3D hydrological simulation of a forested headwater catchment

Abstract

Soil moisture plays a key role in the water and energy balance in soil, vegetation and atmosphere systems. There is a grand need to increase global-scale hyper-resolution water-energy-biogeochemistry land surface modelling capabilities. High reliability simulation of soil water content using various kinds of numerical modeling tools has been studied in recent years. Multiple methods have been applied for more accurate parameterizations. In distributed hydrological modelling one often faces the problem that input data need to be aggregated to match the model resolution. However, aggregated data may be too coarse for the parametrization of the processes represented. This dilemma can be circumvented by the adjustment of certain model parameters. Unfortunately, it is not clear how to parameterize hydrological processes as a function of scale, and how to test deterministic models with regard to epistemic uncertainties. <br /> In this study, high resolution long-term simulations were conducted in the highly instrumented TERENO hydrological observatory of the Wüstebach catchment. Soil hydraulic parameters were derived using inverse modeling with the Hydrus-1D model using the global optimization scheme SCE-UA and soil moisture data from a wireless soil moisture sensor network. The estimated parameters were then used for 3D simulations of water transport using the integrated parallel simulation platform ParFlow-CLM. The simulated soil moisture dynamics, as well as evapotranspiration (ET) and runoff, were compared with long-term field observations to illustrate how well the model was able to reproduce the water budget dynamics. Different anisotropies of hydraulic conductivity were investigated to analyze how fast lateral flow processes above the underlying bedrock affect the simulation results. For a detail investigation of the model results the empirical orthogonal function (EOF) and wavelet coherence methods were applied. The EOF analysis of temporal-spatial patterns of simulated and observed soil moisture revealed that introduction of heterogeneity in the soil porosity effectively improves estimates of soil moisture patterns. The wavelet coherence analysis indicates that wet and dry seasons have significant effect on temporal correlation between observed and simulated soil moisture and ET. This study demonstrates the usefulness of the EOF and wavelet coherence methods for a more in-depth validation of spatially highly resolved hydrological 3D models.<br /> To further investigate how the reduction of local hydraulic gradients due to spatial aggregation can be partially compensated by increasing soil hydraulic conductivity. Information entropy concept was employed for the scale dependent parameterization of soil hydraulic conductivity. The loss of information content of terrain curvature as consequence of spatial aggregation was used to determine an amplification factor for soil hydraulic conductivity to compensate the resulting retardation of water flow. To test the usefulness of this approach, continuous 3D hydrological simulations were conducted with different spatial resolutions in the highly instrumented Wüstebach catchment. The results indicated that the introduction of an amplification factor can effectively improve model performances both in terms of soil moisture and runoff simulation. However, comparing simulated soil moisture pattern with observation indicated that uniform application of an amplification factor can lead to local overcorrection of soil hydraulic conductivity. This problem could be circumvented by applying the amplification factor only to model grid cells that suffer from high information loss. To this end, two schemes were developed to define appropriate location-specific correction factors. Both schemes led to improved model performance both in terms of soil water content and runoff simulation. Thus, the proposed scaling approach is useful for the application of next-generation hyper-resolution global land surface models.<br /> Then the output of Parflow-CLM simulated pressure was put into a particle tracking simulation platform SLIM-FAST to investigate the water particle migration and transit time distribution (TTD) at the Wüstebach testing site. Different model scenarios were conducted to investigate the effect of number of initial particles, dispersion parameters and hydraulic parameter upscaling schemes. A stable isotope tracer model TRANSEP was used in simulation of TTD for comparison purpose. Our results indicate that: 1) the initial number of water particles has no effect on TTD, unless the initial number exceeds a very high amount. 2) Higher αL leads to higher TTD, a 0.002 m αL gives best agreement with model TRANSEP. 3) Global and localized amplification factor applications lead to mixed results in estimates of transit time distribution and central location of particles plume. TTD estimation using particle tracking codes like SLIM-FAST can provide insight of the evaluation of 3D hydrological models like Parflow-CLM with respect to the correct parametrization of hydrological properties and the representation of water flow pathways.