Examining lag time using the landscape, pedoscape and lithoscape metrics of catchments

In order to explain total variations of lag time as a type of hydrological response affected by landscape-related features of catchments, the present study addresses relationships between lag time and the landscape structure-related metrics (perimeter-area ratio, contiguity index, fractal dimension index, shape index, and related circumscribing circle) of land use/land cover, hydrologic soil groups, and geological permeability classes. To explain a given catchment's lag time, a regionalization approach was adopted, applying spatial data from 39 catchments located in the southern basin of the Caspian Sea, with areas varying from 33 to 4800 km(2) and mean discharge ranging from 0.47 to 21 m(3) s(-1). Forest cover (57.4%) accounts for the dominant land cover, while rangeland, farmland, and urban areas comprise the remaining 25.9%, 11.7%, and 1.6%, respectively. Category-level landscape structure metrics were estimated for the catchments' land use/land cover, hydrologic soil groups, and geological permeability classes. These metrics were then applied as independent variables to stepwise multiple linear regression analysis in an attempt to explain the catchment lag time. The regression-based models (0.68 <= r(2) <= 0.80) indicated that the catchment lag time could be explicitly predicted by applying the average values of the related circumscribing circle, fractal dimension, perimeter-area ratio, and shape indices for the landscape categories, hydrologic soil groups, and geologic permeability classes of the catchments. The findings also indicated that shape-related features, implying regularity, irregularity, and elongation of the patches of landscape classes, hydrologic soil groups, and geological permeability classes, can be used to explain total variations in lag time of the catchments. Based on the results of the inter-model comparison, the most appropriate model is the regression model developed using the fractal dimension index, which has a high level of reliability (based on the results of the uncertainty analysis).