Multi-Interval Data Mining of At-A-Station Hydraulic Geometry to Quantify Temporal Stability

Hydraulic geometry, the relationships between a stream's geometry (width and depth) and flow (velocity and discharge), is applicable to flood prediction, analysis of aquatic habitats, water resources management, and modeling point sources of pollution. Various investigators have described hydraulic geometry as stable over long periods of time but unstable over short periods of time. Data was mined in multiple, two-year intervals from an online source to compute and classify hydraulic geometry, as well as to quantify stability. Ordinal data describing bed material and nominal data describing channel evenness, and bank stability (which could affect hydraulic geometry stability) were also mined. In this paper, the temporal stability of at-a-station hydraulic geometry at cross-sections from rivers in mountainous and non-mountainous watersheds in Tennessee was compared by mining data to compute hydraulic geometry from stream flow and channel geometry measurements collected and archived by the U.S. Geological Survey from at least five successive two-year intervals. A variance was computed for each station, which were used to compute pooled variances for mountainous non-mountainous watersheds. An F-test comparing the pooled variances revealed no statistically significant difference, thus, our data suggest that differences in regional topographic setting in Tennessee did not produce differences in the magnitude of temporal stability of at-a-station hydraulic geometry. Data mining can contribute to basic research involving empirical investigations of statistical relationships between environmental variables and hydraulic geometry, as well as theoretical studies that predict hydraulic geometry from physical principles. Furthermore, data mining can provide readily-available hydraulic geometry input for modeling water quality, flooding, and stream habitat. Such applications would benefit from multi-interval hydraulic geometry data for sensitivity and time series analysis, as well as for determining how frequently model output should be updated.