4.3 Investigated methods for the assessment of groundwater recharge
4.4.6 Projection of future groundwater recharge
South. When employing the percolation rates at the soil level from the SWAT model and the recharge rates by the Abusaada& Sauter(2017) regression model, the MODFLOW model reproduces the seasonal behavior well. However, the groundwater model exhibits a less pronounced seasonal behavior in the South with the estimated recharge rates at the control plane groundwater table from the SWAT and PIM approach. Here, the model exhibits the expected seasonality in the northern parts.
In addition, we compare the simulated groundwater response to perennial extreme hydrological periods, i.e., consecutive wet or dry years. Here, the groundwater model can not replicate the hydraulic response to the rainy year of 1991/92 for all presented recharge estimations, especially for the northern part of the aquifer. Here, the groundwater flow model with the PIM recharge underestimates the rapid increase of hydraulic heads in 1991/92. In contrast, the groundwater flow model, in conjunction with recharge estimates ofAbusaada &Sauter(2017), reproduces the seasonal and multi-annual changes to the groundwater table well for the entire aquifer. However, also the estimate from Abusaada
&Sauter (2017) does underestimate the hydraulic response of 1991/92. In general, many groundwater models can not entirely replicate the event of 1991/92. This was also observed by Weinbergeret al. (2012), who concluded that all five groundwater flow models from the last 27 years failed to reproduce the hydraulic response to the extremely wet winter of 1991/92.
In summary, the presented groundwater model performs better with recharge estimates that assume shorter delay times through the vadose zone (i.e., SWAT at the soil level and the regression model fromAbusaada& Sauter (2017). While estimates that consider longer delay times in the vadose zone produce lead to dampened seasonal variations of the simulated groundwater table than observed (i.e., recharge at the groundwater table from the SWAT and PIM approach). It should be noted that the fully-saturated MODFLOW model was calibrated to reproduce the hydraulic response based on a recharge assessment that disregards the delaying effect of the vadose zone. Therefore, it is expected that the dissipated recharge rates at the groundwater table from SWAT and PIM slightly impair the groundwater model performance.
and SWAT. Here, similar storage effects and recharge dynamics, as in Figure 4.6, can be observed. The PIM approach indicates that high precipitation events, such as 12/2039 and 01/2062, produce strong percolation events and elevated recharge rates at the groundwater table that last several years (i.e., up to 4 years). This delayed recharge results from the temporary storage in a thick vadose zone that is considered discreetly with the PIM approach. However, both methods disagree regarding the predicted percolation rates in exceptionally rainy years since the SWAT approach considers the storage effects at the soil and surface more detailed (i.e., soil moisture storage, vegetation storage, and surface runoff) and does not explicitly consider the higher infiltration capacity of the exposed carbonate rocks. Here, the SWAT approach predicts increased surface runoff after the extreme rainfall in 01/2062. However, the projections are subject to uncertainty due to respective process abstractions in both approaches, and the correct percolation is probably in the middle of the two approaches. In future studies, the process representation of both methods may be improved and validated through measurements of extreme dry and wet conditions under present-day climate. Nevertheless, the PIM and SWAT approaches produce similar annual percolation rates for years with average and low precipitation depths. Therefore, both may provide a valid estimate of decadal recharge volumes.
Figure 4.10: Projection of future daily and annual percolation and groundwater recharge depths from 2020 to 2070 obtained from the PIM
and SWAT approach.
Finally, this study assesses the changes to average recharge volumes evaluated over a 20-year control period under present and future conditions. The applied climate model indicates a decrease in precipitation by14:5 %. Here, the empirical regression models byAbusaada & Sauter(2017),Guttman &Zukerman (1995),SUSMAQ(2004), and Zukerman (1999), and the PIM approach predict a mitigated effect of climate change (see Tab. 4.9), with the approach from Abusaada& Sauter(2017) and PIM projecting the lowest recharge decrease of 9 %.
Table 4.9: Projected decrease of precipitation and recharge in 2050-2070 compared to the reference period 1981–2001.
Decrease of recharge (%)
Decrease of recharge excluding 2062/01
Precipitation 14.1 27.2
Guttman& Zukerman(1995) 12.9 30.2
Zukerman(1999) 14.5 31.2
SUSMAQ (2004) 12.3 21.2
Abusaada& Sauter(2017) 8.9 14.0
SWAT 23.3 23.3
PIM 9 27.6
On the other hand, the projected recharge of the SWAT approach suggests an enhanced effect of climate change, with a decrease in recharge volumes by23 %. This results from the more detailed consideration of surface storage effects and surface runoff and less detailed consideration of infiltration along karst features, generating much lower recharge volumes in the exceptionally rainy month 01/2062. For instance, when excluding the unusually high recharge volumes of 2061/62, all methods suggest a much more pronounced effect of climate change (see Tab. 4.9). Nevertheless, in the end, all methods have shown a significant decrease in recharge. Also, as presented in Table 4.2, extreme rainfall may generate substantial surface runoff (i.e., runoff coefficients of up to 10 %), giving rise to the necessity to account for surface runoff as such events become more frequent under a climate with more extreme rainfall. However, the results above underline the high uncertainty associated with the recharge assessment under novel climate conditions and demonstrate the need for model ensembles in both the climate projection and the recharge assessment. For instance, the return time of exceptionally rainy years substantially influences the averaged values over a climatological normal. Therefore, reliable estimates of future recharge volumes require employing several methods and model intercomparisons.
This study lays the foundation for a comprehensive assessment of future recharge volumes.