Lengthening plus shortening of river-to-lake connection times in the Mackenzie River Delta respectively via two global change mechanisms along the arctic coast



[1] River deltas along the circumpolar arctic coast are lake-rich and poorly understood ecosystems, set in a region expected to change rapidly. Over the past 30+ years annual river-to-lake connection times in the Mackenzie Delta have lengthened (>30 days) in the lowest elevation lakes and may have shortened in the highest elevation lakes, respectively via sea level rise and declining effects of river-ice breakup. Lengthened connection times indicate summer low-water levels in the delta have increased by an amount (0.3 m) equivalent to three times local sea level rise (0.1 m) over the same period. Such an amplification effect of recent sea level rise has been completely unexpected and may be a result of enhanced storm surges in response to receding arctic sea ice or coastal backwater effects on the river flow. Shortened connection times are consistent with other work showing a decline in river-ice breakup effects, an important control on annual peak water levels.

1. Introduction

[2] The circumpolar arctic is a region expected to change dramatically over the next 50 years in response to global change [Arctic Climate Impact Assessment (ACIA), 2005]. One global change mechanism is sea level rise, driven by ocean warming plus glacier melting [ACIA, 2005; Overpeck et al., 2006]. Prior work has documented local net rise in sea level at various locations around the arctic, but thus far, the total recent increase has been small [ACIA, 2005; Manson and Solomon, 2007]. The major river deltas around the circumpolar region are low-elevation areas (Figure 1) poised to be among the first major ecosystems affected by such future sea level increases. A second global change mechanism is the role of river-ice breakup effects and ice-jamming in controlling peak annual water levels in such deltas [Marsh and Hey, 1989; Marsh and Lesack, 1996]. In the Mackenzie Delta, flood pulses [Junk et al., 1989] driven by ice breakup control the degree to which river water moves off-channel to replenish the waters of this extremely lake-rich [Emmerton et al., 2007] ecosystem, which presently exists because of such replenishment [Marsh and Lesack, 1996; Prowse and Conly, 2002]. Lakes sustained on brief annual flood pulses are small, very abundant, and diverse, but cannot sustain fish because they are shallow and lack of water renewal over winter causes anoxia. Fish habitat in this system requires lengthy to near-continuous annual connection to the river, and is thus controlled by average low-water levels. The other arctic deltas are linked to north-flowing rivers, where peak water levels should be influenced by ice breakup effects, as in the Mackenzie. Lake-richness varies among them, though the Lena is comparably lake-rich to the Mackenzie. Our understanding of the linkage between river-flooding hydrology and aquatic ecology in these systems is presently derived from work largely limited to the Mackenzie Delta (L. F. W. Lesack and P. Marsh, unpublished manuscript, 2007, hereinafter referred to as Lesack and Marsh, submitted manuscript, 2007) and Peace-Athabasca Delta [Prowse and Conly, 2002], with little comparable information available for the eurasian deltas [ACIA, 2005]. We previously postulated that if the arctic region warms as expected, river ice jamming may become reduced [Rouse et al., 1997] and would lead to reduced flood peaks in rivers, and the possibility that the higher elevation lakes in circumpolar deltas might dry up [Marsh and Lesack, 1996]. More recently, Prowse and Beltaos [2002] have argued that the effect of a warmer climate on ice jamming is less clear because of numerous complicating factors. Here we show, based on analyses of 30+ years of water levels for East Channel at Inuvik (Figure 1) that annual river-to-lake connection times in the Mackenzie Delta have lengthened in the lowest elevation lakes and may have shortened in the highest elevation lakes, respectively via the two global change mechanisms described above. We also assess hypotheses to explain the unexpected early rise in low-water levels that are substantially higher than direct sea level rise thus far observed [Manson and Solomon, 2007] at this location.

Figure 1.

Map of (top) location in circumpolar arctic (modified from Telang et al. [1991]) and (bottom) Mackenzie Delta.

2. Study Area and Methods

[3] The Mackenzie Delta is 13,000 km2 of productive aquatic habitat with a high biodiversity that was created as a result of the historical regime of peak- and low-water (Lesack and Marsh, unpublished manuscript, 2007). It extends (Figure 1) 200 km upstream from the coast and contains 45,000 lakes [Emmerton et al., 2007]. Prior work [Marsh and Hey, 1988, 1989, 1991] has established that these lakes are perched in varying degrees above the distributary channels, with the lakes being flooded with river-water (primary source of water renewal) only as water levels in the channels rise in response to changes in Mackenzie discharge, river-ice breakup effects, and storm surges from the Beaufort Sea. These lakes have been partitioned into three flooding regimes (see auxiliary material). No-closure lakes (1) remain in connection with the river channel for the entire summer and have sill elevations lower than average low-water levels in the river. Low-closure lakes (2) are flooded each spring, but are cut off from the river for some portion of the summer. High-closure lakes (3) are not necessarily flooded every spring and never during the summer. Based on investigations in areas- A, B, and C, plus a cross-delta swath of 2,800 lakes (Figure 1), no-closure lakes represent about 60% of total lake area in the delta, low-closure lakes 25%, and high-closure lakes 15% [Marsh and Hey, 1989; Marsh et al., 1999; Marsh and Hey, 1991]. Lake 129, Lake 80, and Lake 527a are located near Inuvik (area-A, Figure 1) and respectively represent long (no-closure), intermediate (low-closure), and short (high-closure) annual connection times with the river. Lake 129 (2.363 m asl) and Lake 80 (2.631 m asl) are the two lowest elevation lakes, and Lake 527a (5.169 m asl) is the highest elevation lake from a suite of 9 well-studied lakes (subset from 132 total lakes in area-A) of differing sill elevations that span the general range of river levels. A description of the methods of quantifying river-to-lake connection times is included in the auxiliary material. All references to Mackenzie River discharge are based on the gauging station at Tsiigehtchic (Figure 1).

[4] The time-series analysis in this paper is exploratory and similar to the approach of Marsh et al. [2002]. A 5-year running mean was fit to annual river-to-lake connection times and a least-squares regression line was fit to the running means to visually explore possible trending [e.g., Ludwig et al., 2004]. Fitted lines are shown only where the p-value of the fit <0.05, but such lines may not necessarily represent an unambiguously significant trend. Statistical trend significance was based on the more conservative Mann-Kendall (MK) test of the unsmoothed data [Hipel and McLeod, 1994]. In all cases of MK-tests, the p-values > p-values from the corresponding least-squares trend lines. The trend of annual connection times for Lake 129 and for Lake 80 was assessed over the complete water level record available for Inuvik (1973 to 2005). Trend assessment for Lake 527a was extended from 1964 to 2005 because this lake is of sufficiently high elevation that it is affected only by river levels during the ice-breakup, and such levels are available from 1964 to 1972 [Marsh and Hey, 1989].

3. Results

[5] The annual river-to-lake connection times of the lower-elevation lakes in the delta (no-closure plus some low-closure) have generally lengthened (Figure 2). The connection time of Lake 80 has increased from 101 days in 1973 to 138 in 2005 (37 day increase), based on regression of the running-means (i.e. starting from the 5-yr mean of 1973–1977, ending with the 5-yr mean of 2001–2005), and the significance level of the MK-test is p = 0.05. Analogously, the connection time of Lake 129 has increased from 145 days in 1973 to 169 in 2005 (24 day increase), and the significance level of the MK-test is p = 0.07. These increases are not a result of increases in Mackenzie River discharge. Annual discharge has not significantly changed over its record and there is no evidence of discharge increases at more recent points within the record [ACIA, 2005; Woo and Thorne, 2003]. The lengthened connection times also are not a result of possible increases in Mackenzie discharge during the latter open-water period. Our studies have shown that the interannual pattern of river discharge (km3 per yr) from day 200 to day 300 (summer discharge period as defined by L. F. W. Lesack et al., unpublished manuscript, 2007) has no significant trend (Figures 2c and 2d, MK-test p = 0.66). However, the relation between summer river discharge and lake connection times (days per yr) has changed. There is good correspondence between summer discharge levels and connection times, particularly in Lake 80, from 1973 until 1985, then marked differences between discharges and connection times repeatedly occur from 1986 to 2005. From 1996 to 2005, in particular, 7 of the 10 years represent connection times averaging 44 days longer than what would be expected from the relation between summer discharge versus connection time during 1973 to 1985 (see Figure S1 of the auxiliary material), and 5 of these 10 years represent the longest connection times in the 33 year record.

Figure 2.

River-lake connection times (days per yr) for (a) Lake 129 and (b) Lake 80 from 1973 to 2005, overlaid with annual summer discharge (day 200 to 300) of the Mackenzie River. The 5-yr running mean of summer discharge shown with 5-yr running means of river-lake connection time, for each of (c) Lake 129 and (d) Lake 80, and with least-squares regression lines fit to the running means (see methods). Connection times for (e) Lake 527a from 1964 to 2005 in comparison with (f) water level exceedance times (days per yr) for a reference elevation of 5.500 m asl (just below the average peak water level from 1964 to 1986 [Marsh and Hey, 1989]). A 5-yr running mean is shown (thick curved lines) in Figures 2e and 2f with least-squares lines fit to the running means.

[6] Based on the above lengthening of connection times in Lake 80, it has effectively changed from a low-closure lake to a no-closure lake with a present connection time (138 days per year) roughly equivalent to the connection time of Lake 129 (145 days per year) 30 years ago. Given that the summer sill elevation [Marsh and Hey, 1989] represents the threshold point where lake-river connection time is terminated, and that Lake 80 and Lake 129 respectively have summer sills of 1.620 and 1.272 m asl [Marsh and Hey, 1988], the 0.348 m difference in sill elevation between these two lakes (Figure S2) can be taken as an estimate of how much average water levels in latter-summer would have to rise to account for the average connection time of Lake 80 being lengthened to the former connection time of Lake 129. The elevation difference between the two lakes represent fixed points that objectively reflect how much water levels have risen.

[7] Similar analysis suggests the annual connection times of the higher-elevation lakes may have declined, though this trend remains uncertain. Based on regression of the running-means, the connection time of Lake 527a has declined from 5.5 days in 1964 to 3.5 in 2005 (Figures 2e and 2f) and the water level exceedance time (days per yr) at a reference elevation of 5.500 m asl (just below the mean peak level from 1964–1986 [Marsh and Hey, 1989]) has declined from 3.6 to 1.0 days (Figures 2e and 2f). However, the significance level of the MK-test is weak (p = 0.15) in the latter case and outside the range of statistical significance (p = 0.36) in the former. If restricted to the period 1973 to 2005 (same as Lake 80 and Lake 129), the significance level of the MK-tests are stronger (p = 0.09 and 0.16, respectively). If these changes are real, the connection time of Lake 527a would now be roughly equivalent to the exceedance time of average peak water levels about 40 years ago. Given that Lake 527a has a spring sill of 5.169 m asl [Marsh and Hey, 1988], the elevation difference of 0.331 m from 5.500 m asl (Figure S2) could be taken as an estimate of the decline in average water levels during the breakup period when high-closure lakes are flooded.

[8] Although we had postulated connection times for high sill lakes could decline as a result of reduced river-ice breakup effects [Marsh and Lesack, 1996], lengthened connection times in no- and low-closure lakes at this early stage of sea level rise was a surprise. Average net rise in sea level at Tuktoyaktuk has consistently been 3.5 mm per year since the gauging station was established in 1961 [Manson and Solomon, 2007]. From 1973 to 2005 this can account for a net rise in water level of only 0.115 m of the 0.348 m needed to account for the lengthened connection time of Lake 80. By difference, an average increase of an additional 0.230 m is needed to reconcile the rise in delta water levels (Figure S2). Whereas river discharge cannot account for the rise (Figure 2), coastal storm surges are known to propagate completely through the delta to the Tsiigehtchic gauging station upstream of it [Marsh and Schmidt, 1993], and the pattern of increases in lake connection times from 1986 to 2005 (Figures 2a and 2b) appears episodic. Thus, a conspicuous possibility is that the effects of coastal storm surges from the Beaufort Sea on the delta may have increased in magnitude and duration in response to recent sea ice recession associated with arctic warming.

[9] The combination of data needed to assess the possibility that enhanced storm surges are driving the increased water levels in the delta is presently limited. The statistical relation between water levels at Inuvik and levels at Tuktoyaktuk appear to have strengthened from the 1970's to the present, whereas the relation between levels at Inuvik versus Mackenzie summer discharge appears to have weakened over the same interval (see auxiliary material for details). Visual inspection of summer variations in water levels at Inuvik versus Tuktoyaktuk versus Mackenzie discharge suggests Inuvik levels now appear more strongly affected by storm surges of comparable height than 25–30 years ago. Moreover, Inuvik levels may now respond to storm surges of even modest height, suggesting backwater effects associated with river discharge variations may have become enhanced.

[10] It thus appears that not only are the higher elevation lakes now at risk of drying up from declining water level peaks, but the lower elevation lakes now contain more water than can be accounted for by the small amount by which sea-level has recently risen.

4. Discussion

[11] The uncertainty associated with linking the connection time increases in Lake 80 and Lake 129 to comparable behavior in lower-elevation lakes of the delta, more generally, is relatively low. The gauging station at Inuvik is immediately adjacent to area-A. Precise geodetic benchmarks have been installed at the Inuvik station, plus directly beside Lake 129, that has allowed ground survey confirmation of the summer sill elevation for a selection of lakes in this area relative to the Inuvik gauge. The partitioning of the 132 lakes in area-A into closure classes and estimation of total lake area represented in the classes is also relatively precise. Less precise is the statistic of no-closure lakes representing 60% of the total delta lake area. That statistic is consistent with the cross-delta swath [Marsh et al., 1999], but the abundance of no-closure lakes declines in the up-delta direction (area-B), and increases in the down-delta direction (area-C) [Marsh and Hey, 1991]. We have based our assessment on the approximation that the up- and down- delta effect balance each other. Even if the total area of the delta affected by the enhanced low-water levels is overestimated by 10–15%, this is still a large portion of a massive and very lake-rich arctic delta. Additional years of data are needed to fully resolve whether the suggested trend of declining connection times in the higher-elevation lakes is real. Such a trend would be consistent with other work (L. F. W. Lesack et al., unpublished manuscript, 2007) indicating a significant shortening of the ice breakup period in the delta, though not yet unambiguous declines in annual water level peaks.

[12] Prior to our results here, we postulated that reduction of river-ice breakup effects with climate warming would be the strongest and earliest effect in the Mackenzie Delta [Rouse et al., 1997]. It is now clear that sea level effects are equally strong, or stronger, even at this early point in time. Such effects will become much stronger over the next 50 years, given that sea level is expected to rise in the Beaufort Shelf region by another 30 cm within the next 50 years [Manson and Solomon, 2007], and sea ice may be greatly reduced [ACIA, 2005; Nghiem et al., 2006]. The rise in low-water levels means the delta now contains more water than historically. Given that the mean depth of no-closure lakes is about 1.7 m [Emmerton et al., 2007], the increase of low-water depth by 0.3 m is substantial. This raises a variety of important questions, including: (1) the degree to which thermokarst erosion of the permafrost adjacent to affected lakes [Mackay, 1963; Johnston and Brown, 1966] will become enhanced by the larger water volumes and cause total area of such lakes to expand [Mackay, 1992]; (2) whether sediment erosion from such permafrost-affected lake-shores may become enhanced [Rouse et al., 1997]; (3) whether expanding lake areas may lead to enhanced input of riverine sediments into the delta [Rouse et al., 1997]; and (4) to what degree the overall sediment balance of the delta may be changing [Hill and Solomon, 1999; Carson et al., 1999]. A crucial question from an aquatic habitat perspective, is the degree to which the lengthening lake connection times and increased water-depths have expanded the area of the delta able to sustain fish, and changed this habitat in other ways (Lesack and Marsh, unpublished manuscript, 2007).

[13] The increased channel depths during low-water ought to modestly improve channel navigability for boat traffic. However, the possibility that enhanced storm surges have been affecting the area since about the mid 1980's, and may be a strong amplifier of the sea level rise that has occurred so far, suggests coastal flooding and erosion around the circumpolar arctic may generally become enhanced earlier than expected. A significant cause for concern is development plans to extract natural gas and oil from substantial areas of the outer Mackenzie Delta and Beaufort Shelf over the next 30 years [Kavik-AXYS, 2001]. Such extractions are expected to cause local subsidence that will add to the natural subsidence already occurring in this region. Risk of coastal flooding in the town of Tuktoyaktuk and other areas affected by oil and gas extraction should be reassessed in light of our new information.

[14] We caution that our results should be considered preliminary, and that further investigation plus observation time are required to confirm they represent long-term global change in the circumpolar arctic. However, there are strong arguments [ACIA, 2005] that the two mechanisms addressed here will become important if arctic warming continues on its present trajectory. Our results suggest that declines in peak-water may be a weaker global change effect, but it is not yet clear that is necessarily the case. While lakes affected by declining peak-water represent only 15% of total delta lake area, this includes a very large number of the population of delta lakes [Emmerton et al., 2007]. Intermittent connection between these diverse higher elevation lakes with the less diverse lower elevation lakes may be a crucial historical mechanism in creating and sustaining the overall biodiversity of circumpolar river deltas (Lesack and Marsh, unpublished manuscript, 2007). For example, recent work in the Mackenzie Delta has documented via DNA-based methods surprisingly high microbial diversity in Mackenzie River water [Galand et al., 2006].

[15] The apparent changes in sea level and river ice breakup occurring in the Mackenzie represent the first case we are aware of where two global change mechanisms may be simultaneously forcing a major arctic ecosystem in differing ways. Our results suggest enhanced storm surges and river backwater effects as hypotheses to explain the unexpected amplification effect of sea level rise in this system. The on-going forcing of the Mackenzie Delta should be treated as an important ecosystem “experiment”, with levels of research effort increased to confirm and better understand the system response. The Mackenzie response may be earlier than elsewhere because of local coastal subsidence [Hill et al., 1993], which is not widespread along the circumpolar coast. If enhanced storm surges via sea ice recession is responsible for the amplification, this effect will not be limited to river deltas, but will apply to all low-elevation areas of the circumpolar coast that will eventually be affected by sea level rise.


[16] Financial support was received from NSERC, the Polar Continental Shelf Project, the Northern Scientific Training Program Canada, and the Aurora Research Institute. Comments by W.M. Lewis Jr. and one anonymous reviewer improved the paper. Data used in this paper can be accessed from Fisheries and Oceans Canada, Marine environmental data service, http://www.meds-sdmm.dfo-mpo.gc.ca; and Water Survey Canada, Online data products and services, http://scitech.pyr.ec.gc.ca/waterweb.