Prediction of stream-flow regime using ecological classification zones

Hydrologic classification is useful for data organization, transfer of model parameters and estimation of hydrologic sensitivity to disturbance and climatic change. Stream-flow regime has frequently been used as a basis for classification, typically by mapping regimes defined by stream-flow data from a gauging network. As an alternative, we hypothesized that ecological classification systems can predict stream-flow regime because they are based on the same characteristics that control run-off generation (soils, climate and topography). A multivariate regression tree (MRT) was used to relate stream-flow regime to the fractional coverages of the Biogeoclimatic Ecological Classification (BEC) zones within the catchment for gauged streams in British Columbia, Canada. Although the MRT identified a realistic set of regimes, only a small number of BEC zones were used as predictors, reflecting bias in the gauging network. To avoid this bias, we used a water balance model to compute mean monthly stream flow for 932 ungauged basins in British Columbia that were generated with areas between 10 and 1000km2; these monthly stream flows were used to train an MRT model based on BEC zone coverages. This model predicted the regime at gauged basins nearly as accurately as the water balance model for pluvial, nival and glacier-supported nival regimes. Difficulties occurred in smaller basins and in specific regions where the local BEC zones were not included as predictors. Coastal hybrid nivo-pluvial regimes were poorly predicted. With further development, ecological classification systems could have great value as a tool for hydrologic classification for both research and operational applications. Copyright (c) 2013 John Wiley & Sons, Ltd.