Glaciers and Climate of the Upper Susitna Basin, Alaska

. As part of a proposed hydropower facility, extensive ﬁeld observations were conducted in the Upper Susitna basin, a 13,289 km 2 glacierized catchment in central Alaska in 2012-2014. This paper describes a comprehensive data set of meteorological, glacier mass balance, snow cover and soil measurements, as well as the data collection and processing. Results are compared to similar observations from the 1980s. Environmental lapse rates measured with weather stations between about 1000 and 2000 m a.s.l. were signiﬁcantly lower over the glaciers than the non-glaciated areas. Glacier-wide mass balances 5 shifted from close to balanced in the 1980s (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) 1981-1983 to less than -1.5 m w.e. yr − 1 in 2012-2014. Winter snow accumulation measured with ablation stakes on the glaciers closely matched observations from helicopter-borne radar. Soil temperature measurements across the basin showed that there was no of dominant inﬂuence on SWE at the Upper Susitna basin with snow accumulation on the glaciers m measuring 2-3 times higher than at 1000 m. A notable south-north decrease in total SWE and accumulation gradient indicates a strong orographic inﬂuence. Over short spatial scales in the ablation zone, surface roughness is responsible for high spatial variability in SWE.

I think the dataset presentation can be more thorough and better structured than is presented now. I miss at various places context of statements and the naming of the different stations can be structured better, such that it is clear to the reader which data you are referring to. I specified this in the specific comments below. Addressed specifics below I suggest major comments, predominantly since I think the focus of this paper should be on the data and not too strongly framed to dam-implication work and/or climate change work, since this work does not address that.

Specific comments:
The introduction does not focus on the relevant subjects. You present solely a measurement dataset and you focus in the introduction on climate change and (modelled) river runoff, which you did not do in the paper. Please restructure the introduction, remove this information or at least shift focus to the data. You could include more information about previous field works/data sets in this region?

We believe this larger context will be important to many readers. For people unfamiliar with the area, this gives them enough information that they don't have to search elsewhere for an understanding of why our data might be interesting or useful.
In general: update the captions of figures, such that those are complete. In Figure 1 I miss for example explanation of the subpanels. Also update the figure labels and text inside the figures such that those are readable (mainly Figure1) and resolution is high enough (also for the tables).

Added to caption of figure 1: "The main map focuses on the glacierized portion of the basin, the large inset shows the whole Upper Susitna basin which drains to the proposed dam site, and the small inset shows the basin in the context of the state of Alaska."
P1L16: You did not raise any questions yet. Please rephrase. Done P1L16: Your introduction is focused on climate change and river runoff, however I do not think 1 data set is sufficient to solve that. I would focus more on fundamental understanding rather than climate change. Please shift the focus of the importance of the data or even present it only as a dataset.

It is true that this data set won't "solve" climate change or changes to river runoff, but the introduction is meant to connect the paper to the larger topics that it relates to. That's what we're doing here. The wording makes it clear that the paper focuses on the data, this section of the intro is providing the context for why the data was collected. We added a sentence about data scarcity to bolster the motivations.
P10L31: rephrase "in the ice rather than the snowpack" to "the layer consists of ice instead of snow" or similar Done P11 Figure 5: explain the reversed patter of ice temperature (red lines) with depth in the text. At 21 April lines are ordered from light to dark lines with depth, while in June this pattern is reversed. With other words explain why the temperature gradient reverses. "As air temperatures rose, the subsurface temperatures of the upper layers increased but with a time lag that increased with 25 depth (Figures 5 and 6)." P11 Figure 5: the lightest lines (Ice2.5m) are not clearly visible and Ice3m not present at all. Please make those lines more clear.

With 15 lines we did our best to have distinct colors that show the short term variability. Perhaps the confusion comes because we do not have data from 4 m depth so there is a visual gap between the data from 3 m and 5 m, particularly at the beginning of the record. The legend clearly indicates the depths.
P12 Figure6: the reversed temperature gradient is here not visible why (not)? Go more in depth in the data (general comment)

The reversed temperature gradient is visible. On 28 April the coldest temperatures are at the surface. On 26 May the coldest temperatures are at -1 m from the initial snow-ice interface, with warmer temperatures at the surface. The pattern continues through the end of the season.
P12L4: "simple weather stations"? Why is station type in Table A1 than indicated by "HOBO glacier"? please make naming consistent throughout the paper. Edited appendix to remove "AWS" and "HOBO" in favor of "energy balance" "simple glacier" and "simple tundra." P13L7: which of the two sensors are more trustworthy? And why the comparison? Please explain in text.

The manufacturer's stated accuracies are 0.1 C for Rotronic and 0.2 C for HOBO as shown in Tables 1 and 3. Whether that equates to "trustworthiness" is a matter of opinion. The comparison is done for calibration (as stated in the section header). Added a new first sentence: "To ensure the validity of data collected from many individual sensors scattered across our study area, we set them up side-by-side before deploying them to the field."
P13L9-11: I think these argumentations do not match: The HOBO sensor is slower than the Campbell, but coefficient of 1, and then conclusion is that there is a lack of consistent pattern. I do not follow this, please explain and rephrase As stated in the text: 5 minute data show that the HOBO sensors are slower, but hourly averages have a correlation of 1. Figure 7

shows the lack of consistent pattern. Added references to figures in the conclusion sentence: "The lack of a consistent pattern in these comparisons (Figure 7) prevented us from adjusting the HOBO temperature data to match the Campbell data. The high correlation and low temperature offsets (Figure 8) among sensors gave us confidence that using HOBO stations to assess temperature patterns across the basin was valid."
P13L20: add some explanation/conclusion. I miss in this whole section why you do the comparison between the sensors and eventually the physical interpretation or conclusion from your statements.

Added explanation to the top of the section and this physical interpretation to this paragraph: "Again, this gave us confidence that we could use our data to assess humidity variations across the basin."
P14L1-2: did more people had this problem? Is it a random tip that can also occur during dry periods (since this can not be filtered out)? Or is the tip sometimes 'stuck'?

We don't know of others that have run into this problem. It is not a random tip because when the data show rain at one station, there is almost always also rain at other stations. We never observed the bucket getting stuck in our calibration.
P14L4: or conclusion is the HOBO has a sensor problem.

Yes that is the conclusion of the previous paragraph. This paragraph focuses on the Campbell sensor and the conclusion is that if the internal electronics create a double tip, the logger is filtering it out.
P14L16: is this katabatic flow measured or a assumption it develops?

We measured wind speeds and directions at the On-Ice station consistent with a katabatic flow in summer, as the text now notes.
P15 Figure 7: include the colours in the caption Added "Each colored line represents a HOBO sensor, blue dots are for the reference Rotronic sensor, and red dots are for a Campbell 107 Temperature Probe." P16-P17: Section 3.3 I do not think this section is a great addition to your purpose of the paper and not supported by any in depth discussion, please remove. We included this section to help explain differences in glacier mass balance discussed later. For the revisions, I added text to the mass balance section to make the connection more direct. "These mass balance patterns are consistent with the warming temperatures and relatively stable precipitation measured at Talkeetna Airport." P18 Figure 11: is this the same transect measured every week at same location? What do you mean with "plotted relative to a reference station"? Does this mean steepness in line is varying in time? Please explain in caption. Mention in caption what upper stations in winter are not operating/measurement problems.

The transect is not the same in every week. The stations are fixed in place, and whenever the station has enough data to calculate a weekly average, it gets plotted. Changed caption to: "Weekly air temperature profiles show the winter inversions and summer differences between glacier and tundra temperature. For each week, the reference station (Windy Creek Lower, 940 m a.s.l., triangles and black dots) was plotted on the horizontal axis according to the date. The other stations (triangles or circles) were plotted to the left or right of the reference station according to their temperature difference."
P19 figure 12: Add coloured lines for each of the dots to show whether the gradients change in time/how sensitive they are. Given the inconsistent recovery of data from different sites in different years, we chose not to display best fit lines. It doesn't make statistical sense to try to compare these lines from year to year. P19 figure 12: Insert in caption how the mass balance in computed (from the "HOBOglacier" station in Table A1? Inserted "...measured with the glaciological method" P21L2-3 please rephrase. Absolute difference at 2000 and 1000m or did you do some averaging? This is an approximate estimate based on binned radar data and point measurements for multiple years and multiple glaciers. We're trying to convey the big picture here, and the reader can refer to the figure for the numerical details.
P21L4: how do you know surface roughness is responsible? Please add explanation or supporting material for this statement. Added: "Over short spatial scales in the ablation zone, surface roughness is responsible for high spatial variability in SWE. The end-of-summer glacier surface is rough due to streams, crevasses, melt ponds, and moraine material. The end-of-winter snow surface tends to be relatively smooth compared to the summer surface, but can also have wind-derived roughness features that contribute to the variability in SWE over small distances."

P21L6-8 I am not convinced What do you think it is then? We tried our best to adequately account for errors in our methods and this lists what we couldn't account for.
P22L9: again, why the roughness of the ice surface? See above P22L16: add explanation why data become noisier. Why does ice give more noise signal?

Added: This was likely caused when the acoustic signal bounced off different elements of the rough ice surface in successive measurements.
P22L20: you assumed constant density? What are the implications of this assumption? Changed "assumed" to "used" since this is based on measurements. New sentence reads: "To calculate ablation from the observed distance change, we used a density of 350 kg m$^{-3}$, based on the average of 5 snow density measurements at the site." P24 Table 6: increase fontsize  Table doesn't fit on the page with a larger font. The copy editors may be able to fudge this somehow.
Section 5.2: please remove, I do not think is Section is of additional value Readers who are familiar with the prior work or the field site will be interested in this information.
P24L11: "though we did not do a detailed texture analysis", But still you know it matches with the STATSGO soil map? I am not convinced.

Soil texture is relatively easy to observe in the field and the soils we encountered were largely sandy loam. This matches with the STATSGO map. Sites with very little soil development (Windy Creek Upper, Valdez, Two Plate) show as bedrock on the STATSGO map. We also looked into some of the STATSGO attributes such as organic content and drainage characteristics, which aligned broadly with our field observations. As mentioned, we did not do a more detailed texture analysis (e.g. taking samples back to the lab), so what more do you want to convince you? Changed: "Though we did not do a detailed texture analysis on the soils, the characteristics we observed generally matched up with the State Soil Geographic (STATSGO) soil map (https://datagateway.nrcs.usda.gov/)." To: "Though we did not do a detailed texture analysis on the soils, our observations of soil texture, organic content, and drainage characteristics generally matched up with the State Soil Geographic (STATSGO) soil map (https://datagateway.nrcs.usda.gov/)."
P27L27: this is only a minor section and for me not strong part of your paper and now you present the climate change numbers as one of your main conclusions. I think it is totally fair to present our data as a baseline for future measurements. The point of concluding statements is to highlight how this paper and the data within it are relevant to the broadest audience possible.
P28 Figure 18: air temperature is gray colour? Yes, as indicated by the legend. P28L6: not new conclusion, snow amounts are generally higher at high elevations True, but it has not been shown for this study area and time period, so it is still a valid conclusion.
P28L7: you did not measure the soil and your conclusion is that these match with the mapped soil descriptions. Please remove this statement out of your conclusions and preferably also out of the text Detail added above to support these conclusions.
Conclusion in general: please do not focus on climate change and dam implications, but give conclusions about the data you found in the field. What did you find and why is it special? We do give conclusions about our data and we highlight how the data are relevant to the broadest audience possible (people who are interested in climate change and runoff). *****

RC2:
General Comments: This study presents, validates and interprets a comprehensive and impressive data set, which covers a range of parameters in the variable environments of the Susitna Basin, Alaska. The data set includes meteorological, glaciological and soil parameters. The data set is unique, as many of the measurements were done in complex terrain where measurements generally are sparse. It is effortful and requires extensive planning to acquire meaningful data in this terrain. Problems in the data are addressed and generally, implications that arise with these problems are described in detail. Overall, the manuscript is well structured and provides a good overview of the data. The data set itself could be extremely valuable for model validation or comparison with future field studies. Thanks! However, the manuscript is not always coherent and suffers from redundant information (e.g. section 3.3, section 5.2, figure 9), None of these sections or figure are redundant. Perhaps the reviewer meant to say relevant? That critique was addressed in response to reviewer 1 and can be summarized with: "Readers who are familiar with the prior work or the field site will be interested in this information." which distracts the reader from following the key points and weakens the focus of the paper (see specific comments).

Addressed below
The introduction does not adequately motivate the manuscript, as it does not really make clear why the data set is important and what the purpose of the data set is (see specific comments).

Addressed below
In addition, some of the presented data appear isolated and need to be put into context better (e.g. section 4.4., how do continuous mass balance measurements compare to stake measurements nearby?) See P22L10: "The net depth of snow lost at the sonic distance sensor site was 1.97 m (17 April -30 June), which is comparable to the 2.15 m snow depth measured in a snow pit about 6 m away on 14 April 2013." And see figure 5 which shows the ablation stake and continuous measurements together. Added reference to figure 5 in section 4.4.
Therefore, I suggest a number of minor revisions to focus the main messages of the manuscript and to emphasize the uniqueness of the data set.

Specific Comments:
Introduction: Mentioning climate change in the beginning of the introduction is not convincing, since you only acquire three years of data. Either remove the link to climate change, or emphasize that the data set is meant to be used for comparison with future studies (as you do in the conclusions). We also compare our data to data from 30 years ago and see evidence of climate changes, so we assert that climate change is in fact relevant to the paper.
P2, l9-10: using changes in river flow on dam operations as a motivation here seems misplaced, since you mention in the beginning that the dam was not built. In addition, river flow is not covered in this study. Please rephrase or remove. You could motivate each of the data types (meteorological/climatological, glaciological, snow, soil) individually, as you do later in the manuscript. E.g., p4,l23-26 motivates meteorological/climatological measurements, p25,4-5 motivates soil measurements and should be placed in the introduction This was the main motivation for the project (and the reason the project was funded) so we believe it to be relevant context. The wording makes it clear that the paper focuses on the data, this section of the intro is providing the context for why the data was collected. We added a sentence about data scarcity to bolster the motivations.
-p1, l6: since only the years 1981-83 were investigated, please do not write 1980s here but refer specifically to the years 1981-83 DONE -p2, l5: state precise number instead of "more than 120 glaciers" Changed to "more than 100", see notes to reviewer 1 -p3, l5: please add reference here -p3, l10-p4,l3: detailed description of surge history seems unnecessary here, please shorten We focus on mass balance, which can be affected by surging (and vice versa). The latest surge of West Fork occurred shortly after the 1981-83 mass balance measurements to which we are comparing our data. While the effect of a surge on the mass balance may not be quantifiable given the data that is available, the surge still needs to be mentioned.
-p6, l4: what does the number 3 in brackets mean? Added " Figure" to the 3. p13, l8: reference to figure 4: is this right or do you mean figure 7? Good catch, thank you -p13, l11: rather say "very close to 1.0" Point taken that it is not exactly 1.0000000. But with two significant digits, I think it is appropriate to report it as 1.0 rather than "close to 1.0" implying it could have been 0.99.
-p13, l12/13: "The lack of a consistent pattern in these comparisons prevented us from adjusting the HOBO temperature data to match the Campbell data." This is confusing since you mention an average offset the sentence before. Can you clarify this? Added references to figures to clarify: "The lack of a consistent pattern in these comparisons (Figure 7) prevented us from adjusting the HOBO temperature data to match the Campbell data. The high correlation and low temperature offsets ( Figure  8) among sensors gave us confidence that using HOBO stations to assess temperature patterns across the basin was valid." -p.14, l1/2: What is your confidence that no double tips are missed or that normal tips are identified as double tips? For a complete discussion, please see Wolken et at 2015. Reference added here.
-p15, figure 7: how do you explain very high RH offset of some HOBO-sensor at higher RH, especially in 2013? Simply different sensor performance, as mentioned in the text. It is notoriously difficult to accurately measure humidity at cold temperatures (e.g. -10 C).
-p14, l11: "Precipitation amounts did not correlate significantly with elevation, slope, aspect, or location." How do you define "location"? What drives the variations in precipitation amounts? Location can be defined with lat/long coordinates, or by logical grouping in mountain ranges. We chose not to speculate on what drives variation, but we can rule out the factors mentioned.
-p14, l15-17: katabatic wind flow: Can you back this with references or add a more thorough analysis based on your data, e.g. wind direction analyses? Or is this just an assumption you make? Added a paragraph on wind direction to the "Meteorological data" section.

Added to caption: Each colored line represents a HOBO sensor, blue dots are for the reference Rotronic sensor, and red dots are for a Campbell 107 Temperature Probe.
-p16, figure 9: what is the purpose of figure 9? There is no in-depth analysis provided in the text and patterns are trivial (yearly temperature cycle, lower temperatures with higher elevation). In addition, it figure again suffers from relatively low resolution. Please either remove figure or provide more detailed analysis We believe the patterns are not trivial, but chose to spend more space in the paper describing mass balance data. Future papers may delve deeper into the spatial and temporal variations introduced here. We also feel it is best to present this "raw" data before looking at it in terms of lapse rates (figures 10 and 11). Fixed resolution issue.
-p16, section 3.3: this section does not provide a thorough analysis and is not useful for the manuscript since it is not based on your data. Either please remove or transform; rather than a trend analysis, the section could provide an assessment whether the years 2013-2014 were exceptional (in terms of temperature) or normal. We later reference this section when discussing mass balance differences between the 1980s and recent periods. -p19, l1/2: Since you used only point measurements, it is also possible that the stake measurements do not fully represent the area that was covered by satellite. In addition, you mentioned earlier that most of the stakes were placed on the centreline of the glacier, which is typically higher than the margins (which are included in the satellite estimation?), potentially leading to higher estimation of the equilibrium line altitude Reworded -p19, l7/8: "Glacier-wide mass balance estimates were then calculated by summing the distributed mass balance over the whole glacier." Did you use hypsometry of each individual glacier? Or did you use hypsometry of the entire glacier area for the calculation of the individual glacier mass balances? If so, the numbers you get probably have very high uncertainties. Please clarify. Changed "based on the glacier hypsometry " to "based on the individual glacier's hypsometry" -p20, l2: "East Fork Glacier had a similar mass balance as Maclaren Glacier." Why do you stress this here? Seems misplaced, please delete Done -p20, l17: "To robustly validate model simulations of snow accumulation…"; please move to introduction, since this provides a motivation for your measurements and you are presenting results in this section Removed reference to modeling as suggested by reviewer 1.
-p21, l4: "A notable south-north decrease in total SWE and accumulation gradient indicates a strong orographic influence." Please remove: sentence is redundant since you mention elevation dependence the sentence before. In addition, this is not always necessarily a north-south gradient Done -p22, figure 14 caption: "In early August 2013, the sensor's mounting pole began to tip over and give bad readings. On 1-2 September 2014, 18 cm of snow accumulation was recorded, consistent with observations during a site visit. The sensor pinged off falling snow, so some points in that window are labeled as bad data." please remove or move to text I think it is valid to leave this short description of the bad data in the figure caption, so the viewer doesn't have to search the text for the description of the bad data.
-p22, l16: "data became noisier as the surface transitioned from snow to ice." can this be seen in figure 14? Yes, when zoomed in.
-p22, l21/22: "This leads to an average melt rate of 0.016 m w.e. d−1 for the summer of 2013 and 0.012 m w.e. d−1 for 2014." This information seems a bit isolated from the previous, interesting mass balance investigations. Can you provide a comparison here? How does this compare to nearby ablation stakes summer mass balances?

The previous paragraph mentions this comparison, and it is plotted in figure 5 (added reference here). "The net depth of snow lost at the sonic distance sensor site was 1.97 m (17 April -30 June), which is comparable to the 2.15 m snow depth measured in a snow pit about 6 m away on 14 April 2013."
-p23, l7: "Snow water equivalent (SWE)": abbreviation has been initialized before Deleted "SWE" -p23, l14: "…generally showed a strong elevation dependence." dependence of what type? Maybe just write "increased with elevation" Changed -p23, l16/17: "At the Lower Windy Cr. site, about 40 mm of SWE (25%) was lost due to melt of the end-of-winter snowpack between 9 April and 22 April 2014 (Table 6)" Why is this stressed here? Please remove

Added context: "Lower Windy Creek was the only site where we collected SWE data more than once in a year. About 40 mm of SWE (25\%) was lost due to melt of the end-of-winter snowpack between 9 April and 22 April 2014 (Table \ref{tab:Snow})."
-p24, section 5.2: This section does not add any value to the paper but is very distracting; please remove Readers who are familiar with the prior work or the field site will be interested in this information.
-p25, l1: "characteristics we observed"; please specify these characteristics so the agreement between soil pits and STATSGO becomes clearer Done: "our observations of soil texture, organic content, and drainage characteristics" -p25, l4/5: "Understanding the distribution of permafrost and seasonally-frozen ground across the basin is important for modeling of water moving across the landscape"; again, this provides a motivation for your measurements and you are presenting results in this section, so please move to introduction -p27, l28/29: "Summer air temperatures in 2012-2014 were 1.1•C warmer than 1981-1983. Annual temperatures were 0.5°C warmer in the recent period." Why do you add this here? This is not based on your own measurements and thus not a significant outcome. Please consider removing.

It is significant because it helps explain the glacial mass balance changes we observe. Since we do not do temperature index modeling in this paper, we do not directly attribute the mass balance change to the temperature change, though we assume (hope!) the reader will see the connection.
-p28, l2: since only the years 2012-14 vs. 1981-83 were investigated, please do not generally say 2010s vs. 1980s since you have no information on the other years Done Technical Corrections: -p8, l12: move "On-Ice" to the end of this sentence Done -p6, l33: please remove "instead of every minute" Done -p16, l4: "most distinct" instead of "least complicated" Done -p18, l2: "refers to the period October…" instead of "refers to October…" Done -p20, l11/12: "Therefore, lower annual balance in the latter period (-1.72±0.87 m w.e.) compared to the former period (0.04±0.25 m w.e.) were driven by the more negative summer balances." should be "balances… were driven…" or "balance… was driven" Done -p26, figure 17 caption: "soil" instead of "soils"; remove "," after data Done

Introduction
Climate change is projected to have significant impacts on future water resources. In snow-and glacier dominated :::::::::::::: glacier-dominated catchments the response is strongly affected by changes in snow and glacier storage Huss and Hock, 2018).
More than 120 ::: 100 glaciers flow down the southern flanks of the central Alaska Range into the three forks of the Upper Susitna River (Figure 1). The glaciers provide a significant portion of the total runoff within the Upper Susitna drainage. It is well documented that glaciers across Alaska are currently retreating (Gardner et al., 2013;Luthcke et al., 2013). Changes to the 10 timing and amount of runoff due to continued melting of glaciers have been projected to occur worldwide Huss and Hock, 2018). Therefore, it is important to understand how changes to the Upper Susitna basin glaciers and river flow could affect dam operations and environmental resources.
This paper describes the data collected during the 2012-2014 field campaign detailing the instrumentation, method of deployment, and results for each set of data. Observations included meteorological variables, glacier mass balance, snow depth 15 and density, and soil type and temperature. Where possible we also compare the data with the results from the 1980s field campaign.

Study area
The watershed above the proposed Susitna-Watana dam (62.822523 • N, 148.538986 • W; henceforth referred to as the Upper Susitna basin) covers an area of 13,289 km 2 with elevation spanning from 450 to 4,200 m above sea level (a.s.l., Figure 1).

20
About 4% of the basin is glacierized. The total glacier area is 678.4 km 2 according to the Randolph Glacier Inventory version 6.0 (Pfeffer et al., 2014;Kienholz et al., 2015), which is based on satellite imagery from 3 July 2009. Modern glaciers are well within the limit of the Late Wisconsinan glacial advance (20)(21)(22)(23)(24)(25), when this part of the Alaska Range hosted the northern extent of the Cordilleran Ice Sheet (Kauman and Manley, 2004).
Almost all of the basin's glacier area is found in the Alaska Range whose highest ridges and peaks form the basin's northern 25 boundary. This area is characterized by high relief (Figure 1). Most glaciers (in total 127) in the study area are located in the Alaska Range, but a few small glaciers exist in the Talkeetna Mountains which form the southwest boundary of the basin.
ice density of 900 kg m −3 , this represents 123 Gt of ice. Some of the larger glacier termini reach elevations between 800 and 900 m a.s.l.
The nine glaciers in the Talkeetna Mountains draining to the Susitna river have a combined area of 8.9 km 2 . The largest glacier, located at the head of the Black River, is 7.3 km 2 . The total glacier volume is less than 0.6 km 3 (0.5 Gt).
The station records all the variables listed in Table 1 except for snow and ice temperatures and snow surface elevation changes.

25
The station was installed on 16 July 2012 and continues to operate under DGGS stewardship ::: the ::::::::: stewardship ::: of ::::: Alaska :::::::: Division  On-Ice station on top of the instrument arm at 2 m above the surface and unshielded (Figure 3). Gauges were not heated and hence gave most accurate results when precipitation was liquid rather than solid.
*** Only at the On-Ice station

Meteorological data
The daily meteorological data for both the On-Ice and Off-Ice station are shown in Figure 4, and seasonally averaged correlation coefficients between the two station's daily data are given in Table 2. Air temperatures at each station were significantly more variable in winter than in summer due to frequent winter weather systems. Both station's daily temperatures correlated well in all seasons except summer (June-August), when temperatures at the On-Ice station exhibited considerably less day-to-day 5 variability than the Off-Ice station. In addition, albeit 118 m lower in elevation, the On-Ice station's temperatures were also more than 4 • C lower than at ::::: cooler :::: than the Off-Ice station. These differences are attributed to the fixed glacier surface temperature of 0 • C during the extended periods of glacier melting during summer. The cold glacier surface cooled the air above it. Though differences were not as pronounced, lower air temperatures at the On-Ice station were also observed during other seasons. This is due, in part, to the high albedo at the On-Ice station.

10
Relative humidity at the On-Ice station was typically higher than at the Off-Ice station, consistent with lower air temperatures On-Ice and greater availability of moisture for evaporation ::: over ::: the :::::: glacier. The On-Ice station also displayed less day-to-day Daily mean wind speeds were typically higher at the On-Ice station than at the Off-Ice station which can be attributed to the relatively smooth glacier surface, longer fetch, and summer-time katabatic wind. Highest wind speeds occurred during winter with maximum wind gusts (3-second averages) up to 30.5 m s −1 (11 December 2013).
Total precipitation at the Off-Ice station was slightly higher (9%) than at the On-Ice station during the periods both stations were functional, however, direct comparison is difficult since the instrumental set-up was different with no wind shield and installation higher above the surface at the On-Ice station.
Incoming solar radiation displayed pronounced seasonal variability consistent with the site's latitude close to the polar circle.
Daily mean values varied between just 16 W m −2 in winter and approximately 400 W m −2 in summer. Cloudy days in summer 5 are clearly discernible due to their lower shortwave radiation compared to neighboring sunny days. High correlation (r=0.97) between both daily time series indicates relatively homogeneous cloud conditions at those sites. A portion of the difference between the two stations may be due to topographic shading.
In In both summers (2013 and 2014), outgoing longwave radiation increased through the spring to just below 320 W m −2 in mid-June and then plateaued through the end of August. Blackbody radiation from an object at 0 • C would be expected to be 316 W m −2 , indicating that the effective radiative temperature of the ice surface and air between the surface and the sensor 20 was just above 0 • C, and the surface was melting uninterruptedly for extended periods in summer.

Snow and ice temperatures
The thermistor string we deployed in a hot-water-drilled borehole needed time to equilibrate to its surroundings after installation. We consider the thermistor ice temperature measurements to be reliable starting on 25 April 2013, about 8 days after installation. By that time, the 5 m deep thermistor temperatures were within 0.02 • C of the trend they held for the subsequent 25 4 weeks.
Temperatures within the upper 10 meters of the glacier surface ranged from approximately -10 • C in the upper layers of snow pack in early May 2013 to close to the melting point of 0 • C at 10 m below the ice surface ( Figure 5).
:: c). : As air temperatures rose, the subsurface temperatures of the upper layers increased but with a time lag that increased with depth (Figures 5 and 6). When the air temperature rose above 0 • C, surface melt began to occur. As meltwater or rain 30 percolated into the snowpack it refroze, causing abrupt temperature increases of the uppermost thermistors (e.g. 10 and 25 May 2013, Figure 5). The glacier surface lowered with respect to the subsurface sensors by roughly 5 m between late April and early September 2013 as the melt season progressed (Figure 5b). The three snow sensors and uppermost 6 ice sensors were exposed to the air sequentially during the summer melt season. On 24-25 June, the ice temperature sensors at 0.1, 0.5,   Figure 5. Hourly ice and snow temperature (T) measurements from thermistors installed at depth in the snow and ice adjacent to the On-Ice weather station :: in :::: 2013 (c). Panel (a) shows air temperature, (b) shows cumulative surface lowering measured by the sonic ranger and ablation stakes. Initial thermistor depths are listed in the legend given in depth below the initial snow surface for the three snow thermistors and depth below the snow-ice surface interface for the 12 ice thermistors. The snow depth at installation on 19 April 2013 was 2.10 m. Note that the instrumentation depth becomes shallower as the glacier surface ablates during the melt season. The onset of large diurnal temperature fluctuations above 0 • C indicates that the thermistors have melted out and are affected by solar radiation. After a sensor exceeds 0 • C, we do not plot the remaining data. 1, and 1.5 m below the snow-ice interface all experienced rapid warming to about 0 • C. The 1.5 m sensor then cooled back down to -0.85 • C. We interpret this to be another meltwater event but this time in the :::: upper :::::: layers :: of :::::: glacier ice rather than the snowpack. It is difficult to know whether the event was representative of conditions in the glacier (i.e. water moving through cracks and along grain boundaries in the ice) or simply conditions along the thermistor cable. By 27 June, all 2.1 m of snow at the site had melted and the ice surface was exposed, as measured by the SR50 and confirmed with the time lapse imagery. Height relative to initial snow-ice interface (m)  After 27 June, sensors at 2, 2.5, and 3 m depth exhibited diurnal temperature wiggles with an amplitude of up to 0.2 • C. By mid-July the diurnal wiggles appeared in the sensors at 5, 6, and 7 m depth too. Some of the ice temperature sensors at depths 5 m recorded temperatures greater than 0 • C starting in late August 2013. These measurements are likely errors, perhaps due to faulty voltage sensor outputs. The data-logger continued to record reasonable results for other variables.

Instrumentation
To supplement the multi-variable weather stations described above and constrain the spatial patterns of temperature and precipitation within the basin, we installed 26 simple weather stations across the basin both on and off the glaciers (Figure 1; Table   A1). The 14 stations on or very near the glaciers (EF1, EF2, EF3, Mac1, Mac2, Mac3, Repeater HOBO, NWTrib1, Off-Ice HOBO, SU1, SU3, WF1, WFTranB, WF5; letters refer to glacier names) measured air temperature and relative humidity at a 10 nominal height of 1.75 m above the glacier surface. We refer to these stations collectively as the "glacier" weather stations. The sensor mounts for the glacier stations were designed to maintain approximately the same sensor height relative to the glacier surface throughout the ablation season. This was accomplished by allowing the mount to slide down the ablation stake as the glacier surface melted (Figure 3). The other 12 simple weather stations are referred to as the "tundra" stations. The typical tundra station measured temperature, relative humidity, rainfall, and soil temperature at 10 cm and 1 m depths. Simple weather 15 station instruments are listed in Table 3.
Temperature and relative humidity were recorded every 10, 15, 30 minutes (most stations), or 60 minutes, depending on the station. Each station's data was then averaged in post-processing to hourly and daily values. Hourly and daily precipitation sums were calculated for each tip of the rain gauge tipping bucket.
The temperature offset of hourly-mean HOBO sensors relative to an arbitrarily-chosen reference HOBO station was typically within ±0.1 • C and rarely beyond ±0.3 • C. Over both periods, the temperature offset had a mean of 0.02 • C and a standard 10 deviation of 0.07 • C (Table 4).
Compared to the Campbell/Rotronic sensor, HOBO temperatures were lower by 0.2-0.3 • C on average • C (Table 4) but differences exhibited a diurnal cycle with temperatures more than 1 • C lower during mid-day in many cases ( Figure 4 : 7). Fiveminute data showed that the response time of the HOBO sensors was slower than the Campbell sensors, but this can not explain the differences in hourly or daily means (Figure 8). The correlation coefficient between hourly values of temperature from the 15 Rotronic sensor and HOBO sensors was 1.0.
The lack of a consistent pattern in these comparisons :::::: (Figure :: 7) prevented us from adjusting the HOBO temperature data to match the Campbell data. The high correlation and low temperature offsets :::::: (Figure :: 8) : among sensors gave us confidence that using HOBO stations to assess temperature patterns across the basin was valid.
Measured relative humidity ranged from approximately 25% to 90% during the April 2013 calibration period and about 20 20% to 90% in April 2014 (Figure 7). HOBO sensors generally recorded higher relative humidity compared to the Rotronic sensor (Table 4). The offset in temperature (HOBO was colder than Campbell) explains part of that difference, though absolute humidity calculations show that the HOBO sensors are registering higher total moisture content for both the cold (2013) calibration period and the warmer one (2014). Relative humidity values from the HOBO sensors were well-correlated (r=0.99) For each year the differences in hourly means between each HOBO station and a reference station were calculated; then the differences from all stations were concatenated before calculating the mean, standard deviation, range, and skewness of the distribution. Two reference stations were used: a Rotronic HygroClip2 Temperature/RH Probe measured by a Campbell datalogger and an arbitrarily chosen HOBO station (data in parenthesis).
We looked for similar double tips in the On-Ice and Off-Ice station data, but did not find any. The HOBO and Campbell 10 sensors use the same internal electronics to detect tips, so we surmise that the Campbell logger is filtering out any double tips before recording the data.

Spatio-temporal variability of air temperature, humidity, and precipitation
Time series of daily air temperature, humidity and precipitation of all stations (On-Ice, Off-Ice : , and tundra stations) are shown in Figure 9. while the timing of events was generally consistent among the stations (Figure 9). Precipitation amounts did not correlate significantly with elevation, slope, aspect, or location.
Given the difficulty of calculating monthly lapse rates due to missing data (Figure 10), we also calculated weekly average 5 temperature and plotted it to illustrate the changing lapse rate with the seasons across the basin (Figure 11). The summer glacier/tundra pattern transitioned back to a single lapse rate when the air temperature fell below 0 m a.s.l. was buried by snow for part of each winter and therefore stands out as a warm outlier during those times. Springtime lapse rates are the least complicated :::: most :::::: distinct : -most stations fall along the best fit gradient.

Glaciological data
Glacier mass changes were determined from in-situ point observations in spring and fall of each year, snow radar measurements in spring and continuous measurements of relative surface elevation change at the On-Ice Weather station.

Radar-derived accumulation in glacierized terrain
Unlike ablation, which tends to be spatially coherent and well correlated with elevation, snow accumulation typically shows 5 pronounced small-scale variability, making it difficult to accurately measure and model (Sold et al., 2013). This is especially true in complex terrain, where topography and meteorological processes vary over short distances (McGrath et al., 2015).
There is good correspondence between the radar measurements and the traditional method ( Figure 13). A few points at high 10 elevation where the discrepancies are largest might be explained by a misinterpretation of the location of the previous summer surface, either by manual measurement (probe) of a shallow ice layer, or by selection of a deeper firn layer in the radar data.

Continuous point mass balance measurements
Adjacent to the West Fork Glacier weather station (1398 m a.s.l.) we installed an acoustic distance sensor and a Wingscapes time lapse camera to measure snow accumulation and melt with high temporal resolution. The distance sensor was fixed 15 vertically by mounting it on a pole drilled a few meters into the glacier, allowing the distance measured from the sensor to the surface to be directly related to melt or accumulation. The time lapse camera was mounted to the On-Ice weather station and  Figure 15). There was very little summer snowfall in 2013. There is ::: was : a distinct increase in measurement noise after the transition from snow to ice. This is likely due to the roughness of the ice surface ::: was ::::: likely :::::: caused ::::: when ::: the ::::::: acoustic :::::: signal ::::::: bounced ::: off ::::::: different :::::::: elements :: of ::: the ::::: rough ::: ice :::::: surface :: in ::::::::: successive :::::::::::: measurements. The height change rate while the surface In 2014, the distance sensor gave good readings from 26 April to 1 September when the sensor was removed after a two-day snowstorm. The 2014 measured net surface lowering from the last significant spring snowfall (28 April) to the first fall snowfall (1 September) was 3.8 m; summer snowstorms added 0.86 m of snow to the glacier which also melted away, for a total summer Decreasing distance is due to snow accumulation. In early August 2013, the sensor's mounting pole began to tip over and give bad readings.
On 1-2 September 2014, 18 cm of snow accumulation was recorded, consistent with observations during a site visit. The sensor pinged off falling snow, so some points in that window are labeled as bad data. September). Although the rate of surface lowering is larger for the snow surface than the ice surface, the rate of mass change is lower due to lower density.
The field measurements showed variations in SWE according to elevation, region, and vegetation type but with the small number of sites, definitive statistics were not feasible. The SWE measurements illustrated a general increase with elevation in 2014, which became more significant in late April. Basin-wide SWE data distinguished three major regions (Maclaren, Clearwater and Talkeetna), where the Maclaren sites represented the highest SWE and the Talkeetna region the lowest. Within each region, the SWE data generally showed a strong elevation dependence :::::: increase ::::: with :::::::: elevation. Among the two main vegetation types, shrubs presented larger SWE than the spruce locations (Table 6).

Soil pit characterization
At each of 9 tundra sites, we dug a soil pit from the surface down to the top of the mineral soil between 1-3 October 2013 (see Figures 1, 16). We recorded the type of vegetation growing in the soil, visual characteristics of each soil horizon, as well as our estimation of the soil texture (Table A2). Two sites were too rocky to dig a pit (Two Plate Creek and Valdez Creek).
Broad patterns that we observed in the data include: no permafrost in the upper layers of soil at the sites we sampled although it may persist in deeper layers. The annual range of temperature at the shallow soil sensors is less than the air temperature  April the shallow soil temperature sensor recorded a mean temperature of -1.9 • C and a diurnal cycle variation of 2 • C while the air temperature diurnal amplitude was about 8 • C. As the air temperature warmed over the week of 27 April, the soil warmed 5 to the freezing point and held steady until most of the soil had thawed. On 10 May, the shallow soil temperature exceeded 0 • C for the first time that year and thereafter resumed a diurnal cycle. The 1 m deep soils warmed from -6.3 • C to -1.7 • C over this interval.

Conclusions
Comprehensive observations of meteorology, snow cover, glacier mass change, and soil properties are important for assessing basin-wide changes and providing input to hydrological modeling. In this study we focused on the Upper Susitna watershedthe source area that would feed a proposed hydroelectric dam. Our measurements reoccupied many of the same sites used by an initial study of the region 30 years ago. The 1980's measurements in combination with those presented here provide a baseline for future studies in the area.
Summer air temperatures in 2012-2014 were 1.1 • C warmer than 1981-1983. Annual temperatures were 0.5 • C warmer in the recent period. We found lapse rates to be significantly lower over glacierized surfaces in summer than over non-glaciated areas.
Our meteorological stations filled a large gap in observations (spatially and elevation). Through correlations with long-running 5 NOAA sites, we can better estimate past conditions within the basin.
Glacier surface mass balance measurements showed that during the melt season the glaciers were losing mass more than 3 times more rapidly in the 2010s than in the 1980s :::::::: 2012-2014 :::: than ::: in ::::::::: 1981-1983. Winter snow accumulation measured by traditional methods closely matched measurements gathered from a helicopter-borne snow radar. Annual glacier-wide mass balance went from being close to 0 m w.e. (balanced) in the 1980s to losing more than 1.5 m w.e. yr −1 in recent years.
10 Snow depth in non-glacierized areas showed wide variability from site to site, reflecting complex deposition and redistribution patterns. Within local areas, higher elevations received more snow than lower elevations.
Our observations of soils in the basin generally match up with mapped soil descriptions. Soil temperature measurements revealed that none of the sites had permafrost in the upper 1 meter of soil. Most sites froze in the winter, though three sites remained at the freezing point despite air temperatures of -20 • C.

15
The data sets described here provide new data in an extremely data scarce region. The data are valuable as baseline to assess future changes and will aid calibration and validation of hydrological, glaciological and other environmental models. processed the raw radar data. All authors but WH and JZ contributed to the field work in the 2010s. WH provided information on the 1980s data. CA and AH contributed to some initial data analyses. RH, GW, AL, and JZ secured funding from the Alaska Energy Authority.

Competing interests
The authors declare that they have no conflict of interest.