Investigating variability in small multi-scale catchment response using lumped conceptual modelling of storm-based runoff and stable isotope data

In this paper we use a lumped conceptual modelling approach to further examine variability in streamflow generation from a series of small nested catchments (7-147 ha) within the Westcreek watershed of Mont-Saint Hilaire (MSH), Quebec, Canada. Experimental study of these catchments shows variability of hydrologic response (e.g runoff ratios and event-based total new water delivered) that under some environmental conditions (e.g. large storms on dry antecedent moisture conditions) can correspond in part to morphological characterization (e.g. total catchment area or sub-catchment area) and estimates of maximum dynamic storage capacity. However, there is significant variation in small catchment response that remains unexplained. Further, the MSH catchments show the nonlinear change in hydrologic response with antecedent moisture conditions (AMCs) that studies have identified as attributable to a shift in hydrologic connectivity and dominant runoff mechanisms and we aim to investigate the ability of lumped conceptual models to represent this behavior for the MSH catchments. For each catchment, a series model structures and parameter sets are tested for their abilities to reproduce measured streamflow and δ18O stable isotope compositions during five separate storm events collected across wet to dry AMCs. Model performance was evaluated with various statistical summaries (objective functions) for each of the time series used (i.e. runoff, δ18O compositions). A GLUE (generalized likelihood uncertainty estimation) approach is used to identify behavioral parameter sets and provide information about their uncertainties. Preliminary analysis of the largest catchment (147 ha) has tested six model structures, each comprised of a soil moisture accounting module that estimates effective rainfall (i.e. the part of precipitation available to runoff) and a routing module with two reservoirs in parallel. The number of behavioral parameter sets decreased significantly with the additional use of streamflow δ18O data in model evaluation. We found well-defined and similar behavioral routing parameters common to all model structures (i.e. time constants for both reservoirs and quick flow fraction) but some parameters (e.g. quickflow fraction) did change with storm size and antecedent moisture conditions, consistent with experimental observations. Based on these preliminary results, model structure is currently being refined and will be applied across the additional catchments to further examine the variability in hydrologic response under a range of environmental conditions.