Recent Trends in Freshwater Inﬂux to the Arctic Ocean from Four Major Arctic-Draining Rivers

: Runo ﬀ from Arctic rivers constitutes a major freshwater inﬂux to the Arctic Ocean. In these nival-dominated river systems, the majority of annual discharge is released during the spring snowmelt period. The circulation regime of the salinity-stratiﬁed Arctic Ocean is connected to global earth–ocean dynamics through thermohaline circulation; hence, variability in freshwater input from the Arctic ﬂowing rivers has important implications for the global climate system. Daily discharge data from each of the four largest Arctic-draining river watersheds (Mackenzie, Ob, Lena and Yenisei; herein referred to as MOLY) are analyzed to identify historic changes in the magnitude and timing of freshwater input to the Arctic Ocean with emphasis on the spring freshet. Results show that the total freshwater inﬂux to the Arctic Ocean increased by 89 km 3 / decade, amounting to a 14% increase during the 30-year period from 1980 to 2009. A distinct shift towards earlier melt timing is also indicated by proportional increases in fall, winter and spring discharges (by 2.5%, 1.3% and 2.5% respectively) followed by a decrease (by 5.8%) in summer discharge as a percentage of the mean annual ﬂow. This seasonal increase in discharge and earlier pulse onset dates indicates a general shift towards a ﬂatter, broad-based hydrograph with earlier peak discharges. The study also reveals that the increasing trend in freshwater discharge to the Arctic Ocean is not solely due to increased spring freshet discharge, but is a combination of increases in all seasons except that of the summer.


Introduction
Terrestrial freshwater contribution from Arctic-draining rivers to the Arctic Ocean plays an important role in several oceanic processes, affecting systems on both global and regional scales. Variability in this contribution can have wide-ranging effects on global feedback interactions, hydrological extremes and contaminant and nutrient pathways [1]. For example, runoff from Arctic-draining rivers influences salinity stratification within the Arctic Ocean. This stratification regime governs freshwater export from the Arctic Ocean through the northern North Atlantic Ocean, and is an integral part of the global ocean circulation regime. A change in the stratification of surface waters can affect North Atlantic Deep Water (NADW) formation which, coupled with Atlantic Meridional Overturning Circulation (AMOC), is a critical driving force in the global thermohaline circulation [2][3][4][5][6][7]. Arctic runoff plays an important role as a nutrient supplier to near-shore and estuarine ecosystems, providing an influx of organic carbon during the spring freshet, yet diluting waters with respect to inorganic nitrate and silica compounds [8][9][10][11].

Basin Characteristics
The pan-Arctic drainage basins and outlet stations of the Mackenzie, Ob, Yenisei and Lena rivers are shown in Figure 1, with station characteristics provided in Table 1. Total contributing areas of the four major river systems, including ungauged drainage areas, are as follows: Mackenzie 1,800,000 km 2 [24]; Ob 2,975,000 km 2 [25]; Lena 2,488,000 km 2 [26]; and Yenisei 2,554,482 km 2 [27]. The pan-Arctic region contains nearly half of the global alpine and sub-polar glacial area [28]. Meanwhile, some major Eurasian Arctic basins extend below 50 • N, further south than what is traditionally considered within the Arctic region [6] (see Figure 1). As a result, discharge behaviour at each of the four major drainage outlets is influenced along its course by sub-basin tributaries which may adhere to variety of hydrological regimes, such as nival, pluvial, prolacustrine, hybrid or other. For example, hydrologic retention due to extensive wetland coverage or large lakes within a catchment, such as is found in the Ob or Mackenzie basins, will lead to a more moderated seasonal discharge characteristic than basins without such retention [29].
Water 2020, 12, x FOR PEER REVIEW 3 of 13 hydrologic retention due to extensive wetland coverage or large lakes within a catchment, such as is found in the Ob or Mackenzie basins, will lead to a more moderated seasonal discharge characteristic than basins without such retention [29].   1963-1965, 1968-1974, and 1977-1979. Records are not infilled.
Reservoir regulation is known to impact the seasonal distribution of discharge [30,31]. Each of the MOLY watersheds experiences some degree of flow regulation within their catchments, ranging from only one major reservoir in each of the Mackenzie and Lena basins, to four or more major reservoirs in the Ob and Yenisei basins [25,31,32]. In terms of flow regulation, the Yenisei basin is the most substantially regulated, with at least six major reservoirs having a capacity greater than 25 km 3 located along the Yenisei and Angara stems [31,32]. It is considered "strongly affected" by flow regulation and fragmentation according to an assessment of anthropogenic changes in river flow and river channel continuity of large river systems [33]. The next-most regulated is the Ob basin, containing one major reservoir with a capacity greater than 25 km 3 and three midsize dams [25]. The Ob basin is moderately affected based on its classification of flow regulation and fragmentation. Of the Asian basins, the Lena is least affected by flow regulation, with only one major reservoir located along the Vilyuy tributary. It is moderately affected in terms of regulation and fragmentation [33]. The Mackenzie basin is also moderately affected, despite having only one major reservoir located along the Peace tributary. Large lakes in the Mackenzie basin (e.g., Great Slave Lake and Great Bear  Reservoir regulation is known to impact the seasonal distribution of discharge [30,31]. Each of the MOLY watersheds experiences some degree of flow regulation within their catchments, ranging from only one major reservoir in each of the Mackenzie and Lena basins, to four or more major reservoirs in the Ob and Yenisei basins [25,31,32]. In terms of flow regulation, the Yenisei basin is the most substantially regulated, with at least six major reservoirs having a capacity greater than 25 km 3 located along the Yenisei and Angara stems [31,32]. It is considered "strongly affected" by flow regulation and fragmentation according to an assessment of anthropogenic changes in river flow and river channel continuity of large river systems [33]. The next-most regulated is the Ob basin, containing one major reservoir with a capacity greater than 25 km 3 and three midsize dams [25]. The Ob basin is moderately affected based on its classification of flow regulation and fragmentation. Of the Asian basins, the Lena is least affected by flow regulation, with only one major reservoir located along the Vilyuy tributary. It is moderately affected in terms of regulation and fragmentation [33]. The Mackenzie basin is also moderately affected, despite having only one major reservoir located along the Peace tributary. Large lakes in the Mackenzie basin (e.g., Great Slave Lake and Great Bear Lake) provide substantial storage capacity, acting to reduce high spring peaks and sustain lower flows resulting in a more consistent runoff pattern throughout the year, similar to the effect of flow regulation [34]. The percentage of each basin's area that is located directly upstream of a major reservoir (obtained by delineating the drainage areas of the reservoirs) is as follows: Mackenzie 3.9%; Ob 11.6%; Yenisei 46.5% and Lena 4.2%. See Figure 1 for the locations of major reservoirs.

Data Sources
Daily discharge data were obtained from the Environment and Climate Change Canada Hydrometric Database (HYDAT) for stations in the Mackenzie basin and from the Regional, Hydrometeorological Data Network for Russia (R-ArcticNET Russia v4.0) [35] for the Ob, Lena and Yenisei basins. Availability of Arctic hydrometric data is temporally limited, with all outlet stations having published records to 2009 only, and records in many smaller basins not extending past 2000. Complete records for the Mackenzie outlet station begin in 1973, while the Yenisei outlet station has several extensive gaps during the period 1963-1979. As a result, the period 1980-2009 was chosen for analysis of combined MOLY flow, while individual stations were assessed for the entire available record. Available record periods are given in Table 1.

Spring Freshet Definition
Two methods were used to define the volume of discharge released during the spring freshet period: (i) flows occurring during the period April through July (AMJJ), referred to as V 1 , and (ii) integrated flow from the date of the spring pulse onset to the hydrograph centre of mass, calculated from pulse onset to the last day of the calendar year, referred to as V 2 . July was used as the end-date of the V 1 period, since some basins display high discharge rates well into the summer months. The date of the spring pulse onset was determined as the date at which cumulative departure from mean annual flow was most negative. This yields the date when flows on subsequent days are greater than the year average [36,37]. Visual inspection of the results verified that this is a reliable method for identifying the start date of the spring freshet. Choosing the freshet end date by visual means is subjective and influenced by precipitation, temperature and other factors; therefore, the hydrograph centre of mass adjusted by pulse onset as the freshet end date was used as a consistent method for determination of the freshet end date. Other descriptors used to analyze freshet characteristics are given in Table 2. Table 2. Metrics used to describe freshet characteristics. See text for definitions.

Symbol Description
September-November volume

Trend Analysis
The Mann-Kendall test was applied to assess temporal trends in freshet timing and magnitude [38,39]. This non-parametric test is often used for detecting trends in hydrologic time series that may be affected by seasonal climatic variability, missing data or extremes and makes no prior assumptions about the normality of data [40]. In addition, a Trend-Free Pre-Whitening (TFPW) approach [41] was used to correct data for serial autocorrelation following the methods of Burn et al. [42]. This approach first fits a monotonic trend for a data series which is then removed prior to pre-whitening the data series. The monotonic trend is then re-added to the residual de-trended and pre-whitened data series, whereby the Mann-Kendall test statistic and local significance are calculated. To reject the null hypothesis H o , which says there is no significant trend, the p-value must be smaller than α. All trends in this study were considered for their statistical significance at α = 5% and α = 10% level.

Freshet Characteristics
Over the period 1980-2009, the average freshet start dates are May 12, 14, 28 and 19 for the Mackenzie, Ob, Lena and Yenisei rivers, respectively. Based on the freshet definition V 2 , during the period of 1980-2009, averages of 48%, 51%, 57% and 52% of the total annual flows in the Mackenzie, Ob, Lena and Yenisei Rivers, respectively, were released during the freshet period. Table 3 shows the percentage of total MOLY freshwater volume released by each of the four rivers based on the V 1 and V 2 freshet definitions as well as during the months of April through July. Overall, total proportional freshet volume contributions (based on the V 1 or V 2 freshet definitions) were greatest from the Lena and Yenisei, with the Yenisei reaching its peak proportional contribution in the month of April and the Lena slightly delayed, reaching its proportional peak contribution in June and July. The Lena River is largely unregulated and therefore characterized by a sharp spring peak and low winter flows typical of a naival basin with extensive permafrost coverage [43]. By contrast, extensive regulation of the Yenisei River dampens the spring freshet with flows being enhanced from storage releases at other times of the year, such as late fall and mid-winter [31]. The Mackenzie and Ob stations exhibit a more consistent spring contribution characteristic of flow regimes moderated by the existence of large lakes or wetland areas.

Changes in Timing and Magnitude
Observed changes in the timing of freshet discharge are generally most notable during the shorter period of 1980-2009 versus the entire length of available records. All outlets show either a decreasing (i.e., earlier) trend or no trend in pulse onset date (Table 4 and Figure 2), although only the Mackenzie and Lena stations exhibit a significant trend in earlier pulse onset dates of 1.2 to 1.4 days per decade, respectively, over their longer records. However, freshet duration over the longer record (Table 4) shows either a slight decreasing trend (Ob) or no trend, while all outlets have an increasing but non-significant trend in freshet length over the shorter period of 1980-2009 (Table 4). Although peak freshet magnitudes are generally decreasing, no significant changes were detected (Table 4)        * denotes trend is significant at the 10% level and ** denotes trend is significant at the 5% level To assess whether discharge seasonality has shifted for individual stations, the fraction of flows released during the freshet and winter (VDJF), spring (VMAM), summer (VJJA) and fall (VSON) were calculated and are shown as percentage of total annual flow for each station. Table 5 gives trends in the percentage changes during each time window. Over both periods, the percentage of flow released during the freshet (V2) decreases for all stations (except the 0.3% increase for the Ob), although none of those trends are significant. Meanwhile, winter and spring percentages generally increase, while summer proportions decrease. An increase in VMAM percentage coupled with a decrease in VJJA percentage is notable, since it indicates a shift in the timing of overall peak discharges. Although pulse onset occurs in May for all stations, overall discharges typically peak in June. In addition to the decrease in the proportion of freshet discharge, peak discharges are also shifting towards earlier To assess whether discharge seasonality has shifted for individual stations, the fraction of flows released during the freshet and winter (V DJF ), spring (V MAM ), summer (V JJA ) and fall (V SON ) were calculated and are shown as percentage of total annual flow for each station. Table 5 gives trends in the percentage changes during each time window. Over both periods, the percentage of flow released during the freshet (V 2 ) decreases for all stations (except the 0.3% increase for the Ob), although none of those trends are significant. Meanwhile, winter and spring percentages generally increase, while summer proportions decrease. An increase in V MAM percentage coupled with a decrease in V JJA percentage is notable, since it indicates a shift in the timing of overall peak discharges. Although pulse onset occurs in May for all stations, overall discharges typically peak in June. In addition to the decrease in the proportion of freshet discharge, peak discharges are also shifting towards earlier dates. Fall discharges show either a slight increase (Mackenzie) or decrease (Ob, Lena) during the longer records, while during 1980-2009 all stations indicate an increase in fall discharge.

Changes in Combined Circumpolar Discharge
Total annual discharge from all four basins increased significantly by approximately 89 km 3 /decade over the 1980-2009 period ( Figure 5A). To better assess the seasonal contributions to this overall annual increase, trends for all seasonal measures (V 1 , V 2 , V DJF , V MAM , V JJA , and V SON ) were determined. From Figure 5B,C, it is apparent that, while freshet discharge V 2 shows a statistically significant increase of up to 33 km 3 /decade, other seasons also display increasing discharges over the same period. With the exception of summer (+16 km 3 /decade, p = 0.254), all increases are statistically significant at the 5% level. Spring, fall and winter show increases of 29 km 3 /decade, 35 km 3 /decade and 16 km 3 /decade, respectively. Trends in combined MOLY seasonal flow were also investigated to determine whether this annual increase could be attributed to a rising freshet, rising winter low-flows, or some other combination of seasonal increases. While freshet discharge shows a significant increase of 30 to 33 km 3 /decade depending on the definition of freshet ( Figure 5), this change is complemented by corresponding increases in winter, spring and fall. In fact, compared to other seasons, fall exhibited Despite variation in the individual proportions of seasonal flow, there is consistency in the sequencing of the combined discharge compared to individual flows. Freshet contribution as a fraction of combined annual flow for MOLY stations decreases by approximately 1.7% during 1980-2009, although this trend is not significant. Winter proportional contribution increases significantly by 1.3%, while spring fraction shows a non-significant increase. Combined summer fractional flows display a significant decrease of up to 5.8%, which is consistent with earlier findings indicating highly decreased summer proportions for individual outlet stations. Fall fractions show a statistically significant increase of approximately 2.5%.

Summary and Conclusions
Analysis of discharge at the outlet stations of the four largest Arctic-draining rivers indicates that the combined annual discharge from these rivers has increased by 89 km 3 /decade over the period 1980-2009, amounting to an approximate 14% increase over the 30-year period. This estimate is comparatively larger than the 7% increase found in a previous study using longer records from the six largest Eurasian rivers during 1936-2009 [15]. This is consistent with the results found here, which are consistently greater during the shorter, more recent period of analysis. As Figure 4 and Table 4 indicate, trends over the longer periods tended to occur at a much slower rate than those over the relatively shorter 1980-2009 period. This apparent rapid increase in freshwater volume contribution during 1980-2009 may be an effect of the shorter period of analysis, but could also be attributable to accelerated high-latitude warming in recent decades. Similar results from many Eurasian and North American pan-Arctic basins over the recent periods have been attributed to intensification of hydrological processes that are an expected manifestations of a warming climate [44,45]. In particular, the increases in winter baseflow were found to be caused predominately by increased permafrost thawing, which enhances infiltration and deeper flowpaths resulting in broad-scale mobilization of subsurface water into rivers [46].
Trends in combined MOLY seasonal flow were also investigated to determine whether this annual increase could be attributed to a rising freshet, rising winter low-flows, or some other combination of seasonal increases. While freshet discharge shows a significant increase of 30 to 33 km 3 /decade depending on the definition of freshet ( Figure 5), this change is complemented by corresponding increases in winter, spring and fall. In fact, compared to other seasons, fall exhibited the greatest increase, of up to 35 km 3 /decade. This may be a result of delayed river ice freeze-up dates, or increased late-summer and autumn precipitation. Meanwhile, the fraction of discharge released during the freshet as a percentage of total annual flow decreased by approximately 1.7%, while winter and fall proportions increased. A distinct shift towards earlier melt timing was also indicated by a strong decrease (5.8%) in proportional summer discharge along with a corresponding increase (2.5%) in spring discharge.
Individually, trends in the fraction of flow released seasonally agree with overall trends in circumpolar flow. Individual rivers show varying decreases in portion of flow released during the freshet, coupled with increases in winter, spring and fall fractions and decreases in summer amounts. The only exception to this general tendency is in the Ob River, which shows a decrease in winter and a slight increase in summer proportional flow. These deviations are not substantial enough to affect the combined trends of all four rivers. Pulse onset dates occurred earlier, while freshet durations increased slightly and peak freshet magnitudes generally decreased. Rising winter and fall discharge proportions, combined with lower peak freshet magnitudes, increased freshet durations, and lower summer proportions are indicative of a potential shift to a flatter, more gradual annual hydrograph with an earlier pulse onset. While this apparent shift in seasonality can clearly have important consequences for the Arctic and global feedback systems, it remains yet to be determined how much of this change can be attributed to flow regulation and how much to climatic changes. Despite the recent window of observation used for combined flow, many basins have had some form of flow regulation in place for extended periods, and the establishment of such regulation will likely have impacts on the longer-term records. In addition, studying trends over large, continental-scale basins will obscure any effects of regional climatic variation on smaller-sized basins. It is thus recommended to undertake an analysis of trends and climatic drivers on a sub-basin level to determine potential causes of shifting seasonality in Arctic freshwater influx. Funding: This work was partially supported by a Discovery Grant and ArcticNet funding from the Natural Sciences and Engineering Council of Canada (NSERC) to one of the co-authors.