Introduction
Western and central Siberia's thermokarst (thaw) lakes extend over a
territory spanning over 1 million km2 (half of the western Siberia
lowland, 0.5 million km2, and all of the northern Siberia lowland, 0.84
million km2). They are highly dynamic hydrochemical systems that receive
chemical elements from the atmosphere and surrounding peat soil and exchange
greenhouse gases (GHGs) with the atmosphere, delivering dissolved carbon and
metals to adjacent hydrological systems. Because more permafrost will
continue to thaw due to climate warming, which is heavily intensified in this
region (Frey and Smith, 2005), the directions and magnitude of lakes–rivers
exchange processes may be significantly modified, seriously affecting the
biogeochemical fluxes both on land and in the coastal zone of the Arctic
Ocean.
In contrast to the relatively good understanding of western Siberia
thermokarst lakes functioning during the active (summer) season (Walter et
al., 2006, 2008; Walter Anthony et al., 2012; Audry et al., 2011; Pokrovsky
et al., 2011, 2013, 2014; Shirokova et al., 2009, 2013; Karlsson et al.,
2012, 2014; Manasypov et al., 2014), the intra-annual variations of lake
water chemistry including the glacial period and spring flood have remained,
up to the present time, virtually unexplored. At the same time, the glacial
season is extremely important in GHG regulation in boreal and subarctic
lakes, due to significant accumulation of gases under the ice and their
liberation during the ice melting (Karlsson et al., 2013). Similarly, the ice
formation period provides important insights into the solute concentration
and colloid coagulation mechanisms, given the main particularity of shallow
(0.3 to 1.0 m deep) thermokarst lakes in western Siberia – their full
freezing during winter (Pokrovsky et al., 2011, 2014). The possibility of
complete freezing of western Siberia thermokarst lakes contrasts with the
majority of other circumpolar water bodies, such as from the Kolyma lowland
(Walter Anthony et al., 2014), the Lena Delta (Boike et al., 2013) and the
Mackenzie Delta plain (Tank et al., 2009, 2011; Grosse et al., 2013; Walter
Anthony and Anthony, 2013), which often have a depth of more than 2 m and as
such do not freeze to the bottom, while exhibiting chemical and thermal
stratification of the water column. Another important difference of
thermokarst western Siberia lakes from well-studied river delta/river valley
lakes is the lack of a connection to the underground network in the former.
Therefore, lake connection to the hydrological network may occur only via
surface flow (Kirpotin et al., 2008), without subsurface flow. Among all
studied circumpolar water bodies, shallow ponds of the north of Canada
(Laurion et al., 2010; Negandhi et al., 2013; Bouchard et al., 2014) seem to
be most similar to western Siberian thermokarst lakes. However, in a recent
compilation of studied circumpolar ponds (Rautio et al., 2011), only 1 region
among 16 (thaw ponds of Boniface, Quebec) exhibits a pH of 5.4 ± 0.6, a
Cond. (specific conductivity) of 18 µS cm-1 and DOC of
13.4 ± 4.7, which is comparable to highly diluted
(5 < Cond < 20 µS cm-1), acidic (3.5 ≤ pH ≤ 5.5), and humic (10 ≤ DOC ≤ 40 mg L-1)
western Siberia water bodies.
Towards filling the gap in our knowledge of seasonal variations of
thermokarst lake chemical composition, we present in this work results of
analysis of water and ice sampled in thermokarst lakes of various sizes, from
several m to several km in diameter, during four main hydrological seasons:
June, August, October and February. Our primary goal was to better understand
the thermokarst lake biogeochemical functioning, which should allow us to
constrain the impact of lake water metal and carbon cycling on river water
composition and greenhouse gas exchange with the atmosphere in the course of
a year. From a larger perspective, we aimed at the understanding of seasonal
patterns of dissolved organic carbon and metal micronutrient concentrations
in these shallow but highly abundant water bodies, different from previously
studied glacial and deep thermokarst/yedoma lakes. This knowledge should
allow predictions of phytoplankton activity, sedimentation and microbial
respiration on the annual scale, necessary for evaluation of the net
ecosystem exchange under various climate change scenarios. To this end, we
addressed the following specific questions: (i) is there a statistically
significant difference in major and trace element concentrations between
different seasons, within a broad range of lake sizes? (ii) How significant
is lake size and lake water residence time control for lake water chemical
composition during different seasons? (iii) What are the mechanisms and
degrees of element differentiation during ice formation and full water column
freezing? (iv) How significant is water dilution during spring melt and what
are the consequences for river water feeding by lakes during this period?
Study site area with symbols showing the position of sampled lakes
and small water bodies in different seasons. Different colors correspond to
different elementary ecosystems.
![]()
Study site description, sampling, analytical and statistical
methods
Our study site is located in the central part of western Siberia
(63.5∘ N, 75.4∘ E, Nojabrsk administrative region) within
the northern taiga geographical sub-zone (Fig. 1). It contains discontinous
permafrost over the late Pleistocene sand and clay deposits that are covered
by a layer of peat that is 1–2 m thick. All the lakes in this study are
located at watershed divides between adjacent rivers. The water bodies ranged
from a few m to several km in diameter and had a similar depth of
1.0 ± 0.5 m under normal precipitation/evaporation conditions
(450 mm). During dry summer, the lakes can decrease their depth twofold
(Pokrovsky et al., 2013). The morphology, hydrology and water balance of
thermokarst lakes have been extensively studied during four field campaigns
in June, August, and October 2013 and February 2014. The detailed depth
mapping of ∼ 20 large lakes was performed via GPS-echosounder from a
rubber boat; the denivelation and the direction of the water flow were
measured using a levelling network in several narrow profiles; the depth and
density of snow were measured over a model site of 500 m × 500 m
that included lakes, a palsa bog, some streams and an adjacent forested
riparian zone. The water residence time was calculated from the annual
precipitation and evapotranspiration measured at the neighboring meteostation
of the Russian Hydrometeorological Station, the annual runoff of the
territory (between 200 and 250 mm, Novikov, 2009, and Frey et al., 2007) and
the water volume of the lakes measured on-site using a GPS-echosounder.
Major hydrophysical and hydrochemical parameters of studied lakes.
N
E
S, m2
T, ∘C
O2, mgL-1
O2, % sat
R, µScm-1
pH
DIC, ppm
DOC, ppm
CI, ppm
SO4, ppm
June
Z-43
63∘13,580′
75∘43,476′
1.8
6.5
N.D.
N.D.
N.D.
3.43
0.749
22.5
0.1532
0.9452
Z-44
63∘13,567′
75∘43,465′
0.071
7.5
N.D.
N.D.
N.D.
3.25
0.562
33.9
0.1458
0.9957
Z-45
63∘13,559′
75∘43,467′
4.9
6.0
N.D.
N.D.
N.D.
3.43
0.784
26.5
0.1466
0.8256
Z-46
63∘13,513′
75∘43,527′
31 400
7.2
N.D.
N.D.
N.D.
3.28
0.396
13.6
0.0663
0.3680
Z-47
63∘13,521′
75∘43,608′
1.8
5.5
N.D.
N.D.
N.D.
3.51
0.598
23.6
0.0978
0.5896
Z-48
63∘13,492′
75∘43,726′
78.5
5.9
N.D.
N.D.
N.D.
3.49
0.433
17.2
0.0306
0.4516
Z-49
63∘13,475′
75∘43,737′
1.0
5.2
N.D.
N.D.
N.D.
3.37
0.545
22.1
0.1285
0.7087
Z-50
63∘13,436′
75∘43,707′
96 163
7.5
N.D.
N.D.
N.D.
3.77
0.550
6.5
0.0848
0.4372
Z-51
63∘13,492′
75∘43,584′
25 434
7.2
N.D.
N.D.
N.D.
3.08
0.435
14.7
0.0131
0.448
Z-52
63∘13,505′
75∘43,444′
17 663
6.6
N.D.
N.D.
N.D.
3.36
0.390
13.3
0.0678
0.369
Z-53
63∘13,512′
75∘43,374′
9.6
5.8
N.D.
N.D.
N.D.
3.47
0.447
17.4
0.0289
0.576
Z-54
63∘13,521′
75∘43,389′
0.13
5.5
N.D.
N.D.
N.D.
3.11
0.855
27.2
0.1337
0.752
Z-56
63∘47′55.0′′
75∘31′21.0′′
5671.6
3.8
9.1
86
12
4.2
0.431
14.1
0.1781
0.097
Z-57
63∘47′54.4′′
75∘31′6.55′′
3000
5.7
8.86
82
12
4.35
0.423
9.8
0.1637
0.616
Z-58
63∘47′56.5′′
75∘31′19.9′′
314
6.2
8.19
79
13
4.4
0.438
10.2
0.1616
0.595
Z-59
63∘47′56.2′′
75∘31′14.8′′
5024
6.0
8.3
77
30
4.5
0.437
24.9
0.0501
0.476
Z-60
63∘47′55.0′′
75∘31′14.1′′
314.
6.6
8.23
78
21
4.0
0.459
15.2
0.0657
0.301
Z-61
63∘47′56.1′′
75∘31′12.9′′
176.6
6.5
8.2
77
17
4.05
0.410
12.5
0.0809
0.589
Z-62
63∘48′28.0′′
75∘30′52.1′′
1 766 250
5.9
8.25
79
12
4.2
0.477
13.
0.2440
0.843
Z-63
63∘48′04.7′′
75∘31′13.4′′
28.3
8.5
7.33
74
20
4.5
0.394
24.5
0.024
0.233
Z-64
63∘48′05.0′′
75∘31′14.5′′
19.6
10.5
8.54
88
23
4.3
0.524
24.4
0.040
0.615
Z-65
63∘48′29.5′′
75∘30′50.5′′
11 304
8.3
8.6
86
14
4.4
0.449
17.1
0.157
0.731
Z-66
63∘48′00.2′′
75∘30′53.8′′
314.
9.9
8.55
85
15
4.8
0.429
12.6
0.064
0.964
Z-67
63∘47′59.3′′
75∘30′46.8′′
78.5
11.4
8.38
90
28
5.01
0.402
18.8
0.050
0.385
Z-68
63∘47′59.0′′
75∘30′46.8′′
176.6
10.2
8.46
88
22
4.4
0.378
16.9
0.043
0.725
Z-69
63∘47′58.7′′
75∘30′48.8′′
314.
8.5
9.2
99
14
3.9
0.458
12.9
0.149
0.341
Z-70
63∘47′58.0′′
75∘30′48.3′′
78.5
12.5
8.25
91
18
4.7
0.378
17.9
0.040
0.505
Z-71
63∘47′58.9′′
75∘30′47.9′′
0.0079
N.D.
N.D.
N.D.
N.D.
N.D.
0.610
22.6
1.484
0.843
Z-72
63∘48′04.5′′
75∘30′43.7′′
1.8
16.5
7.5
84
30
3.92
0.390
18.2
0.170
0.860
Z-73
63∘48′15.5′′
75∘31′07.4′′
706.5
12.9
8.7
98
17
3.8
0.372
12.8
0.069
0.339
Z-74**(A)
63∘48′39.7′′
75∘31′59.9′′
176 625
6.3
9.2
94
9
4.0
0.426
7.8
0.146
0.647
Z-75
63∘48′39.9′′
75∘31′59.4′′
0.2
17.2
8.1
82
34
3.7
0.854
35.4
0.102
0.776
Z-77
63∘46′58.0′′
75∘39′12.2′′
113
12.4
7.95
87
32
4.4
0.432
27.5
0.102
0.247
Z-78
63∘47′01.4′′
75∘39′14.0′′
3.1
10.8
7.12
77
41
3.7
0.419
27.7
0.037
0.419
Z-79
63∘47′01.0′′
75∘39′05.4′′
314
12.7
7.6
86
18
3.5
0.304
18.3
0.022
0.201
Z-80
63∘46′56.6′′
75∘39′13.1′′
1256
11.8
7.8
85
23
3.3
0.303
19.4
0.093
0.276
Z-81
63∘47′01.7′′
75∘39′02.2′′
31 400
12
8.3
93
19
4.0
0.323
15.8
0.304
0.163
Z-82
63∘47′07.2′′
75∘39′02.0′′
113
14.4
8.4
97
19
4.2
0.359
15.3
0.009
0.216
Z-83**(B)
63∘47′18.9′′
75∘38′30.1′′
1 766 250
3.4
8.6
82
11
4.6
0.858
6.5
0.117
0.209
Z-84
63∘47′06.7′′
75∘38′10.8′′
314
15.1
7.8
9
19
3.9
0.355
17.2
0.009
0.244
Z-85**(C)
63∘47′05.4′′
75∘38′10.4′′
70 650
1.9
7.99
8
7
4.6
0.433
5.5
0.067
0.471
Z-87
63∘47′41.5′′
75∘33′01.0′′
1256
13.7
8.43
95
12
4.5
0.338
11.6
0.060
0.493
Z-88
63∘47′43.9′′
75∘33′07.5′′
177
2.5
8.02
98
N.D.
4.6
0.325
17.9
0.031
0.260
Z-89
63∘47′44.2′′
75∘33′09.0′′
707
N.D.
7.4
98
N.D.
N.D.
0.285
9.7
0.189
0.215
Z-90**(D)
63∘47′45.5′′
75∘33′09.7′′
159
N.D.
7.9
95
N.D.
N.D.
0.283
12.1
0.138
0.146
Z-91
63∘47′49.2′′
75∘33′10.3′′
380
N.D.
7.8
95
23
4.4
0.329
25.3
0.045
0.247
Z-92
63∘48′01.0′′
75∘33′28.5′′
78.5
20
7.7
95
12
4.7
0.334
16.6
0.014
0.255
Z-93**(E)
63∘48′11.0′′
75∘33′54.4′′
2 268 650
16
8.36
99
9
4.8
0.402
18.5
0.245
0.756
Z-94
63∘48′16.8′′
75∘33′47.2′′
19.6
18
8.36
104
24
3.9
0.304
20.2
0.041
0.272
Z-95
63∘48′29.7′′
75∘33′55.0′′
4.9
17.8
7.4
98
31
4.1
0.305
26.
0.045
0.732
Z-96*
63∘48′30.2′′
75∘33′54.7′′
580 586
9.3
7.9
88
8
4.7
0.550
8.6
0.166
1.011
Z-97
63∘47′39.0′′
75∘33′02.3′′
3
19.6
7.9
91
18
4.34
0.285
21.9
0.032
0.316
Z-98
63∘46′17.2′′
75∘40′58.6′′
0.79
15.6
7.27
92
19
4.5
0.282
20.7
0.108
0.317
Z-99
63∘46′18.5′′
75∘40′42.9′′
314
17.9
6.44
86
N.D.
4.45
0.298
20.3
0.136
0.027
Z-100
63∘46′14.4′′
75∘40′34.2′′
3.1
15.7
6.7
84
N.D.
4.15
0.311
23.7
0.081
0.322
Z-101
63∘46′14.4′′
75∘40′32.5′′
785
13.3
6.8
8
N.D.
4.0
0.347
10.
0.055
0.252
Z-102
63∘46′10.1′′
75∘40′21.5′′
4.9
12.5
5.7
7
N.D.
3.6
0.353
28.8
0.151
0.465
Z-103
63∘46′09.2′′
75∘40′11.4′′
961.6
19.5
6.8
88
N.D.
4.2
0.291
20.9
0.075
0.318
Z-104
63∘46′07.8′′
75∘40′07.0′′
0.13
17.9
5.3
71
N.D.
3.6
0.304
45.
0.182
0.619
Z-105
63∘46′07.7′′
75∘40′04.4′′
706.5
14.6
6.7
80
N.D.
4.2
0.294
18.4
0.037
0.218
Z-106
63∘46′01.4′′
75∘39′57.0′′
12.6
18.2
6.8
90
N.D.
4.1
0.270
22.4
0.043
0.305
Z-107
63∘46′03.7′′
75∘39′54.3′′
196 250
15.6
6.8
85
N.D.
4.31
0.293
25.
N.D.0
N.D.
Continued.
N
E
S, m2
T, ∘C
O2, mgL-1
O2, % sat
R, µScm-1
pH
DIC, ppm
DOC, ppm
Cl, ppm
SO4, ppm
August
Z-1**(E)
63∘47′57.9′′
75∘33′49.2′′
1200 000
N.D.
N.D.
N.D.
N.D.
5.45
0.422
16.55
0.089
0.3259
Z-2*
63∘48′11.1′′
75∘33′23.2′′
580 586
N.D.
N.D.
N.D.
N.D.
5.63
0.440
14.40
0.187
0.6155
Z-3**(A)
63∘48′27.6′′
75∘32′29.3′′
51 300
N.D.
N.D.
N.D.
N.D.
4.88
0.376
13.86
0.111
0.2534
Z-4
63∘51′32.11′′
75∘37′10.6′′
33 000
N.D.
N.D.
N.D.
N.D.
4.88
0.363
7.73
0.039
0.0804
Z-5
63∘49′32.1′′
75∘34′00.6′′
74 000
N.D.
N.D.
N.D.
N.D.
4.92
0.429
12.8
0.423
0.1880
Z-6**(B)
63∘47′15.2′′
75∘37′58.3′′
138 000
N.D.
N.D.
N.D.
N.D.
6.20
2.470
11.7
0.480
0.1711
Z-7**(C)
63∘47′06.7′′
75∘38′12.2′′
60 000
N.D.
N.D.
N.D.
N.D.
4.67
0.385
13.7
0.076
0.1023
Z-8
63∘47′02.3′′
75∘38′02.5′′
6000
N.D.
N.D.
N.D.
N.D.
4.41
0.370
27.7
0.156
0.342
Z-9
63∘47′01.1′′
75∘39′05.5′′
359.8
N.D.
N.D.
N.D.
N.D.
4.20
0.403
27.4
0.098
0.200
Z-10
63∘46′56.8′′
75∘39′10.9′′
1481.9
N.D.
N.D.
N.D.
N.D.
4.21
0.334
51.8
0.197
0.326
Z-11
63∘46′53.8′′
75∘39′18.6′′
1766 250
N.D.
N.D.
N.D.
N.D.
5.28
0.392
8.4
0.182
0.283
Z-12
63∘47′36.2′′
75∘33′10.0′′
1256
N.D.
N.D.
N.D.
N.D.
4.66
0.392
11.6
0.048
0.104
Z-13
63∘47′38.5′′
75∘33′07.1′′
1963
N.D.
N.D.
N.D.
N.D.
4.16
0.570
35.4
0.090
0.306
Z-14
63∘47′43.8′′
75∘33′12.5′′
1963
N.D.
N.D.
N.D.
N.D.
4.33
0.310
18.5
0.037
0.114
Z-15**(D)
63∘47′45.9′′
75∘33′11.2′′
1809
N.D.
N.D.
N.D.
N.D.
4.19
0.301
19.5
0.038
0.072
Z-16
63∘47′49.3′′
75∘33′10.6′′
201.0
N.D.
N.D.
N.D.
N.D.
4.16
0.386
35.7
0.077
0.119
Z-17
63∘47′56.3′′
75∘31′26.1′′
2374.6
N.D.
N.D.
N.D.
N.D.
4.94
0.341
52.9
1.178
0.486
Z-18
63∘47′56.6′′
75∘31′21.9′′
707
N.D.
N.D.
N.D.
N.D.
4.35
0.373
17.6
0.082
0.049
Z-19
63∘47′56.0′′
75∘31′12.7′′
415
N.D.
N.D.
N.D.
N.D.
4.17
0.335
23.4
0.088
0.091
Z-20
63∘48′00.9′′
75∘30′54.8′′
78.5
N.D.
N.D.
N.D.
N.D.
4.65
0.306
38.6
0.242
0.159
Z-21
63∘47′58.9′′
75∘30′50.8′′
491
N.D.
N.D.
N.D.
N.D.
4.50
0.319
35.6
0.118
0.245
Z-22
63∘47′58.2′′
75∘30′53.8′′
7850
N.D.
N.D.
N.D.
N.D.
4.25
0.316
31.0
0.157
3.115
Z-23
63∘47′56.8′′
75∘30′43.7′′
4416
N.D.
N.D.
N.D.
N.D.
4.56
0.300
28.7
0.084
3.116
Z-24
63∘47′57.4′′
75∘30′33.9′′
6359
N.D.
N.D.
N.D.
N.D.
4.35
0.321
18.9
0.142
0.126
Z-25
63∘47′59.3′′
75∘30′35.1′′
5408
N.D.
N.D.
N.D.
N.D.
4.88
0.312
18.6
0.073
0.164
Z-26
63∘47′57.5′′
75∘30′30.8′′
314
N.D.
N.D.
N.D.
N.D.
4.48
0.340
54.7
0.193
0.280
Z-27
63∘43′59.9′′
75∘49′48.3′′
2375
N.D.
N.D.
N.D.
N.D.
5.88
0.865
9.3
0.031
2.016
Z-28
63∘44′12.7′′
75∘48′55.6′′
3524
N.D.
N.D.
N.D.
N.D.
6.10
1.251
15.3
0.197
0.183
Z-29
63∘47′26.1′′
75∘33′50.5′′
152
N.D.
N.D.
N.D.
N.D.
5.54
0.465
22.8
0.766
0.246
Z-30
63∘47′28.2′′
75∘33′51.8′′
31 400
N.D.
N.D.
N.D.
N.D.
5.59
0.360
38.2
0.608
4.797
Z-31
63∘47′42.9′′
75∘31′32.4′′
5024
N.D.
N.D.
N.D.
N.D.
5.66
0.357
10.6
0.55
0.855
October
OZ-1
63∘13′31.6′′
75∘43′32.6′′
96 163
N.D.
N.D.
N.D.
N.D.
4.21
1.552
48.2
0.181
0.110
OZ-2
63∘13′22.6′′
75∘43′22.1′′
17 663
N.D.
N.D.
N.D.
N.D.
4.57
0.713
32.1
0.183
0.379
OZ-3
63∘13′18.8′′
75∘43′22.1′′
491
N.D.
N.D.
N.D.
N.D.
5.29
1.835
52.7
0.282
0.232
OZ-4
63∘13′15.1′′
75∘43′19.0′′
70 650
N.D.
N.D.
N.D.
N.D.
4.69
0.618
12.3
0.128
0.153
OZ-5
63∘13′16.4′′
75∘43′34.6′′
1963
N.D.
N.D.
N.D.
N.D.
3.75
0.536
27.5
0.115
0.128
OZ-6
63∘13′20.8′′
75∘43′44.8′′
12 266
N.D.
N.D.
N.D.
N.D.
4.29
0.445
14.1
0.078
0.133
OZ-7
63∘13′16.5′′
75∘43′17.1′′
49 063
N.D.
N.D.
N.D.
N.D.
4.99
0.533
13.5
0.129
0.164
OZ-8
63∘13′06.5′′
75∘43′22.4′′
707
N.D.
N.D.
N.D.
N.D.
4.74
1.828
39.1
0.309
0.308
OZ-9
63∘13′03.4′′
75∘43′15.7′′
2826
N.D.
N.D.
N.D.
N.D.
4.33
0.488
17.0
0.106
0.037
OZ-10
63∘13′23.2′′
75∘43′02.6′′
177
N.D.
N.D.
N.D.
N.D.
4.52
3.492
35.4
1.252
0.178
OZ-11
63∘47′54.1′′
75∘31′33.3′′
110
N.D.
N.D.
N.D.
N.D.
3.99
0.723
24.3
0.541
0.036
OZ-12
63∘48′03.0′′
75∘31′33.8′′
196 250
N.D.
9.3
86
N.D.
4.57
1.264
29.2
0.201
0.746
OZ-13
63∘48′18.2′′
75∘31′30.4′′
196 250
N.D.
10.5/8.3
95/74
8
4.77
0.574
24.3
0.165
0.601
OZ-14
63∘48′28.0′′
75∘31′28.7′′
707
N.D.
12.3/1
100/94
6
4.21
0.460
16.8
0.122
0.060
OZ-15
63∘48′36.8′′
75∘31′38.1′′
41 527
N.D.
8.8/7.2
74/64
N.D.
4.84
0.650
26.3
0.224
0.248
OZ-16**(A)
63∘48′37.3′′
75∘32′28.2′′
503 300
N.D.
7.4/6.2
67/56
N.D.
4.87
0.581
13.7
0.194
0.314
OZ-17*
63∘48′21.0′′
75∘33′08.6′′
580 586
N.D.
9.1/6.9
8/66
N.D.
5.01
0.507
15.3
0.284
0.316
OZ-18**(E)
63∘48′15.2′′
75∘34′12.0′′
1207 000
N.D.
8.1/7.3
8/68
N.D.
4.69
0.495
13.7
0.193
0.708
OZ-19
63∘47′52.9′′
75∘33′19.8′′
9499
N.D.
8.6/5.8
78/55.2
N.D.
4.10
0.602
29.6
0.074
0.247
OZ-20**(D)
63∘47′47.0′′
75∘33′11.3′′
3485
N.D.
11.7/10.3
101/97
N.D.
4.01
0.596
24.9
0.179
0.052
OZ-21
63∘47′47.2′′
75∘32′58.9′′
11 304
N.D.
10.1/4.7
92/44
N.D.
4.91
0.530
45.6
0.464
1.410
OZ-22
63∘46′56.1′′
75∘38′53.7′′
2462
N.D.
8.25/5.87
72/53
N.D.
3.96
0.508
28.6
0.246
1.566
OZ-23
63∘46′58.4′′
75∘38′45.2′′
5024
N.D.
4.2/3.34
38/30
N.D.
3.81
0.662
27.7
0.828
0.878
OZ-24
63∘46′59.0′′
75∘38′33.8′′
3215
N.D.
5.1/4.5
46/40
N.D.
3.99
0.595
16.5
0.140
0.121
OZ-25**(C)
63∘47′02.6′′
75∘38′21.2′′
53 630
N.D.
9.2/4.4
85/41
N.D.
4.06
0.556
16.3
0.095
0.204
OZ-26
63∘47′1.8′′
75∘38′24.2′′
5672
N.D.
8.3/4.
75/38
N.D.
4.85
0.820
31.1
0.234
2.909
OZ-27
63∘47′12.1′′
75∘38′34.64′′
17 663
N.D.
7.8/5.4
79/49
N.D.
3.87
0.466
33.1
0.164
0.535
OZ-28**(B)
63∘47′21.9′′
75∘38′28.2′′
1396 000
N.D.
9.75
87
N.D.
6.90
2.920
12.3
0.543
1.686
OZ-29
67∘47′1.2′′
75∘38′51.1′′
28 339
N.D.
4.2/5.4
37/49
N.D.
3.98
0.488
37.5
0.170
0.606
OZ-30
63∘47′04.2′′
75∘38′56.1′′
13 267
N.D.
4.2/5.8
37/51
N.D.
3.98
0.444
34.4
0.187
0.669
OZ-31
63∘46′57.8′′
75∘39′02.1′′
962
N.D.
4.5/3.9
4/35
N.D.
3.77
0.469
23.1
0.229
0.186
February
LF1 bottom
63∘48′19.6′′
75∘33′19.2′′
580 586
N.D.
N.D.
N.D.
N.D.
N.D.
0.496
30.3
0.792
3.440
LF4 bottom
63∘48′44.7′′
75∘34′37.3′′
351 335
N.D.
N.D.
N.D.
19
4.655
0.778
19.0
0.235
1.678
LFS bottom
63∘49′09.4′′
75∘34′44.1′′
1 743 000
N.D.
N.D.
N.D.
31
4.778
1.022
32.9
0.281
2.973
Oxygen concentrations in samples OZ-13 to OZ-31
represent surface/bottom values. Highlighted lakes (A, B, C, D, and E)
** were sampled during all open water seasons. R stands for specific
conductivity and N.D. for non-determined. * stands for a lake sampled
during all four seasons.
Water samples were collected from the PVC boat for large lakes or directly
from the lake center for small (< 50 m diameter) water bodies (thaw
ponds) during June and August 2013 and from the ice during the October 2013
and February 2014 sampling campaigns. A list of sampled water bodies and
their main hydrological and hydrochemical characteristics is presented in
Table 1. The sampling and filtration methods used in this study, as well as
the chemical analysis techniques, are identical to those utilized in our
previous studies (Pokrovsky et al., 2011). An ultraclean sampling procedure
was used for all manipulations in the field (Shirokova et al., 2010). Water
samples were filtered on-site through sterile single-use
Minisart® filter units (Sartorius, acetate
cellulose filters) with a pore size of 0.45 µm. For TE analysis,
samples were acidified to pH = 2 with double-distilled HNO3.
Dissolved oxygen, pH, and Eh were measured on-site with uncertainties of
5 %, 0.02 units, and 2 mV, respectively, using a
WTW® oximeter and a
Hanna® portable pH meter with an Eh/pH
electrode.
Major anion concentrations (Cl- and SO42-) were measured by ion
chromatography (HPLC, Dionex ICS 2000) with an uncertainty of 2 %. DOC
was analyzed using a Carbon Total Analyzer (Shimadzu TOC 6000) with an
uncertainty lower than 3 %. Major and trace elements were determined with
an ICP-MS Agilent ce 7500, routinely used in our laboratory for the analysis
of samples from boreal organic-rich lakes (cf. Pokrovsky et al., 2013).
Indium and rhenium were used as external standards. The international
geostandard SLRS-5 (Riverine Water Reference Material for Trace Metals
certified by the National Research Council of Canada) was used to check the
accuracy and reproducibility of each analysis (Yeghicheyan et al., 2014). We
obtained good agreement between replicate measurements of SLRS and certified
values (relative difference < 10 % SD for repeated measurements),
except for B and P (30 %). While P was discarded for further analyses, B
concentrations in most lakes are a factor of 3 to 7 higher than those in the
SLSRS-5, and thus were retained. In addition to TE analysis using the Agilent
quadropole instrument, 60 % of samples were processed with an
ultrasensitive Element XR ICP-MS instrument operated in a low and medium
resolution mode. Using this ICP-MS greatly increased the detection limits of
a number of elements and improved the precision of the analyses while
avoiding interferences. The uncertainty of the Element XR analysis was
≤ 5 %, while its detection limit was a factor of 100 lower than the
traditional (Agilent) instrument. The average agreement between the two
ICP-MS instruments for the majority of the TE was 10–15 %.
The lake ice was sampled in February 2014 from the central part of the lakes
using titanium circular ice coring. The ice core was cut using a Ti saw into
10 cm slices, melted at room temperature and immediately filtered through
0.45 µm filters. Before and after fieldwork, blank samples were run
by filling the pre-cleaned PVC container with MilliQ water and submerged Ti
blades at neutral pH, and leaving it to react for several hours. No
detectable contamination of Ti or any major and trace elements was observed.
The ice filtrate was processed in the same way as the lake water. In three
lakes (out of ∼ ten studied in February), the bottom layer of unfrozen
water below the ice column was collected for analysis. Chemical composition
of Fe-rich coagulates collected from the ice surface in early June was
measured by ICP-MS after full dissolution of the solid using a microwave acid
digestion procedure (Stepanova et al., 2014). Particulate organic carbon was
quantified using the TOCCHS instrument. The freeze-dried
precipitates were characterized by scanning electron microscopy (SEM) using a
Jeol JSM840a, and by X-ray diffraction using an INEL CPS 120 CoKα.
The element concentrations were analyzed using best fit functions based on
the least squares method, Pearson correlation and one-way ANOVA (SigmaPlot
version 11.0/Systat Software, Inc). Regressions and power functions were used
to examine the relationships between the elemental concentrations and the
lake surface areas. The ANOVA was carried out using Dunn's method because
each sampling period contained a different number of water bodies. Principal
component analysis (PCA) was used for the ensemble of sampled lakes and for
each season individually, to reduce the number of variables and to detect the
structure in the relationships between the elements. The data consisted of
∼ 130 lake water samples grouped into three distinct seasons (spring,
summer and autumn). Statistical PCA analyses were applied in order to derive
a distinctive view of the influence of various parameters, notably the
seasons, on the lake water chemical composition variability. Both normed and
non-normed PCA treatment was attempted. Statistical analysis considered each
chemical element as a variable (35 in total) for all lakes. For this step,
the STATISTICA package (http://www.statsoft.com), which is also
designed to compute and render graphics, was used to interpret the spatial
structures.
Results
Effect of the lake size on dissolved carbon and related elements concentration in the water column
The list of sampled water bodies with their main hydrophysical and
hydrochemical characteristics and their geographical localization are given
in Table 1 and Fig. 1, respectively. Within the range of 100 to
2 × 106 m2 lake size, the effect of lake size on major
and trace element concentration is not strongly pronounced. The correlation
coefficient was below 0.5 for all elements over all seasons, suggesting a
quite weak if not negligible impact of the lake size on lake water chemical
composition (Table S1 of the Supplement).
The highly variable number of sampled lakes during each season was basically
due to the difficulties in sampling logistics and the access to the site.
Only five lakes (labeled A, B, C, D and E in Table 1) could be collected
during all three open water seasons. In winter, we could not sample more than
three lakes (both ice core and bottom water). All small lakes
(< 1000 m2) were frozen solid in October (autumn period). We did not
focus in this work on small-size (< 500 m2) lakes in summer, because
the small water objects were thoroughly studied in our previous work
(Shirokova et al., 2013). Besides, many small ponds (10–100 m2) were
dried in August 2013. For this reason, the size range of the sampled lakes is
different among seasons.
Only in spring, for the range of 0.1 to 2 × 106 m2 lake
size, was Pearson's coefficient statistically significant (p< 0.05) and
suggested a decrease in the following element concentration with the
logarithm of the lake size increase: DOC (-0.69), V (-0.43), Cr
(-0.54), Ni (-0.58), Cu (-0.63), Ga (-0.71), As (-0.53), Cd
(-0.69), and Pb (-0.8). Such a semi-logarithmic negative relationship
suggests a rapid decrease in the concentration within the small water bodies
and its relative stability in large lakes. Indeed, this trend was mainly due
to the increase in certain element concentrations in the smallest
(< 10–100 m2) water bodies rather than a steady evolution of
concentrations in medium-size and large water bodies.
DOC concentration in thermokarst lakes of various sizes during four
observation seasons.
![]()
The DOC concentration exhibited two maxima, at 1000–10 000 m2 surface
area in summer and autumn and in micro-depressions (< 1 m2) in
spring (Fig. 2). The nature of the DOC was different among different
seasons, because the slope of the UV280 – [DOC] dependence increased in
the order spring (0.024) < summer (0.030) < autumn (0.035) < winter
(0.0354). The ratio of UV280 to [DOC] during four sampled seasons
demonstrated the lowest values in spring and the highest in autumn–winter,
being independent of the lake size (Fig. 3).
None of the seasons demonstrated a statistically significant dependence
between the element concentration and the water body size above 100 m2
(p> 0.05). There was quite high variability in the major element
concentration in lakes of the same size range, typically over 2 orders of
magnitude, as can be seen for Ca and Si in Fig. S2a and b in the Supplement,
respectively. There was a general increase in the lake water pH with the
increase in the lake size (statistically significant at p< 0.05), most
visible during summer (Fig. 4). The small-size lakes sampled in spring
exhibited the lowest pH, between 3.5 and 4.5, whereas the larger lakes
(> 1000 m2) in summer demonstrated a pH of 5 to 6. Similar to major
cations, the major anion concentration did not demonstrate any discernable
trend (p > 0.05) with the lake surface area; only Cl- yielded the
minimal concentration for the lakes of ca. 100 m2 surface area.
Dissolved inorganic carbon (DIC) concentration was a factor of 30 lower than
[DOC], without a statistically significant trend with the lake size (not
shown). The illustration of the lake surface area effect on trace element
concentration during all four sampled seasons is given in Fig. S3 for Fe,
Al, Mn, Zn, Cu, Pb, Mo, V, As and Sb, and the primary data are listed in
Table S2.
UV absorbance properties of thermokarst lake waters. The
UV280 / DOC ratio increases in the order spring < summer
< autumn, suggesting the progressive increase in aromaticity that is
independent of the lake size (p > 0.05). The average values of
UV280nm / [DOC] are equal to 0.024 ± 0.0037,
0.030 ± 0.0072, 0.035 ± 0.0069, and
0.0354 ± 0.0068 L mg-1 cm-1 in spring, summer, autumn and
winter, respectively.
![]()
Increase in lake water pH with the increase in the lake size over
four studied seasons. Note that the lowest pH is observed in spring and is
highest in summer. This likely reflects the dominance of allochthonous
organic input in the lake in spring, notably in the smallest water bodies,
and autochthonous processes of some phytoplankton and macrophyte activity at
the end of the summer.
![]()
There was a limited degree of element correlation with either DOC or Fe
considering all seasons of the year, as follows from Pearson pair correlation
(Table S3). Examples of mostly pronounced correlations between Fe and DOC,
As and Fe, and Cd and DOC are illustrated in Fig. S4.
The statistical treatment of the data did not reveal significant (p> 0.05)
links between TE concentration and UV280 nm. Rather poor correlations shown
in Fig. S4 of the manuscript are illustrative for both UV and DOC. The
concentration of TE in peat unfrozen horizon and in ground vegetation (green
and brown mosses) in the discontinuous permafrost zone of western Siberia are
rather similar (Stepanova et al., 2015). As a result, UV280 can help to
distinguish between the element leaching from vegetation as opposed to peat
leaching. Rather, the similarity of specific UV absorbances across the
seasons and lake sizes strongly suggests the dominance of terrestrial (soil
and litter) sources of OC with minimal aquagenic DOM.
Considering all seasons simultaneously, several groups of elements could be
distinguished based on this correlation analysis:
elements correlated with Fe (RFe > RDOC): Cl, SO4, Mg, Ca, Sr, Ba, Rb, Al, Ga, Cr, Mn, Co, Ni, Si, As, REEs and
U,
elements correlated with DOC rather than with Fe (RDOC > RFe): Cu, Zn, Cd and
Pb,
elements strongly correlated with Al (R > 0.6 at p< 0.05): Be, Si, K, Rb, Cs, Sr, Ba, Ti, V, Cr, Co, Cu, Ga, As, Sb, REEs, W and U, and
elements exhibiting high pair correlations (R ≥ 0.9 at p< 0.05): Ga–Al, Cr–Al,
Hf–Th.
Note that all elements exhibited better inter-correlations in spring compared
to other seasons (Table S3).
Lake freezing during the glacial period (October to the end of May) and ice-layered structure formation
Observations during the beginning of the ice formation (October 2013) yielded
important and novel features of the thermokarst lake ice cover evolution
during the glacial period. Given the closed basins of all sampled lakes
underlain by permafrost without a connection to the groundwater source (i.e.,
Pokrovsky et al., 2011, 2014), the water was expected to be under excess
pressure under the ice. This was confirmed by direct observations in October:
the water always rose upward and spread over the ice surface after drilling
the first 10–20 cm of the ice layer. Upon the thickening of the ice, cracks
formed on the ice cover and the water from the deeper layers seeped and
spread over the ice surface, thus forming organic- and Fe-rich multilayered
ice up to 30 cm thick. The wind acting on still unfrozen freshly ejected
water created ice ripples spreading over a surface of 10–50 m2
(Fig. S5). The full freezing of ejected organic-rich water led to browning
of the ice and presumably produced significant coagulation of colloids. The
products of this coagulation could be notably seen at the end of the spring,
during massive ice melt (Fig. S6).
In situ oxygen analysis with a submersible O2 sensor in June and August
demonstrates saturation with atmospheric oxygen close to 90 ± 10 %
(Table 1). In October, under the ice, we found only 40–80 % saturation
of the water layer, with some redox stratification within 20 and 50 cm of
the unfrozen water column. There was a 50 % drop in O2 concentration
at the sediment–water interface relative to the bottom of the ice core.
Generally, the smaller the water body, the lower the oxygen concentration,
although a straightforward relation between the size of the lake and the
oxygen saturation in October could not be established (p> 0.05).
The ice core analysis with the resolution of 5–10 cm demonstrated either an
enrichment of the surface ice layers in dissolved elements, or their
non-systematic variation over the full depth of the ice core (Table S4).
The enrichment, by a factor of 2 to 5, of the first 0–30 ± 10 cm was
detected for DOC and most metals such as Al, Fe, Ni, Co, Cu, Zn and Pb, as
illustrated for selected metals in Fig. 5. Manganese demonstrated the
largest variation along the ice core depth. Its concentration decreased from
3 to 10 ppb at the very surface down to less than 0.1 ppb below 50 cm. In
contrast, many trace elements (Cd, Cs, Sb, Ti, Zr, Hf, Th, U, REEs) did not
demonstrate a statistically significant (p< 0.05) trend of concentration
with depth for three sampled ice cores. Ca2+, SO42-, DIC,
Cl-, and Na+ also demonstrated nonsystematic variation with depth.
In winter, only three (out of ∼ ten) sampled lakes yielded the liquid
water presence above the sediment. The thickness of the unfrozen layer ranged
from 10 to 20 cm, with an ice thickness of 60 to 80 cm.
Distribution coefficient of soluble components between the lake
water and the ice (average of three samples). The lower part of the ice
column, formed in quasi-equilibrium with the remaining bottom water, was used
for this estimation.
Element
Kd (water)/(ice)
DIC
1.64 ± 0.37
DOC
10.3 ± 7.0
Cl-
7.58 ± 3.0
SO4
44.1 ± 10.9
Na
21.2 ± 3.4
Mg
17.9 ± 5.4
Al
93.1 ± 32.9
Si
12.4 ± 5.2
K
45.8 ± 24.1
Ca
10.2 ± 5.2
Ti
50 ± 20
V
22.9 ± 3.5
Cr
20.7 ± 8.9
Mn
117 ± 35
Fe
67.8 ± 12.4
Ni
16.6 ± 2.7
Cu
46.8 ± 21.5
Ga
34.9 ± 12.1
As
46.4 ± 9.8
Rb
41.7 ± 14.4
Sr
20.2 ± 12.7
Cd
7.4 ± 1.7
Sb
10.9 ± 1.9
Cs
5.7 ± 2.1
Ba
27.2 ± 9.3
La
41.3 ± 16.2
Ce
58.8 ± 21.0
Pr
45.8 ± 38.5
Nd
52.4 ± 9.6
Sm
26.7 ± 13.9
Eu
49.0 ± 32.5
Gd
21.6 ± 1.6
Dy
47.5 ± 34.5
Ho
14.4 ± 11.2
Er
23.4 ± 7.9
Yb
19.1 ± 12.4
Pb
41.5 ± 12.2
U
11.5 ± 2.9
DOC, Fe, Mn and Pb concentrations in the ice cores of thermokarst
lakes sampled in February 2014. LF-1 and LF-5 refer to the sampled lakes
(Table 1).
![]()
Using the element concentration measured in the remaining water at the
sediment–ice interface and the bottom 10 cm layer of the ice core, element
distribution coefficients between the ice and the remaining fluid can be
assessed. Note that the obtained values can be considered to be preliminary
estimations, given the very limited number of samples. The agreement in
Kd (water/ice) between three sampled lakes for which both fluid
and bottom ice were available is reasonably good for a number of major
elements, as can be seen from the average with 2 SD (n=3) listed in
Table 2. The lowest Kd (water/ice) is observed for DIC
(1.64 ± 0.37) and Cl (7.6 ± 3.0). The DOC distribution varied
significantly among the lakes (Kd (water/ice) is between 5 and
21). Al, Ti, Mn, Ce and some REE yielded Kd (water/ice) between
50 and 100. The majority of other trace elements exhibit the Kd
(water/ice) between 10 and 40 without any discernable link to element
chemical properties and degree of binding to DOM.
Comparison of the distribution coefficients of a trace element (TE)
normalized to Fe between the solid precipitates collected in June from the
ice surface of a large lake (Fig. S6) and filtered water (black columns) and
those measured between colloids (1 kDa – 0.45 µm) and the
LMW< 1kDa fraction in a large thermokarst lake (Pokrovsky et
al., 2011) of the discontinuous permafrost zone (grey columns).
![]()
The most significant transformation of the lake water chemical composition
presumably occurs during the full freezing of the water column over the
winter period. At the end of the winter, during massive snowmelt, products of
colloidal coagulation are largely present at the surface of the ice blocks
within the lake borders (Fig. S6). The coagulates collected in large
(∼ 106 m2) thermokarst lakes in early June are essentially
composed of iron oxohydroxide and organic matter with a significant amount of
co-precipitated trace elements. The aggregates are XRD-amorphous and contain
17.5 ± 0.5 % of POC.
Based on total chemical analysis of solid co-precipitates, we calculated the
distribution coefficients of a trace element (TE) between the Fe-rich
coagulates and the lake water, normalized to Fe (as the major component of
the coagulates):
Kd=[TE]/[Fe]coagulate/[TE]/[Fe]lake water,
where [TE]/[Fe]lake water represents the ratio of average TE
and Fe concentrations in large lakes at the end of summer, before the
beginning of ice formation. The Kd values of major and trace
elements range from 0.05 to 0.5 (Fig. 6). Although based on a limited
number of samples, the values of the distribution coefficient obtained in the
present study are in reasonable agreement with those reported for colloids
(1 kDa – 0.45 µm) and the LMW (< 1 kDa) fraction of another
large thermokarst lake located in the discontinuous permafrost zone
(Pokrovsky et al., 2011) as well as the rivers and streams of the boreal
non-permafrost zone (Vasyukova et al., 2010). For the majority of dissolved
elements, the Kd of particles are within a factor of 2 different
from the previously reported Kd of colloids.
Average concentrations and standard deviation (±2σ) of
major and trace elements during different seasons of the year for two size
classes of thermokarst lakes, 100–500 000 and >500 000m2.
Considering all elements simultaneously, the difference between the two size
classes is statistically significant (p=0.007, 0.0008 and 0.00007 for
spring, summer and autumn), as follows from the Wilcoxon t test.
Element
Units
100–500 000 m2
>500 000m2
Spring
Summer
Autumn
Winter
Spring
Summer
Autumn
Winter
DOC
mgL-1
15.2 ± 5.1
25.6 ± 13.3
28 ± 10.9
N.D.
13.2 ± 6.6
12.9 ± 2.8
16.8 ± 6.2
27.4 ± 6.03
DIC
mgL-1
0.38 ± 0.07
0.42 ± 0.2
0.79 ± 0.67
N.D.
0.5 ± 0.18
0.49 ± 0.17
1.15 ± 0.93
0.77 ± 0.21
Cl-
mgL-1
0.09 ± 0.06
0.13 ± 0.11
0.21 ± 0.16
N.D.
0.18 ± 0.05
0.21 ± 0.14
0.28 ± 0.13
0.44 ± 0.25
SO42-
mgL-1
0.37 ± 0.2
0.44 ± 0.41
0.43 ± 0.38
N.D.
0.69 ± 0.27
0.33 ± 0.15
0.75 ± 0.5
2.33 ± 0.65
Specific cond.
µScm-1
17.9 ± 5.98
N.D.
7 ± 1
N.D.
9.8 ± 1.47
N.D.
N.D.
25 ± 6
pH
4.2 ± 0.34
4.72 ± 0.58
4.17 ± 0.38
N.D.
4.44 ± 0.29
5.49 ± 0.44
5.21 ± 0.86
4.72 ± 0.06
Be
µgL-1
0.004 ± 0.001
0.007 ± 0.004
0.013 ± 0.008
N.D.
0.002 ± 0.001
0.012 ± 0.003
0.013 ± 0.005
B
µgL-1
42.4 ± 41.8
28.7 ± 26.5
76.6 ± 66.3
N.D.
8.22 ± 7.61
30.1 ± 20.3
223 ± 169
776 ± 413
Na
µgL-1
866 ± 854
1220 ± 1197
910 ± 608
N.D.
997 ± 782
896 ± 351
1396 ± 925
6519 ± 3818
Mg
µgL-1
99.5 ± 35
229 ± 186
150 ± 76.6
N.D.
62.2 ± 38.4
125 ± 12.7
199 ± 48.3
316 ± 137
Al
µgL-1
34.7 ± 18.3
101 ± 69.4
181 ± 114
N.D.
41.7 ± 15.4
137 ± 77.7
238 ± 110
484 ± 169
K
µgL-1
124 ± 123
207 ± 203
302 ± 227
N.D.
256 ± 144
272 ± 95.6
447 ± 253
2251 ± 1254
Ca
µgL-1
211.4 ± 92.7
289 ± 160
309 ± 152
N.D.
213 ± 167
470 ± 367
661 ± 438
893 ± 406
Ti
µgL-1
0.512 ± 0.27
1.16 ± 0.664
1.76 ± 1.3
N.D.
0.569 ± 0.229
4.46 ± 2.75
4.44 ± 4.01
3.63 ± 1.24
V
µgL-1
0.272 ± 0.088
0.363 ± 0.252
0.484 ± 0.361
N.D.
0.415 ± 0.205
0.848 ± 0.374
0.882 ± 0.402
1.33 ± 0.298
Cr
µgL-1
0.138 ± 0.075
0.366 ± 0.166
0.477 ± 0.274
N.D.
0.157 ± 0.045
0.457 ± 0.125
0.651 ± 0.238
1.14 ± 0.423
Mn
µgL-1
6.25 ± 4.39
15.7 ± 9.92
14.6 ± 8.95
N.D.
4.14 ± 3.52
17.9 ± 4.02
21.1 ± 8.77
26.3 ± 11.7
Fe
µgL-1
84.4 ± 57.3
251 ± 162
230 ± 111
N.D.
127 ± 63
273 ± 240
336 ± 146
447 ± 108
Co
µgL-1
0.037 ± 0.018
0.103 ± 0.05
0.141 ± 0.064
N.D.
0.029 ± 0.016
0.095 ± 0.035
0.093 ± 0.039
0.215 ± 0.034
Ni
µgL-1
0.108 ± 0.067
0.316 ± 0.149
0.299 ± 0.171
N.D.
0.106 ± 0.04
0.557 ± 0.425
0.351 ± 0.141
0.477 ± 0.145
Cu
µgL-1
0.252 ± 0.207
0.47 ± 0.289
0.789 ± 0.691
N.D.
0.208 ± 0.07
0.634 ± 0.439
1.2 ± 0.768
3.56 ± 2.01
Zn
µgL-1
27.1 ± 23.3
62.2 ± 47
114 ± 97.8
N.D.
27.8 ± 26.2
108 ± 97.3
314 ± 276
794 ± 423
Ga
µgL-1
0.01 ± 0.004
0.023 ± 0.016
0.042 ± 0.033
N.D.
0.007 ± 0.004
0.037 ± 0.023
0.06 ± 0.031
0.146 ± 0.069
As
µgL-1
0.371 ± 0.093
0.636 ± 0.191
0.612 ± 0.288
N.D.
0.371 ± 0.069
0.626 ± 0.057
0.658 ± 0.182
0.944 ± 0.3
Rb
µgL-1
0.147 ± 0.141
0.213 ± 0.205
0.328 ± 0.294
N.D.
0.275 ± 0.069
0.388 ± 0.166
0.92 ± 0.536
2.07 ± 0.963
Sr
µgL-1
2.84 ± 1.42
5.91 ± 3.96
8 ± 5.53
N.D.
3.12 ± 1.6
8.94 ± 4.72
13 ± 6.22
29.5 ± 12.6
Zr
µgL-1
0.243 ± 0.2
0.1 ± 0.068
0.216 ± 0.165
N.D.
0.488 ± 0.011
0.309 ± 0.138
0.612 ± 0.359
2.06 ± 1.8
Nb
µgL-1
0.015 ± 0.011
0.014 ± 0.012
0.024 ± 0.023
N.D.
0.013 ± 0.002
0.024 ± 0.012
0.02 ± 0.015
0.024
Mo
µgL-1
0.012 ± 0.007
0.01 ± 0.009
0.021 ± 0.015
N.D.
0.023 ± 0.008
0.021 ± 0.003
0.021 ± 0.009
0.033 ± 0.017
Cd
µgL-1
0.01 ± 0.007
0.024 ± 0.015
0.028 ± 0.013
N.D.
0.0005
0.017 ± 0.01
0.025 ± 0.019
0.04 ± 0.008
Sb
µgL-1
0.04 ± 0.014
0.046 ± 0.012
0.077 ± 0.031
N.D.
0.037 ± 0.01
0.064 ± 0.015
0.08 ± 0.027
0.14 ± 0.044
Cs
µgL-1
0.008 ± 0.007
0.008 ± 0.007
0.007 ± 0.006
N.D.
0.003 ± 0.002
0.013 ± 0.007
0.023 ± 0.013
0.027 ± 0.011
Ba
µgL-1
49.5 ± 48.9
126.4 ± 81.2
410 ± 393
N.D.
63.5 ± 61.8
139 ± 77.3
466 ± 367
1107 ± 446
La
µgL-1
0.01 ± 0.007
0.037 ± 0.031
0.052 ± 0.041
N.D.
0.003 ± 0.002
0.072 ± 0.02
0.071 ± 0.016
0.221 ± 0.016
Ce
µgL-1
0.021 ± 0.014
0.09 ± 0.083
0.122 ± 0.114
N.D.
0.001 ± 0.0009
0.15 ± 0.052
0.14 ± 0.02
0.561 ± 0.04
Pr
µgL-1
0.004 ± 0.003
0.011 ± 0.01
0.016 ± 0.015
N.D.
N.D.
0.018 ± 0.006
0.017 ± 0.003
0.065 ± 0.007
Nd
µgL-1
0.016 ± 0.011
0.047 ± 0.045
0.059 ± 0.055
N.D.
0.028 ± 0.007
0.078 ± 0.026
0.074 ± 0.011
0.215 ± 0.022
Sm
µgL-1
0.007 ± 0.005
0.012 ± 0.011
0.018 ± 0.017
N.D.
0.011 ± 0.005
0.019 ± 0.007
0.018 ± 0.006
0.063 ± 0.005
Eu
µgL-1
0.013 ± 0.012
0.019 ± 0.018
0.036 ± 0.034
N.D.
0.022 ± 0.018
0.014 ± 0.005
0.035 ± 0.032
0.142 ± 0.052
Gd
µgL-1
0.009 ± 0.006
0.013 ± 0.011
0.019 ± 0.017
N.D.
0.015 ± 0.003
0.019 ± 0.007
0.021 ± 0.005
0.068 ± 0.008
Dy
µgL-1
0.011 ± 0.009
0.012 ± 0.009
0.016 ± 0.015
N.D.
0.023 ± 0.004
0.019 ± 0.005
0.02 ± 0.005
0.061 ± 0.012
Ho
µgL-1
0.002 ± 0.001
0.003 ± 0.002
0.005 ± 0.004
N.D.
N.D.
0.004 ± 0.002
0.004 ± 0.0008
0.018 ± 0.004
Er
µgL-1
0.015 ± 0.013
0.006 ± 0.005
0.01 ± 0.009
N.D.
0.029 ± 0.002
0.012 ± 0.003
0.013 ± 0.002
0.041 ± 0.008
Tm
µgL-1
0.002 ± 0.001
0.002 ± 0.001
0.002 ± 0.001
N.D.
N.D.
0.002 ± 0.001
0.002 ± 0.0009
0.009 ± 0.001
Yb
µgL-1
0.011 ± 0.009
0.006 ± 0.005
0.007 ± 0.005
N.D.
0.021 ± 0.003
0.01 ± 0.003
0.013 ± 0.004
0.04 ± 0.007
Lu
µgL-1
0.002 ± 0.001
0.002 ± 0.001
0.002 ± 0.001
N.D.
N.D.
0.002 ± 0.001
0.002 ± 0.0006
0.01 ± 0.005
Hf
µgL-1
0.0065 ± 0.0015
0.007 ± 0.006
0.01 ± 0.008
N.D.
0.0034 ± 0.0015
0.012 ± 0.003
0.012 ± 0.007
0.034 ± 0.026
W
µgL-1
0.016 ± 0.009
0.083 ± 0.081
0.115 ± 0.112
N.D.
0.018 ± 0.004
0.059 ± 0.046
0.068 ± 0.054
0.026 ± 0.019
Pb
µgL-1
0.199 ± 0.165
0.266 ± 0.181
0.487 ± 0.394
N.D.
0.094 ± 0.057
0.184 ± 0.067
0.256 ± 0.094
0.985 ± 0.124
Th
µgL-1
0.015 ± 0.048
0.011 ± 0.009
0.015 ± 0.01
N.D.
0.011 ± 0.007
0.023 ± 0.007
0.018 ± 0.009
0.026 ± 0.006
U
µgL-1
0.003 ± 0.002
0.006 ± 0.005
0.009 ± 0.008
N.D.
N.D.
0.01 ± 0.005
0.008 ± 0.004
0.029 ± 0.01
Seasonal trend in element concentration in the thermokarst lake water: progressive increase from June to February
The first statistical test aimed to assess the difference in element
concentration between different seasons for the full range of the lake size.
It was found, using the H-criterion of the Kruskal–Wallis and Mann–Whitney
U tests, which allowed one to estimate the difference between two independent
sets of data based on one given parameter, that all elements are different
between three main seasons at the significance level of p< 0.05. This
allowed one to calculate the average element concentration with ±2σ standard deviation or all studied lake sizes from 100 to
> 500 000 m2 for spring, summer and autumn, as listed in Table 3.
Note that the winter data are of low representativity, given that only three
unfrozen lakes could be used for this calculation. The median was similar to
the average within the 2σ standard deviation of the average.
Size-averaged concentrations of DOC (a), Si (b),
Fe (c), Al (d), Ti (e), K (f), Zn
(g), Cr (h), Ni (i), Cu (j), Cd
(k) and Pb (l) during different seasons. The points are the
average and the error bars are the standard deviation. Only the medium and
large lakes (> 100 m2 surface area) were used for this estimation.
The dotted line represents conservative behavior for a hypothetical lake of
75 cm depth in summer, 20 cm flooding in spring, surface 20 cm ice
formation in October, and almost full freezing (10 cm bottom water left) in
February.
![]()
Continued.
![]()
DOC (a) and Fe (b) concentrations during different
seasons plotted as a function of water residence time in the water body.
Vertical dashed line marks the beginning of concentration–residence time
independence.
![]()
The majority of dissolved elements exhibited a statistically significant
increase in their average concentration from spring to autumn (20 to
50 %). These are DOC, Mg, K, Ca, Al, Ti, V, Cr, Mn, Fe, Ni, Co, Cu, Ga,
Rb, Sr, As, Cd, Sb, Mo, Cs, Ba, all REEs, Pb and U. Examples of the average
element concentration evolution grouped by several families of elements with
similar chemical properties and affinity with DOC and Fe colloids (i.e.,
Pokrovsky et al., 2013) are shown in Fig. 7. The trends of element
concentration in the lake water from spring to autumn are quite similar for
large (> 500 000 m2) and medium (100–500 000 m2) water
bodies. We estimated the minimal impact of lake water freezing, assuming a
conservative scenario of element concentration evolution in thermokarst
lakes. Based on thorough hydrological observations, we accepted, for
1 m2 of the average lake depth of 75 ± 25 cm in summer, to which
we added 20 cm flooding in spring and from which we subtracted 20 cm of the
water layer for ice formation in October. The three sampled lakes exhibited
almost full freezing in February, with ca. 10 cm of bottom water left and
65 cm of the ice. The concentration factor was calculated as the ratio of
the water volume under 1 m2 in a given season to that in spring. It was
equal to 1, 1.3, 2.0 and 9.5 for June, August, October and February,
respectively.
Conservative behavior of selected major and trace elements is illustrated in
Fig. 7 as a dashed line. Most major elements (Si, K) and a number of trace
metals (Al, Ti, Cr, Cu) exhibited close to conservative behavior. In
contrast, DOC demonstrated significant depletion during winter, between
October and February. Fe, Ni, Cd and Pb were slightly non-conservative,
showing a depletion in winter. The effect of the season can be quantified as
the ratio of the average summer and autumn to spring concentration
(Rsummer/spring). This ratio is highest for Ba (6.7) and light REEs
(7 to 3), followed by Zn (4.3), Si (4.3), Al, Fe and trivalent hydrolysates
(4.2 to 2.8), tetravalent hydralysates, Cr and divalent heavy metals such as
Co, Cu, Ni, Cd and Mn (2.5 to 3.3). A number of elements did not demonstrate
a significant increase in the concentration during the baseflow season
compared to the spring period, exhibiting Rsummer/spring values
below 2 (B, Mg, Na, Ca, Cs, Pb, V, As, Sb, Th and Zr). This empirical result
helps to distinguish two groups of elements: elements whose concentrations
are affected by the seasons by a factor of 2 or higher and elements that do
not significantly increase their concentration in summer relative to the
spring seasons. For example, the increase in Si and metal micronutrients (Zn,
Co, Ni, Cu, Cd, Ba and Mn) might indicate some preferential release of these
elements during active plant and upper moss litter leaching in summer. High
concentrations of B, Na, Mg, Ca, Cs and Pb in early spring relative to the
end of summer may indicate their input with atmospheric precipitates.
Despite significant uncertainty in the average summer-period element
concentration in thermokarst lakes (see Table 3), the values obtained in
the present study are consistent with the thermokarst lake water chemical
composition across the gradient of the permafrost zone, from
discontinuous/sporadic to continuous permafrost. The values of DOC and all
major and trace element concentrations agree, within the uncertainty of the
average, with the values of element concentration interpolated from the
latitude–lake water concentration dependence (Manasypov et al., 2014) for
the latitude of the study site. It follows that the relatively small area
investigated in this work can be a good surrogate for a much larger region of
western Siberia (described in Manasypov et al., 2014) in terms of seasonal
evolution of thermokarst lake chemical composition.
Water residence time effect on lake water chemical composition
The effect of lake water residence time (WRT) on the summer period and the
average spring–autumn chemical composition of thermokarst lakes exhibits two
clusters of the data points (Fig. 8). Statistical analysis of observed
dependencies confirmed a significant link between the water residence time
and both summer and season-average DOC concentrations, with the Spearman
correlation coefficient (RS) equal to -0.65 (p< 0.05). The
most important result is that, at WRT < 8 months, an almost five-fold
increase in DOC concentration occurs. A similar impact is observed for Fe
with RS=-0.55 (p< 0.05). This result can be understood given
that the Fe is likely to be present in the form of organic colloids. The
alkaline and alkaline–earth major cations and other metals, however, did not
demonstrate any statistically significant (p > 0.05) link to the WRT.
The estimation of WRT for the smallest (< 10–100 m2) water bodies
was impossible due to the ephemeral nature of these small thaw ponds.
However, considering their typical existence between the start of the
snowmelt (May) and summer drought (July–August), the water resides in these
water bodies between 0.1 and 0.3 years. The highest concentration of DOC
observed in spring in < 10 m2 depressions (Fig. 2) is therefore
consistent with the trend shown for large lakes in Fig. 8.
PCA analysis of possible TE sources in thermokarst lakes
The data consisted of ∼ 130 lake water samples grouped into three
distinct seasons (spring, summer and autumn). Statistical analysis considered
each chemical element as a variable (35 in total) for all lakes. Considering
all seasons together, the first factor was responsible for 16 % variation
and included B, Na, Si, K, Ca, Ti, V, Cr, Ni, Zn, Rb, Mo, Sb, Cs, Ba, La, Ce
and U, whereas the second factor (6.3 %) included Al, Fe, Co, As and Cd
(Fig. S7). Separation of F1 and F2 factors in June and August was uncertain
since they provided only 9 and 6.7 % (June) and 11.8 and 5.8 %
(August) variations. In contrast, October's data could be explained by an
18.8 % variation of F1 (B, Na, Si, K, Ca, Ti, V, Cr, Ni, Zn, Rb, Sr, Zr,
Mo, Sb, Cs, Ba, La, Ce and U) and 4.4 % F2 (SO4, Cu, Co, Fe and As). The
list of elements and corresponding factors for each season are given in
Table S5.
This PCA treatment demonstrated rather high variability of lake chemical
composition, mostly pronounced during June and August. In June, multiple
factors are responsible for element enrichment in the lake water, namely the
lateral input from thaw snow and lake ice, dissolution of colloid coagulation
products, and leaching of plant litter. The F1 factor in June is very poorly
pronounced. Nevertheless, it may mark the colloidal transport of TE (organic
complexes), whereas the second factor may correspond to the degree of snow
input (typical atmospheric aerosol-originated elements). In August, internal
(autochthonous) processes and subsurface feeding are likely to strongly
modify lake water chemical composition both for major (pH and DIC of the F2
factor) and TE. In addition, atmospheric precipitation in the form of rain
that dilutes lithogenic TE but also delivers marine aerosols and dust
(subjected to dissolution) can strongly modify the role of individual
correlations. Finally, October represents the period when the first factor is
mostly pronounced, as also translated in the PCA results of all seasons
together. Presumably, the period of the beginning of ice formation
corresponds to the maximal stability of the F1 × F2 structure;
during this time, the influence of both allochthonous (lateral and subsurface
TE influx from peat, mineral soil horizons and ground vegetation) and
autochthonous (bio- and photo-destruction of organic colloids, primary
productivity) processes is minimal.
Element speciation in thermokarst lake waters
We used geochemical program Visual MINTEQ (Gustafsson, 1999), version 3.1
(October 2014) for Windows (see Unsworth et al., 2006, for a vMINTEQ
application example), in conjunction with a database, the NICA-Donnan humic
ion binding model (Benedetti et al., 1995; Milne et al., 2003) and the
Stockholm Humic Model (SHM). Speciation calculations were performed for Ba,
Ca, Cd, Co, Cu, K, Mg, Mn, Na, Ni, Pb, Sr, Zn, Al, FeIII,
ThIV and UVIO2 for average lake water composition in
spring, summer, autumn and winter as well as the average August composition
of large (> 500 000 m2) and small (100–500 000 m2) lakes
(Table 3). In addition to the NICA-Donna approach, the Stockholm Humic Model
(SHM) of visual MINTEQ was used to calculate metal speciation in the lake
water.
Results of the calculation of metal degree of complexation with DOM are
listed in Table S6 and illustrated in Figs. S8 and S9. According to the
NICA-Donnan model, the majority of divalent major and trace elements are
bound to organic, Donnan-like complexes. Only Na and K exhibit a decrease of
80 to 10 % complexed fraction from spring to winter. Extremely high
complexation of metals with DOM (up to 80 to 90 %) within the NICA-Donnan
concept is supported by dialysis experiments conducted in various thaw lakes
of the western Siberian subarctic (Pokrovsky et al., 2013; Shirokova et al.,
2013). In contrast to the NICA-Donnan, the Stockholm Humic Model (SHM)
predicts moderate and quite variable association of divalent metals with DOM,
with the lowest values for Ba and Mn (40 % in spring and 20 % in
winter) and the highest values for Cu and Pb (80–90 % for all seasons).
It is important to note that the highest proportion of organic complexes is
observed in summer, presumably due to the highest pH recorded during this
season, which favors the deprotonation of functional groups of the DOM. The
lowest proportion of organic complexes in winter could be due to competition
between metals for organic ligand binding sites, given the significant
increase in Me2+ compared to DOC during solute concentration by
freezing.
The lake size has a pronounced impact on some metal complexation with DOM as
illustrated in Fig. S9. Even within the “conservative” SHM model, the TE
in small thaw ponds exhibits 100 % complexation with DOM. The degree of
complexation is smaller in large lakes with a factor of 2 lower DOC
concentration, despite the fact that their pH is higher than that of small
lakes (5.49 and 4.72, respectively). Interestingly, this may indicate higher
bioavailability of metal micronutrients in large lakes, in line with the
hydrobiological evolution of western Siberia thermokarst lakes, namely the
elevated productivity and the presence of macrophytes and phytoplankton
blooms in large mature thermokarst lakes (Pokrovsky et al., 2014). Another
important conclusion from the speciation modeling is that the DOC
concentration has a more pronounced impact on metal speciation (notably the
% of organic complexes) than the pH – both across the seasons and
different lake size groups.
Discussion
Lake water composition control by snow and ice melt in spring
All trends described in this work on numerous lake analysis are confirmed by
the data of five lakes sampled throughout the year. This is illustrated for
pH, DOC, Al and Cu for the same five lakes sampled in spring, summer and
autumn. For a large range of lake surface area, the trend spring < summer
< autumn is clearly visible and statistically significant (p< 0.05).
The lack of a trend in DOC and metal concentration with the lake surface area
increase above 1000 m2 (Fig. 2, Table S1, R2 = 0.17,
Rs = -0.33) in June implies that
the lakes are essentially influenced by allochthonous (surface runoff from
snowmelt) sources of dissolved material with minimal transformation of DOC
and metal complexes by intra-lake autochthonous processes. In the opposite
case, one would expect a tendency of DOC and chemical element increase with
the decrease in the size of the water body, notably below 100 m2, as it
is encountered in summer periods in the sporadic and discontinuous and
permafrost zone of thermokarst lake development (Shirokova et al., 2013).
Given that the upper part of the peat column is still frozen in the beginning
of June, the source of the solutes to the lake water is most likely to be
ground vegetation, comprising dwarf shrubs, mosses and lichens, as well as
the litter fall of the previous year. The input of the vegetation leaching
products should be similar for the lakes of different sizes, given the very
homogeneous palsa bog dominant landscape at the study site.
The spring flood period is known to be the most important in terms of land
coverage by water, on the annual scale (Zakharova et al., 2009, 2014). As a
result, the average chemical composition of the dominant surface waters in
the discontinuous permafrost zone of western Siberia can be approximated by
that of the spring period as listed in Table 3. Elevated DOC concentration
in thermokarst lakes during the spring period, a factor of 2 to 3 higher than
the typical 10–15 mg L-1 of boreal waters (Dillon and Molot, 1997),
suggests a high potential for releasing CO2 into the atmosphere, notably
in numerous small depressions filled by thaw water.
A systematic (p< 0.05) increase in the average slope of dependence
UV280nm – [DOC] in the course of the season (Fig. 3), from
0.024 in spring to 0.0354 in winter (the latter with only three samples),
suggests an increase in the input of aromatic compounds (presumably from peat
lixiviation) to the end of the vegetative season relative to the leaching of
fresh vegetation products (low aromatic, plant and litter exudates) mostly
visible after the snowmelt, when the surrounding peat is still frozen.
Correcting the UV absorbance for dissolved Fe3+ (Weichaars et al., 2003)
by subtracting a term of 0.08 × [Fe, ppm] will change the
UV280nm value by less than 10 %, which is beyond the
variability of the seasons and lake size. Note also that a recent study of
DOC-UV properties at a boreal site demonstrated the absence of the influence
of nitrate and iron on UV and visual absorbance of up to 2.2 mg L-1 of
Fe(III) (Avagyan et al., 2014). The lack of any trend (p > 0.05) in the
UV / DOC ratio as a function of the lake size (Fig. 3) strongly
suggests the similarity of the DOC sources in the full range of the lake size
in each season. As a result, the season rather than the lake size at
S > 100 m2 seems to be the most important factor controlling both
the average concentration and the chemical nature of DOC and related
elements.
Comparison of TE concentrations in summer relative to spring (Sect. 3.3)
distinguishes two groups of elements: elements that do not significantly
increase their concentration in summer relative to the spring and elements
whose concentrations are affected by the seasons by a factor of 2 or higher.
For example, the increase in Si and metal micronutrients (Zn, Co, Ni, Cu, Cd,
Ba and Mn) in summer relative to spring might indicate some preferential
release of these elements during active plant and upper moss litter leaching
in relatively warm waters. In contrast, high concentrations of B, Na, Mg, Ca,
Cs, and Pb in early spring compared to the end of the summer may indicate
their input with atmospheric deposition after snowmelt.
Sequence of ice crystallization events during the glacial period on
the shallow (< 1 m depth) thermokarst lakes of western Siberia.
(a) Start of the ice formation; (b) squeezing water towards
the surface via seeps; (c) freezing of water pockets and seeps and
multi-layer ice formation.
![]()
Concentration mechanism during freezing period; enhanced colloidal coagulation
The unusual distribution of organic carbon and related trace elements over
the ice core profile, with an accumulation in the upper 20–40 cm of the ice
column, stems from a sequence of events during lake freezing, linked to (i)
shallow depth and the possibility of the full freezing of the water column
and (ii) the lack of any outlet or hydraulic connection to the groundwater
of these confined water bodies underlain by the impermeable permafrost layer
in the form of the frozen sand/peat. The hypothetical scheme of a
crystallization process is illustrated in Fig. 9. The squeezing of remaining
water from the bottom towards the surface through the ice cracks starts as
early as October. This produces the layered, organic- and Fe-rich secondary
(allochthonous) ice that is crystallized at the already frozen lake surface.
During winter, the progressive freezing downward from the water column and
the decrease in the connectivity between the remaining water and the lake
surface bring about the decrease in element concentration in the bottom ice
while concentrating solutes in the remaining fluid at the ice–sediment
interface. The progressive lake freezing produces layered ice with contrasted
content of major and trace elements among different layers, reflecting the
full freezing of water pockets formed via seeping the bottom water at the
lake surface. The DOC and many related elements exhibit a general decrease in
concentration downward from the ice core surface. The freezing sequence of
the thermokarst lakes contrasts with the freezing of non-permafrost (glacial)
lakes, which are generally deeper than 1 m and have an inlet/outlet and
underground connection. To which degree the scheme of ice formation suggested
for shallow thermokarst lakes in the discontinuous permafrost zone can be
applied to other, sporadic and continuous permafrost regions of western
Siberia, remains unknown. However, given the shallow depths and closed-basin
settings of most thermokarst lakes, the extrapolation of obtained ice
chemistry data and partitioning coefficients to other thermokarst lakes of
western Siberia should be possible.
To our knowledge, the information on TE partitioning between the lake water
and the ice for other boreal settings is lacking. The similar range of DOC
and many major and trace metal distribution coefficients between the ice and
the remaining solution encountered in the present study (Table 2) is
remarkable and contrasts with the recent results of DOM incorporation during
sea ice formation (Muller et al., 2013). The latter authors argued that DOM
is incorporated into sea ice relatively more than inorganic solutes. In the
case of low total dissolved solid (TDS), DOM-rich thermokarst lakes of the
present study, the majority of solutes may follow the DOC being present as
organic colloids (1 kDa – 0.45 µm), as follows from vMINTEQ
speciation calculation (Sect. 3.6) and as also demonstrated by dialysis experiments (Pokrovsky et al.,
2011, 2013).
Various physico-chemical processes may be involved in the transformation of
solutes and DOM and metal-rich water pockets incorporated into the ice,
notably within its most surface layers. Among them, photoreductive
dissolution of iron oxy(hydr)oxide trapped in the ice (i.e., Kim et al.,
2010) can occur both in winter and spring, during massive ice melting. The
enrichment of both Fe and Mn in the first 20 cm of the ice column relative
to the deeper ice layers (Fig. 5) can certainly promote enhanced reductive
dissolution of both Mn (Kim et al., 2012) and Fe (Kim et al., 2010) under ice
thaw during polar day in early June. However, the Fe(II) fraction was
estimated to be below 10 % in June's lake sampling, as the
LMW< 1kDa fraction containing essentially uncomplexed
Fe2+ (aq) was ≤ 5 % of the total dissolved Fe (see the method
of Fe(II) analysis in the lake water by the in situ dialysis technique in
Pokrovsky et al., 2012). Therefore, oxygenated spring waters and shallow
depths of thermokarst lakes provide favorable conditions of Fe(III) colloid
formation in these organic-rich water bodies. Highly non-conservative
behavior of DOC during the winter time demonstrating significant depletion in
February (Fig. 7a) may suggest some heterotrophic respiration of DOM under
the ice. Alternatively, the DOC could be transformed into POC and
precipitated to the lake bottom or be trapped as particles in the growing
ice.
Note that the Zn concentration in winter becomes very high, up to
1 mg L-1 (Fig. 7g). The main reason for such an increase in Zn
concentration could be its high mobility in acidic, organic-rich aquatic
systems. The additional mobilization of Zn from moss cover at the lake border
during its freezing can be linked to the release of Zn from the cell
cytoplasm during freeze-induced submerged vegetation cell damage. However,
the concentration factor for Zn is not dramatically different from that of
the other elements: it is similar to that of Cu, Zr, B, Na, K, REEs and U,
and is even lower than that of Pb. As such, we consider high element
concentration during winter to be a consequence of solute concentration
during freezing rather than a specific mechanism of metal mobilization from
ground and bottom vegetation subjected to freezing.
During the massive ice melting, a significant number of coagulated
organo-ferric particles remain on the ice surface (Fig. S6) and eventually
precipitate to the lake bottom. Although under anaerobic conditions in the
thermokarst lake sediments, Fe(III) may be reduced and return to the water
column (Audry et al., 2011), the majority of Fe oxyhydroxide is likely to
remain in particulate form, thus preventing Fe from moving to the
hydrological network during spring floods. We hypothesize that one of the
most powerful mechanisms of organic and organo-mineral colloid coagulation in
thermokarst water bodies is annual freezing of the whole water column leading
to significant concentrations of Fe, Corg and related divalent
metals and trivalent and tetravalent hydrolysates in the particulate phase.
In the northern part of the discontunuous permafrost zone, the lake sediments
are indeed enriched in Fe and trivalent and tetravalent hydrolysates (Audry
et al., 2011). It is remarkable that Fe-normalized distribution coefficients
of TE between the organo-ferric coagulates and filtered
(< 0.45 µm) lake water (Fig. 6) are of the same order of
magnitude or within a factor of 2 different from the distribution
coefficients of TE between Fe-rich colloids (1 kDa – 0.45 µm) and
the LMW< 1kDa fraction assessed in previous studies in
thermokarst lakes (Pokrovsky et al., 2013) and boreal surface streams
(Vasyukova et al., 2010). This similarity strongly suggests that the
elementary mechanisms of TE incorporation into organic matter-stabilized Fe
oxyhydroxides include mainly co-precipitation and that it is generally
similar for colloids and particles.
Allochthonous versus autochthonous processes forming lake water chemical composition
Allochthonous processes forming lake water chemical composition include the
input of fresh vegetation products and the leaching of the upper peat layer.
At the beginning of the active season, in June, there is a lateral input of
melting snow, reflecting the interaction of water with ground vegetation such
as mosses and lichens as well as with the litter fall of the dwarf shrubs
from the previous year. During the summer (baseflow) season, there is a peat
leaching at the lake border via mainly wave abrasion and elements and DOC
release from moss and lichen coverage via lateral flow fed by rains.
The DOM entering the lake ecosystem with the snowmelt and surface inflow
during summer rain is subjected to two processes of autochthonous
transformation during the open water period: photo- and bio-degradation,
mostly pronounced in June–September (i.e., Larouche et al., 2015). During
summer, the productivity of phytoplankton including possible exometabolite
release represents less than 10 % of heterotrophic bacterioplankton
respiration in western Siberia thermokarst lakes of the
discontinuous/sporadic permafrost zone (Shirokova et al., 2013). The
heterotrophic bacterioplankton activity brings about the conversion of
colloidal DOM and low molecular weight (LMW) organic ligands to particulate
organic matter (POM) in the form of coagulates or the bacterioplankton
biomass; in both forms, the POM is likely to precipitate to the lake bottom
in the course of the active season. During the glacial period, the processes
leading to sedimentation of POC to the lake bottom are cryoconcentration and
colloid coagulation (see Sect. 4.2 and Fig. S6). The physico-chemical
coagulation of DOC may become especially important during progressive lake
freezing in winter. Highly non-conservative behavior of DOC (Fig. 7a)
relative to the other major components (Si, K, Na) strongly suggests the
preferential removal of DOC from the remaining water under the ice cover. The
other elements such as the metals Fe, Ni, Cd and Pb (Fig. 7c, i, k and l,
respectively) follow the non-conservative behavior of DOM, suggesting their
massive removal from the water column in the form of organic coagulates. In
contrast, Al, Ti, Zn, Cr and Cu (Fig. 7d, e, g, h and j, respectively) remain
conservatively in the bottom water or even accumulated in the water. It is
possible that they are less dependent on large-size DOM colloids mostly
subjected to coagulation during cryoconcentration. Indeed, these metals could
be bound to LMW< 1kDa organic complexes and thus remain in
unfrozen water in the lake bottom layer. Additional input of Al and Ti from
mineral clay sediments of the lake (i.e., Audry et al., 2011) cannot be
excluded.
The behavior of DIC is almost conservative during lake freezing, without any
excess of DIC over the theoretical value in February. As a result, there is
no significant accumulation of inorganic carbon in the form of CO2 or
HCO3- under the ice cover. This result strongly suggests the
dominance of physico-chemical coagulation rather than microbial respiration
in DOC removal from the water column in winter. The latter process dominates
the open water period, when heterotrophic aerobic bacterioplankton
respiration of allochthonous DOM produces gaseous CO2 that is released
into the atmosphere in western Siberia thermokarst lakes that are strongly
supersaturated with respect to the atmosphere (Shirokova et al., 2009, 2013),
similar to the other boreal lakes (Jansson et al., 2008; Rautio et al.,
2011). The response of this microbial process to the DOC input in the lake
may be fast, on the order of days, given that (i) we observed elevated
(> 20 mg L-1) DOC concentrations in the smallest (< 2 m2)
ground depressions formed within hours after snowmelt, and (ii) that the
leaching of DOM from plant litter is very fast, on the order of hours (Berg,
2000; Fraysse et al., 2010).
A plot of DOC concentration versus water residence time in medium and large
lakes (300–1200 000 m2) revealed two clusters of the data points,
with maximal DOC concentration observed in lakes and ponds of short-term
water circulation and a stable and rather low DOC concentration
(10 ± 5 mg L-1) in lakes of slow water turnover (Fig. 8a).
Since the allochthonous input increases DOC in the water body and
autochthonous microbial respiration and physico-chemical coagulation remove
DOC from the water column, the former process is certainly dominant for
short-lived water bodies. The hydrological balance of the smallest water
depressions (< 200 m2) could not be quantitatively assessed, but it
can be suggested that the lowest water residence time in these, partially
ephemeral, water bodies is consistent with the highest DOC measured in this
study. In the lakes of slow water turnover, the input and removal of DOC are
presumably balanced. Note, however, the similarity of the intensity of
autochthonous processes in larger lakes (i.e., ≥ 100 000 m2),
regardless of their size and the water residence time.
Seasonal evolution of stock of carbon and TE in thermokarst lakes
The average of the six smallest (< 1 m2) depressions sampled in
spring shows a factor of 2.3 (p< 0.05) higher concentration of dissolved
organic carbon (30.8 ± 7.3 mg L-1) compared to the larger water
bodies (13.2 ± 6.6 mg L-1) during this period. This strongly
suggests the importance of short-term plant debris (litter) and submerged
ground vegetation leaching occurring right after the snowmelt. On the western
Siberia lowland, the water stock is highest during the spring period, notably
in terms of water coverage of the land depressions (60 to 70 % of the
overall watershed area; Zakharova et al., 2014). According to satellite
observations in western Siberia, the area subjected to the spring flood
period is 55–65 % higher than that of the summer period (Zakharova et
al., 2014). Taking into account these observations, and the DOC concentration
in thermokarst water bodies, we estimate that the overall increase in the
soluble stock of DOC and related metals in surface waters in June relative to
August may be as high as 200–500 %. This value stems from (i) a factor
of 2–3 increase in the wetland flooding in June compared to the summer
baseflow season multiplied by (ii) a factor of 1.5–2 higher DOC and metal
concentrations in small (< 1–10 m2) water bodies compared to the
larger lakes. This dissolved fraction may be easily mobilized from the
watershed to the river and further transported to the ocean.
In shallow thermokarst lakes, progressive ice melt from the surface towards
the bottom slowly liberates inorganic carbon trapped or dissolved in the ice.
The Kd(water/ice) for DIC (1.2–2.2) is much lower compared to
other elements including DOC. As such, the ice is not particularly enriched
in DIC relative to the bottom water and both winter and spring period DIC and
CO2 concentrations are not significantly higher than those in summer. At
present, we do not know of any seasonal measurements of CO2 and CH4
in thermokarst water bodies of western Siberia. Our unpublished data
(Fig. S11) do not demonstrate any significant enrichment in CO2 or
CH4 during spring flood. Only in very small depressions
(< 10 m2) does the CO2 level increase. Given that the
concentration of methane is 1 to 2 orders of magnitude lower than that of
CO2 and that [CH4] is independent of the depression size, the role
of methane oxidation in CO2 enrichment is negligible. Overall, we do not
expect significant buildup of CO2 under ice and CO2 release from
the lake water into the atmosphere during spring melt, contrasting with
well-known phenomena on deep boreal lakes (Karlsson et al., 2013). The
reasons for this contrast could be low volume of the thermokarst lake water
and a relatively short period suitable for this accumulation in western
Siberia, since already in February, there is a lack of liquid water under the
ice, yielding a very low fraction (between 10 and 20 %) of this unfrozen
water stock. Altogether these arguments suggest that the only mechanism
capable of enriching the lake water in CO2 during spring melt is
heterotrophic respiration of “fresh” allochthonous DOM, as confirmed by the
elevated DOC concentration during this period, notably in small water bodies.