Quantifying the impact of synoptic weather types, patterns, 1 and trends on energy fluxes of a marginal snowpack

. 10 Synoptic weather patterns and teleconnection relationships across a 39 year climatology are investigated for their 11 impact on energy fluxes driving ablation of a marginal snowpack in the Snowy Mountains, southeast Australia. 12 K-means clustering applied to ECMWF ERA-Interim data identified common synoptic types and patterns that 13 were then associated with in-situ snowpack energy flux measurements. The analysis showed that the largest 14 contribution of energy to the snowpack occurred immediately prior to the passage of cold fronts through increased 15 sensible heat flux as a result of warm air advection (WAA) ahead of the front. Indian Ocean Dipole and Southern 16 Oscillation Index phase combination had a strong relationship with energy flux, with eight of the ten highest 17 annual snowpack energy fluxes occurring during a negative IOD phase and positive SOI phase. Overall, seasonal 18 snowpack energy flux over the 39 year period had a decreasing trend that is likely due to a reduction in the number 19 of precipitation generating cold fronts and associated preceding WAA ahead of precipitation. This research is an 20 important step towards understanding changes in surface energy flux as a result of shifts to the global atmospheric 21 circulation as anthropogenic climate change continues. understanding of the energetics of Australia’s snowpack as they pertain to the 100 influences of shifting synoptic-scale circulations. Significant work has been conducted on identification of patterns and trends in Australian synoptic climatology 102 as it pertains to precipitation variability (Chubb Pook al., 2006; 2010; 2012, 2014; Theobald et al., 103 2016). However, impacts on surface energy fluxes as a result of synoptic types and changing climatological 104 conditions have not been explored as they have in other regions. The objective of this study is to identify the 105 synoptic weather types that contribute the highest amounts of energy fluxes to the Australian snowpack, and their 106 climatological trends. This is accomplished through: 1) the identification and classification of common synoptic 107 types during periods of homogeneous snow cover, 2) attribution of snowpack energy flux characteristics to each 108 synoptic type, 3) construction of energy balance patterns as they pertain to common synoptic 109 patterns/progressions, and 4) investigation of relationships between trends in snowpack energy flux and 110 teleconnections. at the Pipers Creek tower underwent a series of measurement corrections prior 218 to analysis. Coordinate rotation based on the methods of Wilczak et al. (2001) was applied on the 10 Hz EC data to remove levelling errors in sonic anemometer mounting when calculating fluxes. In addition, frequency corrections were made to the EC data to account for sensor response delay, volume averaging, and the separation 221 distance of the sonic anemometer and gas analyser when calculating fluxes (Campbell Scientific, 2018b). Finally, 222 WPL air density corrections (Webb et al., 1980) were made by the software to account for vertical velocities that exist as a result of changing air mass density through fluxes of heat and water vapour.


Synoptic weather influences on snow and glacier processes 24
varied largely depending on the synoptic type and its meteorology. A common characteristic between these studies 48 and others in various regions is that they focused primarily on the surface meteorology for synoptic classifications 49 rather than multiple level analysis, which enables insight to the potential influence of mid and upper-level 50 atmospheric conditions on surfaceatmosphere energy exchanges. Regardless, no analysis at any level exists on 51 synoptic type influence on snowpack ablation within Australia. 52

Synoptic weather types and trends in the Australian Alps 53
Precipitation in the Australian Alps is crucial to agriculture, the generation of hydroelectric energy, and recreation 54 and was estimated to be worth $9. temperatures above 1.5°C as measured by the SI-111 infrared radiometer that did not correspond to rain-on-snow 157 events and periods with albedo measurements less than 0.40 (Robock, 1980) were considered not to have 158 heterogeneous snow cover and were eliminated. Days within the ERA-Interim data that matched snow cover days were extracted and analysed using the k-means 178 clustering algorithm developed by Theobald et al. (2015). The algorithm was tested for 1-20 clusters and an elbow 179 plot of the cluster distances was used to identify the optimum number of clusters (Theobald et al., 2015), which 180 was seven. The identification of an elbow in the plot (Figure 3) at seven clusters indicates a reduction to the benefit 181 of adding additional clusters as the sum of distances for additional clusters fails to yield significant reductions 182 beyond that point (Wilks, 2011). 183 Clustering of the synoptic conditions for each day was verified through manual analysis of MSLP and 500 hPa 184 charts from the Australian Bureau of Meteorology (BOM) (Bureau of Meteorology, 2018). Cloud cover for each 185 type was investigated and verified through the use of visible band Himawari-8 satellite data 186 (https://www.ncdc.noaa.gov/gibbs/) at 03:00 UTC (13:00 local time) with one of three categories assigned to each 187 day studied; 1) no cloud cover, 2) partial cloud cover, or 3) complete cloud cover. Cloud cover was investigated 188 at midday to avoid misclassification due to short-lived clouds that appear over the area during the dawn and dusk 189 periods.
where the energy available for snow melt ( ) is equal to the sum of net radiation exchange ( * ), sensible ( ℎ ) 203 and latent ( ) heat flux, ground heat flux ( ), and the energy flux to the snowpack from liquid precipitation 204 ( ) (Male and Granger, 1981;McKay and Thurtell, 1978). 205 While net all-wave radiation exchange ( * ) is used for basic analysis of the snowpack energy balance, a 206 decomposition into its individual components is necessary to understand the role of short and longwave radiation 207 exchange in snowpack energetics (Bilish et al., 2018). Therefore, net radiation should be broken into its net flux 208 terms: 209 * = * + * (2) 210 that quantify the net shortwave ( * ) and net longwave ( * ) components. 211 The approach taken within this paper is to examine net radiative flux components individually, similar to the 212 methods used by Bilish et al. (2018), to be precise in the identification of synoptic-scale effects on snowpack 213 energy fluxes through differences in temperature, relative humidity, cloud cover. calculation and comparisons 214 of snowpack energy flux terms were performed using the terms in Eq. (1), but with the net radiation terms ( * and 215 Ltd, 2018). However, for climatological comparisons of synoptic types and snowpack energy flux the period of 269 most-likely snow cover, June through October, will be considered as the snowpack/snow cover season as times 270 outside of this period are not consistent in their snow cover properties. This season has been used to minimize 271 error introduced into the analysis through the incorporation of periods without snow cover. cover, but they became intermittent and fewer classifiable days were in each of the months. This led to fewer 293 periods of study at the beginning and end of the snow seasons when snowpack was variable, with more in the late 294 winter and early spring months when snow cover was more consistent. Mean surface, cloud, and energy flux 295 characteristics of synoptic types identified during the two seasons are presented in Table 2. southwesterly winds from T1 are the result of the high pressure centre being located to the northwest of the study 302 area. T2 has a predominantly zonal flow resulting from an elongated high to the north-northeast. T5 and T7 are 303 characterized by north-northwesterly flow from high pressure centres over the New South Wales 304 (NSW)/Queensland (QLD) coast and directly over the Snowy Mountains region, respectively. 305 T3 is characterized as having dominant northwest winds along a trough axis that is positioned over SEA with a 306 secondary coastal trough extending from southern NSW to the NSW/QLD border. T4 shows a transition from a 307 surface trough that has moved to the east of the study region to a high pressure system that is moving into the area 308 with winds from both features that converge over the Snowy Mountains region. The only synoptic type to have 309 dominant influence from a surface low was T6 that had weak south-southwesterly flow over the region from a 310 weak cut-off low to the east. Manual identification of cloud cover agreed with the mean RH characteristics of T4 and T6 with both types having 334 100% cloud cover between partial and complete cloud cover days (Table 3). T6 showed the highest RH values of 335 any type with values greater than 90% over the region at the 700 and 500 hPa levels. While not definitive, this 336 would suggest that T6 has deeper or more cloud layers than T4, which likely only has clouds at lower altitudes. 337 T2 and T7 had the lowest percentage (both 76%) of days with any cloud cover, which is confirmed by their low 338 RH values at 700 hPa (<20% & <30%) and 500 hPa (<30% & <40%), respectively. In addition, they also had the 339

Frequency and duration 357
The frequency of each synoptic type during the 2016 and 2017 snowpack seasons is shown in Table 4. T3 and T7 358 occurred most frequently with 26.74% (46 days) and 19.77% (34 days) respectively. The higher number of days 359 in T3 and T7 is reflected in the mean type duration that shows these types with the longest duration, which is 360 likely due to these synoptic types occurring in a more stagnant synoptic pattern over multiple days as seen in the 361 mean type duration (Table 2). 362 Transition probabilities for the 2016 and 2017 seasons were developed similar to those used by Kidson (2000)  363 that detail the likelihood of a synoptic type occurring on the following day given an initial type (Table 5). The 364 highest transition probabilities were identified for each type and a flowchart was developed based on the most 365 likely synoptic type progressions (Figure 9). If the highest transition probabilities were within < 0.05 of each 366 other, two paths were plotted. The flowchart shows what would be expected for a basic synoptic-scale circulation 367 at mid-latitudes; a trough propagating eastward into the Snowy Mountains region in T7, T5, and T3; either 368 continued eastward movement of the surface trough (T4) or the development of a weak cut-off low (T6); then 369 transitioning to dominant high pressure over the region again (T2, T1, or T7). 370

Energy flux characteristics of synoptic types 371
It is important to consider the effects of synoptic type frequency when determining primary sources of energy 372 fluxes over long periods as synoptic types that contribute the most to snowpack ablation may simply have a higher 373 rate of occurrence and lower daily energy flux values than other types. In order to obtain a more detailed 374 understanding of each type's energy flux, mean daily energy flux calculated for each type was determined to be a 375 better method of comparison. Therefore, both mean daily ( Figure 10 day -1 ) followed by T7 (1.11 MJ m -2 day -1 ) and T2 (0.97 MJ m -2 day -1 ). 396

Radiation flux 397
The largest contribution of radiative energy to the snowpack from all synoptic types was * which accounted for 398 57-81% of total positive flux. By comparison, * accounted for 66-90% of negative energy flux from the snowpack 399 with the highest amounts of loss belonging to the types with the lowest percentage of cloud cover (T1, T2, and 400 T3). Total radiation flux varied largely by synoptic type and was found to be positive in types T3, T6, & T5 and 401 negative for the rest of the types. The three types with positive net radiation had the highest incoming longwave 402 radiation flux values that allowed for greater cancellation of outgoing longwave values and allowed for incoming 403 shortwave radiation to dominate the net radiation flux. The largest loss in net radiation energy was exhibited by 404 T1 that was 3% higher than the next closest type (T2). The types with net radiation loss (T1,T2, T4, and T7) had 405 values that ranged from -0.66 MJ m -2 day -1 (T4) to -1.44 MJ m -2 day -1 (T1). However, T4 had dissimilar cloud and 406 relative humidity characteristics to T2 and T7, which had the two lowest cloud cover percentages and two of the 407 lowest RH values. T4 had 100% cloud cover and had an associated reduction in incoming shortwave radiation 408 that allowed the outgoing longwave radiation term to become more dominant than in T2 or T7 and, therefore, 409 gave it the highest net radiative energy loss of the three. when examined as a daily mean value, it does show a high degree of variation that was associated with T5 and 415 T3. This is due to several large rain events that occurred during 2016 (July 18; July 21-22; and August 31) and 416 one during 2017 (August 15). Despite relatively low energy flux contributions by rainfall, it is interesting to note 417 that the ten days with the highest rainfall fluxes (>0.05 MJ m -2 day -1 ) consisted of four T5 days, three T3 days, 418 two T7 days, and one T6 day showing a significant clustering of high precipitation days in types T5 and T3. 419

3.2.4
Total daily net energy flux 420 Overall, four synoptic types (T3, T5, T6, and T7) had positive mean daily net energy flux to the snowpack (Figure  421 12). Of these, T5 had the largest energy flux that was related to its relatively high temperatures that contributed 422 to the highest ℎ value of any synoptic type and increased solar radiation from less cloud cover. Contrary to the 423 reduction in cloud cover that aided T5 in having the highest total energy flux contributions, T6 had the highest 424 cloud cover and yet had the second highest energy flux to the snowpack that was primarily due to increased 425 incoming longwave radiation. Similarly, positive net radiation flux associated with T3 gave it a net positive daily 426 net energy flux. Positive net energy flux from T7 is a result of relatively low percentage of cloud cover and the 427 associated increase in ↓ as well as the second highest ℎ term of any type. 428 T1 and T4 showed the greatest negative mean daily net energy flux of all synoptic types (Figure 12), which could 429 be attributed to their low ℎ values as a result of the lowest measured temperatures of any synoptic type and to 430 having low ↓ terms. T2 also had a net negative mean daily energy flux but to a lesser extent than either T1 or 431 T4. Relative humidity values lower than any other type were the primary driver behind T2's negative net value as 432 it resulted in the highest longwave radiation loss from the snowpack through having the lowest cloud cover, as 433 well as loss. 434 T5 contributed the most energy to the snowpack during the two seasons despite T3 having nearly 96% more 435 occurrences in the same period. This was largely due to high ℎ values associated with strong WAA ahead of the 436 passage of cold fronts associated with the T5 synoptic type, which had the second largest overall energy 437 contribution to the snowpack. While mean daily energy flux contributions of T6 to snowpack energy flux are 438 103% higher than those of T3, the high number of occurrences associated with T3 made it the second highest 439 contributor of energy flux during the two seasons with T6 contributing the third highest amount of energy flux.  balance closure was conducted on periods when snow surface temperatures and ambient air temperature were less 453 than 0°C to limit the amount of energy being used in internal snowpack melt processes. Energy balance data was 454 divided into day and night periods following the methods of Helgason and Pomeroy (2012) that used an incoming 455 shortwave radiation threshold of > 200 Wm -2 for day periods and 0 Wm -2 for night periods. Daytime periods 456 showed a slightly higher mean energy balance closure ratio of 0.33 ± 0.55 (n = 761) than night periods that had a 457 mean closure ratio of 0.31 ± 0.51 (n = 2291). Inability to account for nearly constant internal snowpack melt 458 processes has likely resulted in the low closure ratio and high variability in calculated energy balance closure. 459

3.3
Climatological trends of snowpack energy flux 460

Frequency of occurrence and trends 461
Occurrence frequencies were determined for each synoptic type by classifying each day from June-October from 462 1979 to 2017. Statistics on seasonal averages of frequency for each type are displayed in Table 6

Atmospheric teleconnections 491
Investigation into teleconnection impacts on estimated seasonal snowpack energy flux showed relationships with 492 IOD and SOI phase, but little relationship with SAM phase (Figure 15 (Figure 14), which can be attributed to 504 dominance of positive IOD phase and negative SOI phase during these periods. 505 Overall, IOD phase was the dominant climatological control of estimated seasonal energy flux and accounted for 506 an average increase of 31% when changing from a positive to negative phase (Table 7). SOI phase had a similar 507 effect accounting for a mean positive energy flux increase of 28% during seasons that were positive. While SAM 508 phase can account for an increase of 12% energy flux by season when it is negative, its contribution is relatively 509 small compared to the other two terms. Negative IOD and positive SOI can account for a 48% increase in energy 510 flux to the snowpack when compared to years with opposite phases. The highest estimated energy flux to the 511 snowpack occurs when IOD is negative, SAM is negative, and SOI is positive, which results in up to 56% more 512 energy flux to the snowpack than other periods. While the highest difference in energy flux is seen in the correct 513 combination of IOD -, SAM -, and SOI + periods when compared to others, only 40% of the top ten seasons with 514 the largest energy flux had this characteristic. However, 80% (including three of the top five seasons) had a 515 negative IOD and positive SOI suggesting that strength of phase is also an important consideration. positive energy flux during these periods that include: an increase in temperatures due to WAA and the associated 555 increase in positive ℎ ; decrease in negative * due to an increase in cloud cover; a decrease in following 556 frontal passage and associated increase in RH; and progressively increasing as the trough approaches and 557 immediately after passage. 558 Synoptic types characterized by surface high pressure as their primary influence (T1, T2, T4, and T7) showed the 559 lowest contributions to snowpack energy flux. In T1, T2, and T7, net shortwave radiation terms ( * ) were positive 560 and varied by ~2-24% for these types, however, low RH and cloud cover allowed for highly negative * terms 561 that were not compensated by change in * . In contrast, T4 had higher cloud cover and increased RH that were 562 due to advection of moisture from the Tasman Sea. The higher RH in T4 and low mean air temperature (-2.06°C) 563 resulted in and ℎ terms of similar magnitudes, but opposite signs that nearly cancelled out. This resulted in a 564 * term that was of lesser magnitude than those of T1, T2, and T7, but still the dominant term in its energy 565