employing distributive modeling approaches
Configuration 2: Dendritic conduit simulation
Conduit System Fissured System Porous System Spring discharge Limit of significance
expectancies. For assessing the difference between the “local” and the average values, deviations from the average value are provided in per cent. Figure 4.5 shows the range of deviations for each parameter. For the reference simulation (Chapter 4.3.1), the difference between Af and Ef at the observation well and the average values are 3.3% and 27.3% for the single conduit (configuration 1) and 2.5% and 12.2% for the dendritic simulation (configuration 2). The smaller diameter but more evenly distributed dendritic conduits lead to a more homogeneous distribution of groundwater ages and even more so of groundwater life expectancies. Since life expectancy cannot be measured in the field, the difference between the age values is the more important parameter which lies with 2–4%
well within the range of other uncertainties in the field.
The percentage of direct recharge rdir, which showed the largest overall influence on average values, has the largest influence on the spatial distribution. The maximum differences between the average age and the age in the observation well of 31.5% for configuration 1 and 89.5% for configuration 2 were reached for high rdir values above 90%. For high rdir values the newly recharged water does not enter the fissured system evenly distributed, i.e. as diffuse recharge at the top, but indirectly via the conduit system due to gradient inversion. Therefore, lower velocities are observed in the fissured system and the average age increases. For configuration 2 the observation well is situated significantly closer to the conduit system due to its overall higher lateral extent (Figure 4.1). It receives a high amount of young water from the conduit system which leads to the large differences in groundwater age compared to the average age. For configuration 1, the ages in the observation well tend to higher values compared to the average age until a critical value of 95% direct recharge is reached and the water flux from the conduit system is high enough to reach the observation well.
Further, small εf values of 5 m have a relatively high influence of 13.5% for configuration 2 and 8.9%
for configuration 1. The parameter maf can lead to a significant homogenization due to the smaller variety of different paths the water can take. For small maf values, the differences of Af and Ef between the observation well and the average value decrease to 0.1% and 2.3%, respectively, for configuration 1. For configuration 2, the age difference also decreases to 0.1%. The difference in life expectancies increases with decreasing aquifer thickness, however. Since the observation well is vertically positioned in the middle of the aquifer, it lies closer to the aquifer top and therefore the conduit system for a reduced vertical extent. Therefore, the life expectancy is reduced more strongly in the observation well. For a thickness of 10 m the water in the observation well has less than half the life expectancy of the average groundwater in the aquifer.
As can be expected, the groundwater ages at the spring represent a significantly better approximation of the average values than the individual local observation well age measurements.
For the reference simulation the age difference is only 0.7% for the single conduit and 0.4% for the
dendritic conduit configuration. For configuration 1 the only parameter that induces a significant difference of over 2% between the spring and the average age value is the percentage of direct recharge rdir (Figure 4.5c). The dendritic conduit simulation (configuration 2) shows higher sensitivities to most parameters. The two porosities θp and θf, the conduit system dispersivity εc and the porous-‐fissured exchange coefficient β do not influence the spatial distributions for either setup.
rdir and the total recharge r have a slightly higher influence for configuration 1 since the single conduit transmits water more efficiently (see also Chapter 4.3.3). All other parameters have higher influences on the difference between average fissured system age and spring water age for configuration 2. Only four of these influences are significant, however. A decrease in the vertical conduit position zc or in the fissured-‐conduit exchange coefficient α leads to slightly higher spring water ages and slightly lower average ages in the fissured system (Chapter 4.3.2.1). Decreasing values of the conduit roughness n or the fissured system dispersivity εf both increase the average age inside the fissured system and the spring water age simultaneously. Since the influence of both parameters on the average ages is higher, however, they also increase the difference between spring water and average age significantly.
Figure 4.5. Differences in age and life expectancy between the fissured system average and: (a) the observation well for the single conduit simulation (configuration 1) (b) the observation well for the dendritic
4.3.3 Transient age simulations – influence of recharge events
As described in Chapter 4.2.2 a theoretical recharge event was introduced into the model. Figure 4.6 shows the event and the corresponding groundwater age responses in the fissured system and at the spring. Initially, the spring water age represents the average age of the water in the fissured system.
While the response of the groundwater age at the spring is rapid and intense, the fissured system average age only decreases slightly and with a lag time of several days. For configuration 1 the minimum for the average fissured age will be reached after 26 days and for configuration 2 after 27.4 days (not shown) and only differ from the initial age by ca. 0.16 years. The influence on the age in the porous system (not shown) is in the order of 1x10-‐4 years and is not significant. Diffusive exchange processes are too slow to significantly react to the relatively short-‐term recharge event. It is noticeable that the spring age recovers much faster back to average values than the spring discharge.
For both configurations, the spring water age is higher than the fissured system average after the simulation period of 20 days, even though the spring water age is still ca. 0.1-‐0.2 years younger than at the beginning of the recharge event (Figure 4.6). This is due to a slight decrease in the average age of groundwater in the fissured system. The young recharge water is stored in the upper regions of the aquifer, while the conduits drain an important amount of water from greater depths (Figure 4.3).
Therefore, this effect is stronger for the single conduit than for the dendritic system.
The change of groundwater ages at the spring is significantly less pronounced for the dendritic conduit than for the single conduit configuration (Figure 4.6b). The dendritic conduit pattern results in flow regimes where conduit water flows into the fissured system for head gradient conditions directed towards the fissured system immediately after direct recharge events. The direct recharge pulse is not transmitted completely and directly to the spring, as is the case for the single conduit.
This effect also dampens spring discharge but this dampening is comparatively less well developed.
The variation in spring water age for the dendritic system is less than 50%, while the variation in discharge is almost 80% of that of the single conduit configuration. This shows the relatively large amount of fissured matrix water mobilized by the new recharge mixed into the spring discharge.
system, the amplitude of the maximum is well met, but the timing of the peak, which is a consequence of the spatial aquifer extent, is not reproduced (Figure 4.7). The calibrated parameter η (Eq. 4.9) has in all cases the value 1, which implies that the piston flow component of the model is switched off and a simple exponential model would have given the same goodness of fit. The simulated average transit time is with ca. 1.8 years significantly shorter than that of the numerical model with ca. 3.6 years. The exponential curves drop below those of the numerical model after ca.
70 years and do, therefore, not include the very old components of the porous system significantly contributing to longer average ages. The PFM fails in all cases to simulate the transit time curves because even the first peak could not be described by piston flow behaviour. The fit of the first peak can possibly be improved if direct recharge is considered in the simulation.
The DM fails to approximate the transit time curves at the spring. The fit is improved considerably if only the last part of the curve is considered for calibration, i.e. water with ages higher than 2.5 years as already suggested by Maloszewski et al. (2002). Nevertheless, average transit times are still underestimated by more than one year. For the fissured system ages monitored at the observation well, the statistical fit of the DM is with a difference of ca. 4×10-‐4 years only slightly worse than that of the EPM. The DM can nonetheless be viewed as superior since it approximates the average transit time at the observation well very well with only differences of 0.06 and 0.09 years for the dendritic and the single conduit system, respectively. The apparent dispersion parameter PD (Chapter 4.2.3) is in both cases ca. 1.9 showing a high importance of the dispersive component.
The transit time distribution for the porous system monitored at the observation well is very flat and broad due to the slow diffusion processes. Both, the EPM and the DM estimate the average transit time of the distributed model of 101.1 years almost correctly with a difference of 1.1 years for DM and only 0.3 years for EPM.