Quantifying and understanding the source of recharge for alluvial systems in arid environments through the development of a seepage model

Conceptual differences in reach routing approaches where a) represents the kinematic wave equation used to simulate gaining rivers (blue line) and b) a seepage model used to simulate losing rivers (green line), where alluvial water is subject to evapotranspiration in the alluvial bed as well as percolates further to the groundwater store. [Display omitted] •Seepage model used to determine the rate of transmission loss.•Simulated seepage mainly controlled by reach slope.•Seepage the main source water for alluvial aquifers recharging bi-annually.•High uncertainty in the input precipitation data which is spatially variable.•Seepage modelling has the potential to improve simulations for arid conditions. In rural settings, groundwater is often the most utilised freshwater resource and can be subject to a number of factors that impact its availability and distribution. River seepage plays an important role in recharging alluvial groundwater systems, especially for arid environments where ephemeral rivers dominate. In this study, a river seepage component was developed for a spatially distributed rainfall/runoff model, the J2000 model. To represent river seepage, physical channel properties such as channel width, slope, length as well as a lumped riverbed hydraulic conductivity under ‘dry’ and ‘wet’ conditions were used. To facilitate and account for enhanced evapotranspiration in the alluvial channel, seepage firstly satisfied the potential evaporation demand and thereafter contributed to simulated recharge. The results show that by accounting for seepage, observed low flows were well simulated (Log E2: 0.74). Although, the input precipitation data and quality of the measured precipitation was associated with a high uncertainty and further impacted by a change in station density between the headwater and valley region (3 times higher in the headwaters). As a result, peak flows were less well represented (E2: 0.39) and a second downstream calibration was not possible. On average, river/reach seepage was estimated to be 8 mm d−1, where 150–300 mm yr−1 of seepage was simulated during wet years, exceeding simulated percolation on a reach level and a significant source over the catchment area. Given streamflow only comprised 1.3% of the total water balance, the efficiencies attained by this approach were good considering the extremely harsh climatic conditions. This approach is likely transferable to other arid environments and rivers with intermittent flow, where ri