Warming temperatures are impacting the hydrometeorological regime of Russian rivers in the zone of continuous permafrost

Large Arctic river basins experience substantial variability in climatic, landscape, and permafrost conditions. However, the processes behind the observed changes at the scale of these basins are relatively poorly understood. While most studies have been focused on the “Big 6” Arctic rivers – the Ob’, Yenisey, Lena, Mackenzie, Yukon, and Kolyma – few or no assessments exist for small and medium-sized river basins, such as the Yana and Indigirka River basins. Here, we provide a detailed analysis of streamflow data from 22 hydrological gauges in the Yana and Indigirka River basins with a period of observation ranging from 35 to 79 years up to 2015. These river basins are fully located in the zone of continuous permafrost. Our analysis reveals statistically significant (p < 0.05) positive trends in the monthly streamflow time series during the autumn–winter period for most of the gauges. The streamflow increases in a stepwise pattern (post-1981) for 17 out of 22 gauges in September (average trend value for the period of record is 58 % or 9.8 mm) and 15 out of 22 gauges in October (61 % or 2.0 mm). The positive trends are seen in 9 out of 19 rivers that do not freeze in November (54 %, 0.4 mm) and 6 out of 17 rivers that do not freeze in December (95 %, 0.15 mm). Precipitation is shown to decrease in late winter by up to 15 mm over the observational period. Additionally, about 10 mm of precipitation that used to fall as snow at the beginning of winter now falls as rain. Despite the decrease in winter precipitation, no decrease in streamflow has been observed during the spring freshet in May and June in the last 50 years (from 1966); moreover, five gauges show an increase of 86 % or 12.2 mm in spring floods via an abrupt change in 1987–1993. The changes in spring freshet start date are identified for 10 gauges; the earlier onset in May varies from 4 to 10 d over the observational period. We conclude that warmer temperatures due to climate change are impacting the hydrological regime of these rivers via changes in precipitation type (rain replacing snow). Other factors, such as the melting of permafrost, glaciers, and aufeis or changes in groundwater conditions, are likely to contribute as well; however, no direct observations of these changes are available. The changes in streamflow can have a significant impact on the ecology of the zone of continuous permafrost, while the increasing freshwater fluxes to the Arctic Ocean can impact the Arctic thermohaline circulation.


Introduction
Numerous studies have shown that river streamflow in Northern Eurasia and North America is increasing (Holland et al, 2007;White et al., 2007;Lammers, 2009, 2013;Rawlins et al., 2010). Most of them are focused exclusively on the "Big 6" Arctic rivers -the Ob', Yenisey, Lena, Mackenzie, Yukon, and Kolyma (Peterson et al. 2002;Rawlins et al., 2009;Holmes et al., 2013;Rood et al. 2017). Reported streamflow changes are 45 not homogeneous in terms of different runoff characteristics and time scales. While basin-averaged annual and seasonal discharges of the Lena River basin for the period from 1936 to 1995 were increasing (Rawlins et al., 2009), spring peak discharges show no significant changes within the same time frame (Shiklomanov et al., 2007).
Large Arctic river basins are characterized by a great variety of climatic, landscape and permafrost conditions and the mechanisms of the observed changes could hardly be understood on the large scale of the "Big 6". Although 50 climate model simulations project increased net precipitation over the pan-Arctic watershed, this is not always supported by ground meteorological data analysis (Rawlins et al., 2010). Runoff change does not necessarily coincide with precipitation and potential evaporation changes (Gao et al., 2016) but usually agrees with increase of air and soil temperature. In some cases there is an opposite change direction in runoff from that in precipitation (Karlsson et al., 2015). There are also the studies of glacier retreat and their input into increased streamflow of the 55 Arctic Rivers (Dyurgerov and Carter, 2004). Bliss et al. (2014) assessingassessed the global-scale response of glacier runoff to climate change based scenario modelling, and have proposed that the Canadian and Russian Arctic region exhibits steady increases of glacier runoff and affectconsequently hydrological regime.
River runoff change estimates in North-Eastern Siberia are limited and contradictory. Magritsky et al. (2013) reported an increase in the total runoff of Yana and Indigirka during the period 1976-2006 by 1.5-3% compared to 60 the period before 1976 and noted that the runoff of these rivers increased in summer and autumn by 20-25% and did not change in winter. According to Georgievsky (2016), on the contrary, there is an increase of Yana and Indigirka rivers winter runoff by 40% over the period 1978-2012 compared with the period 1946-1977, and an increase in the spring flood. Majhi and Daqing (2011) concluded that the Yana River monthly flow rises at the Jubileynaya gauge (224,000 km 2 ) for the period of 1972-1999 for June, August, September, October and April, while May, July and Although air temperature rise and associated permafrost degradation have been reported for the whole Arctic and subarctic territories (Nelson, 2003;Frauenfeld et al., 2004), local anomalies could differ due to specific 80 environmental settings. Three periods with different air temperature tendencies were reported for continental regions of North-Eastern Siberia: warming period 1916-1945, cooling period 1946 and modern warming period that started in 1976 (Kirillina, 2013). Analysis of Kirillina (2013) precipitation in the region shows a positive trend with the exception of the Oymyakon meteorological station.
Permafrost temperature in Russia has been increasing and the active layer has been deepening for the last 20-30 85 years (Romanovsky et al., 2010;Sherstyukov and Sherstyukov, 2015) but reported changes are spatially heterogeneous and do not necessarily follow air temperature trends. Permafrost degradation could cause greater connectivity between surface and subsurface water (Walvoord and Kurylyk, 2016), talik development (Yoshikawa and Hinzman, 2003;Smith et al., 2005;Jepsen et al., 2013) and other complex hydrological consequences consequences (Quinton et al., 2011;Connon et al., 2014). such as land-cover changes that can alter basin runoff 90 production in the permafrost region (Quinton et al., 2011), and the changes in the wetland-dominated basins that characterize the southern margin of the permafrost region (Connon et al., 2014).
Existing assessments of maximum and minimum flow across the cold regions is are contradictory. Minimum flow decrease is reported for the Tibetan Plateau (Gao et al., 2016) and Finland (Ashraf et al., 2017) but an increase is observed at Mackenzie River (Yang et al., 2015), at the Lena river and its tributaries (Tananaev et al., 2016) and in 95 most of the Arctic (Rennermalm and Wood, 2010). Shiklomanov et al. (2007) found no widespread significant change in spring maximum discharge, except for the Lena River, among the 139 gauging stations in the Russian Arctic. Tananaev et al. (2016) reported trends in maximum discharge only in nine time series (negative at three and positive at six gauges) out of the 105 in the Lena River basin. The Ob' and Lena rivers showed uncorrelated changes in spring peak with those aggregated from small natural watersheds located within these larger river basins 100 (Shiklomanov et al., 2007). There are significant negative trends (p < 0.1) over 1950-2001 from aggregated maximum discharge for Kolyma river basin (Shiklomanov et al., 2007). According to recent projections, the annual maximum river discharge could almost double by the mid-21st century in the outlets of major Siberian rivers (Shkolnik et al., 2017).
While some studies have examined flows at the outlet of the Yana and Indigirka river basins, Eestimation of 105 streamflow changes of small and medium-sized rivers in the Yana and Indigirka river basins do not exist. The objective of this research is therefore a quantitative assessment of current changes of hydrometeorological regime within two large arctic river basins -the Yana and Indigirka, with both basins completely located within the continuous permafrost zone, and which have long-term runoff observations along the main rivers and their tributaries at smaller scales.

110
In section 2 of this paper, we will describe the study area. In section 3, we will describe the data to be examined and the methods that we will use. Results of the analysis will be presented in section 4, while these results will be put into context and potential causal factors influencing the results described in section 5. Finally, some conclusions will be drawn in section 6.

General description, relief
The Yana and Indigirka rivers are two of the few large Arctic rivers whose basins are completely located in the zone of continuous permafrost (Figure 1). Mainly tThe terrain is mainly mountainous. There is a high elevation of the 120 Verkhoyansk (Orulgan, 2 389 m), Cherskiy (Pobeda, 3 003 m) and Suntar-Khayata (Mus-Khaya, 2 959 m) ranges, as well as wide river valleys. The average altitude of the research basins ranges from 320 to 1410 m and the values of the outlet gauges from -1.55 to 833 m respectively. The lower parts of the Yana and Indigirka basin area are presented by the Yano-Indigirskaya Lowland. The average height is above 30-80 m. Some places of the lowland are raised up to more than 500 m.
Steady cold weather begins in the first week of October; spring starts in the second part of May or at the beginning 135 of June, when seasonal snow cover starts to melt. Snow accumulation is relatively small and accounts for 25-30 cm depth.
Annual average precipitation at Verkhoyansk meteorological station (137 m, 1966-2012) is about 180 mm, while at Vostochnaya station  it is 280 mm. Most precipitation (over 60%) occurs in summer, with a peak in July of up to 70 mm per month (Vostochnaya, 1966(Vostochnaya, -2012.

Soil, vegetation and landscapes
The research basins are situated in the transitional zone between forest-tundra and coniferous taiga ( Figure 2). For high altitude mountain areas above 1900 m a.s.l., goltsy (bald mountain) and small glaciers with minimum and maximum area defined as 0.024 and 5.76 km 2 (GLIMS andNSIDC, 2005, updated 2017) are typical. Vegetation in 145 this landscape is absent. Broken stone is present in the form of glacial frost-split boulders as well as diluvia soil of the valley slopes with admixed loam material.
Below is the tundra, which is characterized by distribution of a tight and depressed layer of grass and moss with bushes under which there is rock formation with some ice with admixtures. Most of the research areas are covered larch woodland with moss-lichen cover. In river valleys, grass-moss larch forest and swampy sparse growth forest 150 are typical. Soil types at these landscapes are clayey podzol with partially decayed organic material underlain by frozen soil and bedrock (Makarieva et al., 2019). The updated map of permafrost landscapes (Fedorov et al., 2018) can be used for more detailed description of the studied basins.

Permafrost
Permafrost distribution controls the hydrological regime, especially baseflow formation and ratio of maximum to minimum discharge (Niu et al., 2016). In the studied river basins, permafrost thickness can reach over 450 m at watershed divides and up to 180 m in river valleys and depressions. There are highly dynamic cryogenic features such as retrogressive thaw slumps found in the Yana river basin that indicate on-going permafrost degradation processes (Günther et al., 2015). Permafrost temperature at a depth of zero annual amplitude in the studied region 160 typically vary from -3 to -11 °С, the active layer depth -from 0.3 to 2 m (Explanatory note …, 1991).

Hydrological regime and water balance
The hydrological regime is characterized by spring freshet, high summer-autumn rainfall floods and low winter flow. In winter, small and medium-sized rivers freeze thoroughly. Spring freshet starts in May-June (on average the 165 20 th of May) and lasts approximately for а month and a half. In summer, deglaciation watersglacier runoff as well as ones from melting aufeises and snowfields add to rainfallscontribute to streamflow (Hydrological Yearbooks, 1936-1980State Water Cadastre, 1981 (Grave, 1964). Annual precipitation amount at the mountain peaks is estimated to be as high as 800 mm (Vasiliev, Torgovkin, 2002;Hydrological Yearbook, 1983).
The difference by six times at two watersheds of similar area shows the impact of local conditions on runoff 175 formation. Annual values of flow at the outlet gauges of the Indigirka (ID 3871, 305000 km 2 ) and Yana (ID 3861, 224000 km 2 ) Rivers are 166 and 156 mm respectively.
According to observational data (Makarieva et al., 2018a) annual evapotranspiration depends on the type of underlying surface and the distribution of landscapes over the catchment area. On average, it ranges from 90 mm at the goltsy landscape to 140 mm at the sparse growth of larch trees (Lebedeva et al., 2017).

Groundwater and taliks
Since the whole Yana and Indigirka river basins are located in continuous permafrost, groundwater can be found in a seasonally developed active layer, underneath permafrost (supra-permafrost water) and in taliks (Shepelev, 2011).
Depending on texture, infiltration and other soil properties, the active layer stores and transmits water to rivers in 185 summer and early autumn. Water in the active layer could sustain recession river flow in autumn only before the freezing front reaches the permafrost table. Taliks with thicknesses of several meters typically exist under small and middle-sized rivers even if they the rivers may freeze in winter (Mikhailov, 2013). Through-taliks are typically found along the river channels with water depth exceeding 3-5 m. Rivers in continuous permafrost often lose water to channel taliks during summer and gain in winter (Arzhakova, 2001). Through-taliks are also associated with 190 fractured deposits with depth exceeding the permafrost thickness (Glotov, 2015). Rivers with lengths >800 km and basin area >75 000 km 2 typically do not freeze up and flow over the winter. Year-round groundwater springs and winter flow of large rivers suggest suprapermafrost water contribution.

Aufeis and glaciers
195 Aufeis (naleds), which form at mountain foothills, as well as in sub-mountain and intermountain depressions, are another distinguishing feature of the region. According to Simakov and Shilnikovskaya (1958) Tthe research area hasd about 1,281 10,000 aufeises covering an area of about more than 142,00082 km 2 . (Sokolov, 1975). Recent assessment of aufeis area conducted for the Yana and Indigirka River basins based on recent Landsat imagery analysis has shown that the number of aufeis increased up to 1,796 but the total area decreased to 1,713 km 2 200 (Makarieva et al., 2018c).
In winter, the aufeises reduce river stream and underground flow, and in the warm season, melted aufeis waters form an additional source of river runoff. Most significant flow from melting aufeis is observed in May-June. In most cases, the share of the aufeis component does not exceed 3-7% of annual river runoff (Reedyk et al., 1995;Sokolov, 1975), but in May its proportion may exceed 50% of the total runoff (Sokolov, 1975).

210
Few glaciers are found in the research area (GLIMS andNSIDC, 2005, updated 2017). The total area of glaciers is about 2.2 km 2 at the Yana river and 436 km 2 (0.14% of the area) of the Indigirka river basin.
At some studied basins, glaciers reach up to 0.38% of the basin area (3488, Indigirka river -Yurty). The glacier runoff may exceed 3.8% of the overall annual runoff and reach 6.1% of runoff in July and August, as, for example, at the basin of the Agayakan river (the Indigirka basin), where glaciers cover over 1.35% of the catchment (USSR 215 surface waters…, 1972).
On the slopes of the Suntar-Khayata and Cherskiy Ranges, perennial snow fields and rock glaciers are widespread (Lytkin, Galanin, 2016). They, along with the ice of the active layer and summer atmosphere precipitation, may represent a significant source of the local rivers, however in this respect they are poorly studied (Zhizhin et al., 2012;Lytkin, Galanin, 2016).  (Table 1, Fig. 1).
Ten out of 22 examined catchments have areas less than 10 000 km 2 , with five of them being less than 100 km 2 .
Maximum catchment area is 305 000 km 2 (riv. Indigirka -Vorontsovo). Average elevation above sea level of the 230 examined catchments varies from 320 m (Khoptolooh stream -Verkhoyansk) to 1 410 m (riv. Suntar -riv. Sakhariniya mouth). Thus, the study covers not only large, but also small and medium-sized rivers over a broad range of elevations and landscapes typical for the studied mountain region.
We used daily discharge data for the entire observation period from 1936 up to and including 2014-2015, published in Hydrological Yearbooks (Hydrological Yearbooks, 1936-1980State Water Cadastre, 1981 and available 235 for the period 2008-2015 on the website of the Automated information data system for state monitoring of water bodies (AIS SMWB) (URL: https://gmvo.skniivh.ru, reference date: 01.03.2018).
Monthly and annual flow (mm) and spring freshet start dates (counting days from the beginning of the year) were analyzed here. There are different ways to determine a freshet start date. For example, Lesack et al. (2013) defined the initiation of freshet discharge based on a threshold value of 3% increase in discharge per day. In this study where 240 non-freezing and freezing rivers were investigated, a different approach was considered to be more appropriate. A day was considered defined as a freshet flood start date if its discharge reached or exceeded 20% of the average discharge value in the studied year. All the streamflow series were visually checked up for the adequacycorrectness of suchthis assumption.
In some cases when daily discharge data was not available, monthly values of flow were adopted from State Water 245 Cadastre (1979Cadastre ( , 1987. In general the uncertainties associated with discharge determination significantly change from year to year and strongly depend on the computational methods used and frequency of discharge measurements (Shiklomanov et. al, 2006). Possible errors in flow values shown in the database as reliable are as follows for all gauges: average annual flow errors do not exceed 10%, monthly flows errors are 10-15% for the open channel period and 20-25% for winter 250 months. The errors of "approximate" flow, placed in the database in parentheses, can exceed the values indicated above by 2-3 times (State water cadastre, 1979). The errors for streamflow in large rivers streamflow may be slightly lower: monthly flows errors are 4-12 % for the open channel period and 17 % for winter months in the Lena River basin (Shiklomanov et al., 2006).
Daily air temperatures and precipitation data series, observed at thirteen weather stations (the elevation range varies 255 from 20 to 1288 m) located in the studied basins, over different periods from 1935 (but not later than 1966) to 2015 (some stations to 2012) were reduced to average monthly values series (Table 2).
Monthly soil temperatures under natural cover at different depths down to 320 cm from three weather stations over a period of 1966-2015 were also analyzedanalysed. Detailed description of soil temperature data sets and their quality control methods may be found in Sherstyukov (2012а;2012b). Active layer thickness (ALT), used for the analysis, 260 was estimated with polynomial interpolation of soil temperatures at depth (Sherstyukov, 2009;Streletskiy and Sherstyukov, 2015).
The source of the meteorological and soil temperature data is All-Russian Research Institute of Hydrometeorological Information -World Data Centre (http://meteo.ru/data). Combined monthly hydrometeorological data used in this study is available in Makarieva et al. (2018b).

Methods
In this study, we used applied the combination of widely used statistical methods for the trends detection assessment of trends presence and assessment of their values in hydrometeorological data. General guidance with detailed description of different types of tests and their adequate application can be found in Kundzewicz and Robson 270 (2004). as described by Tananaev et al. (2016). Time series of runoff characteristics (monthly flow, estimated flood starting dates and maximum daily flows) and meteorological elements (monthly and annual values of air temperature and precipitation; soil temperature and ALT) were evaluated for stationarity, in relation to presence of monotonic trends, with Mann-Kendall and Spearman rank-correlation tests, at the significance level of р < 0.05 (Mann, 1945;Kendall, 1975;Lehmann, 1975). If both tests proved suggested a trend at the significance level 275 р < 0.05, a serial correlation coefficient was tested at unity lag r(1). With the serial correlation coefficient r < 0.20, the trend was considered reliable. In the case of r≥ 0.20, to eliminate autocorrelation serial correlation in the input series «trend-free pre-whitening» procedure (TFPW), described by Yue et al. (2002), was carried out. «Whitened» time-series were repeatedly tested with Mann-Kendall non-parametric test at the significance level р < 0.05. The Pettitt's test (Pettitt, 1979) was applied to look for the presence of a change point in time series at p ≤ 0.05, the 280 Buishand range test (Buishand, 1982) was used to search for numerous discontinuities at the same significance level.
Trend values were estimated with Theil-Sen estimator (Sen, 1968) and are given in the relevant data units along with the percentage change since the beginning of observations. Note that the trends are presented for the entire period of observations, and not for the period after the change point was identified, as there can be multiple trends within the period of observations. In some (specified) cases, the significance level was relaxed to the value р > 285 0.05.

Results
The results of the trend analysis are presented in the Tables 3-8. The Tables have the same structure and designations. The cells filled with grey color correspond to statistically significant trends with p<0.10. If any value 290 is bold, it has significance p<0.05; if a value is in italics, it has significance 0.05<p<0.10. In Tables 4 (precipitation) and 7-9 (streamflow) each cell with significant trend contains three numbers: 1) the value of total change for the whole period of observations in the characteristic unit (for example, mm) 2) percentage of total change (%); 3) where available -the year of change point or letter "m" for monotonical trend. If there is neither year, nor "m", the Pettitt's test was not carried out due to too many gaps in the data. Statistically significant trends values are divided 295 into 4 groups and marked with different colors accordingly: change points around 1966 -magenta, 1970-1985green, 1986-1995 -violet, 1996 and later -yellow. Monotonous trends and where change points were not available due many gaps are in black. For streamflow, the year of change point marked with * indicates that the gauge has a long-term series of more of than 70 years with change point in about 1966 and no significant trend after that period (last 50 years). In some cases, a second year of change point is given in brackets, it was estimated with Buishand 300 range test. We used the same colors as in the Tables 3-8 in Figure 3 showing the percentage change of monthly and annual streamflow and Figure 4, S12-S17 which presents the spatial patterns of change periods.

305
Annual air temperature increase is statistically significant at all 13 studied stations (Table 3, Fig. S1) with an average cumulative value of about +2.1°С (from +0.16 to +0.46 and average +0.35°С per 10 years) for the historical period of observations (minimum, maximum and mean number of years correspond to 47, 80 and 63).
During the period from April to July, there is a significant air temperature trend at most of the weather stations of the region; average trend values in these months account for +3.0°С, 3.1°С, 1.9°С and 2.3°С correspondingly (Table  3). Positive air trend is observed in March in the mountainous part of the Indigirka River (5 stations with elevation > 500 m). In August, air temperature has increased at three Yana basin stations and two lowland stations in the Indigirka basin accounting on average for 3.4 and 2.4°С increase respectively. In September, air temperature has increased at two stations of the lower part of the Indigirka river basin by 2.3°С on average.
There are no significant trends in October at the Yana basin weather stations but the mean positive trend at 5 stations 315 of the Indigirka basin accounting for 3.7°С in the same month. In November, at 7 stations of 13 in both basins positive trends are statistically significant and reach on average 4.7 °С with a maximum value of 6.3°С at Delyankir station (ID 24691).
Positive winter air temperature trends are significant at 4 weather stations in December, 6 -in January and at 1 station -in February. In March, the positive trend value is 3.6 °С on average for 7 meteorological stations of the 320 mountainous part of the Indigirka River basin.
Change point analysis was conducted for 9 stations out of 13 with full data series. Almost all changes occur through abrupt shifts rather than monotonic trends. Most change points are attributed to two periods : 1975-1985 and 1986-1996. The shifts in winter air temperature occurred at the beginning of the millennium (1999)(2000)(2001)(2002)(2003)(2004).

325
Monthly precipitation analysis for 13 meteorological stations in the region, located at elevations from 20 to 1 288 m, 1966 -2015 (some stations until 2012), has shown no evidence of a systematic positive trend (Table 4, Fig. S2).
On the contrary, at 6 weather stations out of 9 in the Indigirka basin, a statistically significant precipitation decrease is observed in January and at 3 stations in February. In the Yana basin, 2 stations have shown decrease of 330 precipitation in December. Average statistically significant decrease in winter period months accounted for the ranges from -3.0 to -6.7 mm per month or -44-92 %.
In lowland areas which are close to the furthest most downstream gauges of the Yana and Indigirka rivers, reliable negative trends were identified in July (-73%, -23 mm) at Chokurdah station (ID 21946, the Indigirka basin) and in August station (-71%, -26 mm) at Kazachie station (-71%, -26 mm) -(ID 21931, the Yana basin). It is unlikely that 335 runoff from these areas significantly impacts the flow at the outlets of the Yana and Indigirka Rivers as most of the runoff is formed in upper mountainous parts of the basins.

340
We The evaluationed of cold season precipitation calculated as the sum from October to April was carried out. Two stations in the Yana River basin have shown a statistically significant (p<0.05) decrease of about 40% or 20 mm.
Three stations in the Indigirka River basin experienced negative trends at the level of significance 0.05<p<0.08 with reduction of precipitation of about -28% or 15 mm per season at the level of significance 0.05<p<0.08. The change points in decreasing tendency for solid precipitation are estimated from 1980 to 1996 with most of them occurring in 345 1986.
A Ppositive trend in warm period precipitation (May -September) is detected at Oymyakon station (ID 24688), the upstreams of the Indigirka river, with the values of 35% or 54 mm. Warm season precipitation has decreased by almost half (-48% or -60 mm) at the outlet of Indigirka river (Chokurdah station, ID 21946).

350
The state and timing of precipitation in May and September (which are the months when the 0°C air temperature threshold is crossed) were analyzed. The results show that meteorological stations in the higher elevations (which mostly are located in the Indigirka river basin) exhibit a shift towards larger amounts of rain rather than snow in both studied months.

355
In May, three stations show an upward trend for a larger fraction of rain (p<0.05). Vostochnaya station (ID 24679, 1288 m) exhibits the strongest changes: mean share for the whole period is 0.43, trend value is 0.45. The stations Oymyakon and Agayakan (ID 24684, 24688) have mean value of rain share equal 0.79 and on average increase by 0.16 (Table 5).
In September, the mean share of rain fluctuates from 0.

Soil temperature and ALT
There are three meteorological stations with soil temperature data available for the entire historical period including the beginning of the 21st century for studied territory with total area of 529 000 km 2 -two in the Indigirka and one in the Yana river basin.

370
We analyzed soil temperature at a depth of 80 cm and estimated maximum active layer thickness ( Table 6) On average, soil temperature at studied depth increased by +3.4°С in summer months for the last 50 years.
Maximum trend values are observed in July and August and account for +4.1 and 4.4 °С respectively. In winter, the corresponding temperature average dropped by -2.2°С for the same period of observations. Winter trends should be 380 viewed with some caution as there was a significant gap in the observations from 1985 to 1999.  (Table 6, Fig. S3). In June -September soil temperature at 80 cm depth dropped in on average by -2.8°С and increased by +4.8°С in October -April.

390
The change of soil temperature occurs in an abrupt manner -the shifts in winter temperature are identified in 1977 (October), 1983(November, April), 1990(May) and 1994-1995. In summer, the shift cannot be estimated due to the data gap from 1990-1999.

Runoff
Positive statistically significant trends (р<0.05) in monthly streamflow were identified during two main periods: in 405 autumn-winter and in the first month of a spring flood -in May for most of the river gauges (Fig. 23, Table 7-8).
Most of the time series with significant trends are nonstationary, where changes are attributed to break points. Single and double change points are common for studied streamflow records and are described below.

410
In May, a statistically significant increase in streamflow is observed at twelve of the twenty-one studied gaugesobserved at 12 gauges of the 21 studied ones. Those twelve may be divided into two groups.
The first group (Fig. S4, Group A) contains 7 gauges with basin areas from 8290 to 89600 km 2 characterized by a longer continuous series of observations beginning from 1937 to 1956. The shifts of streamflow occur in [1964][1965][1966] and follow a similar behaviour. In most cases, a negative tendency changes to an insignificant positive one. Average 415 trend rates are 79% or 8.9 mm for this group. If the assessment of trend is made for the last 50 years after the change point , no significant trend is detected in May for this group of gauges.
The second group (Fig. S5, Group B) consists of 5 gauges with no significant trend, followed by breakpoints during the period from 1980 to 1999 and positive trends thereafter varying from 5.5 to 25.7 mm with average value of 12.2 mm (or from 64 to 103%, 86% in average). Minimum basin area for this group is 16.6 km 2 , maximum is 52800 km 2 .

425
The small river in the Indigirka basin (ID 3510 -644 km2, p<0.057) has shown total decreased trend value during the last 70 years with a break point in 1967. The streamflow shifts from high mean with negative tendency to low mean with slightly positive tendency. If the analysis is made for the period 1967-2015, streamflow in June at this gauge does not present any detectable changes. Total negative trend value accounts for -61% or -14.2 mm.

430
In July, no statistically significant changes of streamflow occur in the Indigirka river basin, but in general one may note a negative tendency (Table 8). In the Yana River basin, two nested gauges show opposite tendencies (Table 7).
The streamflow at gauge ID 3474 (8290 km 2 ) monotonically decreased by 38% (26 mm). At the same time gauge ID 3443 (52800 km 2 ) experienced total increase of streamflow by 17 mm or 36% with a breakpoint in 1995.

435
In August, an increase in streamflow was identified at 7 out of 22 studied gauges with an average increase by 55% or 18 mm. Areas of the catchments, where discharge increases in August, vary from 644 to 89600 km 2 . Two nested gauges (ID 3507 and ID 3489) have a monotonic increasing trend from mid-50s to 2015. The other, including two nested gauges of the Adycha river (ID 3443, ID 3445), exhibit the shift in 1982-1987 (Fig. S6, Fig.2).
In September, positive trends were identified at 17 out of 22 gauging stations (average value is 58% or 9.8 mm); 440 three of those rivers are small ones (catchment areas are less than 100 km 2 ), the others are medium and large rivers (Table 7-

Low flow (October -March)
In October, streamflow increases at 15 out of 22 gauges (oin average by trend rate is 61% or 2.0 mm) and in November at 11 out of 17 non-frozen gauges (oin average value is by 54% or 0.4 mm) (Table 7- and 2001and , in November -1981and -1999and , mainly around 1994 In December, positive trends are found at 6 out of 15 non-frozen gauges (Table 7-8, Fig. S8, Fig. 2). Average positive trend magnitude is 95% or 0.15 mm. Change points occur from 1981 to 1994. A negative monotonic trend in December is found at gauge 3483 with the magnitude -54% or -0.04 mm.

460
Decrease of streamflow at the Indigirka river at the Vorontsovo (ID 3871) is identified in December and January for the period 1936-1996 and accounts for 32 and 25 % or -0.20 and -0.08 mm respectively with change points in 1964-1969 ( Fig. S8-S9, Fig. 2).
In February, no trends are found at the Yana river basin where only 5 gauges stay unfrozen at this month of the year (Table 7). In March, streamflow trend at the Yana in Yubileynaya is negative with change point around 1989 when monthly average value of 0.003 mm declines to zero.
An increasing annual trend is also found at the gauges (ID 3478, 22.6 km 2 and ID 3479, 7570 km 2 ) with an increase by 73% or 79 mm with shift in 1988 and monotonic growth by 48 % or 34 mm correspondingly. Monotonic flow changes are observed at two nested gauges of the Yana River (ID 3414, 45300 km 2 and ID 3861, 224000 km 2 ) with the magnitudes by 22%, 24 mm and 39%, 60 mm correspondingly (Table 7, Fig. S10).

485
In the mountainous part of the Indigirka river basin, positive step trends in annual streamflow are found starting from 1993 at two gauges in the upstream of the Indigirka river (ID 3488, 3489) and its tributary, the Elgi river (ID 3507), with average magnitude by 27% or 49 mm. An increase in annual streamflow is also observed at the smallest basin from the analyzed set (ID 3516, 16.6 km 2 ) with a magnitude of 31% or 115 mm. Runoff in August and September have contributed most significantly of all months to the annual streamflow rate changes at these gauging 490 stations (Table 8, Fig. S10).

Maximum daily streamflow
The analysis of maximum daily streamflow was carried out for the warm period from May to September. In general, the patterns of changes of maximum daily discharges replicate the change of monthly streamflow (Table S1)

Freshet onset dates
In the small rivers with basin area less than 2000 km 2 , freshet starts in the middle of third week of May (May 11-18); for the larger rivers this date shifts to the middle of the fourth week (May 25, on average). Air temperature 14 increase in the last decades has led to significantly earlier freshet starting dates. Freshet starts 4-8 days earlier than 50-70 years ago (the trends are statistically significant) in 8 rivers with the identified streamflow changes in May, as well as in 3 gauges for which monthly streamflow increase in May is statistically insignificant; two of these gauges have catchment areas smaller than 1 000 km 2 (Table 1)

Precipitation
Although precipitation is projected to increase over the pan-Arctic basin, this is not always supported by ground The complex topography of the Arctic region may cause considerable underestimation of precipitation as meteorological stations tend to be located in low elevation areas (Serreze et al., 2003). Analyzed precipitation data 550 did not undergo wind correction which may vary from 10% in summer to 80-120% in winter due to the effect of wind on gauge undercatch of snowfall (Yang et al., 2005). The study confirms high uncertainty of spatial pattern of the precipitation trends in cold regions (Hinzman et al., 2013). Eastern Russia. The soil temperature data for the studied region is controversial.

Soil
Although aAir temperature increases in May-July period at Oymyakon station, soil temperature dropped on average by -2. 8°С (1966-2015) in summer. Therefore we suggest that the data of Oymyakon station is not reliable as the decrease of soil temperature has happenedoccurred in an abrupt manner which would not be possible without artificial reasonsappears to be related to the replacement of an instrument. Additionally, it is hypothesised that the 565 increase of liquid precipitation shwould warm permafrost through advective heat transport.
Partially it could be the consequence of liquid precipitation increase which increased by up to 53 mm (or 36%) in the warm season (May -September). However, precipitation has been growing monotonically and an abrupt drop of soil temperature occurred between 1987 and 2000.

Autumn
September shows the most considerable change of hydrological regime. The increase of streamflow happened at basins of all sizes in 17 more than ¾ of the of the 22 studied gauges. In September, air temperature and precipitation 595 increased only at 2 and 1 meteorological stations out of 13 respectively.. Although precipitation increased at 10 stations out of 13 in September, only one trend among them is statistically significant.
In the Indigirka river basin, the increase in streamflow can be attributed to the shift of precipitation type in September. Investigating the correlation between monthly streamflow and precipitation in September for 4 small watersheds (area <100 km 2 ) where meteorological stations are located nearby, we found that the correlation 600 coefficient varies from 0.16 to 0.59 for liquid precipitation and from 0.14 to 0.60 for total precipitation (Table 9, Fig.S11). We could not explain the reasons forof low correlation at gauge 3433 (Table 9). It is also worth to notenoting that for gauge 3501, the correlation coefficient increases from 0.27 to 0.46 for total and liquid precipitation respectively.

605
The shift of precipitation from solid to liquid It occurs starting from 1993 and manifests in a considerable increase of rain fraction in comparison with snow at meteorological stations located in the mountains. The fact that the changes of streamflow are observed at the gauges regardless of basin size and in general match the breakpoints of streamflow rise and precipitation ratio shift indicate that it is climate rather than other possible factors (permafrost 610 thaw, increased groundwater connectivity, etc.) which drives those changes. Average magnitude of streamflow trend reaches up to 9.8 mm, which is comparable with the mean absolute increase of rain precipitation of 12.2 mm. In the Yana River basin, one would expect a similar linkage between precipitation and streamflow in autumn, but because of the lack of meteorological stations in higher elevations this cannot be confirmed. Spence et al. (2011) have shown that the trend towards more autumn rainfall in the northwestern subarctic Canadian 615 Shield with no sign of significant total precipitation increase has been sufficient to cause late season peaks in discharge and higher winter low flows. Recessional curves of early autumn flood events extend into later autumn and winter season. The streamflow changes in October generally repeat the spatial pattern of changes in September but with lower magnitude.
The hypothesis of Berghuijs et al. (2014) stating that a shift from a snow-towards a rain-dominated regime would 620 likely reduce streamflow is contradicted by the results of this study for the rivers in the continuous permafrost zone.

Winter
Upward trends of low flow are observed in most of the Arctic (Rennermalm and Wood, 2010). The widely found hypothesis is that increased low flow indicates permafrost degradation, aquifer activation and better connectivity 625 between surface and subsurface water. Increasing active layer thickness, enlarged infiltration and sub-surface water contribution to winter discharge by deeper and longer flow pathways sustain increasing winter streamflow in permafrost environments (Karlsson et al., 2015, Niu et al., 2016St. Jacques and Sauchyn, 2009;Tananaev et al., 2016).
Change points for low flow were detected in 1980s, 1990s and 2000s. Similar change points are observed at the 630 Lena River basin (Tananaev et al., 2016), where the major shifts in all nonstationary time-series occur in the 1990s, but when the data from adjacent decades are combined, the major changes in minimum discharge occur between 1985 and 1995, and in mean annual daily flow between 1995 and 2005.
The changes in November-December tend towards basins larger than 17,000 km 2 although many smaller ones do not freeze up until January. Glotova and Glotov (2015) attributes that to the fact that the fraction of groundwater 635 contribution to middle-sized and large basins is higher than to smaller ones in the continuous permafrost zone. Markov and Gurevich (2008), Gurevich (2009), Dzamalov andPotehina (2010) have substantiated the hypothesis on a regulatory impact of river ice cover in the regions with long-lasting winter on ground water feeding into rivers. It suggests that in colder winters, with a significant thickness of ice, the total water discharge decreases in small river basins. In less severe winters, there is a decrease in the thickness of river ice and the preservation of higher runoff of 640 the underground streamflow in the river by the end of winter. (2014)  (Copernicus Climate Change Service, project #C3S_422_LOT1_SMHI). However, Shiklomanov and Lammers (2014) have found that the relationship between annual maximum ice thickness and mean river discharge over 650

Shiklomanov and Lammers
November-April has shown no significant correlation: the highest correlation coefficients were found for the Yenisey (r = −0.63) and Lena (r = −0.54).

Freshet
All The changes of monthly streamflow changes in May occur in an abrupt manner. At 7 gauges (area >=8290 km 2 ) 655 out of 12, the shift of flow happened around 1966, it was accompanied by two consequent years (1967,1968) with extraordinarily high streamflow in May. Monthly flow in May 1967 andor 1968 exceeded the monthly average by 3.8 -6.5 times. Moreover, at 9 gauges out of 22 it was the highest flow through the whole period of observations. At the other 6 gauges, streamflow was the second highest among the observations with three of them differing by less than 5% from the historical maximum, the other 3 -from 11 to 27%.
No significant trends are detected for the period after the change point  alone. The changes of spring freshet start date were identified at 10 gauges with streamflow changes in May. Trend value varies from 4 to 10 earlier days. This agrees with Savelieva et al. (2000) who stated that the final frost in spring has been 12-15 days earlier than the mean value in Yakutia. Smith (2000) estimated the magnitude of time shift of start date of spring ice-cover events for the outlet of the Indigirka river at Vorontsovo (ID 3871) as 8.1 days for the period of 665 observations 1937-1992.
Another 5 basins have shown the shift of flow in May during the period from 1980 to 1999. In 4 cases of 5 , in most cases, that shift can be attributed to earlier freshet. Yang et al. (2014Yang et al. ( , 2015 documented increases in May flows for the Mackenzie and Yukon Rivers, respectively related to earlier melting of snow. Earlier start of freshet agrees with identified significant air temperature increase in May at 9 out of 13 meteorological stations in the region. But in 670 contrast to the Lena river basin, where strong warming in spring led to an advance of snowmelt season into late May and resulted in a lower daily maximum discharge in June (Yang et al., 2002;Tan, 2011), at the study gauges streamflow have not reduced in June (except ID 3510).

Water balance
675

Precipitation and streamflow relationship
Positive discharge trends in the Arctic have been established in recent years Peterson et al. 2002Shiklomanov and Lammers, 2009). Increasing precipitation has been suggested as the one of the drivers of increasing streamflow (Dyurgerov and Carter, 2004;McClelland et al. 2004).

680
The findings of this study largely contradict this, with an increase of streamflow reported at most of the gauges despite no substantial change in absolute amount of precipitation. We estimate that 10 mm of snow is lost in September as rain draining directly into streamflow. Adding the decrease of precipitation in February-March months, the decrease of snow water equivalent during winter may reach up to 15-20 mm in total. Curiously though, no evident decrease of streamflow is observed during freshet or summer months.

685
Inconsistent trends in discharge and precipitation have also been reported for other Arctic rivers. Berezovskaya et al. (2004) have shown that the patterns in increasing trends of runoff from Siberian Rivers cannot be resolved from the apparent lack of consistent positive trends in the considered precipitation datasets. Milliman et al. (2008) suggest that a fluctuating climate in the Yana River watershed precluded delineating a statistically significant change in either precipitation or discharge, but the Indigirka River basin had a statistically 690 significant decrease of precipitation and non-significant increase in runoff.

Evapotranspiration
The issue of increasing streamflow in the absence of increasing precipitation may be explained by decreased evapotranspiration. By increasing the active layer depth and thus potentially lowering the water table, 695 evapotranspiration might decrease, leading to increases in runoff (McClelland et al. 2004). Milliman et al. (2008) proposed that decreased evapotranspiration due to earlier snowmelt and the changes in water storage within the drainage basins could be a potential source of excess streamflow. Based on modelling results Stieglitz et al. (2000) showed that as warming occurs mostly during winter months it does not lead to an increase in evapotranspiration.
On the other hand, Rawlins et al. (2010) argue that evapotranspiration ration is increasinggrowing. This is supported 700 by reanalysis data for the pan-Arctic domain. According to model simulation, evapotranspiration has a significant trend of 0.11 mm yr −2 (VIC model, 1950(VIC model, -1999, annual total evapotranspiration from LSMs model also shows a positive trend (0.40 mm yr −2 ) (Rawlins et al., 2010). A longer growing season, observed as a result of climate warming in the 20th century, is likely to result in continued upward trends in evapotranspiration (Huntington 2004).

5.6
Cryosphere 705 5.6.1 Thawing permafrost Subsurface ice melting along with air temperature increase, widespread over the studied area (Brown et al., 1998;Yang et al., 2002), can contribute to the streamflow increase (USSR surface waters resources, 1972; Frey and 710 Mclelland, 2009). Though McClelland et al. (2004) argue that if water released from thawing permafrost was making a significant contribution to the observed increase in annual discharge from Eurasian Arctic rivers, one might expect that watersheds with the most permafrost cover would show the largest increase in runoff, but no such pattern is apparent.
An analysis using satellite based gravimetry at the Lena basin suggest that the storage of water in these areas has 715 increased over the past decades (Velicogna et al., 2012;Muskett and Romanovsky;), but the causes of this phenomenon are not clear (Velicogna et al., 2012). One reason may be the permafrost degradation, because warming in the active layer will increase the flow of groundwater and, consequently, the discharge of groundwater into the rivers will be increased too (Ge et al., 2011;Walvoord et al., 2012;Michelle et al., 2016;Lamontagne-Hallé, 2018).
However, this phenomenon only holds if sufficient water is available to replenish the increased discharge.

720
Otherwise, there will be an overall lowering of the water table in the recharge portion of the catchment (Ge et al., 2011). Michelle et al. (2016) suggest that permafrost loss is more likely to contribute to baseflow increases in discontinuous permafrost than in continuous permafrost, where permafrost tends to be cold and thick. The results of this study have shown that baseflow is increasing in the zone of continuous permafrost as well.
If the permafrost is degraded, talik zones can be relieved. Most groundwater discharge occurs through areas 725 overlying open taliks so they play the important hydrogeological role in accommodating preferential pathways acting to either recharge the deeper regional aquifer from the supra--permafrost system or facilitating discharge from the deeper aquifer (Bense et al., 2011;Walvoord et al., 2012;Lamontagne-Hallé, 2018). Permafrost thaw can also generate rapid landscape changes. An example may be thermokarsting and plateau subsidence (Quinton et al., 2011) that in turn influence surface water storage, routing, and runoff (Connon et al., 2014;Michelle et al., 2016).

Glaciers, rock glaciers and aufeises
Another possible contribution to streamflow increase could be the meltwater from glaciers at least in the Indigirka river basin where glaciers take up to 0.12% of the total basin area.
In current climate the glaciers of the region melt more intensively in July and August so their input to streamflow 735 would be intuitively expected during those months. . In the headwaters of the Indigirka river (ID 3488, 51100 km 2 ) the glaciers share is the highest among studied rivers and amounts up to 0.30% of the basin area, however no significant trend of streamflow is observed in July and August. The next downstream gauge of the Indigirka River (ID 3489, F = 83500 km 2 ) exhibits monotonical increase of streamflow in August from 1944 to 2015 which is mainly explained by increasing streamflow of its tributary, the Elgi River (ID 3507) where no glaciers are found.

740
Huss and Hock (2018) estimated hydrological response to current and future glacier mass loss; according to their simulations, annual maximum input of glacier runoff to streamflow in the Indigirka River basin have been passed in the period of 1980-2010 and expected to decline in the future.The glaciers of the region melt more intensively in July and August so their input to streamflow would be intuitively expected during those months, It will also be accompanied by the change of .timing of glacier runoff. In the Indigirka River basin glacier runoff is expected to 745 increase by 20-40% in June and drop by 20% in average in other months (Huss and Hock, 2018). In the headwaters of the Indigirka river (ID 3488, 51100 km 2 ) the glaciers share is the highest among studied rivers and amounts up to 0.30%, however no significant trend of streamflow is observed in July and August.
However, Liljedahl et al. (2017)  where glacier impact would be negligible due to their tiny area (about 1.35 km 2 in total).
The melt of rock glaciers and aufeises, which are widespread in the study area, may have a similar effect of 760 additional water input. According to Lytkin and Galanin (2016), 540 rock glaciers are identified within Suntar-Khayata Range with a distribution density of 8.4 objects per 100 km 2 .

765
A significant positive trend of annual streamflow was identified only for four gauges. Arzhakova (2001) emphasizes that riv. Elgi (ID 3507, the Indigirka river basin) and riv. Adycha (ID 3443, 3445, the Yana river basin) cross tectonic dislocation zones. The studied area is characterized by modern seismotectonic activities (Imaeva et al., 2016). Taliks, which form within faults and excessive jointing zones, supply rivers even in extreme winter conditions (Romanovsky, 1983;Piguzova, 1989). The correlation between streamflow and sub-permafrost waters is 770 also confirmed by Glotov et al. (2011), who assumed that in particular years, the Kolyma river winter runoff losses at cross sections downstream in comparison with cross sections upstream depending on the current extensive extensions and compressions of underflow through talik space during sublittoral seismic activity periods. Savelieva et al. (2000) pointed to redistribution of the water budget between the upper (thaw layer) water and ground water beneath the permafrost. According Savelieva et al. (2000), the increases of temperature and thickness of the permafrost active layer may enhance the water inputs from the seasonal thaw layer into the ground waters beneath the permafrost through tectonic trenches and lake's talik zone.

Overall impacts to the ecosystem
This study has identified increases in runoff in rivers discharging to the Arctic Ocean. Increasing volumes of 780 freshwater flows to the Arctic Ocean could lead to a significant weakening of the thermohaline circulation, changing of sea stratification and sea ice formation (Arnell, 2005;Lique et al., 2016, Weatherly andWalsh, 1996). The total volume of water being discharged from the Yana and Indigirka rivers is around 84 km 3 (about 1.7% of total freshwater discharge to the Arctic). As a result, changes in these two basins are unlikely to impact the Arctic by themselves, noting however that Eeven subtle changes of river streamflow may have large implication on the 785 ocean-climate system (Miller and Russel, 2000). However, if the changes in hydroclimate seen in theses basins are being felt across other Arctic basins, impacts on the Arctic Ocean could be significant.
Approximately 85% of the total terrestrial runoff to the Arctic Ocean is supplied by rivers draining from the Russian Federation (Aagaard and Carmack, 1989), therefore analysis of this database covering the Yana and Indigirka rivers, is particularly timely. Forman and Johnson (1996) have previously identified the volume, timing, and natural 790 variability of Russian river discharge as being a major priority in Arctic science. In particular, the Indigirka is one of the two largest rivers flowing into the East Siberian Sea (Anderson et al, 2011) and thus changes in flows from the Indigirka have the potential to impact on acidity, nutrients and carbon cycling in this sea (Semiletov et al, 2016).
In terms of ecological impacts, increases in discharge, velocity, temperature, and concentration of suspended materials exert important effects on primary production and food-web dynamics in rivers and estuaries (Scrimgeour 795 et al., 1994). Flooding associated with increased discharge is a primary control on the exchange of sediment and organic carbon between rivers and Arctic floodplains (Smith and Alsdorf, 1998).

Conclusions
Analysis of data from the Yana and Indigirka river basins has shown increases in annual air temperature for the 800 region of around 2.0 °С over the last 50 years . Much of this increase has occurred in the late autumn and spring. Precipitation has shown a decrease in the late winter (February-March) and exhibits little change in other seasons. However, the increases in air temperature have led to a shift in the rain-snow ratio, with more rain at the expense of snow in May and September in mountainous areas. In total, snow loss can be estimated at around 20-25 mm, half of which now falls as rain and the other half is the decrease in late winter precipitation. The increase in air 805 temperature is also reflected in soil temperature and the length of the thaw season with increases at two stations and decreases at one other.
Despite the slight decrease in overall precipitation, analysis of daily discharge data for 22 gauging stations in the Yana and Indigirka river basins  has revealed, at most of the stations, statistically significant (р< 0.05) positive trends in monthly streamflow during the autumn-winter period (September through December) and in the 810 spring flood (May-June). Changes in spring flow are also seen through the spring freshet occurring 4 to 10 days earlier over the period of record. Total annual flow has increased significantly in the upstream tributaries of the Yana and Indigirka rivers.
Most of the increases in streamflow are seen via break points, rather than showing a monotonic increase over the entire period of record. The structure of the timing of these changes is very complex; however, in general, the 22 changes occur in the Yana River basin 10 years before the Indigirka River basin. Changes in winter streamflow are seen in the larger river basins (lower-elevation gauges) before they are seen in the smaller river basins. Analysis of monthly data is required in order to identify these changes, as in many cases, changes in annual flow are not significant. Additionally, analysis of other datasets such as precipitation and temperature is required in order to develop hypotheses regarding the potential causes of these changes in streamflow.

820
Increases of streamflow in the autumn-winter period are likely mainly caused by a shift in precipitation from snow to rain in September, with consequent increases in streamflow in that month, continuing into October and later months. Decreased depth of river ice and better connection of surface and ground water due to deepening of active layer and prolonging of freeze-free period may also be causing higher winter flow.
There are no changes seen in spring discharge in the last 50 years (from around 1966) except at several upstream, 825 mountainous gauges in both studied river basins. Curiously, snow losses and the earlier timing of the spring freshet in May do not lead to a decrease of streamflow in June. This additional input of water is likely related to the regional warming trend impacting more intensive melting of the cryosphere elements such as permafrost, glaciers, and aufeises.
It is difficult to directly attribute changes in streamflow to definite causes and it is almost certainly a consequence of 830 a complicated interplay between different changing water storages and pathways in warming permafrost. The possibility also exists of additional water input via precipitation higher in the mountains, but this hypothesis cannot be verified with current observational data.
The observed changes agree with reported positive trends in arctic river discharge albeit with low confidence ( -green, 1985-1995 -violet, 1996 and later -yellow. Monotonous trends and where change points were not available due many gaps are in black. The cells filled with grey color correspond to statistically significant trends with p<0.10. If any value is bold, it has significance p<0.05; if a value is in italics, it has significance 0.05<p<0.10. First and second figures mean trend value (mm and %), third (if available) is for the year of change point. Letter "m" is for monotonical trends. If there is neither year, nor "m", the Pettitt's test was not carried out due to many gaps in the data. Statistically significant trends values are divided into 4 groups and marked with different colors accordingly: change points 1985 and before -green, 1985-1995 -violet, 1996 and later -yellow. Monotonous trends and where change points were not available due many gaps are in black.   1995 1995 1990 1983 1977 1983 1994 24382  -green, 1985-1995 -violet, 1996 and later -yellow. Monotonous trends and where change points were not available due many gaps are in black. "m" is for monotonical trends. Statistically significant trends values are divided into 4 groups and marked with different colors accordingly: change points around 1966 -red, 1970-1985 -green, 1986-1995 -violet, 1996 and later -yellow. Year of change point marked with * indicates that the gauge has long-term series more of than 70 years with change point in about 1966 and no significant trend after that period (last 50 years). In some cases second year of change point is given in brackets, it was estimated with Buishand range test. All designations are the same as in Table 6.