The atmospheric science community spend huge efforts on investigating uncertainty in modeling the soil conditions in atmospheric models because errors in simulated soil states and fluxes may propagate into errors in atmospheric state variables and fluxes. The following sources of error have been identified:
• Discretization, vertical and temporal resolution
• Initial and boundary condition
• Subgrid-scale heteorgeneity
• Assumptions and/or parameterization concepts
• Uncertainty in soil physical parameters
• Data on soil type distribution
For a further discussion of the error sources mentioned in the first three bullets see also the respective subsections of section Simulating Frozen Ground.
Input of heat by precipitation, changes in insolation the soil surface due to cloudiness, changes in soil heat flux at the soil surface due to changes in wind speed can affect soil temperature, soil moisture, as well as soil moisture and heat fluxes (e.g., PaiMazumder et al., 2008). Since these changes in meteorological forcing occur on very short time scales, the temporal resolution like the vertical discretization has an impact on the accuracy with which diurnal change of soil temperatures and active layer depths can be predicted by soil models of atmospheric models. Figure 8 exemplarily shows results from simulations with different time steps and illustrates how temporal resolution can affect simulated soil temperature profiles on the long-term.
First of all, uncertainties in simulating soil temperature regimes may results from incorrectly simulated processes in the NWPM, CTM, GCM or ESM itself (e.g., Avissar and Pielke 1989, Calder et al., 1995, Molders et al., 1996, 1997, Niu and Yang 2004).
Various sensitivity studies aimed at detecting error sources related to assumptions and/or parameterization concepts (e.g., Robock et al., 1995, Cuenca et al., 1996, Shao and Irannejad 1999). As aforementioned, the force-restore method, for instance, has only limited ability to resolve soil horizons (see Fig. 4) and to simulate the vertical distributions of soil conditions and processes (e.g., diurnal variation of the freezing line).
In any soil model in NWPMs, CTMs, GCMs and ESMs prescribed soil parameters (e.g., Table 2) represent different soil types. Ideally, the soil characteristics should be mapped as vertical and horizontal three-dimensional continuous distribution to capture the gradients and mixtures in soil type within a grid-cell or a patch of same soil type within a grid-cell. However, soils are spatially heterogeneous for which attributing a single soil type to an area or patch of several square-kilometers as it is required in atmospheric models can be ambiguous, and is a potential error source. Using a wrong soil type, for instance, can cause errors in predicted near-surface air temperatures and humidity of more than 0.5K and 0.5g/kg even in a 24-hour simulation (Molders 2001). Assigning soil physical parameters to an area or patch is also ambiguous in GCMs or ESMs because soil surface properties can vary in time due to various events (e.g., burning of organic soils during wildfires, land avalanches, flooding, volcanic eruptions) or may be influenced by previous weather conditions (weathering) over centuries. Thus, prescribed fixed values of soil parameters for in nature time-dependent quantities may introduce uncertainty in climate and earth system modeling. Furthermore, the variability in some soil parameters is sometimes greater within the same soil type than across soil types (cf. Table 2). There is observational evidence from lysimeter studies that the heterogeneity within the same soil may cause differences in evapotranspiration and recharge of 112mm (14%) and 137mm (4%) in 5.6 years (Molders et al.,, 2003b).
Was this article helpful?