The complete landscape surface of the active Mackenzie River Delta (13,135 km2) was manually partitioned into discrete lakes (3331 km2), channels (1744 km2), wetlands (1614 km2), and dry floodplain area (6446 km2) via GIS analysis of digital topographic maps recently available for the system. The census total of lakes (49,046) is almost twice as large as prior estimates. Using this new information, total lake volume in the delta during the post river flooding period is estimated as 5.4 km3. Total floodwater storage in the delta lakes and floodplain at peak water levels is estimated at 25.8 km3 and thus is equivalent to about 47% of Mackenzie River flow (55.4 km3 yr−1) during the high-discharge period of delta breakup. During this period the stored river water can be envisioned in the form of a thin layer of water (2.3 m thick on average) spread out over 11,200 km2 of lakes and flooded vegetation and exposed to 24 h d−1 solar irradiance. Consequently, this temporarily stored water has significant potential to affect the composition of river water flowing to the Beaufort Shelf as it recedes to the river channels after the flood peak.
 Floodplain lakes are sites of significant seasonal biogeochemical change and variable water storage [Forsberg et al., 1988; Lesack et al., 1998]. Biological uptake, sedimentation and other processes are normally enhanced in lake settings compared to turbulent and light-limited river water. Consequently, river floodwaters entering floodplain lakes can undergo significant nutrient chemistry changes driven primarily by reduced turbidity, increased light penetration and resulting biological activity. Arctic deltas are unique environments where these effects are intense due to substantial lake coverage, nearly continuous open water irradiation and an annual spring flood that is the dominant event to deliver water, nutrients and sediment to lakes [Lesack et al., 1998]. Lakes in large arctic deltas typically number in the tens of thousands and together can play a large role in altering the nutrient chemistry of river floodwaters while in temporary spring storage [Emmerton, 2006]. This interaction is an important process in maintaining the water and nutrient balances of individual lakes as well as the productivity of coastal regions that receive incoming freshwater nutrient fluxes.
 The Mackenzie River Delta, in Canada's western Arctic, is a lake-rich, ecologically sensitive environment [Mackenzie River Basin Committee, 1981]. Its high latitude (67°–70°N, 133°–137°W) induces 7–8 months of ice cover before the open water season (June–October). Since the Mackenzie River flows northward from areas of relative warmth toward frozen northern regions, the spring freshet eventually encounters ice jams in the delta, resulting in peak water levels and wide scale flooding [Prowse, 1986; Marsh and Hey, 1989; Rouse et al., 1997; Lesack et al., 1998; Marsh et al., 1999]. Most delta lakes are subsequently flooded depending on their elevation and distance from river channels [Mackay, 1963; Marsh and Hey, 1989]. This variable flooding (and subsequent recession) provides the primary nutrient supply to the delta plain and floodplain lakes and thus controls their ecological characteristics and ecosystem health [Lesack et al., 1998]. The delta is expected to undergo environmental changes as a result of climate change [Rouse et al., 1997; Marsh and Lesack, 1996; L. F. W. Lesack et al., unpublished manuscript, 2007]. The annual flood pulse [Junk et al., 1989] of freshet water that sustains lakes of the delta [Marsh and Lesack, 1996] may arrive earlier because of earlier snowmelt and may be of lower magnitude if ice breakup and ice-jamming effects decline. A decline in annual peak water levels would completely disconnect some lakes from the river, resulting in a negative water balance and a possible drying out of such lakes [Marsh and Lesack, 1996]. Nutrient chemistry of river water discharging to the coastal Beaufort Sea may also change if peak annual volumes of water stored in the delta during ice breakup decline [Emmerton, 2006].
 Understanding the influence of the Mackenzie River Delta on discharging waters to the western Arctic Ocean will require investigation of the coverage and volume of lakes over the delta plain. Sippel et al.  demonstrated the importance of such information in the lake-rich Amazonian floodplain using aerial imagery and topographic maps. The authors provided estimates of flood inundation area, lake morphometry, lake enumeration and areal coverage within the floodplain. This information subsequently contributed to studies examining the role of the Amazon River floodplain in regional and global water and biogeochemical cycles. Utilization of synthetic aperture radar (SAR) has also proven valuable in such studies involving flood inundation area and lake water storage [Bowling et al., 2003]. Currently, there is limited knowledge of lake water storage and coverage in the Mackenzie River Delta. Mackay  showed highest lake concentrations in the mid delta (30–50% coverage) and lower in the head and mouth areas (15–30% coverage). Marsh and Hey [1989, 1991] compiled an extensive survey of 132 lakes near Inuvik, Canada cataloguing lake sill elevations and lake areas, as well as comparing lake size in the Inuvik area to sites north and south in the eastern delta. Marsh and Hey [1989, 1994] classified the flooding potential of lakes in the eastern delta and determined flooding depths of higher-elevation lake and watershed areas. Lewis  was the only previous study to estimate lake counts in the delta. Using 1:50,000 topographic maps of the delta mostly based on late summer 1950s aerial photography, a minimum lake size threshold was chosen as 0.01 km2 and lakes were counted based on random sampling of quadrants 2.4° latitude by 5.0° longitude. Of the roughly 3600 lakes sampled, an estimate of 25,000 total delta lakes was given, though the author did indicate this value could be underestimated.
 As part of a long-term investigation on the hydrology and limnology of lakes in the Mackenzie River Delta (L. F. W. Lesack et al., unpublished manuscript, 2007), the goal of our present study was to quantify the general habitat composition and fully census the lakes of the delta. This was achieved via a GIS analysis of the 1:50,000 digital map series (similar to Lewis ) available for the delta, based on extensive aerial photography performed during the 1950s during low water conditions. The complete landscape surface of the delta was manually partitioned into discrete lakes, channels, wet floodplain (herein “wetlands”) and dry floodplain (herein “floodplain”) area using standardized criteria developed for this study. These results were used with other published information to assess the magnitude of potential water storage in the delta relative to Mackenzie River flow during the spring break up period and the plausibility of the hypothesis that the volume of river water moving through the delta is sufficiently high to potentially affect nutrient fluxes to the Beaufort Sea [Emmerton, 2006]. This study also represents a record of historical lake coverage in the delta that should facilitate future work addressing climate-induced changes in lake coverage already realized in other parts of the arctic landscape [Smith et al., 2005; Marsh et al., 2005].
2.1. Study Area
 The Mackenzie River Delta lies at the end of the Mackenzie River at the Beaufort Sea in the western Canadian Arctic (Figure 1). The delta is the second largest in the circumpolar arctic (after the Lena) and is characterized by numerous anastomosing channels, small thermokarst lakes and wetlands that dominate the deltaic plain [Mackay, 1963; Marsh et al., 1999]. The floodplain is permafrost-influenced silt and sand covered by species of spruce (Picea), alder (Alnus), willow (Salix), birch (Betula), poplar (Populus), Equisetum and tundra species north of the tree line [Mackay, 1963]. About 90% of the delta's water supply is contributed by the Mackenzie River at Point Separation (Figure 1) with minor contributions by the Peel River in the southwest (∼8%) and others [Burn, 1995]. The annual hydrology of the delta is characterized by a winter low-flow period (November–April), an ice breakup (flooding) period from initial rising water to peak water levels (April–June) and a lengthy recession period (June–October) [Emmerton, 2006]. Floodplain lakes are generally small and shallow and are mostly of thermokarst origin where heat from standing floodwaters melted ice-rich permafrost into taliks and subsidence ensued [Hill et al., 2001]. Prior work by Mackay  and Marsh and Hey  classified delta lakes as being continuously connected to the river (no closure), annually connected during flooding before disconnection (low closure) and connected less than annually (high closure) (Figure 2). Lake flooding is determined by the sill elevation of the lake and water level of adjacent river channels. Most lakes are shallow enough to support substantial macrophyte growth (common species include Potamogeton, Chara and Ceratophyllum) [Squires et al., 2002].
2.2. Software Application and Source Data
 Determination of lake coverage and other delta statistics was achieved using GIS software (ArcMap™, version 9.0 [Environmental Systems Research Institute, Inc., 2004]). Predigitized topographic maps at the 1:50,000 scale, which represented the entire delta area, were used and separated into upper (M,N series), middle (B series) and lower (A,C,D series) delta regions (Figure 1 and Table S1 of the auxiliary material) (Canada Center for Topographic Information, available at http://maps.nrcan.gc.ca/ and http://toporama.cits.rncan.gc.ca/). East-west divisions across the delta were also chosen (Table S1). Files were predigitized into separate layers conforming to the original topographic maps (wetlands, water bodies, floodplain). Most aerial photography used to generate the maps was performed during the early 1950s, with the exception of some areas of interest such as settlements. Several topographic maps were qualitatively cross-referenced with current delta satellite imagery (LANDSAT-5; July 2004) and results suggested that the delta morphology has remained quite stable over the past 50 years. After input into ArcMap™, polylines (wetlands, water bodies and floodplain material as shaped lines) were converted to polygon features with recognized shapes with areas. All polygons were then distinguished as delta and nondelta entities using definitions developed by previous investigations which distinguished delta material from surrounding regions based on depositional histories [Mackay, 1963; Rampton, 1988; Hill et al., 2001].
2.3. Lake and Channel Classification
 Delta environments in the arctic are extremely complex systems of anastomosing rivers and adjacent lakes often directly connected in continuous networks. Thus defining what an individual lake is (i.e., two separate lakes versus one lake with two bays) and what is a channel is not always straightforward. Therefore objective criteria were developed to both separate lakes from channels and separate lakes connected with other lakes. Delineation of channels from lakes was straightforward with most channels showing obvious high length to width ratios and direct connection with other delta channels. Few cases arose where channels were quite short and wide and in such cases these were designated as channels if length was greater than width. Most lakes within the delta were obvious separate, unconnected entities. Lakes that were connected in long chains, had multiple “bays” or had any characteristics not consistent with a completely closed shoreline were subjected to criteria that would determine if lakes were separate or the same water body. These criteria were developed partially based on the classification system as described by Sippel et al.  in similar work in the Amazonian floodplain. Figure 3 shows that adjoining water bodies were considered separate lakes if separated by a single line on a 1:50,000 topographic map (Figure 3a), separated by a connection that was longer than it was wide (i.e., a channel) (Figure 3b), and maximum lake length was more than twice the width of a lake-river confluence (Figure 3c).
 A “connection” between two lakes was defined as where the width of the connection became consistent after reducing from each lake's maximum width (Figure 3b).
2.4. Delta Statistical Evaluation
 Following separation from flowing channels and each other, lakes were counted and calculated for surface area and other descriptive statistics. Enumeration was performed by an inherent database generated by ArcMap™ for all separate polygons in a given layer. Polygon areas were determined through the application of a polygonal area VB script as provided by ArcMap (available at http://support.esri.com). A sample of approximately 40 lakes, from various hard copy topographic maps of the delta, was then taken and measured for area by planimeter and returned an average deviation of roughly +7% compared to the GIS calculation (Table S2). Area calculations from a 132 lake set near Inuvik [Marsh and Hey, 1991] deviated roughly +3% compared to GIS calculations. Other layers were also crosschecked with results similar to the lake layer. Additional areal distribution and descriptive statistics were also generated (JMP® version 18.104.22.168 [SAS Institute, Inc., 2003]). Delta-scale area and coverage measures were achieved with additional measures of total delta area using manual polygon generation within the ArcMap™ software to trace the delta boundaries in accordance with Hill et al. . Lakes split at map borders were matched with its adjoining portion in all cases in order not to skew the enumeration, distribution and areal statistics.
3.1. Composition of Delta Habitat
3.1.1. Active Delta Area and Habitat Coverage
 Analysis in ArcMap™ generated a total delta area of 13,135 km2, in close agreement with estimates by Hill et al.  of 12,995 km2. Composition of this total delta area was diverse between the four classified habitats during low water conditions (Figure 4). Human settlement in the region has remained low from the 1950s to present with Aklavik, Canada (2005 population 600) the largest community within the delta and Inuvik (2001 population 2894) not part of the delta by definition. Open water covered about 40% of the entire area, with floodplain lakes alone covering approximately 25% of the delta surface. Since aerial photographs that were utilized to generate the topographic maps were taken mostly in late summer (Table S1), the spring flooding period would certainly show a more extensive coverage of open water as around 95% of all lake area is inundated during a normal water year [Marsh and Hey, 1994]. Wetland and intermittent water can be dynamic components in this coverage balance, but during lower flows they combined to cover about 12% of the delta. The remaining 49% of delta coverage was floodplain area comprised of levees, mud, sand and vegetated ground.
3.1.2. Lake Abundance
 Floodplain lakes in the delta, as recognized on the digital topographic maps by the GIS software, occur in extraordinary numbers, with our present census of 49,046 lakes being roughly double the number (25,000) estimated by earlier work (Table 1) [Lewis, 1988]. Of this census, there are a significant number of water bodies that are quite small and collectively represent only a small amount of lake area. Starting from the largest lakes and progressively adding smaller ones, 99% of total lake area in the delta is accounted for by about 35,000 lakes 0.005 km2 and larger and 99.9% of total lake area is accounted for by 45,000 lakes 0.0014 km2 and larger (Figure 5). Thus, even though the full number of water bodies is large relative to prior estimates, the number of lakes would still considerably exceed prior numbers if a size threshold cutoff point were applied to these results. A water body of 0.0014 km2, for example, is small but never the less a significant sized water body in this system (around 40 m in diameter for a circle of equivalent area) and typically is sufficiently deep because of thermokarst subsidence of the lake bottom (an important mechanism in generating the lake richness of this system) to not freeze to the bottom during winter. Lesack et al.  have shown that the maximum depth of lake ice that can form in the delta under the Inuvik temperature regime varies from 0.6 to 1.2 m depending on the depth of overlying snow. Because smaller lakes typically have less wind exposure relative to larger lakes, drifting snow accumulates more readily on smaller lakes and such lakes have ice thicknesses in the range of only 0.6 to 0.7 m. We have rarely observed ice thicknesses greater than 1.0 m, and such cases were on large wind-swept lakes with little snow on the ice. Thus most small lakes with a maximum depth deeper than ∼0.7 m do not freeze to the bottom. Average lake size increased seaward through the delta, progressing from 0.055 km2 in the upper region to 0.072 km2 in the middelta to 0.075 km2 in the lower section. Delta lakes averaged 0.058, 0.075 and 0.070 km2 from east to west and mean lake area overall throughout the delta during low water was 0.07 km2, a smaller value than Lewis  who estimated 0.12 km2.
Table 1. Summary of Descriptive Statistics on Lake Counts and Areas in the Mackenzie River Delta
Lake Abundance and Area
Total of lake areas
Mean lake area
Maximum lake area
Minimum lake area
3.0 × 10−6 km2
Median lake area
Variance of lake areas
Standard deviation of lake areas
Standard error of lake areas
Range of lake areas
Quartiles of lake areas
0.004, 0.047 km2
3.1.3. Habitat Variability Within the Delta
 Percent surface area of lakes and other habitat types varied from the head to the mouth in the delta (Figure 4). Lake coverage was highest in the middle delta with a sharp decrease toward the tide-influenced regions. Lake coverage in the east, middle and west sections of the delta were similar at 26.2, 24.2 and 26.4% of delta area, respectively. Channel area was highest in the lower delta while wetland percent area through the delta increased from the upper delta (∼2%) toward the coastal lower delta (∼24%). Dry floodplain coverage was generally unchanged through the delta, though largest percent areas were found in the delta head region.
3.2. Delta Water Storage
 The above information on delta area and areal composition of habitat types, in combination with other published information, provides a means for assessing the potential for Mackenzie River water storage in the delta and facilitating the assessment of the potential effect of the delta on riverine nutrient fluxes to the Arctic Ocean. There are three quantities of particular interest when assessing the effect of the delta on nutrient fluxes. First, is the total volume of water stored in lakes during the postflooding period (v1), which is a measure of the minimum lake water that is held in delta lakes, or baseline storage. Second, is the volume of river water going into temporary storage in the delta during the spring break up period when peak annual river flows occur (v5). This is the quantity that mixes with (v1) and then can change biogeochemically while in temporary storage in the lakes and surrounding floodplain. Third, is the volume of river water flowing through the delta during the break up period (v6). This quantity remixes with the biogeochemically altered drain water (v5) from the delta lakes and floodplain and is ultimately delivered to the Arctic Ocean as a Mackenzie River nutrient flux. Estimates of low water storage in channels were not a required element in this analysis as channel flow during the breakup period was a more important measure and was governed by discharge records from the Mackenzie River. All volumes plus the parameters used for their estimation are summarized in Tables 2 and 3.
Table 2. Results From Areal Partitioning of the Mackenzie River Delta Landscape Surface Into Lakes, Wetlands, Dry Floodplain, and Channelsa
Fraction of Total Lake Area
Total Floodplain Area:Lake Area [(a2 + a3)/a1]
Further partitioning distinguishes lakes and their associated total floodplain areas into no-closure, low-closure, and high-closure classes as defined by Marsh and Hey .
Seasonal Water Levels
Mean of summer low water levels
1.231 m asl [e1]
Mean of spring peak water levels
5.636 m asl [e2]
Threshold Sill Elevations
<1.500 m asl
1.500 m asl [e3]
4.000 m asl [e4]
Total Delta Area
Delta Area Partitioning During Post Flood Period (1.231 m asl)
Wetlands (wet floodplain)
Lake Area Partitioning
Total Floodplain [a2 + a3] Partitioning
No-closure total floodplain
Low-closure total floodplain
High-closure total floodplain
8,060 [a2 + a3]
Table 3. Estimates of Water Volume in Lakes of the Mackenzie River Delta (Postflood Period) Plus Potential Storage of River Flow (Floodwater) in the Lakes and Adjacent Total Floodplain During Annual Peak Water Levels and Parameters Needed to Estimate the Volumesa
Closure classes of lakes and total floodplain areas are based on Marsh and Hey, . Area partitioning (a values) and water elevations (e values) are as in Table 2.
Lake Volume During Post Flood Period (1.231 m asl)b
Lake area, km2
Average depth, m
Lake water volume, km3
Water Volume Added to Lakes at Peak Flood (5.636 m asl)c
Flooded lake area, km2
Base water elevation, m asl
Maximum depth added, m
4.405 [e2 − e1]
4.136 [e2 − e3]
1.636 [e2 − e4]
Minimum depth added, m
4.136 [e2 − e3]
1.636 [e2 − e4]
Average depth added, m
Water volume, km3
Water Volume Added to Total Floodplain at Peak Floodd
Total floodplain area, km2
8,060 [a2 + a3]
Flooded total floodplain area, km2
Base water elevation, m asl
Maximum depth added, m
4.405 [e2 − e1]
4.136 [e2 − e3]
1.636 [e2 − e4]
Minimum depth added, m
4.136 [e2 − e3]
1.636 [e2 − e4]
Average depth added, m
Water volume, km3
Water Volume Added to Channels at Peak Flood
Flooded channel area, km2
Base water elevation, m asl
Average depth added, m
4.405 [e2 − e1]
Water volume, km3
Totals at Average Peak Flood (5.636 m asl)
Floodwater + baseline lake storage in delta [v1 + v2 + v3 + v4], km3
Off-channel storage [v1 + v2 + v3], km3
Floodwater in off-channel storage [v2 + v3], km3
Off-channel area flooded [a6 + a8], km2
Overall mean depth of floodwater [v5/a9], m
Mackenzie River Flow
Mean annual flow, km3yr−1
Fraction of flow during delta breakup
Estimated flow during delta breakup, km3 yr−1
 The postflooding (low water) estimate of total delta lake volume (5.35 km3 (v1) in Table 3 and box cfgh in Figure 6) is based on stratifying the delta into areas of no-closure, low-closure, and high-closure lakes (Table 2), then obtaining the product of total lake area in each closure class times the mean lake depth in each class. The stratification was done based on the relative fraction of total lake areas in each closure class [Pipke, 1996]. Average lake depth for each closure class is based on measurements of 27 representative lakes in each class (81 lakes total), representing a random stratified sample from a complete swath of 2,700 lakes [Marsh et al., 1999] across the middelta.
 River water volume added to existing lake water (11.7 km3 (v2) in Table 3) during an average peak flood (5.636 m asl) [Marsh and Hey, 1989] was approximated as a rectangular water layer added to the average low-water volume of the lakes in each closure class (box befc in Figure 6). The area of this layer was taken as the low-water area of all lakes in the no-closure and low-closure classes (Table 2) plus the low-water area of the high-closure lakes that would be flooded at 5.636 m asl (a5 in Table 3). The fraction of all high-closure lakes that would be flooded (0.84) was estimated based on the number of high-closure lakes with sill elevations less than 5.636 m asl relative to the total number of high-closure lakes in the set of lakes assessed by Marsh and Hey . The difference in depth from the low-water point to the elevation of the next closure class, in each of the classes, was averaged to estimate general mean depth for each closure class as a whole.
 River water volume on the floodplain (i.e., not on top of lakes) (14.1 km3 (v3), Table 3) during an average peak flood was approximated as a rectangular water layer adjacent to the position of the water's edge of the lakes in each closure class (box abcd in Figure 6). The total area of this layer is the combined estimated area of floodplain plus wetlands (herein “total floodplain”) associated with each closure class of lakes (Table 2). Partitioning of total delta floodplain area associated with each closure class was based on the approximation that the overall ratio of floodplain area to lake area is 2.42. This is in line with the finding of others that 25% of the delta area is accounted for by lakes. The flooded area in each class was then adjusted for average levee heights [Marsh and Hey, 1994]. At average peak water (5.636 m asl), the mean height of levee tops are submerged in the case of no-closure and low-closure lakes (i.e., 100% flooded). In the case of total floodplain area associated with high-closure lakes (a7), the fraction of total floodplain area flooded (0.82) was estimated as the relative difference from the sill height of the high-closure class (4.0 m asl) to mean peak water level (5.636 m asl), versus the difference to the mean levee height for high-closure lakes (6.0 m asl) [Marsh and Hey, 1994]. The difference in depth from the low-water point to the elevation of the next closure class, in each of the classes, was averaged to estimate the mean maximum depth for the layer in each closure class. The mean depth for the layers in each closure class was then taken as half the maximum depth to account for the triangular shape of floodplains adjacent to lakes (i.e., half of box abcd).
 River water volume added to existing delta channels (7.7 km3 (v4) in Table 3) during an average peak flood (5.636 m asl) was approximated as a rectangular water layer added to the average low-water level of the channels (1.231 m asl) [Marsh and Hey, 1989]. Channel area was obtained directly from the GIS results and mean depth of the layer was the difference between average low water and peak water (4.405 m).
 The volume of Mackenzie river water flowing through the delta during break up (55.4 km3 yr−1 (v6) in Table 3) is based on the product of the mean annual flow of the Mackenzie River (284 km3 yr−1) times an estimate of the fraction of the annual flow volume that occurs during the delta breakup period (0.20) (L. F. W. Lesack et al., unpublished manuscript, 2007). Mean annual flow of the Mackenzie River was based on annual discharges at Mackenzie River at Arctic Red River (just upstream of the Mackenzie-Arctic Red confluence) from 1973 to 2004 (Water Survey of Canada, available at http://www.wsc.ec.gc.ca).
4.1. Composition of Delta Habitat: Effects of Deltaic Processes
4.1.1. Active Delta Area
 Deltas are dynamic environments that can change on an annual scale, especially between large flooding events when erosion, ice scour and coastal deposition are enhanced. Sea level increase may also affect total area through coastal retreat [Hill et al., 2001] and increased mouth deposition due to the heightened base level. Sediment balance calculations by Carson et al.  showed a continuing building process within the delta through point bar/overbank lake sedimentation and channel mouth deposition. Compared to erosional processes, deposition within the delta is roughly 40% higher, leaving a slow infilling of lakes and deposition at channel mouths as a result. However, changes in the total delta surface area were not considerable in the past half century based on a visual comparison of delta morphology using current satellite photos. Over this relatively short period of geological time, depositional gains in channel mouth regions may have been balanced by observed significant coastal erosion [Macdonald et al., 1998]. Further, two thirds of incoming sediment from the Mackenzie River is transported to the offshore environment [Carson et al., 1999], thus substantial annual deposition is consumed by subaqueous delta and prodelta regions, each not contributing to the surficial calculation of delta area.
 The progression of smaller lakes in the upper delta toward larger lakes in the lower delta is mainly a process related to the levee height of channels and connection types of the lakes. Levees in the upper delta portion are higher due to increased riverine sediment loads in the region and increased distance and elevation from sea level. This situation created a closed system for most lakes [Lewis, 1988] where only annual overbank flooding supplied water, sediment and nutrients, leaving a negative water balance for the remainder of the open water season due to evaporative and outflow losses. Toward the north of the delta, lakes tend to be larger due to lower channel levees and increased exposure to summer storm surges from the Beaufort Sea [Marsh and Schmidt, 1993; Jenner and Hill, 1998]. These mostly no-closure lakes are in continual connection with channels and their water levels reflect those of adjacent channels. Lake coverage north to south was highest in the mid delta region, presumably because this region contains a balance of lakes in each closure class. By contrast, the upper delta is higher in elevation than the mid and lower delta and contains the highest abundance of high-closure lakes that are typically smaller in area than lakes of the other closure classes [Marsh and Hey, 1989]. Lake coverage varied relatively little from east to west across the delta.
4.1.3. Channels and Wetlands
 Channel coverage was relatively similar in the upper and middle portions of the delta and was highest in the lower delta. The prominence of channels in the lower delta is largely due to tidal and coastal processes that restrict lake building while the branching of channels is more intense through the relatively erosion prone landscapes of the less vegetated lower delta. Wetland coverage during low water conditions increased dramatically downstream through the delta. The delta landscape in the lower delta is more easily inundated by floodwaters due to relatively low levee heights and connection with sea surges that allows for pooling of water and development of wetland environments. Sediment deposition and lake connection is also highest in the lower portion of the delta [Carson et al., 1999; Hill et al., 2001], allowing for higher deposition rates and longer flow connection periods, and presumably more wetland creation.
4.1.4. Dry Floodplain
 Dry floodplain coverage was relatively constant throughout the delta likely due to the well-established vegetated coverage in the delta that resists development of new thermokarst lakes. Future disturbance of vegetation through storms or land use changes may jumpstart lake creation through ponding and subsidences due to permafrost melt.
4.2. Lake Abundances and Classification Criteria
 The increase in lake abundance, based on this study's census, is at least partly related to the threshold criteria used to define a “countable lake”. To estimate the number of “new” lakes added to the census due to the application of the criteria, random sampling of 50 lakes in each topographic map was performed (as defined by the GIS database before criteria was applied). Within these 50 lakes, the divisions between lakes, resulting from successful application of the separation criteria, were enumerated which represented the number of additional lakes added (i.e., 1 lake separated into 2 lakes equals 1 additional lake). Results from this analysis indicated a range of roughly 5000–7000 new lakes, or about 10–14% of the lake total, was added to the census. When compared to the 132 lake set near Inuvik from Marsh and Hey , this analysis also enumerated 132 lakes. This, however, included a balance of 5 occasions where the separation criteria altered the original set's lake identification, either adding lakes not recognized originally or consuming lakes that were not consistent with the separation criteria.
 The significant benefit of this census is the complete partitioning of all area classified as “water” and thereby facilitating more precise estimates of water storage in the delta. Though our estimates of lake and floodwater storage in the delta are ultimately derived from 1950s aerial photography, the results appear applicable to present conditions. On the basis of comparisons with 2004 satellite imagery, we found minimal areal change in habitat within selected areas of the delta. Among the habitat types, we recognize that channel area is somewhat tricky to define because channel widths can vary significantly with water levels, particularly as water levels decline following the freshet. Given that the aerial photography in this analysis was typically flown in mid to late summer (Table S1), we expect that the channel areas reported here will underestimate the true area during peak discharge periods and to a lesser degree may overestimate the true area during low-discharge period prior to freezeup.
4.3. Uncertainties in Delta Water Storage
 While our results provide improved estimates of habitat area in the Mackenzie Delta, our estimates of water storage in the delta are somewhat less precise. However, some elements of the analysis are conservative, and indeed adequate to make the case that the role of arctic river deltas on river flows to the Arctic Ocean should to be further investigated [Emmerton, 2006]. Our water storage estimate involves a number of assumptions and approximations, but we emphasize that most approximations are derived from substantial prior work in this delta. These are discussed below, and in each case, future investigations are planned to assess our general approach and the sensitivity of results to the various uncertainties.
4.3.1. Average Lake Depths
 The estimate of average lake depth in the delta, for example, warrants further explanation. The 81-lake set, on which the estimate was based, consisted of three clusters of 27 lakes from each of the western, central and eastern areas of the swath and were selected as a representative sample for investigation of sedimentation rates [Marsh et al., 1999]. Each 27-lake cluster was based on 9 lakes from each of the no-closure, low-closure, and high-closure lake classes. As lake ice forms, solutes are excluded from the ice matrix and are concentrated into the underlying liquid water [Lesack et al., 1991]. The degree of concentration enhancement is in proportion to the volume of lake water that freezes relative to the total lake volume and can be used to derive mean depth if the lake area is known [Pipke, 1996]. In most of Pipke's 81 lakes, average depths were estimated from only a few depth measurements toward the geometric center of each basin. Such measurements were performed during ice cover and again during open water, then were checked against the solute enhancement factor obtained from concurrent measurements of major ions in the liquid water beneath the ice cover relative to concentrations during open water. The lake depths directly measured in this study were indeed consistent with lake depths derived from the degree of solute concentration enhancement observed in these lakes [Pipke, 1996].
 Other estimates of lake depths have been reported in the delta, though not based on a comparably broad spatial area of the delta. Full bathymetry available for a limited number of lakes in our study area [Fee et al., 1988; Marsh and Bigras, 1988] typically show that lake bottoms are relatively simple in this system, dropping off quickly near the shore (thermokarst effects on shorelines) and having limited topography offshore. Squires et al.  estimated lake depths from the average of measurements over the open water season on about 30 lakes in the Inuvik region of the delta. Mackay  averaged 50 lake depths from the midwestern portion of the delta during the July–August period. The average of the lake sets from these two studies is about 1.50 m and is reasonably close to the Pipke  estimates (Table 3). The variety of information on lake depths indicates the estimate used in this analysis should be reasonably close to the true value, but at present we do not have adequate information to assess whether there may be significant differences in lake depths either upstream or downstream from the central delta area where most work has been performed in this system.
4.3.2. Area of Delta Flooded at Peak Flood
 An important parameter in estimating this quantity is the fraction of the total delta area associated with each lake closure class. The 81 lakes investigated by Pipke  showed significant declines in the average lake areas from no-closure lakes (0.71 km2) to low-closure lakes (0.29 km2) to high-closure lakes (0.13 km2). The assumption used in this analysis was that total floodplain area would be calculated as a function of lake area in each closure class (i.e., “Total floodplain area [a2 + a3] partitioning” section of Table 2). Though not ideal, this is the best information available at present. On the basis of the fact that the fraction of the delta area represented by lakes and dry floodplain is relatively constant in differing regions of the delta, this approximation should on average yield a reasonable estimate of overall delta area associated with the differing classes of delta lakes, though this may not be accurate in particular areas of the delta. Possible variation of this parameter in differing regions of the delta will be assessed in future work.
 A second assumption is that the relative abundance of no-closure, low-closure, and high-closure lakes in the central delta can be used to approximate such relative abundance in the full delta. Marsh and Hey  showed that the number of high-closure lakes (sill elevations higher than the average spring peak water levels) clearly declines from 33% in the central delta to 13% in the lower delta, but increases to 44% in the upper delta. This is fully consistent with the elevational gradient from downstream areas to upstream areas of the delta. Thus true flooded area should to be overestimated in the upper delta (percent of no- and low-closure lakes overestimated), but will be underestimated in the lower delta (percent of no- and low-closure lakes underestimated). These over and under estimates ought to offset one another for the purpose of delta-wide water storage estimation, but this approximation can be assessed in future investigations.
 Estimating the extent of floodplain area actually flooded at mean peak water level (5.636 m asl) involves the assumption that the pattern of levee heights measured by Marsh and Hey  in the central delta near Inuvik is representative of other areas of the delta. Important points are that the pattern of levee heights decline consistently from high- to low- to no-closure lakes, and that the average elevations of levee tops for no- and almost all low- closure lakes are less than the mean peak water level. The net result of this is that no- and low-closure lakes can be treated as though the floodplain area is fully inundated at the mean peak water level. Because the average height of levee tops was generally higher than the mean peak water level in high-closure lakes, the extent of floodplain area inundated was estimated via the extent of the rise in water level above the threshold sill elevation for the closure class (4.0 m asl in this case) relative to the mean elevation of the highest levee tops among the high-closure lakes (∼6.0 m asl in this case). The approximation here is that the mean peak water level (5.636 m asl) represents 0.82 of the maximum extent of flooding (i.e., 1.636/2.0) that would occur at 6.0 m asl. This value is very close to an alternative estimate of this parameter (0.84), based simply on the number high-closure lakes that are flooded at the mean peak water level relative to the total number of high-closure lakes [Marsh and Hey, 1989].
 A last important approximation here underlying the geometric representation in Figure 6 of the delta topography involves effectively treating all the lakes as though each has a well-defined local floodplain area, with each associated floodplain area separated from the floodplain areas of adjacent lakes by well-defined levees. This approximation also treats the lakes as though they are all connected along an elevation gradient via a flow path or channel that cuts through the local levee system surrounding each lake. The issue here is that a large number of the lakes in this delta, though not all, are thermokarst depressions that often have a reasonably defined levee boundary that acts to contain the extent of flooding in the floodplain surrounding the lake, relative to adjacent lakes higher in the elevation gradient. In most such lakes, flooding occurs via gradual entry of water along a defined flow path (sometimes permanent channel) substantially lower than the average levee heights surrounding the lake. There are also substantial numbers of lakes that are not as well partitioned from one another, and in these cases, full inundation of the local floodplain may occur after only modest rises in water level in comparison to more well partitioned lakes. The effect of treating all lakes as though they are well partitioned may be underestimating the extent of floodplain flooding and (see section below) the average depth of the floodplain water layer.
4.3.3. Average Depth of the River Water Layer in the Delta
 Our estimate of the river water volume going into temporary delta storage is relatively sensitive to our estimate of the average depth of the river water layer in the delta at the time of peak water levels. While the average difference in water levels (4.405 m) between the spring peak during break up and low river levels during the summer is based on a reliable and long record (25 years) for Inuvik [Marsh and Hey, 1989], the approximations represented in Figure 6 represent significant assumptions. We are not aware of another approach to estimate water depths on the total delta floodplain without an analysis of high-resolution floodplain topography (not presently available). Work by Marsh et al.  suggests that basing the depth estimate on results from the Inuvik gauging station could underestimate the water layer thicknesses, though it is very difficult to estimate by how much. Measurements during the spring break up of 1992 revealed that the water surface across the delta is not flat, showing a bulge of 0.7 to 1.6 m in the water surface of the Middle Channel relative to East and West channels during the period of rising water. It is possible that, on average, a higher portion of lakes in the middle of the delta may flood annually [Marsh and Hey, 1989] and to a greater depth, than on the eastern or western sides. However, we do not know if the water surface “shape” is the same every year, and the degree to which the lake elevations across the delta may be adjusted to that shape. Our flat surface approach is an approximation of what is really happening, but it should lead to a conservative estimate of the stored water volume. We are not convinced the average depth of the overall floodwater layer is substantially deeper than the estimate used in our calculations because the overall water storage estimate is quite large relative to the volume of river discharge during the ice breakup period (see section below). The spatial water level patterns need to be investigated, but this is a very difficult problem that requires modeling the hydraulics of the system during periods of considerable ice jamming and this has not yet been attempted anywhere that we are aware in a system of this scale.
4.3.4. Volume of Mackenzie Flow During Delta Breakup
 For the purpose of this analysis, we defined breakup as the period from when water levels begin to rapidly rise beneath ice cover, in response to the spring freshet, until ice is fully cleared from the river channel, typically just after peak water level. Breakup discharge in the Mackenzie (∼30–40 days per year) over the entire record (1973–2005) for the gauging station at Arctic Red River (primary inflow to the delta) represents about 20% (55.4 km3 yr−1) of the total annual river discharge, but the true fraction in fact is unknown (L. F. W. Lesack et al., unpublished manuscript, 2007). Discharge cannot be directly measured when the channel cross-sectional areas are unknown because they are filled with varying amounts of ice, and relations between water level and discharge cannot be established because ice jamming affects water levels to a greater degree than the volume of river flow. In the case of the Mackenzie, discharge during this period is not measured, but estimated via patterns in upstream river flow and antecedent meteorological and ice conditions. If 0.20 is indeed the proportion of river discharge that, on average, occurs during breakup (i.e., effectively the rising limb of the annual hydrograph), that would mean the average volume of water moving in and eventually out of off-channel storage in the delta (25.8 km3) would be equivalent to about 47% of the river flow (55.4 km3 yr−1) during this period. This is a very large portion of the breakup discharge. On the other hand, true river discharge during the breakup period could be significantly higher than reported in the records.
 Overall, we are convinced the general uncertainties in this analysis are sufficiently constrained to make a case that the effect of the Mackenzie River Delta on its river flow is significant. If river flow to the Beaufort Sea is important, flows during the break up period and flow interactions between the delta and river need to be investigated via hydraulic modeling.
4.4. Implications for Riverine Nutrient Fluxes to the Beaufort Shelf
 While the water storage volumes we have derived are not precise, they represent a useful foundation for assessing the potential scale of the effect that may occur on river-borne nutrients, as river water during the high-flow period of spring break up passes through such a lake-rich delta system. The investigation of several important mechanisms should to be facilitated by our results. First, we have now quantified that roughly 5.4 km3 of lake water has the potential to mix with 25.8 km3 of river water during the break up period. On the basis of these volumes, the nutrient concentrations in the lakes per se would need to significantly differ from the concentration in river water to impart a strong “lake signature” to the nutrient content of the floodwater (either enhanced or diluted) that will eventually recede to the river.
 On the other hand, the interaction of river water with the total floodplain area beyond defined lake boundaries during the break up period represents a relatively large volume of water in contact with an extensive surface area of floodplain landscape. Prior work has documented the potential for floodwaters percolating in flooded vegetation to acquire significant concentrations of nutrients and DOC [Lesack et al., 1998; L. F. W. Lesack et al., unpublished manuscript, 2007]. The present results indicate that 25.8 km3 of river water may spread out over 11,200 km2 during a typical break up. On the basis of 25.8 km3 of floodwater remixing with a likely underestimated 55.4 km3 yr−1 of breakup river flow, the acquisition of nutrients to the floodwaters from the flooded vegetation (or stripping from the floodwater) would not have to be large to affect the overall content of river water when floodwaters recede to the delta channels. In the case of some constituents such as DOC, TDN, and TDP, the leaching effect from flooded vegetation could be substantial [Lesack et al., 1991; Emmerton, 2006; L. F. W. Lesack et al., unpublished manuscript, 2007].
 An additional consideration is that during storage in the delta, river water has considerable opportunity for photochemical reactions to occur using colored dissolved organic carbon as a substrate. The present results indicate 25.8 km3 of floodwater from the river effectively spreads out in a thin layer (2.3 m thick) over an area of 11,200 km2, during a period of 24-hour day length (within a couple weeks of the arctic summer solstice). Prior work has documented significant photobleaching of DOC and build up of H2O2 among lakes of the delta during periods near the solstice [Febria et al., 2006].
 Overall, the water storage volumes derived from our present work can be used to generate testable hypotheses about each of the above mechanisms.
4.5. Climate Change and Potential Responses of the Delta System
 In a review of the effects of climate change on freshwater ecosystems of arctic North America, Rouse et al.  concluded that regional runoff will probably decrease because of lowering of the permafrost table and associated increases in soil moisture capacity plus evapotranspiration would likely be more important than the possibility of modestly enhanced precipitation. The review also concluded that decreased ice cover thicknesses, longer ice-free seasons and smaller temperature gradients between southern and northern portions of drainage basins would lead to reduced ice jamming and reduced flooding in large north-flowing rivers such as the Mackenzie. Taking that analysis further, L. F. W. Lesack et al., (unpublished manuscript, 2007) concluded reduced ice jamming and discharge in the Mackenzie would lead to overall reduced water levels in the Mackenzie Delta and reduced storage of river water during the high-flow period of spring break up. Other reviews [Prowse and Beltaos, 2002; Beltaos, 2002] have reasonably argued the effect of warmer climate on river ice jamming is less clear because of a variety of complicating factors, including the possibility of increased precipitation influencing flow and ice characteristics that could lead to enhanced ice jamming. A couple of recent papers have also suggested that runoff in Eurasian north-flowing rivers discharging into the Arctic Ocean have increased over the last several decades [Peterson et al., 2002; Lammers et al., 2001]. However, a recent analysis (L. F. W. Lesack et al., unpublished manuscript, 2007) suggests that ice breakup effects on water levels in the Mackenzie Delta have declined over the past 40 years. We are also not thus far convinced that Mackenzie discharge is likely to increase over the longer term because there is evidence the that North American streamflow response to climatic changes may differ from that of northern Eurasian streamflow [Arctic Climate Impact Assessment (ACIA), 2005], including evidence [Woo and Thorne, 2003] that runoff in the Mackenzie system has not recently changed.
 If ice-jamming effects and overall water levels in the delta do indeed decline over the long term in response to continued climatic warming and either stable or reduced regional runoff, each habitat type in the delta will be affected, especially lakes of the delta floodplain. Reduction in flood stage will reduce river water replenishment and affect the water balance of high-elevation lakes (high closure lakes represent ∼33% of total lake numbers) [Marsh and Hey, 1989]. Most high-closure lakes are relatively small in area, but they are susceptible to drying up if not flooded for significant periods [Marsh and Lesack, 1996]. Lower-elevation lakes that continue to receive annual river water replenishment during high stage periods may experience increasing sedimentation as a result of elevated sediment transport in arctic rivers responding to enhanced erosion of basin soils as permafrost areas are thawed [Rouse et al., 1997; Syvitski, 2002]. Recent relative sea level rise has been documented for the Tuktoyaktuk region adjacent to the Mackenzie Delta [Manson et al., 2005] and is generally expected to become widespread throughout the arctic [Proshutinsky et al., 2001; ACIA, 2005]. Such change will likely result in complete inundation of some coastal delta lakes and enhanced wave erosion of the lower delta. Warming air temperatures will enhance degradation of permafrost within the delta itself and could lead to eventual drainage of delta lakes, as has occurred in other areas of the arctic [Smith et al., 2005]. However, it is not yet clear that warming would not also enhance thermokarst deepening of existing lakes or perhaps create additional lakes via thermokarst subsidence of the delta surface in areas of ice-rich soils, the process generally thought to be responsible for the lake-rich nature of the present delta [Hill et al., 2001].
 Among the nonlake habitat types of the delta, wetlands are mostly generated through channel abandonment and lake infilling, and generally are in various stages of the process. If lake areas decline from reduced water replenishment and enhanced sedimentation, wetland coverage in the delta could significantly rise. If ice breakup effects and ice jamming become reduced in the lower Mackenzie, delta distributary channels may experience reduced ice scouring and thus reduced bank and levee erosion during the flood period. Dry floodplain (postflooding) may expand, allowing for more coverage of larger terrestrial vegetation and more soil water retention.
 In short, a plausible scenario of climatic warming in the Mackenzie basin is reduction in the proportion of the spring freshet that goes into off channel storage in the delta and expansion in the amount of freshet waters that are delivered directly to the ocean. This may have important implications for aquatic habitat quantity and quality in the delta (L. F. W. Lesack et al., unpublished manuscript, 2007) as well as for aquatic communities in the coastal Beaufort region. Such a change in the river flow regime may also have important implications for fluxes of nutrients to the ocean [Emmerton, 2006].
 Our lake census of the Mackenzie River Delta has shown that lake abundance is nearly twice as high as previously thought, with about 45,000 lakes greater than 0.14 ha in area accounting for 99.9% of total lake area in the system. The total floodwater volume potentially stored in the delta at peak water levels (25.8 km3) is a substantial volume relative to total Mackenzie River flow (55.4 km3 yr−1) during the high-discharge period of delta break up. During this period, the stored river water can be envisioned in the form of a thin layer of water (2.3 m thick on average) spread out over 11,200 km2 of lakes and flooded vegetation, and exposed to 24 hour per day solar irradiance. We thus conclude the volume of river water moving through the delta during the spring break up period appears sufficiently high to affect nutrient fluxes to the Beaufort Sea, with the magnitude and direction dependant on nutrient type. These results are compelling evidence that water exchange between the delta floodplain and the river channel must be better quantified via hydraulic modeling, to correctly quantify riverine nutrient fluxes to the Beaufort Sea. Given that other work has indicated climatic warming will lead to reduced delta lake coverage, and reduced ice jamming and water levels in the Mackenzie River [Rouse et al., 1997], the present work suggests a concomitant reduction in river water stored in the delta during break up could also affect riverine fluxes of nutrients to the Beaufort. Since the water storage volumes of this paper are fully independent of the technical challenges associated with measuring river discharge during the annual ice breakup period [Beltaos, 1995], the results here also provide an opportunity (L. F. W. Lesack et al., unpublished manuscript, 2007) to assess the veracity of historical discharge estimates that have been published for the Mackenzie.
 We appreciate the assistance provided by Catherine Febria, Jolie Gareis, Shannon Turvey, and Nicole Vander Wal in the lab and field and the technical and logistical support provided by Les Kutny and the Inuvik Research Centre/Aurora Research Institute. Financial support was received from NSERC (DGP and NRS programs to L.F.W.L.), from the Polar Continental Shelf Project (helicopter support to L.F.W.L.), from the Northern Scientific Training Program (to Emmerton, Febria, Gareis, Turvey, and Vander Wal), and Department of Indian and Northern Affairs, and facilities use in Inuvik was subsidized by the Aurora Research Institute. We also appreciate the assistance provided by Walter Piovesan, Daniel Say, and Jasper Stoodley in obtaining digital topographic files from the Canada Centre for Topographical Information. Chris Bone and Daniel Stevens (Simon Fraser University) provided helpful advice with the GIS software.