Global Land surface models (LSMs) used in climate models param­eterize the exchanges of water and energy between land surface and the at­mosphere on a physical basis. However, water table depth (WTD), a basic hydrologic quantity, has traditionally been neglected. A rep­resentation of water table dynamics is integrated into an LSM, the Minimal Advanced Treatments of Surface Integration and Runoff (MATSIRO). 

Hereafter, the original MATSIRO and the MATSIRO with groundwater will be referred as MAT-ORI and MAT-GW respectively.

Model Evaluation in Illinois:

Fig. 1: Comparison of the simulations (GW: MAT-GW and Ori: MAT-ORI) against the observations (Obs) of (a) evapotranspiration, (b) total runoff, (c) GW recharge, (d) anomaly of water table depth, and (e) anomaly of moisture in top 2 m soil in the Illinois region.

  • ET is marginally improved compared to original MATSIRO with slight increases in the summer peak of dry years (e.g., 1991, 1994, 1997 and 1999, Figure 1a).
  • The MAT-ORI fails to reproduce observed runoff well (Figure 1b); in wet (dry) season it has higher (lower) peaks. 
  • MAT-GW simulates the seasonal variation of GW recharge fairly well (Figure 1c). The negative recharge in summer is reproduced. 
  • MAT-GW simulates the delay and amplitude of GW fluctuations more realistically. 
  • Summer drying of soil moisture is reduced by capillary flux from GW, and spring soil moisture peaks are also closer to observations.

Global Evaluation:

MAT-GW is applied at global scale to identify the regions where the representation of water table dynamics has a significant impact on hydrological simulations. 

Model Setup and Data:

Fig. 2: Schematic representation of soil column showing moisture flux exchange for (a) model not considering GW capillary flux (NC), and (b) model considering GW capillary flux (WC).

To isolate the effct of GW-supplied capillary flux, a pair of simulations was carried out. The moisture flux exchange between the saturated and unsaturated zones in two simulations is schematically shown in Figure 2.

  • The first model simulation neglects GW capillary flux to the unsaturated zone (hereafter referred to as NC),
  • The second one accounts for it (WC).

Global Runoff and Evapotranspiration:

Fig. 3: Runoff ratio (-) for (a) NC, (b) WC, (c) GSWP-2 [Dirmeyer et al., 2006], and (d) ISLSCP-2 [Hall et al., 2006].
  • The global pattern of runoffratio (instead of runoff) is compared in Figure 3. 
  • The spatial patterns of runoffratio in both simulations correspond well with the GSWP-2 and ISLSCP-2 in most global regions. 
  • However, in mid-to high latitudes of the northern hemisphere, runoffratio differs significantly between NC and WC, GSWP-2, and ISLSCP-2. 
  • Significant differences between the runoffratios in NC and WC are found in the Indian sub-continent, sub-Saharan Sahel, and southeastern Africa, where the influence of GW capillary flux is large. In these regions, runoffratio in WC is much closer to GSWP-2 and ISLSCP-2 than that in NC.

Fig. 4: Mean evapotranspiration (mm/mon) for (a) NC, (b) WC, (c) GSWP-2 [Dirmeyer et al., 2006], and (d) satellite-based estimate [Zhang et al., 2009, 2010].

  • A band of increased ET across central Europe and North America as simulated by both GSWP-2 and Zhang et al. (2010) , can be reproduced only in WC (Figure 4b).
  • However, ET in the Indian subcontinent and southeastern Africa in WC compares well with that by GSWP-2 only. Zhang et al. (2010) report underestimation of the satellite-based ET by~20% in the same regions, possibly causing the mismatch with WC simulation.

Basin-scale Evaluation:

River Discharge:

Fig. 5: Comparison of river discharge (mm/mon) in NC and WC simulations with the GRDC observations. Shaded region indicates values within ± 2 σ.
  • For both NC and WC, the seasonal cycle of river discharge is well reproduced in the majority of basins with NS >0.50In humid basins such as Amazon, Brahmaputra, and Chang-Jiang, the difference between NC and WC simulations is small. 
  • The consideration of capillary flux is more important in the Ganges and Zambezi River basins, where mean discharges decrease from 50 to 33 mm/mon, and 18 to 7 mm/mon, for NC and WC, respectively. 
  • The NS also improve from 0.62 (NC) to 0.85 (WC) in Ganges and from -4.03 to 0.84 in Zambezi.

Terrestrial Water Storage Anomaly (TWSA):

Fig. 6: Comparison of terrestrial water storage anomaly (TWSA) in NC and WC simulations with the GRACE TWSA in 20 selected river basins.
  • The difference between TWSA in NC and WC simulations is relatively small compared to river discharge. 
  • The improvement is significant in the Zambezi River basin for WC simulation. 
  • In humid basins such as Amazon, Chang-Jiang and Danube, both simulations reproduce GRACE data better than in drier basins (such as Colorado, Darling and Orange). 
  • In high latitude basins, TWSA is slightly earlier than GRACE, probably due to the simple snow scheme used in MATSIRO.

Global Groundwater Recharge and Water Table Depth:

Fig. 7: Global distribution of (a) GW recharge (mm/yr), and (b) water table depth (m) in the WC simulation.
  • Globally, mean GW recharge in WC is 29900 km3/yr.
  • The global pattern of WTD is found to be mainly controlled by recharge, baseflow and soil properties. Understandably, the WTD is deeper in dry regions whereas it is shallower in humid regions.
  • Additional smaller-scale heterogeneities in simulated WTD distribution are caused by the difference in soil properties, e.g., in Amazon, the grid cells with a loamy soil usually have a deeper WTD than that in clayey grid cells (Figure 10b).
  • The simulated WTD pattern does not bear any relationship with topography.


Following conclusions can be drawn:

  • Model evaluation conducted in Illinois shows significant improvement over the original MATSIRO in simulations of runoff, evapotranspiration (ET), and WTD. 
  • Global mean ET increases by ~9% when groundwater-supplied capillary flux is considered. 
  • Global groundwater recharge is estimated to be 29900 km3/yr.

  Top   Home 

HTML Comment Box is loading comments...