Supplemental Material
This Supplemental Material contains both text, three figures and one table. The
text explains the procedure of sample collection and laboratory measurement, the
estimate of the slope of the evaporation line, and 3. the procedure of simulating
groundwater discharge after the 2000 Altyn earthquake and the 2001 Kunlun earthquake.
S1. Sample collection and laboratory measurement
At each sampling site five or more water samples were collected at the same time at
locations separated by a few meters. The water samples were analyzed in the State Key
Laboratory of Hydrology, Water Resources and Hydraulic Engineering, Hohai University, using a
Finnigan MAT 253 stable isotope ratio mass spectrometer, calibrated against the IAEA standard.
The standard deviations of the different measurements at each site range from 0.53 to 1.48‰ for
deuterium and from 0.22 to 0.33‰ for 18O. Fig. S1 shows an example of the measured values for
sites at or near the new lakes.
Fig. S1. Diagram shows laboratory measurements of D and 18O for individual water samples at
several sites at or near the new lakes. Also shown is the GMWL for reference.
S2. Slope of the evaporation line
Assuming isotopic equilibrium between the atmospheric moisture and the input water to
the lake, Gat (2010) showed that the slope of the evaporation line on a  D and  18O diagram
may be predicted from the following equation (expressed in terms of the equilibrium fractionation

factor  and the kinetic enrichment factor  [Gonfiantini, 1986]):
S 



( D 1)   D /(1 h)
( 18 O 1)   18 O /(1 h)
where h is the relative humidity and the subscripts D and 18O denote the respective chemical

species.
For  and  we assume the commonly used empirical relations (Gonfiantini, 1986),



 18O exp(5970.2T 2  32.801T 1 0.05223) ,
(S9)
D exp(2408T 2 64.55T 1 0.1687),
(S10)
 18O 14.2(1 h) /1000 ,
(S11)
D 12.5(1 h) /1000,
(S12)

where T is the absolute temperature and h is the relative humidity.
Air temperature at the new lakes during our field session varied between 24 and 42 oC;
the relative humidity was between 0.058 and 0.063. Assuming an average temperature of 30 oC
and an average relative humidity of 0.06, we obtain S = 3.7, which matches the slope of the
regression line given in Fig. 4. This result supports the hypothesis that the regression line in Fig.
4 represents an evaporation line with source compositions at its intersection with the GMWL.
S3. Simulating groundwater discharge after the 2000 Altyn earthquake and the 2001
Kunlun earthquake
Wang et al. (2004) used a simple model (Fig. S2) to calculate the excess discharge of
groundwater from an aquifer in response to coseismic recharge from mountains due to
earthquake-enhanced vertical permeability. The equation for calculating the excess discharge q is

(2r 1)  L'   (2r 1) 2  2 D 
2DQ
q
(1) r1 sin
exp
t

 2L
 
L L' r1
4L2


(S13)
where Q is the coseismic recharge, L the length of the aquifer between the discharge location (i.e.,
the new lakes) and the epicenter of earthquake, L' the length of the recharged section of the
aquifer, t the time since the earthquake, and D the hydraulic diffusivity of the aquifer. Among
these parameters only the epicenter distance and the time of the earthquakes are accurately
known. All the other parameters have large uncertainties. For example, D can vary by more than
ten orders of magnitude and Q and L' are unknown. The problem is further complicated by the
fact that both the Altyn earthquake and the Kunlun earthquake may have contributed to the excess
discharge and each contribution needs to be calculated separately and the parameters involved in
each may be different. Fortunately, the groundwater flow equation is linear and the results of the
two calculations may be algebraically summed.
In view of the large number of parameters and the significant uncertainties in some of
these parameters, we do not intend to embark on any detailed modeling here. Rather, we simply
show that the two basic characteristics of the observation, i.e., the timing of the increase in the
lake area (Figure 3a) and the amount of discharge required to maintain the lakes in the presence
of intense seasonal evaporation, may be satisfied by a reasonable set of parameters.
Fig. S2. Model used in simulating excess groundwater discharge induced by coseismic recharge
of aquifer due to earthquake-enhanced vertical permeability (modified from Wang et al., 2004).
A few words may be needed to justify the choices of parameters listed in Table S1. We
let L' be equal to the approximate width of the mountain ranges. We let the lower bound of Q be
constrained by the annual loss of lake water to evaporation. Finally, we let D be constrained by
the rise time of the increase in the lake area. Considering that the hydraulic conductivity K for
fractured aquifers can be as high as ~ 10-3 m/s (Mouldon et a., 2001) and the specific storage Ss is
generally small, < 10-6 m-1, and since D = K/Ss, we consider that the assumed value for D, 102
m2/s, as listed in Table S1, may be reasonable.
Table S1. Parameters used in simulating groundwater discharge after the Altyn and Kunlun
earthquakes.
Parameters
Time of earthquake
Epicenter distance to lakes: L (km)
Length of saturated aquifer: L’ (km)
Hydraulic diffusivity: D (m2/s)
Coseismic recharge: Q (m3)
Altyn earthquake
January 31, 2000
170
50
50
2x109
Kunlun earthquake
November 14, 2001
450
50
100
4x109
Figure S3 shows the simulated discharge following each earthquake and the algebraic
sum of the two. Time of the Altyn earthquake is zero in this diagram; the delayed onset of the
discharge increase after this earthquake is due to the distance between the recharged aquifer and
the new lakes. The greater delay after the Kunlun earthquake is partly due to its later time of
occurrence and partly due to its greater distance to the lakes. The algebraic sum of two simulated
discharges is also shown in Figure 3a. The figure shows that, if the region is seismically quiet,
discharge to the lake will decrease with time to about a fifth of its presence rate in 35 years.
Fig. S3. Simulated discharge with parameters in Table S1. Curve A shows the increased
discharge following the Altyn earthquake, curve K that following the 2001 Kunlun earthquake,
and top curve the sum of the two.
References
Gat, J.R., 2010, Isotope Hydrology: A Study of the Water Cycle, Imperial College Press, London.
Gonfiantini, R., 1986, Environmental isotopes in lake studies, in Handbook of Environmental
Isotope Geochemistry, eds.: P. Fritz and J.-Ch. Fontes, Elsevier, Amsterdam, 113-168.
Muldoon, M.A., J.A. Simo, K.R. Bradbury, 2001, Correlation of hydraulic conductivity with
stratigraphy in a fractured-dolomite aquifer, northeastern Wisconsin, USA,
Hydrogeology Journal, v. 9, 570–583.
Wang, C.-Y., C.-H. Wang, and M. Manga, 2004, Coseismic release of water from mountains:
Evidence from the 1999 (Mw = 7.5) Chi-Chi, Taiwan, earthquake, Geology, v. 32, 769772.