چكيده به لاتين
Catchment classification is a necessary and basic step toward improving the hydrology-related sciences such as hydrologic modeling. Owing to the complexity and nonlinear relations among the components of a system in the hydrologic processes, the classification based on the complexity level of estimation systems and nonlinear methods, such as the phase space reconstruction concept, has been considered and appears to be highly effective. By using this concept, the system state can be described at any time/space and define the number of variables needed to represent the trajectories of a dynamic system. Although researchers’ efforts have included the development and introduction of some procedures, studies in this area are still in the infancy stage. This study attempts to propose and verify a suitable pattern for catchment classification based on use of the concept of the Chaos theory. To this end, the daily runoffs of sixty stations in Iran were used. The Artificial Neural network (ANN) was employed to identify the classification pattern using the mentioned concept. The results show that using two measures of the Chaos theory (i.e. the lag time and the correlation dimension) one can classify the catchments based on the complexity of the system. The result of data simulation using ANN method was employed to divide the stations into 3 groups based on the lag time (High-lag time, Transition and Low-lag time) and 4 groups based on correlation dimension (low dimension, Transition, high dimension and unidentifiable). While the use of the criteria D2 leads to certain groups also a bit more accurate results in comparison to the use of the lag time in stations classification, use of lag time criteria is easier and requires less information. The results show that although the spatial nearness of two stations does not indicate behavioral similarity, some homogeneity in spatial behavior was observed. The spatial pattern classification of Iran's catchments indicates that catchments with different climate characteristics which are located at a far distance from each other might yield similar responses along with the same level of complexity. Evident from an analysis in this study shows while the runoff dynamic of south-eastern stations exhibit very high complexity more stations located in very wet climates (especially in the north part of Iran) are less complex than the other stations. In general, decrease in amounts of rainfall (drought increasing) leads to increase in the value of complexity. Calculating the mentioned criteria would guide modelers in the identification of the most suitable model based on the complexity of the hydrologic phenomena. This approach saves considerable time and reduces computation requirements.
