Regional Flood Frequency Analysis Using the L-Moments Techniques for the Watersheds in the Province of Gilan

Employing regional flood frequency analysis, a regionalized relationship may be established to estimate flood magnitudes for ungauged and poorly-gauged catchments. Regional flood frequency analysis is usually based on identification of homogeneous regions, selection of suitable regional frequency distribution, and estimation of flood quantiles at watersheds of interest. In this paper, L-moments techniques was used for regional floods frequency analysis in watersheds located in the Province of Gilan. In the conventional technique of regionalization, homogeneous hydrological regions are identified on the basis of watershed characteristics such as physiography, geographical location and flood flow data at the related hydrometric stations. In the present study, using the K-means clustering algorithm, the study area was divided into two sub-regions containing a total of 58 hydrometric stations. Indices of the Average Silhouette Width, Davies-Bouldin and Dunn were used to determine the optimal number of clusters and measurement of the global goodness of clustering. Homogeneity of the region was investigated by the discordancy measure and heterogeneity measure tests. Selection of the best categorization method of the regional distribution was assessed using the goodness of fit measure, and the generalized logistic for the regions 1 and 2 as the best categorization method to identify the regional distribution. Finally, the regional parameters of selected distributions and probability quantiles with different return periods were calculated.