Estimating groundwater dynamics at a Colorado floodplain site using historical hydrological data and climate information

Wednesday, 17 December 2014
Jinsong Chen, Lawrence Berkeley National Lab, Berkeley, CA, United States, Susan S. Hubbard, Lawrence Berkeley National Laboratory, Berkeley, CA, United States and Kenneth Hurst Williams, Colorado School of Mines, Golden, CO, United States
Developing a predictive understanding of how hydrological variations impact biogeochemical cycles in terrestrial environments is challenging due to the wide range of processes occurring across vast spatiotemporal scales and due to the uncertainty associated with future climate conditions. In this study, we develop a multi-scale time series model to estimate groundwater dynamics at a floodplain site using historical hydrological data and climate information. Our study is focused at Rifle, CO, where the US DOE Sustainable Systems 2.0 project is developing a genome-to-watershed reactive transport simulation capability and where snowmelt annually delivers a hydrological pulse to the floodplain system that significantly influences the water table and subsurface cycles of carbon and nitrogen. Although long-term predictions of biogeochemical cycling at the site require estimates of hydrological conditions, hydrological data include only a few years of groundwater elevation measurements, with river gage data available from a station located approximately 26 miles upstream.

To project future hydrological conditions at the site, we developed a multi-scale statistical model to combine both datasets. We first analyzed 47 years of hydrological data from the gage station to identify multi-frequency temporal patterns in the river stage and its relationship to climate factors (e.g., precipitation or temperature). We then developed empirical models to downscale the estimated hydrological information to daily discharge and subsequently transform them to groundwater dynamics at the downstream floodplain site. Our model provides a probabilistic estimation that is conditioned to the multi-scale hydrological and climate information. With the developed approach, we retrospectively estimate groundwater dynamics at the site for the past five decades as well as the associated uncertainty. Based on Colorado River Basin climate projections, we also predict mean and extreme hydrological conditions at the floodplain site over the next five decades. The projections are being used as input to watershed reactive transport models to predict how hydrological perturbations influence biogeochemical cycles from the genome through watershed scales.