The influences of spatially variable rainfall and localized infiltration on groundwater recharge in a water management context
MetadataShow full item record
Water management involves monitoring, predicting, and stewarding the quality and quantity of groundwater recharge at the watershed scale. Recharge sustains baseflow to streams and replenishes water extracted by pumping at wells; it is frequently estimated using numerical models that couple or fully integrate surface water and groundwater domains and use water budgets to partition water into various components of the hydrological cycle. However, uncertainty associated with the input data for large components such as precipitation and evapotranspiration may hinder model accuracy, and preferential flow dynamics such as depression focused recharge (DFR) may not be represented at typical modelling scales (≥10s of sq. km) or with typical approaches. The present study addressed two themes related to groundwater sustainability and vulnerability: 1) the sensitivity of modelled recharge estimates to the spatial variability of rainfall, and 2) the vulnerability of public supply wells to DFR during large-magnitude rainfall or snowmelt events. The region investigated during this research was the Alder Creek watershed (78 sq. km), a typical southern Ontario setting overlying glacial moraine sediments with mostly agricultural land use, some urban and aggregate resource development, and whose recharge supplies multiple municipal well fields for the cities of Kitchener and Waterloo. Rainfall is often the largest component of the water budget and even a small uncertainty percentage may lead to challenges for accurately estimating groundwater recharge as a calculated residual within a water budget approach. However, rainfall monitoring networks typically have widely spaced gauges that are frequently outside the watershed of interest. Assessment of the influence of spatially variable rainfall on annual recharge rate estimates was performed by comparing transient simulations using input data from three different rain gauge networks within a coupled and fully-distributed numerical model. A local network of six weather stations with rain gauges was installed and operated in and around the study watershed for three years, and data from six regional stations (within 30 km of the watershed) and one national station (3 km from the watershed) were obtained from publicly available sources. Time series of distributed, daily rainfall were interpolated via the inverse distance squared method using data from each of the rain gauge networks for three calendar years. The temporal and spatial snowfall distribution was consistent among all scenarios, to maintain focus on differences caused by the rainfall input data. Results showed that annual average recharge rates could differ considerably between scenarios, with differences sometimes greater than the water-budget derived uncertainty for recharge. Differences in overall recharge between pairs of scenarios involving the local rain gauge network were largest, varying by up to 141 mm per year, or 44% of the steady state recharge estimated in a previous study. Streamflow estimates for the local rainfall simulations were closer to observations than those using regional or national rainfall. Because the three scenarios used the same set of underlying soil parameters, the results suggest that the availability of local rainfall measurements has the potential to improve the calibration of transient watershed hydrogeological models. The second theme of the present study was exemplified by the Walkerton tragedy in 2000, where pathogenic microbes were rapidly transported from ground surface to a public supply well during a heavy rainfall event. The vulnerability of such wells to surface-originating contaminants during major hydrological events remains poorly understood and is difficult to quantify. Such events may result in overland flow collecting in low topographic locations, leading to localized infiltration. If focused recharge occurs in the immediate vicinity of a public supply well, the threat to the water quality of that well may significantly increase temporarily. These conditions are frequently encountered within the glaciated landscape of southern Ontario. Conventional approaches for defining the threat of groundwater under the direct influence of surface water (GUDI) do not routinely account for this type of transient infiltration event and instead assume steady state flow fields without localized recharge. The present study combined the monitoring and modelling of a site in southern Ontario where DFR is routinely observed to occur within 50 m of a public supply well. Extensive site characterization and hydrologic monitoring were conducted at the site over a period of 3.5 years, specifically during large-magnitude hydrologic events including heavy rainfall and snowmelt. Integrated surface water – groundwater models employing HydroGeoSphere (HGS) were used to quantify the transport of potential contaminants infiltrating beneath a depression and a creek and the associated risk to the public supply well. Simulated relative concentrations at the well were below “detection” for typical median contaminant concentrations in surface water but > 1 cfu/100 mL with travel times between 118 and 142 days for creek and DFR solutes, respectively, based on maximum initial surface water concentrations. Results suggest that DFR and localized recharge could increase the threat to overburden wells under extreme conditions. Ponding reduced travel time by at least 58 days for the DFR solute. In order to extend the analysis of recharge estimate sensitivity to spatial rainfall variability to the longer term, and to incorporate the influence of actual evapotranspiration (AET) uncertainty, a method was developed to employ stochastic rainfall time series and AET estimates in a Monte Carlo framework to quantify the resulting variability in recharge estimates and three groundwater management metrics. Stochastic rainfall time series were generated via a parametric, mixed exponential method for three virtual stations within the Alder Creek watershed and constrained by field-derived spatial correlation coefficients. Observed snowfall data from one nearby national weather station were used to calculate total precipitation. Stochastic annual AET estimates were generated based on: 1) calculated annual potential evapotranspiration at the national weather station, 2) observed variation about the Budyko curve in 45 US MOPEX watersheds with PET/P ratios within ±0.05 of the average ratio calculated for the national weather station near the watershed, and 3) a correction factor to remove AET from the saturated zone. Recharge rates for the Alder Creek watershed were calculated via a 46-year vadose zone water budget for each of 16,778 realizations. The surface water fraction of streamflow was estimated using hydrograph separation results for the watershed. It was hypothesized that spatially variable precipitation would exert more influence on recharge than AET because it is a larger component of the local water budget. Groundwater recharge results were applied to three different metrics related to water quality, well vulnerability, and water quantity. Results suggest that estimates of non-point source contaminant loadings to the water table could differ by up to ±14% from the average. Worst case changes in capture zone area estimates for a public supply well could be up to ±15% different from the average. The ratio of maximum to minimum cumulative recharge over all realizations was 1.31, though contributions from spatial rainfall variability alone led to a ratio of 1.15. This suggests that AET uncertainty and spatial rainfall variability each contribute nearly the same amount of variability to recharge estimates. This latter ratio is less than the result (~2) from a previous study of a much larger watershed in Spain. The results highlight the importance of AET estimates for recharge rate estimation, and their potential impacts on land use planning and groundwater management. This method could be used to project impacts of climate change on recharge variability at the watershed scale. Overall, results suggest that the spatial variability of rainfall could impact recharge rate estimates in numerical models of small to medium sized watersheds (e.g., 78 sq. km), especially during short simulations. Annual recharge estimates could vary over a range equivalent to 44% of a previously estimated steady state value, though long-term (46-yr) estimates could vary over a range equivalent to 12% of this value due to averaging over time. Non-point source loadings and capture zone areas could vary up to ±7.0% and ±7.4% from the average, respectively, over the long term due to spatial rainfall variability, though uncertainties associated with AET could increase this to ±14% or ±15%, respectively. The hydrological event characterization and well vulnerability modelling of the second research theme suggest that localized recharge could lead to increased microbial risks for wells screened in overburden sediments during large hydrological events (≥ 40 mm rainfall over 4 days) through the phenomenon of temporary ponding. The method developed for the long-term stochastic recharge rate analysis could be applied in other settings as an alternative to, or to complement, large-scale, fully-distributed 3D numerical modelling.
Cite this version of the work
Andrew James Wiebe (2020). The influences of spatially variable rainfall and localized infiltration on groundwater recharge in a water management context. UWSpace. http://hdl.handle.net/10012/16476