چكيده انگليسي :
Estimating daily streamflow in watersheds is essential for various water resource projects, such as hydropower schemes, irrigation system design, river pollution management, reservoir sedimentation, and erosion control. However, the lack of such information in some basins, referred to as basins without statistics, remains a fundamental challenge for the hydrological community. In response to this challenge, hydrologists have addressed it by expanding various tools, including regionalization of studies.
In this research, regression relationships were developed to plot flow duration curves for sub-basins without statistics using a parametric approach. For this purpose, 13 suitable hydrometric stations were selected in the northern Karun basin, considering geographic dispersion, appropriate statistical period length, and area diversity (including Armand, Gedar Kebak, Tagarg Ab, Beheshtabad, Pol Kereh Bas, Tang Darkesh Varkesh, Barz, Dezk Abad, Soolgan, Kooh Sokhte, Gerd Bish, Tang Zardalu, Zarin Darakht, and Morghak). Subsequently, incomplete statistics were reconstructed. After drawing flow duration curves for all sub-basins, flow values of index discharges (Q5, Q10, Q20, Q25, Q50, Q70, Q75, Q80, Q90, Q95) were extracted from these curves. In the next stage, physiographic, climatic, and land use variables (area (A), perimeter (P), average basin height (hm), average basin slope (wsa), main channel slope (rs), concentration time (tc), main channel length (rl), Miller coefficient (mc), Gravelius coefficient (gc), equivalent rectangle length and width (i, l), equivalent circle diameter (cd), and annual average precipitation (pay)) were collected as influential factors in streamflow. Then, multiple-variable regression analysis was performed on the relationship between these values and the index discharge values using both all variables and selected factors based on factor analysis (once with three factors and once with five factors).
In the next step, to ensure the homogeneity of the northern Karun basin watersheds, a cluster analysis method was used for this investigation. The results showed that four basins, Beheshtabad, Barz, Kooh Sokhte, and Morghak, were non-homogeneous selected basins, while nine basins, including Armand, Gedar Kebak, Tagarg Ab, Pol Kereh Bas, Dezk Abad, Soolgan, Gerd Bish, Tang Zardalu, and Zarin Darakht, were identified as a homogeneous region. The previous steps were then repeated with the remaining 9 stations. To evaluate the best regression model, statistical measures such as RMSE, R2, and NS were used. The results indicated that the best model was the one obtained from all variables and non-homogeneous stations according to cluster analysis. Additionally, the model using logarithm of variables and homogeneous stations performed well. The model using all variables and all stations also showed the best performance in some cases, and the model using three factors exhibited the least performance error. In the best methods, the average basin slope (ws) and annual average precipitation (pay) had the most positive impact, while the Gravelius coefficient (gc) had the most negative impact on the models. In regression relationship analysis, entering variables using the Enter method showed better performance.
