Upscaling Hydraulic Conductivity of a Forested Headwater Catchment Based on Information Loss in Terrain Curvature

Abstract:

Soil moisture plays a key role in the water and energy balance in soil, vegetation and atmosphere systems. Its spatial and temporal variability is strongly controlled by spatial and temporal variability in soil hydraulic properties and hydrologic forcings. Unfortunately, it is not clear how to parameterize soil hydraulic properties and hydrological processes as a function of scale, and how to test deterministic models with regard to epistemic uncertainties. In this study, high resolution simulations were conducted in the highly instrumented Wüstebach catchment with the TERENO observatory Eifel/Rur. Digital elevation models (DEMs) with a resolution of 1, 5, 10, and 20m were used to calculate information loss of profile curvature by spatial aggregation. An amplification factor was applied to upscale interpolated hydraulic conductivity to compensate for the loss of information content of curvature distribution caused by spatial aggregation from high resolution case (1m) to lower resolution cases (5m, 10m, and 20m). Four different cases with respect to the use of the amplification factor were analyzed and discussed: 1) no amplification factor applied, 2) amplification factor applied to all aggregated grid cells, 3) amplification factor applied to aggregated grid cells where high standard deviation of slopes exist in the 1m DEM, and 4) amplification factor applied to aggregated grid cells with low values of average slope over maximum slope within each 10x10 block in the 1m DEM. Our results indicate that the introduction of an amplification factor can effectively compensate the information loss leading to significantly improved model performances in different resolution cases. Furthermore, by utilizing cases 3) and 4) that apply the amplification factor only to specific aggregated grid cells, we effectively solved the problem of overcompensation for effective Ks that led to overestimate runoff and underestimate soil water content in case 2), and improved model performance in terms of both soil water content and runoff simulation.