H51H-01
The effects of spatial structure and connectivity on hydrologic response: non-linear response timescales as an emergent property in subsurface hillslope runoff.
Landscapes exhibit spatial structure at all scales. It is increasingly apparent in hydrology that this structure leads to the emergence of processes at larger scales that cannot be modeled using spatially averaged versions of smaller scale process models. The concept of ‘connectivity' has emerged as a potentially useful framework for developing a new understanding of landscape processes. In this work, the effect of spatial structure on lateral flow through a perched hillslope aquifer is explored using a numerical model, and the connectivity of flow pathways is invoked to understand the results. The effect of spatial structure is examined by comparing a two-dimensional distributed model in which the conductivity and bedrock topography are allowed to vary, and a (non)linear reservoir model. The reservoir model has a single parameter, a response timescale which is assumed to vary with the storage. This timescale qualifies as a simple closure relationship for the behavior of the whole hillslope. A dimensionless framework is used to classify regimes of hydrologic response to storms, simplifying the analysis. Three modes of behavior are identified. The first two modes are observed in homogenous hillslopes: an ‘advective' mode in which the response timescale increases with the storage in the hillslope and a ‘diffusive' mode in which it is invariant. The third mode, in which the response timescale decreases with storage, only occurs in hillslopes with spatially variable conductivity or topography. Under this mode, a hillslope aquifer will initially drain rapidly after a storm, but then release water very slowly as the storage decreases. In a hillslope with sufficiently variable bedrock topography, the response timescale becomes infinitely large for non-zero storage, creating a threshold-like ‘spill and fill' effect. Stochastic rainfall is applied to the hillslopes to determine how the three modes of hydrologic response are manifested in the filtering of a rainfall signal into runoff. In hillslopes where spatially variable connectivity creates both relatively fast and slow response timescales, peak runoff can respond to variability at both short timescales (within-storm intensity fluctuations) and the long timescales (the history of recharge events). This ‘preferential' third mode is an example of an emergent process. It occurs at a larger scale (the hillslope) as a result of spatial structure and interactions at a smaller scale. It can be explained physically by the connectivity of fast and slow pathways within the hillslope. Water is rapidly released through the fast pathways that are well connected to the hillslope base. Water that must pass through slow pathways before reaching the outlet remains in the hillslope for longer periods of time. The preferential mode is most pronounced where the connectivity of the fast pathways is highest, and connectivity of the slow pathways to the outlet is lowest.
H51H-02
Patterns, Thresholds and Nonlinearities in Headwater Catchment Storm Response
Hydrological connectivity often describes the ease with which stormwater moves through a landscape. Because the timing and magnitude of peak runoff are controlled by the degree of hydrological connectivity, highly connected systems should be characterized by short response times to precipitations, a steeper rising hydrograph limb and a higher peak discharge than disconnected systems. Hence, processing and analyzing event hydrographs should help us identify patterns in the hydrological behavior of a catchment and examine the variability of the response to precipitation for given antecedent moisture conditions. We have analyzed a three year long record of rainfall and streamflow data collected at a frequency of 15 minutes at the Hermine, a small headwater forested catchment located in the Laurentians near Montreal, Quebec. We have segmented the time series into isolated hydrological in order to derive the characteristics of storm hyetographs and hydrographs. A 10-day antecedent rainfall index was utilized as a surrogate for antecedent moisture conditions. Multivariate statistics techniques, such as redundancy analysis and variance partitioning, were used to assess a vast spectrum of catchment responses. Results indicate that time to peak and lag time values are very large when associated with low antecedent wetness. However, there is no clear tendency when antecedent wetness is high. Under these conditions, there is a set of variable responses illustrating that a highly connected state only occurs under very particular conditions that are not easily predicted by antecedent rainfall patterns. The threshold at which this switch happens is not easy to identify. These observations are confirmed by regression tree analysis that classified runoff events with respect to antecedent moisture conditions, baseflow, runoff coefficient and catchment response times. The model shows that the catchment features a fairly homogeneous dry-state reaction reflecting a disconnected state in the watershed while several wet-state responses are observed. These results show that storm runoff is a very episodic phenomenon triggered by various mechanisms. They also suggest that across a continuum of wetness conditions, the catchment dynamics is driven by a very quick saturation excess response under very wet conditions, as opposed to a longer-lasting "filling and spilling" of the bedrock depressions under dryer conditions. The time-dependent influence of these two mechanisms needs to be better assessed at various temporal and spatial scales.
H51H-03
Thresholds for runoff and sediment transport in Semi-arid areas; implications for connectivity
The concept of connectivity is increasingly being applied within a range of disciplines in the Earth and Environmental sciences as researchers recognize the need to move beyond the traditional view that runoff is generated by either Hortonian infiltration excess or by the variable source area model. In studies which focus on connectivity two key assumptions tend to be made. Firstly, that runoff thresholds must be exceeded for runoff to be produced and secondly, that all factors that influence runoff thresholds are important for hydrological connectivity. It follows that hillslope hydrological connectivity can be initiated by shorter duration, or lower intensity events, whereas catchment-scale hydrological connectivity and flooding, requires prolonged, high intensity storms. Each catchment thus has a base spatial pattern in terms of connectivity, depending on key runoff generating areas, and a response curve as the catchments wets up. In this paper we explore how this base spatial pattern changes according to thresholds in the landscape for runoff generation and sediment transport. By examining a range of events at different spatial scales it is hoped that an understanding can be developed of key thresholds in semi-arid landscapes which will assist in understanding long term landscape development.
H51H-04
Dynamic Surface Connectivity in Semi-Arid Areas: Consequences for Water and Sediment Redistribution.
We analyze the dynamics of surface (runoff) connectivity patterns in semi-arid areas with patchy vegetation. The surface connectivity pattern emerges from the interaction between hydrology, vegetation and erosion. This interaction leads to the development of a spatially variable infiltration field with low infiltration rates in the bare soil areas (due to surface crusting) and high infiltration rates in the vegetated areas (due to improved soil aggregation and macroporosity). We use a modeling framework that couples a landform evolution model with a dynamic vegetation model for water-limited ecosystems. The model captures the dynamics of spatially variable infiltration rates that are responsible for the development of a runoff-runon system, which determines the surface connectivity of the landscape and modulates the resulting sediment erosion and depositional areas. The amount of water and sediments retained by the landscape is related to the dynamic surface connectivity between the upstream and downstream areas. We analyze and compare the patterns of surface connectivity resulting from different initial topographies for hillslopes with varying slope gradients and soil erodibilities. Modeled results agree with experimental observations suggesting that, in disturbed hillslopes, there is a threshold of slope gradient below which runoff and erosion will eventually return to pre-disturbance levels and above which runoff and erosion will remain higher. In addition, we found that this threshold is related to both soil erodibility and vegetation type, which determines the dynamics of surface connectivity patterns. The principles of this dynamics have important consequences for the co-evolution of vegetation cover, hydrology and erosion patterns in arid zones, as changes in surface connectivity can lead to an irreversible loss of resources (water, soil, and nutrients) and to the desertification of these semi-arid areas.
H51H-05
Hydrologic and Geomorphic Connectivity in the Little River Watershed, Tennessee, USA
Characteristics of low-order stream channels, and of the waters they convey, depend on characteristics of contributing area land surfaces, yet land-surface and stream processes are often treated separately. Land uses influence the amount and rate of water delivery to a stream and the chemical, biological, and particulate constituents of streamwater. At the local scale, connections between land and channel are physical and site- specific. For a given stream, then, under what conditions, where, and how frequently (in time and over space) does rainfall runoff enter the channel? What factors control these connections? Answers to these questions provide a first step toward identifying, mapping, and modeling connections and connected contributing areas in a watershed. This paper reports preliminary results of an investigation of visible, land-to-channel connections in low-order tributaries of the Little River, which flows into the Tennessee River in the Southern Appalachian Mountains. Land uses in this 611 km2, humid, temperate watershed are predominantly forest, agriculture, and low-density residential; industrial and urban land uses are also present. Results of site-specific, field-based analyses were used to develop a typology of the ways in which human activities increase hydrologic connections between hillslopes and channels in this watershed and to examine the spatial distribution and size range of each type. In the Little River watershed, roadways and cattle access paths in the riparian zone provide direct routes for runoff and sediment transfer from land to stream, but hundreds of small ponds, primarily in agricultural and residential areas, partially disconnect this system and change the relationships between contributing area land use and streams. The ponds are not apparent in hydrologic networks derived solely from digital elevation models, but they are so widespread that any effort to generalize the connections between land and channel in this area will need to incorporate their presence and define their role.
H51H-06
Scaling in River Corridor Widths Reflects the Signature of Valley Forming Processes
In this paper, we analyze the statistical multiscaling structure of the "river corridor width" (RCW) series of a mountain stream in the Eel river basin, California, from the stream outlet (L= 0 m) to its basin divide (L =56 km) using 1m LIDAR digital elevation model (DEM). The RCW (defined the lateral distance from the centerline of the river to the left and right valley walls at a fixed height above the water surface) was extracted using a weighted cost algorithm in such a way that it delineated a modified DEM that shows the elevation of the river corridor geometry from a newly defined reference, the center line of the main stream. We flooded this DEM at different depths (D= 5m, 10m, 15m etc., up to 50 m) and the two interfaces of the flooded surface and river's right and left extended flood plains were extracted as a geophysical signal. We argue that the extracted signal carries the signature of valley forming processes (fluvial and hillslope processes and their interaction) that have been taking place at depth D above the water surface. These processes are in a transverse direction to the main stream and in a unit normal direction to the extracted signal. Fluvial processes that bring the sediment and debris flows at tributary junctions introduce a localized large-scale disturbance which superimposes on smaller- scale organization created by hillslope diffusion and land wasting processes such as landslides. To extract the embedded signatures of high frequency and low frequency fluctuations that superimpose in this geophysical signal and study its scaling properties, we use Wavelet Transform Modulus Maxima (WTMM) method. The results of this analysis provide a quantitative measure of landscape dissection at different river regimes and different scales and reveals important connections between the observed statistical structure of valley geometry and the physical processes responsible for its formation.
H51H-07
Road Connectivity and Sediment Yields in Forested Landscapes
Unpaved roads are often considered to be the predominant sediment source in forested catchments, but catchment-scale studies generally have not been able to document a road-related increase in runoff or sediment yields. This discrepancy may be due to the proportion of roads that are connected to the stream channel network. This paper summarizes the results of three multi-year studies in California and Colorado that have investigated both road sediment production and the connectivity of roads to the stream channel network. Annual sediment production rates averaged around 3-8 Mg per hectare of road surface in California and 34 Mg ha-1 in Colorado. In both areas sediment production can be predicted from road segment slope, road surface area, and rainfall intensity or erosivity. In California 25 percent of the road length was connected to the stream channel network, and the majority of the connected segments were due to stream crossings. In Colorado only 14 percent of the road segments were connected, and most of the connected road segments were in the valley bottoms and connected by gullies or sediment plumes. An analysis of published data shows that 92 percent of the variability in road connectivity can be explained by mean annual precipitation, and the absence of engineered drainage structures increases the amount of road connectivity by about 40 percent. In Colorado the predicted sediment yields from unpaved roads are comparable to the sediment from high severity wildfires. In the southern Sierra Nevada of California roads accounted for less than 10 percent of the measured sediment yields in two small experimental catchments and 25-50 percent of the sediment yield in a third catchment. We conclude a relatively few road segments are responsible for most of the road-related sediment at the catchment scale, and that information on both sediment production and road connectivity are needed to identify the road segments of greatest concern and highest priority for remediation.
H51H-08
Quantifying Wetland Functions: A Case Study
Wetlands are reputed to reduce peak flows and improve water quality by trapping sediment and phosphorus. However, there are relatively few studies that quantify these wetland functions. This paper reports on a study of a 45-hectare wetland in southern Wisconsin. The wetland is traversed by a stream channel that drains a predominantly agricultural 17.4 km2 watershed. During the spring and summer of 2006, we collected stage data and water samples at stations upstream and downstream of the wetland, with the former accounting for 82% of the contributing area. Continuous measurements of water stage at these stations were used to construct a streamflow record. During storm events water samples were taken automatically at 2-hour intervals for the first 12 samples and 8-hour intervals for the next 12 samples. Samples were analyzed for total suspended solids, total phosphorus, and dissolved reactive phosphorus. Ten events were observed during the observation period; the two largest events were 1 to 2-year storms. One-dimensional unsteady flow routing was used to estimate the maximum extent of wetland inundation for each event. When normalized for flow volume, all peak flows were attenuated by the wetland, with the maximum attenuation corresponding to the intermediate events. The reduced attenuation of the larger events appears to be due to filling of storage, either due to antecedent conditions or the event itself. In the case of sediment, the amount leaving the wetland in the two largest storms, which accounted for 96% of the exported sediment during the period of observation, was twice the amount entering the wetland. The failure of the wetland to trap sediment is apparently due to the role of drainage ditches, which trap sediment during the wetland-filling phase and release it during drainage. The export of sediment during the largest events appears to result from remobilization of sediment deposited in the low-gradient stream channel during smaller events. This hypothesis was supported by the finding that the estimated bed shear during large events exceeded laboratory measurements of the critical shear stress of bed sediment samples. In the case of total phosphorus, the inflow to the wetland about equaled the outflow, although the wetland sequestered 40% of the incoming dissolved reactive phosphorus. The discrepancy is almost certainly due to net export of sediment. Wetlands such as this are very common in the glaciated portion of the U.S., and many contain channels and ditches. The region is dominantly agricultural, and sediment and phosphorus are the primary causes of impaired surface-water quality. Our results suggest that these wetlands are not very effective in mitigating this impairment when flow is concentrated in channels.