Figure 7. Two of the soil water characteristic curves used in this study (a), and mean infiltration at the land surface for 10 years of pumping (b).
However, in G1, the increase of WTD is less than 0.5 m in average after 10 years of pumping (Figure 4a), and thus the WTD in most of the modeling area is still in the critical depth range. Normally, in G1, the continuous increase of WTD with pumping would remain in effect on the variations of GST. Therefore, the attainment of dynamic equilibrium in GST variations (Figure 5a) seems incompatible with the above explanations for G2. In fact, this ‘paradox’ can be explained from the perspective of GW flow system as follows. The GW flow system becomes nonequilibrium with an obvious increase of WTD in the beginning of pumping (Figure 4a). With pumping, the increase of WTD promotes the infiltration at the land surface (or recharge at the water table) (Figure 7b) and decreases the discharge to streams (Condon and Maxwell, 2019), and thus a new dynamic equilibrium of the flow system is gradually achieved. Such a self-adjustment of the GW flow system has also been reported in previous studies (Cao et al., 2013; Condon and Maxwell, 2019). Hence, the GST increases with WTD in the beginning while it becomes dynamic stabilized when a new equilibrium of the GW flow system is achieved. Therefore, at a sustainable pumping rate such as that in G1, the dynamic equilibrium of GST can be achieved due to the self-adjustment of the GW flow system even though the WTD is still in the critical depth range. For an unsustainable pumping rate such as that in G2 and NCP, the combined effects of the self-adjustment of the GW flow system and the critical depth range theory should be responsible for the nonlinear variations of GST. In addition, the slight nonlinearity in Figure 4b indicates, in G2, the effect of the critical depth range is dominant while that of the self-adjustment of the flow system can be almost neglected.
  1. Effects of soil properties on the subsurface buffer
The discussion in section 3.1 indicates the buffer capacity of the subsurface on variations of GST (or the land surface heat fluxes), which is consistent to our hypothesis and other studies from a more general perspective (Condon and Maxwell, 2014b; Cuthbert et al., 2019; Smerdon and Stieglitz, 2006). For example, Condon and Maxwell (2014b) conceptualized the GW system as a buffer on the hydrological variations. Smerdon and Stieglitz (2006) described the damping effect of the subsurface buffer on the propagation of temperature signals from the land surface to the deep subsurface. The same results obtained from the Little Washita basin of the U.S. and the NCP of China (Yang et al., 2019), which are two totally different study areas, indicate the generalizability of our conclusions that deserve more attention in the areas with over-exploitation of GW worldwide. It is noted that the change of the buffer capacity with pumping is mainly due to the change of thermal properties in the subsurface. The thermal conductivity and volumetric heat capacity of the subsurface decrease with the decreasing soil moisture induced by pumping (Figure 8). Figure 8 was plotted based on the source code of ParFlow.CLM. Hence, when a hot or cold signal (e.g., the increase or decrease of the ground heat flux) is input at the land surface, the decreased thermal conductivity cannot propagate the signal to deep subsurface in time while the decreased volumetric heat capacity cannot effectively damp the signal due to the limited storage/release ability of heat. Therefore, the coupling depth (BBCP), which determines the heat capacity of the subsurface in terms of volume, should also be responsible for the buffer capacity of the subsurface.