Diagnostic analysis of runoff partitioning at the catchment scale

Horton overland runoff, Dunne overland runoff and subsurface flow are the three major runoff generation mechanisms contributing to streamflow. The control of the climatic conditions and landscape properties on the occurrence and relative dominance of these different runoff components, i.e., and therefore the runoff partitioning, is captured in a qualitative manner by the famous Dunne Diagram. However, an improved, quantitative understanding of the controls runoff partitioning at the catchment scale is necessary to constrain and improve hydrological predictions at catchment scale, especially in ungauged catchments. This dissertation systematically investigates the controls of climate, soil and topography on runoff partitioning at the catchment scale, and consequent of runoff partitioning on the transportation of water and nutrients, using both the downward and upward approaches. In the downward approach, given observed patterns at some higher level, we look for the most possible process interactions at a lower level that might have led to them. In the upward approach, knowing the processes at a lower level or scale, we explore how their interactions may have lead to the observed patterns at a higher level or scale. In the first example, a large catchment scale distributed model has been applied to two basins in Oklahoma, the Illinois River basin near Tahlequah and Blue River basin near Blue. After validating the model against observed streamflow data, the model is used as a tool to diagnose the controlling factors underlying the temporal and spatial pattern of runoff partitioning. It is found that in both basins there is a competition between Dunne runoff and subsurface flow, and this competition is quantitatively shown to be controlled by the seasonality of climatic forcing, and the relative magnitudes of the saturated hydraulic conductivity of the soils and the topographic slope. The effects of the spatial trend of runoff partitioning are thus examined in terms of signatures of runoff response derived from the predictions of the same hydrological model. The signatures are mainly constructed at the event scale, such as instantaneous response function which describes the advection and dispersion effect of catchment on generated runoff. Dimensionless flood peak and time-to-peak are used to explore the advection and dispersion separately. The results in this work are on the one hand consistent with previous theoretical studies, and on the other hand also somewhat surprising. For example, the power-law relationship between peak of the IRF and drainage area is seen to become flatter under wet conditions than under dry conditions, even though the (faster) saturation excess mechanism is more dominant under wet conditions. This result appears to be caused by partial area runoff generation: under wet conditions, the fraction of saturation area is about 30%, while under dry conditions it is less than 10% for the same input of rainfall. This means travel times associated with overland flow (that mostly contributes to the peak and time to peak) are in fact longer under wet conditions than during dry conditions. We go further to explore the possible controlling factors of runoff partitioning in a broader context with the use of an upward diagnostic approach. In this case, a simple highly distributed hydrologic model based on point scale processes has been built for this purpose, which is comprehensive enough to simulate the effects of different combinations of climate, soil and topography, and generate a diversity of runoff generation mechanisms. With this model numerous virtual experiments have been conducted to produce a variety of runoff responses under various climate, soil and topographic combinations. A small set of dimensionless similarity numbers, which are physically meaningful, have been shown to effectively quantify runoff partitioning at both the annual and catchment scales. A few hypothetical relationships have been proposed. Each combination of these dimensionless numbers could be feasible in theory, but only some of these combinations actually occur in nature. By constraining the predictions of the model with the empirical Budyko curve, we narrow down to these feasible or ???behavioral??? combinations, which are further governed by close interconnections between climate, soil and topography. Based on the above diagnostic analysis of runoff partitioning, we then investigate the transportation of different runoff components from their locations of generation all the way to the catchment outlet, which is shown to be effectively quantified in terms of the mean residence time (to account for the advection effect) and dimensionless catchment instantaneous response function (IRF, to account for the dispersion effect) for each runoff component. In addition, the consequent impacts of runoff partitioning on the transportation of nutrients (nitrogen and phosphorous, and sediments) have been studied at an agricultural basin with extensive tile drains, The Upper Sangamon River basin located in central Illinois. It is found that there is a carry-over of nitrogen storage from dry years to wet years, and this is mainly caused by the loading of NO3-N from the hillslope to the channel by way of tile drainage that is prevalent in this region. From this dissertation, the resulting improved understanding of the controls and subsequence of runoff partitioning at the catchment scale, especially the quantitative description of runoff partitioning in terms of the inter-connections between climate, soil and topography, could be potentially tested in the field, and if deemed reasonable, could also be used to constrain/improve hydrological model predictions, and advance water resources management in a context of both water quantity and quality