Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Numerical modelling of snow and ice thicknesses in Lake Vanajavesi, Finland

Numerical modelling of snow and ice thicknesses in Lake Vanajavesi, Finland SERIES A DYNAMIC ME TEOROLOGY AND OCEANOGRAPHY PUBLISHED BY THE INTERNATIONAL METEOROLOGICAL INSTITUTE IN STOCKHOLM Numerical modelling of snow and ice thicknesses in Lake Vanajavesi, Finland 1, 2 2 3, 1 1 1 By YU YANG *, MATTI LEPPARANTA ,BIN CHENG and ZHIJUN LI , State Key Laboratory of Coastal and Offshore Engineering, Dalian University of Technology, Dalian 116024, China; 2 3 Department of Physics, University of Helsinki, PO Box 48, FI-00014 Helsinki, Finland; Finnish Meteorological Institute, PO Box 503, FI-00101, Helsinki, Finland (Manuscript received 27 March 2011; in final form 7 January 2012) ABSTRACT Snow and ice thermodynamics was simulated applying a one-dimensional model for an individual ice season 20082009 and for the climatological normal period 19712000. Meteorological data were used as the model input. The novel model features were advanced treatment of superimposed ice and turbulent heat fluxes, coupling of snow and ice layers and snow modelled from precipitation. The simulated snow, snowice and ice thickness showed good agreement with observations for 20082009. Modelled ice climatology was also 1 1 reasonable, with 0.5 cm d growth in DecemberMarch and 2 cm d melting in April. Tuned heat flux from waterto ice was 0.5 W m . The diurnal weather cycle gave significant impact on ice thickness in spring. Ice climatology was highly sensitive to snow conditions. Surface temperature showed strong dependency on thickness of thin ice (B0.5 m), supporting the feasibility of thermal remote sensing and showing the importance of lake ice in numerical weather prediction. The lake ice season responded strongly to air temperature: a level increase by 1 or 58C decreased the mean length of the ice season by 13 or 78 d (from 152 d) and the thickness of ice by 6 or 22 cm (from 50 cm), respectively. Keywords: thermodynamic model, sensitivity tests, snow, ice, Lake Vanajavesi cuts off airlake exchange of oxygen and reduces the 1. Introduction production of dissolved oxygen by limiting the light pene- Freshwater lakes cover about 2% of the Earth’s land tration (Livingstone, 1993). Ice also prevents the exchange of surface. In Finland, there are ca. 190 000 lakes larger than momentum between the atmosphere and the lake water 500 m , accounting for10% of the area of the whole (Williams et al., 2004). Lakes affect the local weatherby country, and these lakes freeze every winter. Their ice modifying the air temperature, wind, humidity and precipi- sheets consist of congelation ice and snowice with snow tation in their surroundings (Ellis and Johnson, 2004; covernormally on top. In medium-size and small lakes, the Rouse et al., 2008a, 2008b), and ice coverprotects the heat ice coveris usually stable, while in large lakes mechanical content of lakes by its insulation capacity and by damping displacements may take place, creating piles of ice blocks or turbulent mixing in the water body. Thus, the presence ridges. An ice cover stabilises the thermal structure of a (orabsence) of ice coverhas an impact on both egional lake. The ice bottom is at the freezing point, and there is a climate and weatherevents in the winter spring season. weak heat flux from the water to the ice. In spring, ice Understanding the processes and interactions of lake ice sheets gain heat mainly from solar radiation, and as the ice and atmosphere is essential for numerical weather predic- melts, all impurities contained in the ice are released into tion (NWP) (Brown and Duguay, 2010). As the spatial the waterorto the air. resolution of NWP models becomes higher, presently Lake ice plays an important role in the hydrological, approaching one-kilometer scale, including lakes and lake biological, chemical and socio-economical regimes of boreal ice forforecasting and data assimilation in mesoscale, lakes (Leppa¨ ranta, 2009). For example, a compact ice cover NWP systems have gained more attention (e.g. Eerola et al., 2010; Mironov et al., 2010; Salgado et al., 2010). Analytical, semi-empirical lake ice thickness models were *Corresponding author. email: yangyang-0606@hotmail.com widely used until 1980s, and thereafter full numerical Tellus A 2012. # 2012 Y. Yang et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution-Noncommercial 3.0 1 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Tellus A 2012, 64, 17202, DOI: 10.3402/tellusa.v64i0.17202 (page numbernot forcitation purpose) r 2 Y. YANG ET AL. models have also been employed (e.g. Liston and Hall, 2. Data 1995; Duguay et al., 2003). The basic principle of these models is to solve the heat conduction equation forthe 2.1. Observations during the ice season 20082009 snow and ice layers in the vertical direction, as was first Lake Vanajavesi (61.138N, 24.278E) is located in south- done in sea ice thermodynamic modelling by Maykut and western Finland (Fig. 1). It is a large, shallow and Untersteiner (1971). The Canadian Lake Ice Model is a eutrophic lake. The area is 113 km , and the mean and snow and ice process model (Duguay et al., 2003) adapted maximum depths are 7 and 24 m, respectively. The lake has from a sea ice model of Flato and Brown (1996). a long eutrophication history with very poor water quality In most of the lake ice models snow thickness is in the 1960s and 1970s. The ice season lasts 46 months, on prescribed with a climatological growth rate or with a average from December to April, and the thickness of ice fixed snowfall scenario based on in situ data. In the boreal reaches its annual maximum value of 4560 cm in March. zone, however, the role of snow in the growth and melting The ice and snow thickness may show distinct spatial of lake ice is very important, and snow and ice form a variations due to snowdrift and heat flux from the lake coupled system. In analytical modelling, the influence of water. snow is usually considered by modifying the growth law In the winterof 2008 2009, a field programme was parameters (e.g. Leppa¨ ranta, 1993). Interactive snow and performed in Lake Vanajavesi. Hydrographical surveys snowice layers were included in the numerical quasi- were made regularly, an automatic ice station was set up steady ice thickness model of Leppa¨ ranta (1983). The first and optical investigations were performed (Jaatinen et al., detailed lake model study in Finland was presented by 2010; Lei et al., 2011). The average maximum snow and ice Lepparanta and Uusikivi (2002), based on a Baltic Sea thicknesses reached 15 and 41 cm, respectively, in March. model of Saloranta (2000), where snow metamorphosis and The freezing and breakup dates were 1 January and 27 snowice formation due to flooding were included (see also April, respectively. Shirasawa et al., 2005). In this study, a one-dimensional high-resolution thermo- dynamic snow and ice model (HIGHTSI) was applied for Lake Vanajavesi, located in southern Finland. This model contains congelation ice, snowice and snow layers with full heat conduction equation. Compared with earlier lake ice models, new features were superimposed ice formation, an advanced atmosphereice heat exchange treatment accounting forthe influence of stability of stratification to the turbulent fluxes and coupling of the snow and ice layers. Atmospheric forcing was derived from weather observations and climatology, which also drove the snow coverevolution. The simulation results were compared with measured ice and snow thickness. A case study was performed for the ice season 20082009, forced by daily weather observations. Ice climatology was examined forthe 30-year period 1971 2000; also the correlation between the observed monthly total precipitation and snow accumulation was investigated in order to understand the uncertainties of precipitation as model forcing for climatological simulation. A number of climate sensitivity simulations were carried out for the ice season. The objectives of the present work were to assess the applicability of the HIGHTSI model forlake snow and ice thermodynamics, to find out the most important factors affecting lake ice growth and melting and to evaluate the influence of climate variations on the lake ice season. Section 2 introduces data, the model is described in Fig. 1. Geographical location of Lake Vanajavesi (A); Section 3 and the results are presented and discussed surrounding observation sites Jokioinen meteorological in Section 4. Final conclusions follow in Section 5. observatory (B), Lake Kuivaja¨ rvi (C) and Lake Pa¨ a¨ ja¨ rvi (D). NUMERICAL MODELLING OF SNOW AND ICE THICKNESSES 3 Synoptic-scale weatherconditions and egional climate than in the winter2008 2009. The precipitation was over this area are represented by Jokioinen meteorological 232 mm, about the same that was observed in 20082009, observatory (60.88N, 23.58E; WMO station 02863), located but during the ice growth season, the average snow some 50 km southwest of Lake Vanajavesi (Fig. 1). The accumulation was 145 mm WEQ or58% more than in weatherforcing data forthe lake ice model consist of wind 20082009. speed (V ), air temperature (T ), relative humidity (Rh), The average freezing and ice breakup dates of Lake a a cloudiness (CN) and precipitation (Prec), collected at Vanajavesi were 30 November and 30 April, respectively, in 3-hour time intervals. For the present modelling work, we 19712000 (Korhonen, 2005). A large variation occurred in interpolated data to 1-hour intervals. The snow thickness these dates in individual years. The earliest freezing date on land is also available from Jokioinen. was 28 Octoberand the latest one was 28 January. Forthe The winter2008 2009 was mild (Fig. 2). The early winter ice breakup, the earliest and latest dates were recorded as was warm, with average air temperature of 0.68Cin 1 April and 27 May, respectively. Novemberand Decemberpreventing Lake Vanajavesi Due to the lack of snow and ice thickness data in Lake from freezing. From late December to the end of March, Vanajavesi, the observations of Lake Kuivajarvi (Korhonen, the air temperature was permanently below the freezing 2005), located about 60 km southwest of Lake Vanajavesi, point, on several occasions approaching 208C and were used as the validation data. It can be assumed that the thereafter increased gradually to above 08C. The ice season climatological conditions are the same in these two lakes. 20082009 can be divided into two phases: ice growth (JanuaryMarch) and melting (April). The corresponding 3. The lake ice thermodynamic model average air temperatures were 4.4 and 4.38C, respec- tively. The total precipitation from November to April was 3.1. Surface temperature and surface heat/mass 234 mm. During the ice growth season, however, the snow balance accumulation was merely 92 mm water equivalent (WEQ). The thermodynamic lake ice model HIGHTSI solves the surface temperature (T ) from a detailed surface heat sfc 2.2. Climatological for long-term simulations balance equation using an iterative method: Monthly mean meteorological observations from the Jokioinen observatory in the climatological normal period ð1  aÞQ  I þ Q  Q ðT Þþ Q ðT Þ s 0 d b sfc h sfc 19712000 (http://en.ilmatieteenlaitos.fi/normal-period- (1) þQ ðT Þþ F ðT Þ F ¼ 0 le sfc c sfc m 19712000) were used for the forcing to investigate ice climatology (Table 1, based on Drebs et al., 2002). The mean where a is the albedo, Q is the incoming solarradiation at air temperature was 2.88C in NovemberApril, 1.38C less s the surface, I is the portion of solar radiation penetrating below the surface layer and contributing to internal heating of the snow and ice, Q and Q are the incoming and d b outgoing longwave radiation, respectively, Q is the sensible heat flux, Q is the latent heat flux, F is the conductive heat le c flux coming from below the surface and F stands for melting. All fluxes are positive towards the surface. In HIGHTSI, I is explicitly parameterised based on the fact that penetrating solar radiation is strongly attenuated along the vertical depth immediately below the surface (Grenfell and Maykut, 1977). The term (1a)Q I thus s 0 represents the part of short-wave radiation contributing to the surface heat balance. It depends on the surface layer (i.e. the first layer of snow or ice in model) thickness (Launiainen and Cheng, 1998; Cheng, 2002; Cheng et al., 2008). Equation (1) serves as the upper boundary condition of the HIGHTSI model. If the surface temperature is at the melting point (T ), excessive heat (F ) is used formelting: f m dh/dtF /rL , where h is the thickness of snow orice, m f Fig. 2. Time series of wind speed (a), air temperature (b), r is the density of snow orice and L is the latent heat of relative humidity (c), cloudiness (d) and precipitation (e). The data f were initially observed at 3-hour time interval. fusion. r 4 Y. YANG ET AL. Table 1. The monthly mean meteorological data for 19712000 (Drebs et al., 2002; h? is snow thickness on the 15th of each month; Ph is s s the portion of precipitation contributing to the snow accumulation) 1 1 Month V (m s ) T (8C) CN Rh (%) Precipitation (mm month ) h? (cm) Ph (%) a a s s October3.8 4.6 0.76 88 59 0 November3.9 0.4 0.83 90 57 2 X December3.9 4.1 0.81 90 45 9 46 January 3.8 5.9 0.79 89 41 19 77 February 3.7 6.5 0.74 87 29 29 94 March 3.8 2.7 0.68 82 30 31 23 April 3.7 2.7 0.7 74 32 10 X May 3.7 9.5 0.62 64 35 X X The albedo is critical for snow and ice heat balance. Q ¼ q R C ðq  q ÞV (6) A parameterisation suitable for the Baltic Sea coastal snow le a l E a sfc a and land-fast ice is selected in this study (Pirazzini et al., where T T is the difference between air and surface a sfc 2006): temperature, q q is the difference between air and a sfc surface specific humidities, V is the wind speed, R the a l a ¼ 0:15 h B 0:1 cm enthalpy of sublimation of water, r is the density of air, c h ða a Þ a a s s i a ¼ min a ; a þ h ] 0:1 cm and h 5 10 cm (2) s i i s 0:1 is the specific heat of airand C and C are the bulk : H E a ¼ a h ] 0:1 cm and h > 10 cm s i s transfer coefficients estimated using the MoninObukhov similarity theory (Launiainen and Cheng, 1995). The surface conductive heat flux was determined by: where a and a are snow and ice albedo, a 0.75, s i s 1:5 a ¼ minðÞ 0:55; 0:85  h þ 0:15 and h and h are the s i i i @T snow and ice thickness, respectively. F ¼k (7) @z sfc Incoming solarorlongwave adiation can be either parameterised for estimation from cloudiness, temperature where T is temperature, z is the depth below the surface and and humidity or directly obtained from an NWP model. k is the thermal conductivity in the surface layer. Forthe solarradiation, we applied the clear sky equation of Shine (1984) with cloud effect of Bennett (1982): 3.2. Snow and ice evolution S cos Z Q ¼ Snow coverevolution was modelled taking into account 1:2 cos Z þðcos Z þ 1:0Þe  10 þ 0:0455 several factors: precipitation, packing of snow and melt ð1  0:52 CNÞ (3) freeze metamorphosis. Packing depends on precipitation, where S is the solarconstant, Z is the solarzenith angle, e is air temperature and wind speed, which are obtained from the water vapour pressure (mbar) and CN is the cloudiness observations or from an NWP model. A temperature ranging between 0 and 1. The net longwave radiative flux criterion was set (T B0.58C) to decide whetherprecipita- (Q Q ) was calculated by: d b tion is solid orliquid. The impact of wind on snow pffiffiffi accumulation is regarded as a model uncertainty. In the Q  Q ¼ r T 0:746 þ 0:066  e d b s a climatological sense, however, one may have a rough idea (4) ð1 þ 0:26  CNÞ er T s sfc of the influence of wind on snow accumulation by comparing the measured precipitation and snow thickness where the first term on the right-hand side represents the (cf. Section 4.2). In addition to the surface melting, snow incoming longwave radiation (Efimova, 1961, with cloud and ice temperature may reach the melting point below the effect of Jacobs, 1978), the second term is outgoing surface because of the absorption of the penetrating solar longwave radiation, s is the StefanBoltzmann constant radiation. The subsurface melting is then calculated and o is the surface emissivity, which ranges from 0.96 to according to Cheng et al. (2003). 0.99 (Vihma, 1995). In boreal lakes, snow may contribute significantly to the The surface turbulent heat fluxes were parameterised on total ice thickness. Snow-load can cause a negative ice the basis of atmospheric stratification. The sensible heat Q freeboard and consequent formation of a slush layer by and latent heat Q were calculated by: le mixing of lake waterand snow. A slush layercan also be Q ¼ q c C ðT  T ÞV (5) h a a H a sfc a formed by percolation of snow meltwater or liquid r NUMERICAL MODELLING OF SNOW AND ICE THICKNESSES 5 precipitation. Freezing of slush into ice was calculated by where F is the heat flux from water to ice. The model the heat flux divergence: parameters for this study are summarised in Table 2. The freezing date in HIGHTSI is provided as the initial dh @T @T si s i q L ¼ k j k j (8) condition by a thin ice layer(0.02 m). If the weather si f s z¼hs i z¼hi dt @z @z conditions do not favour ice growth, HIGHTSI will resume the initial ice thickness. The first day when ice starts to where h is snowice thickness and r is the density of si si grow successively is defined as the freezing date. It is clear snowice. Slush formation from flooding was calculated that this is a crude approximation to obtain the freezing according to Lepparanta (1983). After its formation, date, but once the ice growth has started, the model catches snowice becomes an internal layer part of the ice sheet. up well with the nature. The density of snow is a prognostic variable, calculated according to Anderson (1976): 1 @q 4. Results and discussion 0:08ð273:15T Þ c q a 2 s ¼ c e w e (9) 1 s q @t 4.1. Seasonal study for the winter 20082009 where c and c are empirical constants and w is the mass 1 2 s The ice season 20082009 lasted 4 months. Model simula- of snow in the above layerin WEQ. In this study, we set 1 1 3 1 tions were started up in the beginning of January from the c 5m h and c 21 m kg . 1 2 initial snow and ice thicknesses of 0.5 and 2 cm, respec- The snow and ice temperatures were solved from the heat tively. The simulated snow thickness agreed well with the conduction equation: observations (Fig. 3), showing that the model can well @Tðz; tÞ @ @Tðz; tÞ @qðz; tÞ reproduce the total snow accumulation and the snowmelt. qc ¼ k  (10) @t @z @z @z Snow thickness overland, measured in Jokioinen, is also shown in Fig. 3. A previous study by Ka¨ rka¨ s (2000) has where q(z, t) presents the solar radiation penetrating into indicated that the snow coveron land is on average 2.3 the snow and ice based on an exponential attenuation law: times thickerthan on lake ice in this egion, based on q (z, t)(1a )I exp(k z), where k is the extinction s s,i 0 s,i s,i monthly snow measurements from January to April during coefficient of snow orice. a 10-yearperiod (1971 1980) in Lake Pa¨ a¨ ja¨ rvi (61.078N, The integral interpolation method was applied to build 25.148E). In the present case, the ratio between the up the numerical scheme (Cheng, 2002). A high spatial observed mean snow depth in Jokioinen and the modelled resolution (e.g. 10 layers in the snow and 20 layers in the average snow depth over the Lake Vanajavesi ice was 2.1. ice) ensures that the response of the snow and ice The modelled ice thickness fitted well with the field temperature regime to the absorption of solar radiation measurements, and particularly the ice breakup date was near the surface is correctly resolved. In the case of snow on close to the observation. In the beginning of January, the top of the ice, the snowice interface temperature is ice grew rapidly in response to the very cold air tempera- calculated via a flux continuity equation: ture. As soon as the snow depth had reached about 10 cm, @T @T the speed of ice growth reduced significantly because of the s i k ¼ k (11) s i insulating effect of snow. Snow to snowice transformation @z @z also contributed a little to the total ice thickness in the late Heat and mass balances at the ice bottom served as the season. A 9-cm slush layerand a 1-cm snow ice layerwere lowerboundary condition: measured on 18 March (Lei et al., 2011), while the model dh @T prediction was a 5-cm slush layer and no snowice. This i i q L ¼k þF i f i w error is likely due to overestimation of snow density in the dt @z (12) bottom model. Later, on 7 April, a 2-cm slush layer and a 5-cm T ¼ T frozen slush layer were observed (Lei et al., 2011), while in Table 2. Model parameters found in the literature Parameter Value Source Extinction coefficient of lake ice (k ) 1.517 m Heron et al. (1994); Arst et al. (2006); Lei et al. (2011) Extinction coefficient of snow (k)620 m Patterson et al. (1988); Arst et al. (2006); Lei et al. (2011) Lake ice density (r ) 910 kg m Corresponds to 1% gas content Initial snow density (r ) 330 kg m Leppa¨ ranta and Kosloff (2000) Surface emissivity (o) 0.97 Vihma (1995) r 6 Y. YANG ET AL. Fig. 4. Time series of observed and simulated ice thickness based on varying heat flux from water: 0 W m (black dashed), 2 2 0.5 W m (black solid; reference experiment), 2 W m (grey dashed) and 5 W m (grey solid). Measurements of ice thickness are shown as in Fig. 3. Fig. 3. Time series of observed and modelled snow and ice thickness. The dark grey line and the asterisk are the observed snow thickness in Jokioinen and on Lake Vanajavesi, respectively. conduction is reduced and closer to the heat flux from the The black dashed and solid lines are modelled snow and ice water. Ice may then even melt from the bottom. The thickness (reference experiment), respectively. The circles are the examined variations of the heat flux from water can alter observed average ice thickness, and the spatial standard deviation the maximum ice thickness by93.4 cm, as compared to the is indicated by the vertical bar. results from the optimised reference simulation, where the heat flux from water to ice was 0.5 W m . This level was oursimulation these values had become 2.9 and 2.7 cm, quite low, of the same magnitude than the molecular respectively, by the end of March. On 14 April, a 3-cm conduction of heat. It is much less than what has been slush layerwas still observed but snow ice had disap- earlier obtained in Lake Paajarvi, which is deeper and also ¨ ¨ ¨ peared. The melt rate of ice was 1 cm d from 7 April to has more groundwater inflow in the mass balance. A large 14 April (Lei et al., 2011), both in the model and in the heat flux from the water would give less ice thickness and observations. The melting of ice started in the beginning of 2 earlier breakup date. For example, a 1 W m heat flux April, as soon as snow had gone, and it accelerated towards melts about 1 cm ice in a month, and forthe mean melt rate the breakup (Fig. 3). The melting period lasted 30 d, with of 1.3 cm d ice breakup would be about 1 d earlier. mean melt rate of 1.3 cm d . The parameterised solar radiation may be an under- The heat flux from lake water to the ice bottom has been estimate. The uncertainties are due to the effect of clouds as earlier examined in the nearby Lake Pa¨ a¨ ja¨ rvi by Shirasawa well as due to multiple scattering in the air (Shine, 1984; et al. (2006) and Jakkila et al. (2009). The magnitude was Wendlerand Eaton, 1990). Fig. 5 gives the observations found to be 310 W m based on field data. In freshwater and calculations forJokioinen in January April 2009. The lakes, before freezing the water column is well mixed down largest differences, 200 W m , appeared mostly in to 148C depending on the autumn wind conditions, lake MarchApril. Then, on average, underestimation of solar size and land area around. Thereafter, inverse winter radiation may account for 7 W m of the total net stratification is set up, and once ice is formed, further radiative flux. This amount of energy can offset 3 cm of the mixing is strongly limited. The water temperature structure ice thickness and lead to a 5-d shift in the ice breakup date. underthe ice is usually quite stable with a very small heat The surface temperature reflects the ice thickness during flux out from the water body. the cold season, because the ice thickness impacts the To examine the sensitivity of ice thickness to the heat strength of the insulation between the lake water body and flux from the water, model runs were carried out with the atmosphere. By applying the heat conduction law different fixed heat flux levels: 0, 0.5, 2 and 5 W m (Eq. 1), we can solve the ice thickness as a function of (Fig. 4). In the early winter, there was a very large the surface temperature or vice versa. The inverse proce- temperature gradient across the ice sheet, and the heat dure is used in thermal remote sensing for mapping the sea conduction through the ice and the consequent ice growth ice thickness (Rothrock et al., 1999; Leppa¨ ranta and rate were large. As soon as the ice thickness has reached Lewis, 2007; Ma¨ kynen et al., 2010; Wang et al., 2010). about 20 cm and there is also snow on the ice, the heat In weather forecasting, the surface temperature is needed as NUMERICAL MODELLING OF SNOW AND ICE THICKNESSES 7 Fig. 5. Calculated and observed incoming short-wave radiative flux underclear and cloudy sky conditions. The data coverthe winter2008 2009 from January to April. Goodness of fit for the whole ice season’s data was R 0.92 (n2880). a boundary condition for the atmosphere. Due to ice atmosphere coupling, NWP needs an ice model to obtain the surface temperature of lakes in winter. To examine the sensitivity of surface temperature on ice thickness, a number of seasonal model runs were per- formed with different initial ice thicknesses. The results are Fig. 6. Surface temperature versus ice thickness: for a cold given in Fig. 6. In a cold period (Fig. 6a), thin ice grows period 0000 hr 3 January through 2300 hr 5 January (a); for a fasterthan thick ice because of stronger heat conduction warm period 0000 hr 8 April through 1300 hr 11 April (b). In both from warm water to the cold atmosphere, and as a cases, the snow was set to be zero for simplicity. consequence, its surface is warmer. If we look into the average situation, we can see that the surface temperature snow and ice thicknesses were assumed to be 0.5 and 2 cm, dependence on ice thickness is more pronounced for ice respectively. thickness B0.5 m, as noted by Leppa¨ ranta and Lewis The first simulation suggested that snow thickness was (2007). But the daily cycle is seen in the surface tempera- likely overestimated by the model, particularly in early ture, and therefore inversion of surface temperature to ice winter, if the monthly mean precipitation rate was directly thickness would need an ice model with full heat conduc- used as the model input. This led to underestimation of ice tion equation. In a warm period (Fig. 6b), the ice warms up thickness. In the ice growth season, one can estimate the and the temperature approaches isothermal conditions at ratio of snow accumulation from precipitation to the 08C. Next, heat conduction through the ice stops, and the measured snow depth on ground by using climatological surface does not feel the thickness of ice any more. Then, snow thickness and precipitation data (Table 1). In the nighttime cooling reduces the surface temperature down, early winter, the coherence was weak, probably because a but the time scale is too short for cooling to reach the lower part of the precipitation had fallen as liquid water. boundary of the ice sheet to provide thickness information. Additionally, snowdrift may transport snow away from the lake surface. The best coherence between precipitation 4.2. Climatological simulation for the normal period and snow thickness was obtained in February, when 94% 19712000 of the precipitation contributed to snow accumulation. In early spring, due to the melting effect, the discrepancy Ice climatology was examined using the mean monthly between precipitation and snow depth became again larger. atmospheric conditions as the forcing data. The model runs were started up in the beginning of November. The initial We therefore took a simple calculation procedure to define 8 Y. YANG ET AL. a solid portion of the precipitation in each winter month statistics, snow and ice conditions and model results are (Table 1). The result was used as the input into the given in Table 3. Foreach month, a sinusoidal daily air climatological simulations. temperature cycle (model experiment I) was created by the Figure 7 shows the result using the modified snow mean value of the climatological normal period (model accumulation algorithm. The simulated freezing and ice experiment II; cf. Table 1) 9 the standard deviation of breakup dates were 8 November and 12 May, and the 20082009 (Table 3). For comparison, the model runs with duration of ice season was 184 d. Compared to observa- monthly mean temperature were also given. The snow tions (from 30 November to 30 April, altogether 152 d), the thickness remained as the climatological monthly mean for the sake of simplicity. The initial ice thickness (on the first simulated ice season was considerably longer than the mean day of each month) was taken as the monthly mean. ice season in 19712000. The modelled freezing date turned The time series of simulated ice thickness are shown in out to be 22 d too early and the breakup date 12 d too late. Fig. 8. In the model experiment I, the ice growth showed a In the simulations, the treatment of the freeze-up was maximum rate of 6 cm in 10 d in December. This rapid biased, since in the model the cooling process of lake water growth was because the snow is relatively thin. From was missing, which led to the too early freezing. Differences January to March, the snow was thicker and the ice between the simulated and observed ice breakup dates may thickness could still increase by 45 cm in 10 d. This be due to two reasons. First, if the initialisation of melting growth rate is commonly seen as the climatological ice season is delayed, this shifts the breakup date. Secondly, in growth rate in the winter season. In April, the decrease of the melting season, the ice strength decreases due to ice thickness was evidently due to the combined effect internal deterioration, and wind forcing is able to break of the warm air temperature and diurnal variation the ice. Ice may then drift, which further induces turbulent of the incoming solar radiation. The ice melt rate was mixing and heat transfer in the water body (Wake and ca. 0.6 cm d . Rumer, 1983; Burda, 1999; Lepparanta, 2009). Model runs with constant air temperature resulted in The dots in Fig. 7 depict the mean observed ice thickness quite similar ice growth compared to model runs with daily in Lake Kuivajarvi during 19612000 (Korhonen, 2005). cycle in NovemberFebruary (Fig. 8). This may be Compared with the model simulation, the HIGHTS bias anticipated since the daily cycle is not strong in mid-winter was 5 cm, the correlation coefficient was 0.97 and the in latitudes this high (618N). In March, the mean tempera- standard deviation was 3 cm. We can conclude that ture was 2.78C with daily amplitude of 4.58C, and more HIGHTSI can produce climatological ice thickness in ice was produced when the daily cycle was included reasonably good agreement with the regional measure- (Table 3). In April, the ice started to melt in both model ments. runs. However, with the daily cycle much more ice could be The role of the daily cycle in the monthly lake ice melted. The freezing point temperature (08C) can be seen as climatology was investigated with 10-d model runs for each the threshold value to determine the impact of the daily month in NovemberApril. The monthly air temperature cycle. In early spring, the nighttime temperature often drops below 08C, helping to maintain the ice growth. However, if the diurnal temperature range is dominantly positive, the ice melting will accelerate. Overall, lake-ice processes are closely associated with weather conditions from autumn through spring. The freezing of lake surface largely depends on the lake heat storage and the cooling rate of the air temperature during the autumn. Ice breakup can be explained mainly by the net solarradiation (e.g. Leppa ¨ ranta, 2009). Figure 9 shows a comparison of the results of the HIGHTSI simulations using climatological forcing (Table 1), with the air tem- perature artificially shifted by 91or 958C. Compared to the reference run (Fig. 8), shifting by 918C may lead to about 5 d change of freezing date and 8 d change of breakup date. These values are close to 5 d for both dates, obtained by linear regression on lake phenology time series Fig. 7. Time series of simulated snow (grey line) and ice (black for lakes in southern Finland (Palecki and Barry, 1986). line) thickness. The circles show the mean ice thickness observed in The breakup date seems to be more sensitive to the air the middle of each month in Lake Kuivaja¨ rvi (60.768N, 23.888E; cf. Fig. 1) located southwest (60 km) from Lake Vanajavesi. temperature in the model. NUMERICAL MODELLING OF SNOW AND ICE THICKNESSES 9 Table 3. Initial setup (T , h , h ) for 10-d simulations and simulated ice growth rate (dh /dt) a s i i T (8C) dh /dt (cm d ) a i Month I II Initial h (m) Initial h (m) I II s i November[ 4.2, 3.4] 0.4 0 0.01 0.42 0.47 December[ 6.5, 1.7] 4.1 0.03 0.10 0.60 0.58 January [11.7, 0.1] 5.9 0.05 0.25 0.45 0.49 February [10, 3] 6.5 0.10 0.35 0.37 0.37 March [7, 1.6] 2.7 0.10 0.45 0.35 0.20 April [2.5, 7.8] 2.7 0 0.45 0.59 0.03 Changes in the freezing date depend on the local atmo- 4.3. Surface heat balance spheric autumn cooling rate and are therefore different in The surface heat fluxes reflect the ice growth and melt. maritime and continental climate zones (Leppa¨ ranta, Components of the surface energy balance (Eq. 1) were 2009). But the relation between the ice breakup date and calculated forthe winter 2008 2009 and forthe climatolo- spring warming rate is unclear. Nonaka et al. (2007) found gical normal period 19712000 (Table 4). The results are in that a 18C increase in air temperature can lead to an earlier agreement with Jakkila et al. (2009) for Lake Pa¨ a¨ ja¨ rvi, ice breakup by 310 d. Here, model runs suggested an located in a nearby region. overall ice season length to decrease by 13 d for 18C change In both cases, the lake surface lost heat from December in climate. The change of seasonal average ice thickness to March and gained heat in April. In December 2008, heat was 917% or 96 cm from 36 cm for 918C air loss from the surface was less intensive compared with temperature change. If the air temperature shifted by January 2009. This may partly explain the delayed freezing 58C, the ice season would start in the middle of date. In contrast, we can see that a strong heat loss December and end in early April. In contrast, if the air occurred in December for the climatological normal period temperature shifted by 58C, the ice season would start in 19712000. This agrees with the fact that the freezing date early October and end in mid-June. In these cases, the is normally in December. A strong heat loss in January maximum ice thickness would change by 22 and 41 cm. 2009 occurred during the cold period (Fig. 2b). Changes with 958C were not symmetric, because the air Analysing the net surface heat flux for February and temperature change also influenced the length of the ice March, it is seen that in 2009 the negative surface heat season, and ice grew to a first order in proportion to the balance was weakerthan in 1971 2000. In April 2009 a square root of the freezing-degree days. larger positive surface heat balance could have led to an Fig. 8. Ten-day modelled ice thickness based on air temperature Fig. 9. Model sensitivity to the air temperature. The black solid varying sinusoidally (red lines) and constant mean values (black line is the reference of present climate. The grey solid (dashed) line lines) foreach wintermonth: November (a); December (b); and light grey solid (dashed) line are modelled ice thickness based January (c); February (d); March (e) and April (f). on air temperature decreasing (increasing) 5 and 18C, respectively. 10 Y. YANG ET AL. Table 4. The daily mean heat fluxes contributing to the surface heat balance (W m ). The columns are surface layer absorption of solar radiation, net longwave radiation, sensible heat flux, latent heat flux, conductive heat flux from below and heat flux for melting Month (1a )Q  I Q Q Q Q F  F s,i s 0 b d h le c m 20082009 December0.94 18.38 6.37 0.55 10.52 January 0.55 24.05 5.71 6.77 35.98 February 2.30 17.99 0.84 4.55 19.41 March 5.73 16.91 3.94 6.71 13.95 April 21.36 7.26 21.84 8.48 27.46 19712000 December0.17 21.58 5.53 10.84 37.78 January 0.60 22.25 2.07 4.83 24.41 February 2.17 33.93 7.78 2.27 26.25 March 4.45 25.26 5.43 8.91 24.29 April 19.27 8.93 18.68 4.65 24.37 earlier breakup date than during 19712000. It is clearthat is the prevailing factor that affects the lake ice season. the weather conditions in early winter and in early spring The diurnal variation of air temperature has no significant are critical for the determination of the length of the lake influence on ice thickness in the growth season but gives ice season. The net longwave radiation dominates the large impact on the ice thickness in late winter and in the surface heat balance in the winter season, while the net melting season. solar radiation is critical for melting. The surface temperature and the thickness of ice are closely related. The model simulations showed that inver- sion of the surface temperature for ice thickness by thermal 5. Conclusions remote sensing is feasible for thin ice (B 0.5 m), but then the timing of measurement is critical and the full heat The ice coverin Lake Vanajavesi has been investigated with conduction equation needs to be used. The model simula- a one-dimensional thermodynamic model HIGHTSI. tions showed that in NWP and climate modelling, inclusion A case study was performed for one ice season, and a of a proper lake ice model would produce a better surface 30-yearperiod (1971 2000) was examined forice climatol- temperature and consequently improve the quality of the ogy. The results were good as compared with the validation predictions. data. The model was then applied to examine the response Climate change may have a dramatic influence on the of the lake ice season to climate change. The novel features course of the ice season. The annual maximum ice thickness of the HIGHSTI forlake investigations are the advanced was 53 cm in Lake Vanajavesi in 19712000, about 11 cm treatments of superimposed ice, stability-dependent turbu- more than in winter 20082009. The corresponding mean lent heat fluxes and coupling of snow and ice layers. air temperature for 19712000 was about 1.58C lowerthan The modelled snow thickness was produced from the temperature in winter 20082009. Ourmodel experi- measured precipitation and a simple model for snow ments, based on changing the air temperature level in the metamorphosis. This made snow a free model variable. climatological simulations, showed that the air temperature Snow accumulates during cold periods, while it may shifts affect the freezing date by 5 d 8C and the breakup decrease due to transformation into snowice ormelting. date by 8 d 8C . In the case study ice season, snow thickness and snowice Future studies are needed to couple the present model formation were reproduced well. But due to lack of long- with a lake waterbody model to predict better the autumn term data on ice stratigraphy, the modelled snowice cooling of the waterand the consequent freezing date. The cannot be validated quantitatively forthe climatological simulation. Ourcalculations suggested that on an average timing of the onset of snow and ice melting in the model is 16% of the snowfall contributed to the total ice thickness the most critical point for further improvements. (53 cm) with a 10-cm snowice layer. The model experiments produced an estimate for the 6. Acknowledgements heat flux from water to ice by tuning. The result was a This research was supported by YMPANA (Development small value, 0.5 W m , corresponding to the magnitude of of automatic monitoring system for Lake Vanajavesi) molecularheat conduction. The annual maximum ice project, National Natural Science Foundation of China thickness could change up to 10 cm in response to a (Grant No. 51079021, 40930848 and 50879008) and the change of the level of the heat flux to 5 W m . The ice breakup date would be affected by 5 d. The air temperature Norwegian Research Council (Grant No. 193592/S30). The NUMERICAL MODELLING OF SNOW AND ICE THICKNESSES 11 Jaatinen, E., Leppa¨ ranta, M., Erm, A., Lei, R., Pa¨ rn, O. and CIMO scholarship from Finnish Ministry of Education is co-authors. 2010. Light conditions and ice cover structure in acknowledged for funding the first author to work in Lake Vanajavesi. In: Proceedings of the 20th IAHR Ice Finland. Comments from the anonymous reviewers were Symposium, Paper #79. Department of Physics, University of very helpful to improve this paper. We are very grateful to Helsinki, Finland. Dr. Laura Rontu and Dr. Timo Vihma for providing Jacobs, J. D. 1978. Radiation climate of Broughton Island. In: various inspiring comments and suggestions during the Energy Budget Studies in Relation to Fast-ice Breakup Processes preparation of this manuscript. in Davis Strait, Occasional Paper (eds. R. G. Barry and J. D. Jacobs). Institute of Arctic and Alpine Research, University of Colorado, Boulder, 26, pp. 105120. References Jakkila, J., Lepparanta, M., Kawamura, T., Shirasawa, K. and Anderson, E. 1976. A point energy and mass balance model for a Salonen, K. 2009. Radiation transfer and heat budget snow cover. NOAA Technical Report, NWS 19, 150 pp. during the melting season in Lake Pa¨ a¨ ja¨ rvi. Aquat. Ecol. 43, Arst, H., Erm, A., Leppa¨ ranta, M. and Reinart, A. 2006. 681692. Radiative characteristics of ice-covered freshwater and brackish Ka¨ rka¨ s, E. 2000. The ice season of Lake Pa¨ a¨ ja¨ rvi in southern waterbodies. Proc. Estonian Acad. Sci. Geol. 55,323. Finland. Geophysica 36,8594. Bennett, T. J. 1982. A coupled atmosphere-sea-ice model study of Korhonen, J. 2005 Suomen vesisto¨ jen ja¨ a¨ olot [Ice Conditions in the role of sea ice in climatic predictability. J. Atmos. Sci. 39, Lakes and Reviers in Finland]. Suomen ympa¨ risto¨ 751. Finnish 14561465. Environment Institute. (in Finnish with English summary). Brown, L. C. and Duguay, C. R. 2010. The response and role of ice Launiainen, J. and Cheng, B. 1995. A simple non-iterative coverin lake-climate interactions. Prog. Phys. Geogr. 34, 671 algorithm for calculating turbulent bulk fluxes in diabatic 704. conditions over water, snow/ice and ground surface. Report Burda, N. Y. 1999. Computer simulation for Ladoga Lake ice Series in Geophysics, Department of Geophysics, University of dynamics based on remotely sensed data. Int. Geosci. Remote Helsinki, 33 pp. Sensing Symp. 2, 937939. Launiainen, J. and Cheng, B. 1998. Modelling of ice thermo- Cheng, B. 2002. On modeling of sea ice thermodynamics and air- dynamics in nature water bodies. Cold Reg. Sci. Technol. 27, ice coupling in the Bohai Sea and the Baltic Sea. Finnish Institute 153178. of Marine Research  Contributions 5, 38 pp. (Ph.D thesis) Lei, R., Lepparanta, M., Erm, A., Jaatinen, E. and Parn, O. 2011. ¨ ¨ Cheng, B., Vihma, T. and Launiainen, J. 2003. Modelling of the Field investigations of apparent optical properties of ice cover in superimposed ice formation and sub-surface melting in the Finnish and Estonian lakes in winter2009. Est. J. Earth Sci. 60, Baltic Sea. Geophysica 39,3150. 5064. Cheng, B., Zhang, Z., Vihma, T., Johansson, M., Bian, L. and Leppa¨ ranta, M. 1983. A growth model for black ice, snow ice and co-authors. 2008. Model experiments on snow and ice thermo- snow thickness in subarctic basins. Nordic Hydrol. 14,5970. dynamics in the Arctic Ocean with CHINAREN 2003 data. Leppa¨ ranta, M. 1993. A review of analytical models of sea-ice J. Geophys. Res. 113, C09020, doi: 10.1029/2007JC004654. growth. Atmos. Ocean. 31, 123138. Drebs, A., Nordlund, A., Karlsson, P., Helminen, J. and Lepparanta, M. 2009. Modelling the formation and decay Rissanen, P. 2002. Climatological Statistics of Finland 1971 of lake ice. In: Climate change impact on European lakes 2000. Finnish Meteorological Institute, Helsinki. (ed. G. George). Springer, Berlin. Aquatic Ecology Series 4, Duguay, C. R., Flato, G. M., Jeffries, M. O., Menard, P., ´ pp. 6383. Morris, K. and co-authors. 2003. Ice-cover variability on Leppa¨ ranta, M. and Kosloff, P. 2000. The thickness and structure shallow lakes at high latitudes: model simulations and of Lake Pa¨ a¨ ja¨ rvi ice. Geophysica 36, 233248. observations. Hydrol. Proces. 17, 34643483. Leppa¨ ranta, M. and Lewis, J. E. 2007. Observations of ice surface Eerola, K., Rontu, L., Kourzeneva, E. and Shcherbak, E. 2010. temperature and thickness in the Baltic Sea. Int. J. Remote Sens. A study on effects of lake temperature and ice cover in 28, 39633977. HIRLAM. Boreal Env. Res. 15, 130142. Lepparanta, M. and Uusikivi, J. 2002. The annual cycle of the Efimova, N. A. 1961. On methods of calculating monthly values of Lake Paajarvi ice. Lammi Notes 29. Lammi Biological Station, ¨ ¨ ¨ net longwave radiation. Meteor. Gidrol. 10,2833. University of Helsinki. Ellis, A. W. and Johnson, J. J. 2004. Hydroclimatic analysis of Liston, G. E. and Hall, D. K. 1995. An energy-balance model of snowfall trends associated with the North American Great lake-ice evolution. J. Glaciol. 10, 373382. Lakes. J. Hydrol. 5, 471486. Livingstone, D. M. 1993. Lake oxygenation: application of a Flato, G. M. and Brown, R. D. 1996. Variablility and climate one-box model with ice cover. Int. Rev. Ges. Hydrol. 78, sensitivity of landfast Arctic sea ice. J. Geophys. Res. 101, 465480. 2576725777. Maykut, G. A. and Untersteiner, N. 1971. Some results from a Grenfell, T. C. and Maykut, G. A. 1977. The optical properties of time-dependent thermodynamic model of sea ice. J. Geophys. ice and snow in the Arctic Basin. J. Glaciol. 18, 445463. Res. 76, 15501575. Heron, R. and Woo, M. K. 1994. Decay of a High Arctic lake-ice Makynen, M., Simila, M., Cheng, B., Laine, V. and Karvonen, J. ¨ ¨ cover: observations and modelling. J. Glaciol. 40, 283292. 2010. Sea ice thickness retrieval in the Baltic Sea using MODIS 12 Y. YANG ET AL. and SAR data. In: Proceeding of SeaSAR2010, Frascati, Italy, Salgado, R. and LE Moigne, P. 2010. Coupling of the Flake model ESA SP-679. to the Surfex externalized surface model. Boreal Env. Res. 15, Mironov, D., Heise, E., Kourzeneva, E., Ritter, B., Schneider, N. 231244. and co-authors. 2010. Implementation of the lake parameterisa- Saloranta, T. 2000. Modeling the evolution of snow, snow ice and tion scheme Flake into the numerical weather prediction model ice in the Baltic Sea. Tellus 52A,93108. COSMO. Boreal Env. Res. 15, 218230. Shine, K. P. 1984. Parameterization of shortwave flux over high Nonaka, T., Matsunaga, T. and Hoyano, A. 2007. Estimating ice albedo surfaces as a function of cloud thickness and surface breakup dates on Eurasian lakes using water temperature trends albedo. Quart. J. Roy. Meteor. Soc. 110, 747764. and threshold surface temperatures derived from MODIS data. Shirasawa, K., Leppa¨ ranta, M., Kawamura, T., Ishikawa, M. and Int. J. Remote Sens. 28, 21632179. Takatsuka, T. 2006. Measurements and modelling of the water Palecki, M. A. and Barry, R. G. 1986. Freeze-up and break-up of ice heat flux in natural waters. In: Proceedings of the 18th IAHR lakes as an index of temperature changes during the transition International Symposium on Ice, Hokkaido University, Sapporo, seasons: a case study forFinland. J. Appl. Meteorol. Climatol. Japan, pp. 8591. 25, 893902. Shirasawa, K., Leppa¨ ranta, M., Saloranta, T., Kawamura, T., Patterson, J. C. and Hamblin, P. F. 1988. Thermal simulation of a Polomoshnov, A. and co-authors. 2005. The thickness of coastal lake with winterice cover. Limnol. Oceanogr. 33, 323338. fast ice in the Sea of Okhotsk. Cold Reg. Sci. Technol. 42,2540. Pirazzini, R., Vihma, T., Granskog, M. A. and Cheng, B. 2006. Vihma, T. 1995. Subgrid parameterization of surface heat and Surface albedo measurements over sea ice in the Baltic Sea momentum fluxes overpolaroceans. J. Geophys. Res. 100, during the spring snowmelt period. Ann. Glaciol. 44,714. 2262522646. Rothrock, D. A., Yu, Y. and Maykut, G. A. 1999. Thinning of the Wake, A. and Rumer, R. R. 1983. Great lakes ice dynamics Arctic sea-ice cover. Geophys. Res. Lett. 26, 34693472. simulation. J. Waterway Port Coast. Ocean Eng. 109,86102. Rouse, W. R., Blanken, P. D., Duguay, C. R., Oswald, C. J. and Wang, X., Key, J. R. and Liu, Y. 2010. A thermodynamic model Schertzer, W. M. 2008b. Climatelake interations. In: Cold forestimating sea and lake ice thickness with optical satellite Region Atmospheric and Hydrologic Studies: The Mackenzie data. J. Geophys. Res. 115, C12035, doi: 10.1029/2009JC005857. GEWEX Experience, (ed. MK. Woo). Vol. 2, Springer, Berlin, Wendler, G. and Eaton, F. 1990. Surface radiation budget at pp. 139160. Barrow, Alaska. Theor. Appl. Climatol. 41, 107115. Rouse, W. R., Binyamin, J., Blanken, P. D., Bussie` res, N., Duguay Williams, G., Layman, K. L. and Stefan, H. G. 2004. Dependence C. R. and co-authors. 2008a. The influence of lakes on the of lake ice covers on climatic, geographic and bathymetric regional energy and water balance of the central Mackenzie. In: variables. Cold Reg. Sci. Technol. 40, 145164. Cold Region Atmospheric and Hydrologic Studies: The Mack- enzie GEWEX Experience, (ed. MK. Woo). Vol. 1, Springer, Berlin, pp. 309325. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Tellus A: Dynamic Meteorology and Oceanography Taylor & Francis

Numerical modelling of snow and ice thicknesses in Lake Vanajavesi, Finland

Loading next page...
 
/lp/taylor-francis/numerical-modelling-of-snow-and-ice-thicknesses-in-lake-vanajavesi-0oMcB1G79H

References (54)

Publisher
Taylor & Francis
Copyright
© 2012 Yu Yang et al.
ISSN
1600-0870
DOI
10.3402/tellusa.v64i0.17202
Publisher site
See Article on Publisher Site

Abstract

SERIES A DYNAMIC ME TEOROLOGY AND OCEANOGRAPHY PUBLISHED BY THE INTERNATIONAL METEOROLOGICAL INSTITUTE IN STOCKHOLM Numerical modelling of snow and ice thicknesses in Lake Vanajavesi, Finland 1, 2 2 3, 1 1 1 By YU YANG *, MATTI LEPPARANTA ,BIN CHENG and ZHIJUN LI , State Key Laboratory of Coastal and Offshore Engineering, Dalian University of Technology, Dalian 116024, China; 2 3 Department of Physics, University of Helsinki, PO Box 48, FI-00014 Helsinki, Finland; Finnish Meteorological Institute, PO Box 503, FI-00101, Helsinki, Finland (Manuscript received 27 March 2011; in final form 7 January 2012) ABSTRACT Snow and ice thermodynamics was simulated applying a one-dimensional model for an individual ice season 20082009 and for the climatological normal period 19712000. Meteorological data were used as the model input. The novel model features were advanced treatment of superimposed ice and turbulent heat fluxes, coupling of snow and ice layers and snow modelled from precipitation. The simulated snow, snowice and ice thickness showed good agreement with observations for 20082009. Modelled ice climatology was also 1 1 reasonable, with 0.5 cm d growth in DecemberMarch and 2 cm d melting in April. Tuned heat flux from waterto ice was 0.5 W m . The diurnal weather cycle gave significant impact on ice thickness in spring. Ice climatology was highly sensitive to snow conditions. Surface temperature showed strong dependency on thickness of thin ice (B0.5 m), supporting the feasibility of thermal remote sensing and showing the importance of lake ice in numerical weather prediction. The lake ice season responded strongly to air temperature: a level increase by 1 or 58C decreased the mean length of the ice season by 13 or 78 d (from 152 d) and the thickness of ice by 6 or 22 cm (from 50 cm), respectively. Keywords: thermodynamic model, sensitivity tests, snow, ice, Lake Vanajavesi cuts off airlake exchange of oxygen and reduces the 1. Introduction production of dissolved oxygen by limiting the light pene- Freshwater lakes cover about 2% of the Earth’s land tration (Livingstone, 1993). Ice also prevents the exchange of surface. In Finland, there are ca. 190 000 lakes larger than momentum between the atmosphere and the lake water 500 m , accounting for10% of the area of the whole (Williams et al., 2004). Lakes affect the local weatherby country, and these lakes freeze every winter. Their ice modifying the air temperature, wind, humidity and precipi- sheets consist of congelation ice and snowice with snow tation in their surroundings (Ellis and Johnson, 2004; covernormally on top. In medium-size and small lakes, the Rouse et al., 2008a, 2008b), and ice coverprotects the heat ice coveris usually stable, while in large lakes mechanical content of lakes by its insulation capacity and by damping displacements may take place, creating piles of ice blocks or turbulent mixing in the water body. Thus, the presence ridges. An ice cover stabilises the thermal structure of a (orabsence) of ice coverhas an impact on both egional lake. The ice bottom is at the freezing point, and there is a climate and weatherevents in the winter spring season. weak heat flux from the water to the ice. In spring, ice Understanding the processes and interactions of lake ice sheets gain heat mainly from solar radiation, and as the ice and atmosphere is essential for numerical weather predic- melts, all impurities contained in the ice are released into tion (NWP) (Brown and Duguay, 2010). As the spatial the waterorto the air. resolution of NWP models becomes higher, presently Lake ice plays an important role in the hydrological, approaching one-kilometer scale, including lakes and lake biological, chemical and socio-economical regimes of boreal ice forforecasting and data assimilation in mesoscale, lakes (Leppa¨ ranta, 2009). For example, a compact ice cover NWP systems have gained more attention (e.g. Eerola et al., 2010; Mironov et al., 2010; Salgado et al., 2010). Analytical, semi-empirical lake ice thickness models were *Corresponding author. email: yangyang-0606@hotmail.com widely used until 1980s, and thereafter full numerical Tellus A 2012. # 2012 Y. Yang et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution-Noncommercial 3.0 1 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Tellus A 2012, 64, 17202, DOI: 10.3402/tellusa.v64i0.17202 (page numbernot forcitation purpose) r 2 Y. YANG ET AL. models have also been employed (e.g. Liston and Hall, 2. Data 1995; Duguay et al., 2003). The basic principle of these models is to solve the heat conduction equation forthe 2.1. Observations during the ice season 20082009 snow and ice layers in the vertical direction, as was first Lake Vanajavesi (61.138N, 24.278E) is located in south- done in sea ice thermodynamic modelling by Maykut and western Finland (Fig. 1). It is a large, shallow and Untersteiner (1971). The Canadian Lake Ice Model is a eutrophic lake. The area is 113 km , and the mean and snow and ice process model (Duguay et al., 2003) adapted maximum depths are 7 and 24 m, respectively. The lake has from a sea ice model of Flato and Brown (1996). a long eutrophication history with very poor water quality In most of the lake ice models snow thickness is in the 1960s and 1970s. The ice season lasts 46 months, on prescribed with a climatological growth rate or with a average from December to April, and the thickness of ice fixed snowfall scenario based on in situ data. In the boreal reaches its annual maximum value of 4560 cm in March. zone, however, the role of snow in the growth and melting The ice and snow thickness may show distinct spatial of lake ice is very important, and snow and ice form a variations due to snowdrift and heat flux from the lake coupled system. In analytical modelling, the influence of water. snow is usually considered by modifying the growth law In the winterof 2008 2009, a field programme was parameters (e.g. Leppa¨ ranta, 1993). Interactive snow and performed in Lake Vanajavesi. Hydrographical surveys snowice layers were included in the numerical quasi- were made regularly, an automatic ice station was set up steady ice thickness model of Leppa¨ ranta (1983). The first and optical investigations were performed (Jaatinen et al., detailed lake model study in Finland was presented by 2010; Lei et al., 2011). The average maximum snow and ice Lepparanta and Uusikivi (2002), based on a Baltic Sea thicknesses reached 15 and 41 cm, respectively, in March. model of Saloranta (2000), where snow metamorphosis and The freezing and breakup dates were 1 January and 27 snowice formation due to flooding were included (see also April, respectively. Shirasawa et al., 2005). In this study, a one-dimensional high-resolution thermo- dynamic snow and ice model (HIGHTSI) was applied for Lake Vanajavesi, located in southern Finland. This model contains congelation ice, snowice and snow layers with full heat conduction equation. Compared with earlier lake ice models, new features were superimposed ice formation, an advanced atmosphereice heat exchange treatment accounting forthe influence of stability of stratification to the turbulent fluxes and coupling of the snow and ice layers. Atmospheric forcing was derived from weather observations and climatology, which also drove the snow coverevolution. The simulation results were compared with measured ice and snow thickness. A case study was performed for the ice season 20082009, forced by daily weather observations. Ice climatology was examined forthe 30-year period 1971 2000; also the correlation between the observed monthly total precipitation and snow accumulation was investigated in order to understand the uncertainties of precipitation as model forcing for climatological simulation. A number of climate sensitivity simulations were carried out for the ice season. The objectives of the present work were to assess the applicability of the HIGHTSI model forlake snow and ice thermodynamics, to find out the most important factors affecting lake ice growth and melting and to evaluate the influence of climate variations on the lake ice season. Section 2 introduces data, the model is described in Fig. 1. Geographical location of Lake Vanajavesi (A); Section 3 and the results are presented and discussed surrounding observation sites Jokioinen meteorological in Section 4. Final conclusions follow in Section 5. observatory (B), Lake Kuivaja¨ rvi (C) and Lake Pa¨ a¨ ja¨ rvi (D). NUMERICAL MODELLING OF SNOW AND ICE THICKNESSES 3 Synoptic-scale weatherconditions and egional climate than in the winter2008 2009. The precipitation was over this area are represented by Jokioinen meteorological 232 mm, about the same that was observed in 20082009, observatory (60.88N, 23.58E; WMO station 02863), located but during the ice growth season, the average snow some 50 km southwest of Lake Vanajavesi (Fig. 1). The accumulation was 145 mm WEQ or58% more than in weatherforcing data forthe lake ice model consist of wind 20082009. speed (V ), air temperature (T ), relative humidity (Rh), The average freezing and ice breakup dates of Lake a a cloudiness (CN) and precipitation (Prec), collected at Vanajavesi were 30 November and 30 April, respectively, in 3-hour time intervals. For the present modelling work, we 19712000 (Korhonen, 2005). A large variation occurred in interpolated data to 1-hour intervals. The snow thickness these dates in individual years. The earliest freezing date on land is also available from Jokioinen. was 28 Octoberand the latest one was 28 January. Forthe The winter2008 2009 was mild (Fig. 2). The early winter ice breakup, the earliest and latest dates were recorded as was warm, with average air temperature of 0.68Cin 1 April and 27 May, respectively. Novemberand Decemberpreventing Lake Vanajavesi Due to the lack of snow and ice thickness data in Lake from freezing. From late December to the end of March, Vanajavesi, the observations of Lake Kuivajarvi (Korhonen, the air temperature was permanently below the freezing 2005), located about 60 km southwest of Lake Vanajavesi, point, on several occasions approaching 208C and were used as the validation data. It can be assumed that the thereafter increased gradually to above 08C. The ice season climatological conditions are the same in these two lakes. 20082009 can be divided into two phases: ice growth (JanuaryMarch) and melting (April). The corresponding 3. The lake ice thermodynamic model average air temperatures were 4.4 and 4.38C, respec- tively. The total precipitation from November to April was 3.1. Surface temperature and surface heat/mass 234 mm. During the ice growth season, however, the snow balance accumulation was merely 92 mm water equivalent (WEQ). The thermodynamic lake ice model HIGHTSI solves the surface temperature (T ) from a detailed surface heat sfc 2.2. Climatological for long-term simulations balance equation using an iterative method: Monthly mean meteorological observations from the Jokioinen observatory in the climatological normal period ð1  aÞQ  I þ Q  Q ðT Þþ Q ðT Þ s 0 d b sfc h sfc 19712000 (http://en.ilmatieteenlaitos.fi/normal-period- (1) þQ ðT Þþ F ðT Þ F ¼ 0 le sfc c sfc m 19712000) were used for the forcing to investigate ice climatology (Table 1, based on Drebs et al., 2002). The mean where a is the albedo, Q is the incoming solarradiation at air temperature was 2.88C in NovemberApril, 1.38C less s the surface, I is the portion of solar radiation penetrating below the surface layer and contributing to internal heating of the snow and ice, Q and Q are the incoming and d b outgoing longwave radiation, respectively, Q is the sensible heat flux, Q is the latent heat flux, F is the conductive heat le c flux coming from below the surface and F stands for melting. All fluxes are positive towards the surface. In HIGHTSI, I is explicitly parameterised based on the fact that penetrating solar radiation is strongly attenuated along the vertical depth immediately below the surface (Grenfell and Maykut, 1977). The term (1a)Q I thus s 0 represents the part of short-wave radiation contributing to the surface heat balance. It depends on the surface layer (i.e. the first layer of snow or ice in model) thickness (Launiainen and Cheng, 1998; Cheng, 2002; Cheng et al., 2008). Equation (1) serves as the upper boundary condition of the HIGHTSI model. If the surface temperature is at the melting point (T ), excessive heat (F ) is used formelting: f m dh/dtF /rL , where h is the thickness of snow orice, m f Fig. 2. Time series of wind speed (a), air temperature (b), r is the density of snow orice and L is the latent heat of relative humidity (c), cloudiness (d) and precipitation (e). The data f were initially observed at 3-hour time interval. fusion. r 4 Y. YANG ET AL. Table 1. The monthly mean meteorological data for 19712000 (Drebs et al., 2002; h? is snow thickness on the 15th of each month; Ph is s s the portion of precipitation contributing to the snow accumulation) 1 1 Month V (m s ) T (8C) CN Rh (%) Precipitation (mm month ) h? (cm) Ph (%) a a s s October3.8 4.6 0.76 88 59 0 November3.9 0.4 0.83 90 57 2 X December3.9 4.1 0.81 90 45 9 46 January 3.8 5.9 0.79 89 41 19 77 February 3.7 6.5 0.74 87 29 29 94 March 3.8 2.7 0.68 82 30 31 23 April 3.7 2.7 0.7 74 32 10 X May 3.7 9.5 0.62 64 35 X X The albedo is critical for snow and ice heat balance. Q ¼ q R C ðq  q ÞV (6) A parameterisation suitable for the Baltic Sea coastal snow le a l E a sfc a and land-fast ice is selected in this study (Pirazzini et al., where T T is the difference between air and surface a sfc 2006): temperature, q q is the difference between air and a sfc surface specific humidities, V is the wind speed, R the a l a ¼ 0:15 h B 0:1 cm enthalpy of sublimation of water, r is the density of air, c h ða a Þ a a s s i a ¼ min a ; a þ h ] 0:1 cm and h 5 10 cm (2) s i i s 0:1 is the specific heat of airand C and C are the bulk : H E a ¼ a h ] 0:1 cm and h > 10 cm s i s transfer coefficients estimated using the MoninObukhov similarity theory (Launiainen and Cheng, 1995). The surface conductive heat flux was determined by: where a and a are snow and ice albedo, a 0.75, s i s 1:5 a ¼ minðÞ 0:55; 0:85  h þ 0:15 and h and h are the s i i i @T snow and ice thickness, respectively. F ¼k (7) @z sfc Incoming solarorlongwave adiation can be either parameterised for estimation from cloudiness, temperature where T is temperature, z is the depth below the surface and and humidity or directly obtained from an NWP model. k is the thermal conductivity in the surface layer. Forthe solarradiation, we applied the clear sky equation of Shine (1984) with cloud effect of Bennett (1982): 3.2. Snow and ice evolution S cos Z Q ¼ Snow coverevolution was modelled taking into account 1:2 cos Z þðcos Z þ 1:0Þe  10 þ 0:0455 several factors: precipitation, packing of snow and melt ð1  0:52 CNÞ (3) freeze metamorphosis. Packing depends on precipitation, where S is the solarconstant, Z is the solarzenith angle, e is air temperature and wind speed, which are obtained from the water vapour pressure (mbar) and CN is the cloudiness observations or from an NWP model. A temperature ranging between 0 and 1. The net longwave radiative flux criterion was set (T B0.58C) to decide whetherprecipita- (Q Q ) was calculated by: d b tion is solid orliquid. The impact of wind on snow pffiffiffi accumulation is regarded as a model uncertainty. In the Q  Q ¼ r T 0:746 þ 0:066  e d b s a climatological sense, however, one may have a rough idea (4) ð1 þ 0:26  CNÞ er T s sfc of the influence of wind on snow accumulation by comparing the measured precipitation and snow thickness where the first term on the right-hand side represents the (cf. Section 4.2). In addition to the surface melting, snow incoming longwave radiation (Efimova, 1961, with cloud and ice temperature may reach the melting point below the effect of Jacobs, 1978), the second term is outgoing surface because of the absorption of the penetrating solar longwave radiation, s is the StefanBoltzmann constant radiation. The subsurface melting is then calculated and o is the surface emissivity, which ranges from 0.96 to according to Cheng et al. (2003). 0.99 (Vihma, 1995). In boreal lakes, snow may contribute significantly to the The surface turbulent heat fluxes were parameterised on total ice thickness. Snow-load can cause a negative ice the basis of atmospheric stratification. The sensible heat Q freeboard and consequent formation of a slush layer by and latent heat Q were calculated by: le mixing of lake waterand snow. A slush layercan also be Q ¼ q c C ðT  T ÞV (5) h a a H a sfc a formed by percolation of snow meltwater or liquid r NUMERICAL MODELLING OF SNOW AND ICE THICKNESSES 5 precipitation. Freezing of slush into ice was calculated by where F is the heat flux from water to ice. The model the heat flux divergence: parameters for this study are summarised in Table 2. The freezing date in HIGHTSI is provided as the initial dh @T @T si s i q L ¼ k j k j (8) condition by a thin ice layer(0.02 m). If the weather si f s z¼hs i z¼hi dt @z @z conditions do not favour ice growth, HIGHTSI will resume the initial ice thickness. The first day when ice starts to where h is snowice thickness and r is the density of si si grow successively is defined as the freezing date. It is clear snowice. Slush formation from flooding was calculated that this is a crude approximation to obtain the freezing according to Lepparanta (1983). After its formation, date, but once the ice growth has started, the model catches snowice becomes an internal layer part of the ice sheet. up well with the nature. The density of snow is a prognostic variable, calculated according to Anderson (1976): 1 @q 4. Results and discussion 0:08ð273:15T Þ c q a 2 s ¼ c e w e (9) 1 s q @t 4.1. Seasonal study for the winter 20082009 where c and c are empirical constants and w is the mass 1 2 s The ice season 20082009 lasted 4 months. Model simula- of snow in the above layerin WEQ. In this study, we set 1 1 3 1 tions were started up in the beginning of January from the c 5m h and c 21 m kg . 1 2 initial snow and ice thicknesses of 0.5 and 2 cm, respec- The snow and ice temperatures were solved from the heat tively. The simulated snow thickness agreed well with the conduction equation: observations (Fig. 3), showing that the model can well @Tðz; tÞ @ @Tðz; tÞ @qðz; tÞ reproduce the total snow accumulation and the snowmelt. qc ¼ k  (10) @t @z @z @z Snow thickness overland, measured in Jokioinen, is also shown in Fig. 3. A previous study by Ka¨ rka¨ s (2000) has where q(z, t) presents the solar radiation penetrating into indicated that the snow coveron land is on average 2.3 the snow and ice based on an exponential attenuation law: times thickerthan on lake ice in this egion, based on q (z, t)(1a )I exp(k z), where k is the extinction s s,i 0 s,i s,i monthly snow measurements from January to April during coefficient of snow orice. a 10-yearperiod (1971 1980) in Lake Pa¨ a¨ ja¨ rvi (61.078N, The integral interpolation method was applied to build 25.148E). In the present case, the ratio between the up the numerical scheme (Cheng, 2002). A high spatial observed mean snow depth in Jokioinen and the modelled resolution (e.g. 10 layers in the snow and 20 layers in the average snow depth over the Lake Vanajavesi ice was 2.1. ice) ensures that the response of the snow and ice The modelled ice thickness fitted well with the field temperature regime to the absorption of solar radiation measurements, and particularly the ice breakup date was near the surface is correctly resolved. In the case of snow on close to the observation. In the beginning of January, the top of the ice, the snowice interface temperature is ice grew rapidly in response to the very cold air tempera- calculated via a flux continuity equation: ture. As soon as the snow depth had reached about 10 cm, @T @T the speed of ice growth reduced significantly because of the s i k ¼ k (11) s i insulating effect of snow. Snow to snowice transformation @z @z also contributed a little to the total ice thickness in the late Heat and mass balances at the ice bottom served as the season. A 9-cm slush layerand a 1-cm snow ice layerwere lowerboundary condition: measured on 18 March (Lei et al., 2011), while the model dh @T prediction was a 5-cm slush layer and no snowice. This i i q L ¼k þF i f i w error is likely due to overestimation of snow density in the dt @z (12) bottom model. Later, on 7 April, a 2-cm slush layer and a 5-cm T ¼ T frozen slush layer were observed (Lei et al., 2011), while in Table 2. Model parameters found in the literature Parameter Value Source Extinction coefficient of lake ice (k ) 1.517 m Heron et al. (1994); Arst et al. (2006); Lei et al. (2011) Extinction coefficient of snow (k)620 m Patterson et al. (1988); Arst et al. (2006); Lei et al. (2011) Lake ice density (r ) 910 kg m Corresponds to 1% gas content Initial snow density (r ) 330 kg m Leppa¨ ranta and Kosloff (2000) Surface emissivity (o) 0.97 Vihma (1995) r 6 Y. YANG ET AL. Fig. 4. Time series of observed and simulated ice thickness based on varying heat flux from water: 0 W m (black dashed), 2 2 0.5 W m (black solid; reference experiment), 2 W m (grey dashed) and 5 W m (grey solid). Measurements of ice thickness are shown as in Fig. 3. Fig. 3. Time series of observed and modelled snow and ice thickness. The dark grey line and the asterisk are the observed snow thickness in Jokioinen and on Lake Vanajavesi, respectively. conduction is reduced and closer to the heat flux from the The black dashed and solid lines are modelled snow and ice water. Ice may then even melt from the bottom. The thickness (reference experiment), respectively. The circles are the examined variations of the heat flux from water can alter observed average ice thickness, and the spatial standard deviation the maximum ice thickness by93.4 cm, as compared to the is indicated by the vertical bar. results from the optimised reference simulation, where the heat flux from water to ice was 0.5 W m . This level was oursimulation these values had become 2.9 and 2.7 cm, quite low, of the same magnitude than the molecular respectively, by the end of March. On 14 April, a 3-cm conduction of heat. It is much less than what has been slush layerwas still observed but snow ice had disap- earlier obtained in Lake Paajarvi, which is deeper and also ¨ ¨ ¨ peared. The melt rate of ice was 1 cm d from 7 April to has more groundwater inflow in the mass balance. A large 14 April (Lei et al., 2011), both in the model and in the heat flux from the water would give less ice thickness and observations. The melting of ice started in the beginning of 2 earlier breakup date. For example, a 1 W m heat flux April, as soon as snow had gone, and it accelerated towards melts about 1 cm ice in a month, and forthe mean melt rate the breakup (Fig. 3). The melting period lasted 30 d, with of 1.3 cm d ice breakup would be about 1 d earlier. mean melt rate of 1.3 cm d . The parameterised solar radiation may be an under- The heat flux from lake water to the ice bottom has been estimate. The uncertainties are due to the effect of clouds as earlier examined in the nearby Lake Pa¨ a¨ ja¨ rvi by Shirasawa well as due to multiple scattering in the air (Shine, 1984; et al. (2006) and Jakkila et al. (2009). The magnitude was Wendlerand Eaton, 1990). Fig. 5 gives the observations found to be 310 W m based on field data. In freshwater and calculations forJokioinen in January April 2009. The lakes, before freezing the water column is well mixed down largest differences, 200 W m , appeared mostly in to 148C depending on the autumn wind conditions, lake MarchApril. Then, on average, underestimation of solar size and land area around. Thereafter, inverse winter radiation may account for 7 W m of the total net stratification is set up, and once ice is formed, further radiative flux. This amount of energy can offset 3 cm of the mixing is strongly limited. The water temperature structure ice thickness and lead to a 5-d shift in the ice breakup date. underthe ice is usually quite stable with a very small heat The surface temperature reflects the ice thickness during flux out from the water body. the cold season, because the ice thickness impacts the To examine the sensitivity of ice thickness to the heat strength of the insulation between the lake water body and flux from the water, model runs were carried out with the atmosphere. By applying the heat conduction law different fixed heat flux levels: 0, 0.5, 2 and 5 W m (Eq. 1), we can solve the ice thickness as a function of (Fig. 4). In the early winter, there was a very large the surface temperature or vice versa. The inverse proce- temperature gradient across the ice sheet, and the heat dure is used in thermal remote sensing for mapping the sea conduction through the ice and the consequent ice growth ice thickness (Rothrock et al., 1999; Leppa¨ ranta and rate were large. As soon as the ice thickness has reached Lewis, 2007; Ma¨ kynen et al., 2010; Wang et al., 2010). about 20 cm and there is also snow on the ice, the heat In weather forecasting, the surface temperature is needed as NUMERICAL MODELLING OF SNOW AND ICE THICKNESSES 7 Fig. 5. Calculated and observed incoming short-wave radiative flux underclear and cloudy sky conditions. The data coverthe winter2008 2009 from January to April. Goodness of fit for the whole ice season’s data was R 0.92 (n2880). a boundary condition for the atmosphere. Due to ice atmosphere coupling, NWP needs an ice model to obtain the surface temperature of lakes in winter. To examine the sensitivity of surface temperature on ice thickness, a number of seasonal model runs were per- formed with different initial ice thicknesses. The results are Fig. 6. Surface temperature versus ice thickness: for a cold given in Fig. 6. In a cold period (Fig. 6a), thin ice grows period 0000 hr 3 January through 2300 hr 5 January (a); for a fasterthan thick ice because of stronger heat conduction warm period 0000 hr 8 April through 1300 hr 11 April (b). In both from warm water to the cold atmosphere, and as a cases, the snow was set to be zero for simplicity. consequence, its surface is warmer. If we look into the average situation, we can see that the surface temperature snow and ice thicknesses were assumed to be 0.5 and 2 cm, dependence on ice thickness is more pronounced for ice respectively. thickness B0.5 m, as noted by Leppa¨ ranta and Lewis The first simulation suggested that snow thickness was (2007). But the daily cycle is seen in the surface tempera- likely overestimated by the model, particularly in early ture, and therefore inversion of surface temperature to ice winter, if the monthly mean precipitation rate was directly thickness would need an ice model with full heat conduc- used as the model input. This led to underestimation of ice tion equation. In a warm period (Fig. 6b), the ice warms up thickness. In the ice growth season, one can estimate the and the temperature approaches isothermal conditions at ratio of snow accumulation from precipitation to the 08C. Next, heat conduction through the ice stops, and the measured snow depth on ground by using climatological surface does not feel the thickness of ice any more. Then, snow thickness and precipitation data (Table 1). In the nighttime cooling reduces the surface temperature down, early winter, the coherence was weak, probably because a but the time scale is too short for cooling to reach the lower part of the precipitation had fallen as liquid water. boundary of the ice sheet to provide thickness information. Additionally, snowdrift may transport snow away from the lake surface. The best coherence between precipitation 4.2. Climatological simulation for the normal period and snow thickness was obtained in February, when 94% 19712000 of the precipitation contributed to snow accumulation. In early spring, due to the melting effect, the discrepancy Ice climatology was examined using the mean monthly between precipitation and snow depth became again larger. atmospheric conditions as the forcing data. The model runs were started up in the beginning of November. The initial We therefore took a simple calculation procedure to define 8 Y. YANG ET AL. a solid portion of the precipitation in each winter month statistics, snow and ice conditions and model results are (Table 1). The result was used as the input into the given in Table 3. Foreach month, a sinusoidal daily air climatological simulations. temperature cycle (model experiment I) was created by the Figure 7 shows the result using the modified snow mean value of the climatological normal period (model accumulation algorithm. The simulated freezing and ice experiment II; cf. Table 1) 9 the standard deviation of breakup dates were 8 November and 12 May, and the 20082009 (Table 3). For comparison, the model runs with duration of ice season was 184 d. Compared to observa- monthly mean temperature were also given. The snow tions (from 30 November to 30 April, altogether 152 d), the thickness remained as the climatological monthly mean for the sake of simplicity. The initial ice thickness (on the first simulated ice season was considerably longer than the mean day of each month) was taken as the monthly mean. ice season in 19712000. The modelled freezing date turned The time series of simulated ice thickness are shown in out to be 22 d too early and the breakup date 12 d too late. Fig. 8. In the model experiment I, the ice growth showed a In the simulations, the treatment of the freeze-up was maximum rate of 6 cm in 10 d in December. This rapid biased, since in the model the cooling process of lake water growth was because the snow is relatively thin. From was missing, which led to the too early freezing. Differences January to March, the snow was thicker and the ice between the simulated and observed ice breakup dates may thickness could still increase by 45 cm in 10 d. This be due to two reasons. First, if the initialisation of melting growth rate is commonly seen as the climatological ice season is delayed, this shifts the breakup date. Secondly, in growth rate in the winter season. In April, the decrease of the melting season, the ice strength decreases due to ice thickness was evidently due to the combined effect internal deterioration, and wind forcing is able to break of the warm air temperature and diurnal variation the ice. Ice may then drift, which further induces turbulent of the incoming solar radiation. The ice melt rate was mixing and heat transfer in the water body (Wake and ca. 0.6 cm d . Rumer, 1983; Burda, 1999; Lepparanta, 2009). Model runs with constant air temperature resulted in The dots in Fig. 7 depict the mean observed ice thickness quite similar ice growth compared to model runs with daily in Lake Kuivajarvi during 19612000 (Korhonen, 2005). cycle in NovemberFebruary (Fig. 8). This may be Compared with the model simulation, the HIGHTS bias anticipated since the daily cycle is not strong in mid-winter was 5 cm, the correlation coefficient was 0.97 and the in latitudes this high (618N). In March, the mean tempera- standard deviation was 3 cm. We can conclude that ture was 2.78C with daily amplitude of 4.58C, and more HIGHTSI can produce climatological ice thickness in ice was produced when the daily cycle was included reasonably good agreement with the regional measure- (Table 3). In April, the ice started to melt in both model ments. runs. However, with the daily cycle much more ice could be The role of the daily cycle in the monthly lake ice melted. The freezing point temperature (08C) can be seen as climatology was investigated with 10-d model runs for each the threshold value to determine the impact of the daily month in NovemberApril. The monthly air temperature cycle. In early spring, the nighttime temperature often drops below 08C, helping to maintain the ice growth. However, if the diurnal temperature range is dominantly positive, the ice melting will accelerate. Overall, lake-ice processes are closely associated with weather conditions from autumn through spring. The freezing of lake surface largely depends on the lake heat storage and the cooling rate of the air temperature during the autumn. Ice breakup can be explained mainly by the net solarradiation (e.g. Leppa ¨ ranta, 2009). Figure 9 shows a comparison of the results of the HIGHTSI simulations using climatological forcing (Table 1), with the air tem- perature artificially shifted by 91or 958C. Compared to the reference run (Fig. 8), shifting by 918C may lead to about 5 d change of freezing date and 8 d change of breakup date. These values are close to 5 d for both dates, obtained by linear regression on lake phenology time series Fig. 7. Time series of simulated snow (grey line) and ice (black for lakes in southern Finland (Palecki and Barry, 1986). line) thickness. The circles show the mean ice thickness observed in The breakup date seems to be more sensitive to the air the middle of each month in Lake Kuivaja¨ rvi (60.768N, 23.888E; cf. Fig. 1) located southwest (60 km) from Lake Vanajavesi. temperature in the model. NUMERICAL MODELLING OF SNOW AND ICE THICKNESSES 9 Table 3. Initial setup (T , h , h ) for 10-d simulations and simulated ice growth rate (dh /dt) a s i i T (8C) dh /dt (cm d ) a i Month I II Initial h (m) Initial h (m) I II s i November[ 4.2, 3.4] 0.4 0 0.01 0.42 0.47 December[ 6.5, 1.7] 4.1 0.03 0.10 0.60 0.58 January [11.7, 0.1] 5.9 0.05 0.25 0.45 0.49 February [10, 3] 6.5 0.10 0.35 0.37 0.37 March [7, 1.6] 2.7 0.10 0.45 0.35 0.20 April [2.5, 7.8] 2.7 0 0.45 0.59 0.03 Changes in the freezing date depend on the local atmo- 4.3. Surface heat balance spheric autumn cooling rate and are therefore different in The surface heat fluxes reflect the ice growth and melt. maritime and continental climate zones (Leppa¨ ranta, Components of the surface energy balance (Eq. 1) were 2009). But the relation between the ice breakup date and calculated forthe winter 2008 2009 and forthe climatolo- spring warming rate is unclear. Nonaka et al. (2007) found gical normal period 19712000 (Table 4). The results are in that a 18C increase in air temperature can lead to an earlier agreement with Jakkila et al. (2009) for Lake Pa¨ a¨ ja¨ rvi, ice breakup by 310 d. Here, model runs suggested an located in a nearby region. overall ice season length to decrease by 13 d for 18C change In both cases, the lake surface lost heat from December in climate. The change of seasonal average ice thickness to March and gained heat in April. In December 2008, heat was 917% or 96 cm from 36 cm for 918C air loss from the surface was less intensive compared with temperature change. If the air temperature shifted by January 2009. This may partly explain the delayed freezing 58C, the ice season would start in the middle of date. In contrast, we can see that a strong heat loss December and end in early April. In contrast, if the air occurred in December for the climatological normal period temperature shifted by 58C, the ice season would start in 19712000. This agrees with the fact that the freezing date early October and end in mid-June. In these cases, the is normally in December. A strong heat loss in January maximum ice thickness would change by 22 and 41 cm. 2009 occurred during the cold period (Fig. 2b). Changes with 958C were not symmetric, because the air Analysing the net surface heat flux for February and temperature change also influenced the length of the ice March, it is seen that in 2009 the negative surface heat season, and ice grew to a first order in proportion to the balance was weakerthan in 1971 2000. In April 2009 a square root of the freezing-degree days. larger positive surface heat balance could have led to an Fig. 8. Ten-day modelled ice thickness based on air temperature Fig. 9. Model sensitivity to the air temperature. The black solid varying sinusoidally (red lines) and constant mean values (black line is the reference of present climate. The grey solid (dashed) line lines) foreach wintermonth: November (a); December (b); and light grey solid (dashed) line are modelled ice thickness based January (c); February (d); March (e) and April (f). on air temperature decreasing (increasing) 5 and 18C, respectively. 10 Y. YANG ET AL. Table 4. The daily mean heat fluxes contributing to the surface heat balance (W m ). The columns are surface layer absorption of solar radiation, net longwave radiation, sensible heat flux, latent heat flux, conductive heat flux from below and heat flux for melting Month (1a )Q  I Q Q Q Q F  F s,i s 0 b d h le c m 20082009 December0.94 18.38 6.37 0.55 10.52 January 0.55 24.05 5.71 6.77 35.98 February 2.30 17.99 0.84 4.55 19.41 March 5.73 16.91 3.94 6.71 13.95 April 21.36 7.26 21.84 8.48 27.46 19712000 December0.17 21.58 5.53 10.84 37.78 January 0.60 22.25 2.07 4.83 24.41 February 2.17 33.93 7.78 2.27 26.25 March 4.45 25.26 5.43 8.91 24.29 April 19.27 8.93 18.68 4.65 24.37 earlier breakup date than during 19712000. It is clearthat is the prevailing factor that affects the lake ice season. the weather conditions in early winter and in early spring The diurnal variation of air temperature has no significant are critical for the determination of the length of the lake influence on ice thickness in the growth season but gives ice season. The net longwave radiation dominates the large impact on the ice thickness in late winter and in the surface heat balance in the winter season, while the net melting season. solar radiation is critical for melting. The surface temperature and the thickness of ice are closely related. The model simulations showed that inver- sion of the surface temperature for ice thickness by thermal 5. Conclusions remote sensing is feasible for thin ice (B 0.5 m), but then the timing of measurement is critical and the full heat The ice coverin Lake Vanajavesi has been investigated with conduction equation needs to be used. The model simula- a one-dimensional thermodynamic model HIGHTSI. tions showed that in NWP and climate modelling, inclusion A case study was performed for one ice season, and a of a proper lake ice model would produce a better surface 30-yearperiod (1971 2000) was examined forice climatol- temperature and consequently improve the quality of the ogy. The results were good as compared with the validation predictions. data. The model was then applied to examine the response Climate change may have a dramatic influence on the of the lake ice season to climate change. The novel features course of the ice season. The annual maximum ice thickness of the HIGHSTI forlake investigations are the advanced was 53 cm in Lake Vanajavesi in 19712000, about 11 cm treatments of superimposed ice, stability-dependent turbu- more than in winter 20082009. The corresponding mean lent heat fluxes and coupling of snow and ice layers. air temperature for 19712000 was about 1.58C lowerthan The modelled snow thickness was produced from the temperature in winter 20082009. Ourmodel experi- measured precipitation and a simple model for snow ments, based on changing the air temperature level in the metamorphosis. This made snow a free model variable. climatological simulations, showed that the air temperature Snow accumulates during cold periods, while it may shifts affect the freezing date by 5 d 8C and the breakup decrease due to transformation into snowice ormelting. date by 8 d 8C . In the case study ice season, snow thickness and snowice Future studies are needed to couple the present model formation were reproduced well. But due to lack of long- with a lake waterbody model to predict better the autumn term data on ice stratigraphy, the modelled snowice cooling of the waterand the consequent freezing date. The cannot be validated quantitatively forthe climatological simulation. Ourcalculations suggested that on an average timing of the onset of snow and ice melting in the model is 16% of the snowfall contributed to the total ice thickness the most critical point for further improvements. (53 cm) with a 10-cm snowice layer. The model experiments produced an estimate for the 6. Acknowledgements heat flux from water to ice by tuning. The result was a This research was supported by YMPANA (Development small value, 0.5 W m , corresponding to the magnitude of of automatic monitoring system for Lake Vanajavesi) molecularheat conduction. The annual maximum ice project, National Natural Science Foundation of China thickness could change up to 10 cm in response to a (Grant No. 51079021, 40930848 and 50879008) and the change of the level of the heat flux to 5 W m . The ice breakup date would be affected by 5 d. The air temperature Norwegian Research Council (Grant No. 193592/S30). The NUMERICAL MODELLING OF SNOW AND ICE THICKNESSES 11 Jaatinen, E., Leppa¨ ranta, M., Erm, A., Lei, R., Pa¨ rn, O. and CIMO scholarship from Finnish Ministry of Education is co-authors. 2010. Light conditions and ice cover structure in acknowledged for funding the first author to work in Lake Vanajavesi. In: Proceedings of the 20th IAHR Ice Finland. Comments from the anonymous reviewers were Symposium, Paper #79. Department of Physics, University of very helpful to improve this paper. We are very grateful to Helsinki, Finland. Dr. Laura Rontu and Dr. Timo Vihma for providing Jacobs, J. D. 1978. Radiation climate of Broughton Island. In: various inspiring comments and suggestions during the Energy Budget Studies in Relation to Fast-ice Breakup Processes preparation of this manuscript. in Davis Strait, Occasional Paper (eds. R. G. Barry and J. D. Jacobs). Institute of Arctic and Alpine Research, University of Colorado, Boulder, 26, pp. 105120. References Jakkila, J., Lepparanta, M., Kawamura, T., Shirasawa, K. and Anderson, E. 1976. A point energy and mass balance model for a Salonen, K. 2009. Radiation transfer and heat budget snow cover. NOAA Technical Report, NWS 19, 150 pp. during the melting season in Lake Pa¨ a¨ ja¨ rvi. Aquat. Ecol. 43, Arst, H., Erm, A., Leppa¨ ranta, M. and Reinart, A. 2006. 681692. Radiative characteristics of ice-covered freshwater and brackish Ka¨ rka¨ s, E. 2000. The ice season of Lake Pa¨ a¨ ja¨ rvi in southern waterbodies. Proc. Estonian Acad. Sci. Geol. 55,323. Finland. Geophysica 36,8594. Bennett, T. J. 1982. A coupled atmosphere-sea-ice model study of Korhonen, J. 2005 Suomen vesisto¨ jen ja¨ a¨ olot [Ice Conditions in the role of sea ice in climatic predictability. J. Atmos. Sci. 39, Lakes and Reviers in Finland]. Suomen ympa¨ risto¨ 751. Finnish 14561465. Environment Institute. (in Finnish with English summary). Brown, L. C. and Duguay, C. R. 2010. The response and role of ice Launiainen, J. and Cheng, B. 1995. A simple non-iterative coverin lake-climate interactions. Prog. Phys. Geogr. 34, 671 algorithm for calculating turbulent bulk fluxes in diabatic 704. conditions over water, snow/ice and ground surface. Report Burda, N. Y. 1999. Computer simulation for Ladoga Lake ice Series in Geophysics, Department of Geophysics, University of dynamics based on remotely sensed data. Int. Geosci. Remote Helsinki, 33 pp. Sensing Symp. 2, 937939. Launiainen, J. and Cheng, B. 1998. Modelling of ice thermo- Cheng, B. 2002. On modeling of sea ice thermodynamics and air- dynamics in nature water bodies. Cold Reg. Sci. Technol. 27, ice coupling in the Bohai Sea and the Baltic Sea. Finnish Institute 153178. of Marine Research  Contributions 5, 38 pp. (Ph.D thesis) Lei, R., Lepparanta, M., Erm, A., Jaatinen, E. and Parn, O. 2011. ¨ ¨ Cheng, B., Vihma, T. and Launiainen, J. 2003. Modelling of the Field investigations of apparent optical properties of ice cover in superimposed ice formation and sub-surface melting in the Finnish and Estonian lakes in winter2009. Est. J. Earth Sci. 60, Baltic Sea. Geophysica 39,3150. 5064. Cheng, B., Zhang, Z., Vihma, T., Johansson, M., Bian, L. and Leppa¨ ranta, M. 1983. A growth model for black ice, snow ice and co-authors. 2008. Model experiments on snow and ice thermo- snow thickness in subarctic basins. Nordic Hydrol. 14,5970. dynamics in the Arctic Ocean with CHINAREN 2003 data. Leppa¨ ranta, M. 1993. A review of analytical models of sea-ice J. Geophys. Res. 113, C09020, doi: 10.1029/2007JC004654. growth. Atmos. Ocean. 31, 123138. Drebs, A., Nordlund, A., Karlsson, P., Helminen, J. and Lepparanta, M. 2009. Modelling the formation and decay Rissanen, P. 2002. Climatological Statistics of Finland 1971 of lake ice. In: Climate change impact on European lakes 2000. Finnish Meteorological Institute, Helsinki. (ed. G. George). Springer, Berlin. Aquatic Ecology Series 4, Duguay, C. R., Flato, G. M., Jeffries, M. O., Menard, P., ´ pp. 6383. Morris, K. and co-authors. 2003. Ice-cover variability on Leppa¨ ranta, M. and Kosloff, P. 2000. The thickness and structure shallow lakes at high latitudes: model simulations and of Lake Pa¨ a¨ ja¨ rvi ice. Geophysica 36, 233248. observations. Hydrol. Proces. 17, 34643483. Leppa¨ ranta, M. and Lewis, J. E. 2007. Observations of ice surface Eerola, K., Rontu, L., Kourzeneva, E. and Shcherbak, E. 2010. temperature and thickness in the Baltic Sea. Int. J. Remote Sens. A study on effects of lake temperature and ice cover in 28, 39633977. HIRLAM. Boreal Env. Res. 15, 130142. Lepparanta, M. and Uusikivi, J. 2002. The annual cycle of the Efimova, N. A. 1961. On methods of calculating monthly values of Lake Paajarvi ice. Lammi Notes 29. Lammi Biological Station, ¨ ¨ ¨ net longwave radiation. Meteor. Gidrol. 10,2833. University of Helsinki. Ellis, A. W. and Johnson, J. J. 2004. Hydroclimatic analysis of Liston, G. E. and Hall, D. K. 1995. An energy-balance model of snowfall trends associated with the North American Great lake-ice evolution. J. Glaciol. 10, 373382. Lakes. J. Hydrol. 5, 471486. Livingstone, D. M. 1993. Lake oxygenation: application of a Flato, G. M. and Brown, R. D. 1996. Variablility and climate one-box model with ice cover. Int. Rev. Ges. Hydrol. 78, sensitivity of landfast Arctic sea ice. J. Geophys. Res. 101, 465480. 2576725777. Maykut, G. A. and Untersteiner, N. 1971. Some results from a Grenfell, T. C. and Maykut, G. A. 1977. The optical properties of time-dependent thermodynamic model of sea ice. J. Geophys. ice and snow in the Arctic Basin. J. Glaciol. 18, 445463. Res. 76, 15501575. Heron, R. and Woo, M. K. 1994. Decay of a High Arctic lake-ice Makynen, M., Simila, M., Cheng, B., Laine, V. and Karvonen, J. ¨ ¨ cover: observations and modelling. J. Glaciol. 40, 283292. 2010. Sea ice thickness retrieval in the Baltic Sea using MODIS 12 Y. YANG ET AL. and SAR data. In: Proceeding of SeaSAR2010, Frascati, Italy, Salgado, R. and LE Moigne, P. 2010. Coupling of the Flake model ESA SP-679. to the Surfex externalized surface model. Boreal Env. Res. 15, Mironov, D., Heise, E., Kourzeneva, E., Ritter, B., Schneider, N. 231244. and co-authors. 2010. Implementation of the lake parameterisa- Saloranta, T. 2000. Modeling the evolution of snow, snow ice and tion scheme Flake into the numerical weather prediction model ice in the Baltic Sea. Tellus 52A,93108. COSMO. Boreal Env. Res. 15, 218230. Shine, K. P. 1984. Parameterization of shortwave flux over high Nonaka, T., Matsunaga, T. and Hoyano, A. 2007. Estimating ice albedo surfaces as a function of cloud thickness and surface breakup dates on Eurasian lakes using water temperature trends albedo. Quart. J. Roy. Meteor. Soc. 110, 747764. and threshold surface temperatures derived from MODIS data. Shirasawa, K., Leppa¨ ranta, M., Kawamura, T., Ishikawa, M. and Int. J. Remote Sens. 28, 21632179. Takatsuka, T. 2006. Measurements and modelling of the water Palecki, M. A. and Barry, R. G. 1986. Freeze-up and break-up of ice heat flux in natural waters. In: Proceedings of the 18th IAHR lakes as an index of temperature changes during the transition International Symposium on Ice, Hokkaido University, Sapporo, seasons: a case study forFinland. J. Appl. Meteorol. Climatol. Japan, pp. 8591. 25, 893902. Shirasawa, K., Leppa¨ ranta, M., Saloranta, T., Kawamura, T., Patterson, J. C. and Hamblin, P. F. 1988. Thermal simulation of a Polomoshnov, A. and co-authors. 2005. The thickness of coastal lake with winterice cover. Limnol. Oceanogr. 33, 323338. fast ice in the Sea of Okhotsk. Cold Reg. Sci. Technol. 42,2540. Pirazzini, R., Vihma, T., Granskog, M. A. and Cheng, B. 2006. Vihma, T. 1995. Subgrid parameterization of surface heat and Surface albedo measurements over sea ice in the Baltic Sea momentum fluxes overpolaroceans. J. Geophys. Res. 100, during the spring snowmelt period. Ann. Glaciol. 44,714. 2262522646. Rothrock, D. A., Yu, Y. and Maykut, G. A. 1999. Thinning of the Wake, A. and Rumer, R. R. 1983. Great lakes ice dynamics Arctic sea-ice cover. Geophys. Res. Lett. 26, 34693472. simulation. J. Waterway Port Coast. Ocean Eng. 109,86102. Rouse, W. R., Blanken, P. D., Duguay, C. R., Oswald, C. J. and Wang, X., Key, J. R. and Liu, Y. 2010. A thermodynamic model Schertzer, W. M. 2008b. Climatelake interations. In: Cold forestimating sea and lake ice thickness with optical satellite Region Atmospheric and Hydrologic Studies: The Mackenzie data. J. Geophys. Res. 115, C12035, doi: 10.1029/2009JC005857. GEWEX Experience, (ed. MK. Woo). Vol. 2, Springer, Berlin, Wendler, G. and Eaton, F. 1990. Surface radiation budget at pp. 139160. Barrow, Alaska. Theor. Appl. Climatol. 41, 107115. Rouse, W. R., Binyamin, J., Blanken, P. D., Bussie` res, N., Duguay Williams, G., Layman, K. L. and Stefan, H. G. 2004. Dependence C. R. and co-authors. 2008a. The influence of lakes on the of lake ice covers on climatic, geographic and bathymetric regional energy and water balance of the central Mackenzie. In: variables. Cold Reg. Sci. Technol. 40, 145164. Cold Region Atmospheric and Hydrologic Studies: The Mack- enzie GEWEX Experience, (ed. MK. Woo). Vol. 1, Springer, Berlin, pp. 309325.

Journal

Tellus A: Dynamic Meteorology and OceanographyTaylor & Francis

Published: Dec 1, 2012

Keywords: thermodynamic model; sensitivity tests; snow; ice; Lake Vanajavesi

There are no references for this article.