21st Century hydrological modeling for optimizing ancient water harvesting techniques

In order to increase dryland productivity, water harvesting techniques (WHT) have received renewed attention, leading to their massive implementation in marginal drylands. However, versatile tools to evaluate their efficiency under a wide range of conditions are often lacking. For two case studies in the arid and semi-arid central-northern zone of Chile, a fully coupled 3D surface-subsurface hydrological model based on the Richards’ and the Saint Venant equations was used to evaluate and improve existing water harvesting techniques using infiltration trenches (locally called ‘zanjas’). The model was parameterized with detailed runoff and soil-water content data collected during simulated rainfall from a 6 x 2 m experimental plot including a catchment area and infiltration trench at the arid site. Using seven responsive parameters identified by a global sensitivity analysis, surface and subsurface flow processes were calibrated simultaneously. The calibrated model accurately reproduced observed soil moisture contents (R2 0.92) and runoff amounts (R2 0.97), and represented the overflowing infiltration trench, which is a clear improvement over existing frameworks. A comparative analysis with a natural slope demonstrated that the trench was efficient in capturing runoff under high rainfall intensities, such as the one simulated, resulting in a significant decrease (46%) of runoff. However, when extended to natural rainfall seasons, runoff water harvesting was insufficient in dry, normal and wet years, while only under very wet conditions 55% of the potential runoff was effectively harvested and stored in the soil profile. As such, this test case shows the importance of correct water harvesting design to become an effective tool in dryland management, taking both soil physical and climatic constraints into account. The model was further tested on a much larger scale of two ca. 3 ha large watersheds at the semi-arid site, one with infiltration trenches and one without. Good agreement was observed between measured and simulated runoff at the watershed outlet.