processes in the study region.

Bibliography of Canadian Geomorphology