Combining remote sensing datasets and ensemble-based data assimilation to reduce model non-uniqueness and enhance implementation of cascade routing and re-infiltration concept

Re-infiltration is a process that occurs in concert with direct runoff and evaporation processes, typically during and after rain showers. The latest version of MODFLOW (MODFLOW 6) allows representing re-infiltration process in the novel cascade routing and re-infiltration (CRR) concept [1]. The CRR controls transfer of rejected rainfall infiltration and groundwater exfiltration (Fig. 1) from the upslope areas to the adjacent downslope areas, where that water can be: (i) evaporated, (ii) re-infiltrated back to the subsurface, or (iii) discharged as direct runoff into a surface water body (e.g. streams).

The re-infiltration process is influenced by vertical hydraulic conductivity of topsoil, while the water partitioning between evaporation and direct runoff is uncertain and non-unique. To reduce that non-uniqueness, the remote sensing MODIS actual evapotranspiration (MODIS-ETa) product was applied as an additional source of information to complement the traditional hydraulic heads in the transient Sardon catchment MODFLOW 6 model with CRR concept.

The MODIS-ETa product and hydraulic head measurements were assimilated into the Sardon catchment model (~80 km2) over a 3-year daily simulation. The hydraulic heads from 14 piezometers and the accumulated 8-daily MODIS-ETa, distributed over 359 pixels (500x500m size), were available for assimilation. Model input uncertainty was represented by grid-cell-scale parameterization of hydraulic properties, yielding thousands of unknowns to be conditioned through data assimilation. The data assimilation was carried out with an iterative ensemble smoother available in the latest version of the PEST code (PEST++, [2]).

The value of including the MODIS-ETa dataset into the data assimilation scheme allowed for reduction of model non-uniqueness and better implementation of the CRR concept in the simulation. The results of the data assimilation indicate the combined MODIS-ETa and hydraulic head measurements provided substantial conditioning of the uncertain model inputs and ultimately reduced the uncertainty in the model simulated outcomes.

References
[1] M. G. Daoud, M. W. Lubczynski, Z. Vekerdy, and A. P. Francés, “Application of a novel cascade-routing and reinfiltration concept with a Voronoi unstructured grid in MODFLOW 6, for an assessment of surface-water/groundwater interactions in a hard rock catchment (Sardon, Spain),” Hydrogeol. J., 2022, doi: 10.1007/s10040-021-02430-z.
[2] J. T. White, R. J. Hunt, M. N. Fienen, and J. E. Doherty, “Approaches to highly parameterized inversion: PEST++ Version 5, a software suite for parameter estimation, uncertainty analysis, management optimization and sensitivity analysis,” 2020. doi: 10.3133/TM7C26.