The retreat pattern of glaciers controls the occurrence of turbidity currents on 1 high-latitude fjord deltas 2

28 Glacier and ice sheet mass loss as a result of climate change is driving important coastal changes 29 in Arctic fjords. Yet, limited information exists for Arctic coasts regarding the influence of 30 glacial erosion and ice mass loss on the occurrence and character of turbidity currents in fjords 31 which themselves affect delta dynamics. Here, we show how glacial erosion and the production 32 of meltwaters and sediments associated with the melting of retreating glaciers control the 33 generation of turbidity currents in fjords of eastern Baffin Island (Canada). The subaqueous 34 parts of 31 river mouths were mapped by high-resolution swath bathymetry along eastern Baffin 35 Island in order to assess the presence or absence of sediment waves formed by turbidity currents 36 on delta fronts. By extracting glaciological and hydrological watershed characteristics of these 37 river mouths, we demonstrate that the presence and areal extent of glaciers is a key control for 38 generating turbidity currents in fjords. However, lakes formed upstream during glacial retreat 39 significantly alter the course of sediment routing to the deltas by forming temporary sinks, 40 leading to the cessation of turbidity currents in the fjords. Due to the different deglaciation 41 stages of watersheds in eastern Baffin Island, we put these results into a temporal framework 42 of watershed deglaciation to demonstrate how the retreat pattern of glaciers, through the 43 formation and filling of proglacial lakes, affects the activity of deltas. 44


Introduction
as sediment distribution, benthos development , or the burial of organic 57 carbon, which plays a crucial role in controlling O2 and CO2 concentrations (Smith et al., 2015).

58
This article is a non-reviewed preprint published at EarthArXiv Despite the importance of turbidity currents in the transfer of sediment and carbon to deeper-59 water ecosystems (Biscara et al., 2011), remarkably little information exists on sediment 60 transport processes on high-latitude deltas due to a lack of high-resolution bathymetric data and 61 in-situ monitoring. In the eastern Baffin Island region (Canada) (Fig. 1), the links between the 62 pattern of ongoing glacial retreat (Lenaerts et al., 2013) and sediment transport to the coast 63 provide a complete understanding of the consequences of deglaciation on sediment fluxes and 64 partitioning in fjord systems. Therefore, the factors responsible for the presence of turbidity 65 currents during deglaciation can be precisely identified. 66 The effect of deglaciation on the progradation and activity of deltas is often limited to the 67 stratigraphic record since these processes occur over hundreds to thousands of years (Dietrich  Digital Elevation Model (CDEM) (Fig. 2). Rivers were classified using the Strahler 103 classification and a threshold of 100 pixel was used to define a class 1 river (Fig. 2F).

104
For each watershed, glaciological and hydrological characteristics were extracted using zonal 105 statistics in ArcGIS (Fig. 2). The areas (m 2 ) of the watersheds were calculated along with glacial   Watersheds were also created for all the lakes located within the fjord-delta watersheds and 117 data extracted from these sub-watersheds were removed from the total watershed values, 118 producing new adjusted watershed values that exclude lake sediment trapping (Fig. 2G).

Statistical analyses
120 This article is a non-reviewed preprint published at EarthArXiv Shapiro-Wilk normality tests and QQ normal plots were used to assess normality of 121 distributions. Since distributions were non-normal for most of the extracted parameters, a 122 Wilcoxon-Mann-Whitney test was used to determine if active and inactive deltas had significant 123 differences in watershed characteristic. This non-parametric test was used to test differences 124 between two conditions (active vs inactive deltas) and glaciological and hydrological 125 characteristics of watersheds (presence of glacial ice, ice velocity, river classification, etc.). The 126 test was done using the independent Wilcoxon's rank-sum test in R, which is equivalent to the 127 Mann-Whitney test. A p-value < 0.05 indicates a statistical difference between the two 128 distributions (active and inactive deltas). In most instances, ties in the datasets were present and 129 thus, the values were slightly modified (using jitter in R) in order to compute exact p-values. 130 An effect size was then calculated in R to estimate the size of the effect observed following   The main terrestrial parameter driving nearshore fjord hydrodynamics is river inflow, which 139 controls submarine delta activity by generating turbidity currents (Syvitski, 1989; Hughes 140 Clarke, 2016). Submarine delta activity is here viewed through the prism of subaqueous 141 sediment wave organization. In this study, submarine delta activity is therefore defined as the 142 presence of recurring and highly energetic turbidity currents, triggered at the delta front, and 143 flowing downslope. These turbidity currents typically form sediment waves, the presence of 144 which along delta slope is used to assess if a particular delta is active. In the absence of direct migrated over these two-year periods (Fig. 3). Figure  This article is a non-reviewed preprint published at EarthArXiv in their total watershed (Fig. 4H). Additionally, a Fligner-Killeen test shows that there is no 185 significant difference between the variances of glacial ice in active and inactive delta 186 watersheds when comparing their total watershed (P = 0.58) but that there is when comparing 187 glacial ice in the adjusted watershed (P = 0.03) (Table 1). This significant difference between 188 the variances indicates that watersheds of active deltas have higher variance of glacial ice than 189 the inactive deltas, as expressed in Figure 4H where the percentage of glacial ice in adjusted 190 watersheds for inactive deltas largely remains below 10%, but varies between 30-90% for active 191 deltas. Furthermore, glacial ice velocity within adjusted watersheds (Fig. 4J), which is a proxy  (Fig. 4A), which is a proxy for river discharge (Strahler, 1957)  The results presented here clearly demonstrate the critical role played by glacial erosion and 240 the retreat pattern of glaciers across watersheds in modifying the type of sediment supply to 241 fjords (Fig. 5). The supply of sediment from glacial erosion is assumed to remain relatively 242 constant during glacier retreat (Fig 5A), as suggested by the presence of turbidity currents on 243 deltas with watersheds comprising from 30% to 90% glacial ice. Glacial erosion provides large This article is a non-reviewed preprint published at EarthArXiv of sediment to the ocean. When lakes form, sediment supply to the fjord-head delta shuts down 249 as sediment is trapped upstream in lakes, drastically modifying the hydrodynamics of the 250 marine nearshore environment due to severe sediment starvation (Fig. 5A, C). Both small and 251 large lakes act the same way in trapping sediment upstream of the delta. Sediment starvation is 252 not due to reduced sediment supply from the glaciers but is due to sediment not reaching the 253 coast. Because of sediment starvation, some deltas appear to have been significantly eroded, 254 forming bays while upstream lakes in the watershed are being filled with sediment (Fig. 5C). 255 However, once sediment completely fills the lakes, which appears to have occurred in some 256 watersheds (Fig. 5D), deltas can be reactivated on the long term since the course of the river 257 down to the fjord is re-established (Fig. 5). Hence, although all sizes of lakes are efficient in   (Table 1). Therefore, this parameter was used to predict the location of 278 active and inactive deltas for 644 fjord deltas of eastern Baffin Island (Fig. 6) where 1) less than 279 10% glacial ice in adjusted watershed suggests that the deltas are inactive (unlikely in Fig. 6B); 280 This article is a non-reviewed preprint published at EarthArXiv 2) between 10 and 20% glacial ice suggests that they are possibly active (possible in Fig. 6B); 281 and 3) more than 20% glacial ice in adjusted watersheds suggest that the deltas are active (very 282 likely in Fig. 6B). These thresholds applied to the 31 known deltas yields a 6.5% error where 283 two inactive deltas were mistakenly interpreted as active. In these two cases, other parameters 284 such as moraine damming or storing of sediment within the sediment-routing system appears 285 to play a role but could not be quantified. Using percent glacial ice in adjusted watersheds is 286 thus a strong proxy for predicting where turbidity currents occur in high-latitude fjords.

287
Although recent studies have suggested that glacier-derived sediment flux control the because of the effect of lake trapping that prevents sediment delivery to the fjords. It is however 296 important to note that this likelihood of the occurrence of turbidity currents (Fig. 6B) is only 297 applicable in the modern configuration of lake distribution, which inherently evolve through 298 time.

299
As retreat of glaciers is ongoing, the pattern of retreat may modify the future hydrological and 300 glaciological characteristics, which will then have a direct impact on the occurrence of turbidity 301 currents in fjords. For example, if there is a stillstand during the retreat of glaciers, it is likely 302 to construct a frontal moraine, which will then form a moraine-damned proglacial lake that . If this accelerated ice mass loss continues as predicted, we speculate that moraine-309 damned lake will be less likely to form, thus enhancing in the short term the occurrence of 310 turbidity currents in Baffin fjords. Some lakes may be filled which will allow some deltas to be 311 reactivated. However, if ice-mass loss continues until completely melting, the occurrence of 312 turbidity currents will cease and will have an abrupt effect on the hydrodynamics of fjords.

485
This article is a non-reviewed preprint published at EarthArXiv occurrence of turbidity currents during glacier retreat: 1) A direct connection between glacial 487 erosion and the delta will lead to the occurrence of turbidity currents (B, E). 2) The presence of 488 a lake caused by glacial retreat (e.g., by moraine damming or glacial overdeepening) will alter 489 the delivery of sediment to the delta (C, F). 3) However, if the lake is filled, the connection will 490 be re-established, leading to the reactivation of turbidity currents (D, G).

497
This article is a non-reviewed preprint published at EarthArXiv