Temporal Variability of Alpine Solifluction: a modelling approach

Abstract. Holocene periods of enhanced solifluction offer new paleoclimatic information. Long-term observations of present solifluction variability and process studies on movement mechanisms, as well as model simulations of the soil heat and water regimes, show the dependence of solifluction on ground freezing. The annual variability of both processes is strongly controlled by weather and resulting snow conditions immediately before and at the beginning of the winter frost period. Simulated long-term variations during different paleoclimatic scenarios are regulated by both mean preeipitation and temperature changes. Quantitative reconstruetion of Holocene preeipitation during maximum altitudinal depression of solifluction is shown.


Introduction
Solifluction, the slow downslope movement of thawing soils, is a widespread phenomenon in the alpine eco¬ tone of high mountain areas (Photo 1). It results from the combined, but variable, interaction of frost creep and gelifluction mechanisms (French 1996;Washburn 1979). In the Alps, the distribution of landforms gener¬ ated by solifluction has been documented by, among others,FuRRER (1954others,FuRRER ( , 1965,Höllermann (1967,1977) and Stingl (1969).The variable altitudinal distribution of solifluction lobes or sheets, either across a Single mountain ränge or between different regions, early led to the concept of a climatically determined «periglacial altitudinal belt» (Büdel 1937;Troll 1944).
Dating of relict solifluction landforms shows, that sev¬ eral periods of enhanced movement occurred during the Holocene (cf. Steinmann 1978;Gamper 1982Gamper ,1985Veit 1989Veit ,1993 their work in the Alps; Matthews et al. 1993 for a review of studies in Europe ;Smith 1993 for North American approaches). Such «solifluc¬ tion phases» frequently did not occur simultaneously with periods of glacial advance or timberline depression. It seems solifluction responds to climatic changes in a different way than glaciers or the timberline. Hence, an analysis of the processes and relevant Controlling factors involved in solifluction may lead to the formulation of an approach suitable for the identification and quantification of Holocene climatic fluctuations.
There is a widespread agreement that enhanced soli¬ fluction in the past reflects climatic cooling, although knowledge about the precise nature of the dependence of solifluction on climate is still limited (Matthews et al. 1993). As far as we know, there are only three longterm studies investigating the present nature of this relation in non-permafrosl, mid-latitude areas -two in the Swiss Alps (Gamper 1981(Gamper , 1987Krummenacher et al. 1998), one in the Austrian Alps (Veit 1988;Veit & Höfner 1993;Veit et al. 1995). They suggest that solifluction intensily is strongly related to the annual depth of soil freezing. The studies indicate the rel¬ evance of the insulating snow cover in terms of its autumnal onset as well as total duration. Early snow cover may delay or even prevent ground freezing and thus reduce soil movements. At the Swiss site, low winter temperatures enhance frost penetration. At the Austrian site, low summer temperatures additionally induce higher soil movement rates. The expla¬ nation offered is that the lower temperatures delay ground thawing prolonging the solifluction period. Findings of such long-term field studies are supported by process studies of solifluction in large-scale laboratory simulations, showing the importance of ground freezing in promoting soil movements (Harris 1996). Measurements by Matsuoka et al. (1997) in the Swiss Alps also underline the dependence of frost-creep type soil movements on the thawing of initially frozen ground.
To clarify the close relation between snow cover, soil freezing and solifluction intensity, a follow-up study was carried out at the Austrian site in 1995 (Jaesche & Huwe 1997;. A soil physical approach was chosen, water and heat regimes as well as frost heave and subsurface soil movement being monitored on two solifluction lobes. The importance of ground freezing for triggering solifluction move¬ ments could be confirmed. The strongest movements generally occur during a period of a few days when the ground ice is melting. They attenuate but continue for several weeks as long as lateral water supply from higher snow patches persists (Jaesche 1999).
In the study area, present solifluction occurs above 2600 m a.s.l., and leads to the formation of small, mainly nonvegetated solifluction lobes. Densely vegetated and obviously inactive lobes may be found as low as 2300 m a.s.l., representing several Holocene phases of intensified solifluction (Veit 1989(Veit ,1993.This strong 300 m depression of the periglacial belt may be explained in different ways. It might be caused by a marked decrease of mean annual air tempera¬ ture by 2°C can reach a length of more than 100 m, with faces about 0.5-2.0 m high. The lower distribution limit of solifluction lobes is at 2300 m a.s.l. (Stingl 1969).
Radiocarbon dating of fossil soils buried by solifluc¬ tion determined that the big lobes originated mainly during two activity periods, 3350-2800 l4C yr B.P. and around 1250 NC yr B.P.. following an Earlyto Mid-Holocenc period with stable slopes and soil evolution (Veit 1993).
Solifluction processes are observed on a monitoring plot at about 2640-2670 m a.s.l. (Stingl & Veit 1998;Veit et al. 1995). presumably at their present lower limit of significant activity. The site, with inclinations between 10-20°, faces east to north and is heavily shaded by a steep mountain ridge, especially during winter. The mean annual temperature is about -2°C , preeipitation reaches approx. 1200 mm a"'. Observa¬ tions were initiated in 1985 and have resulted in a 16-year data set of annual surface movement so far. Movement rates are obtained from regulär taehymetric measurements at 7 surface markers. consisting of 20-cm-long PVC rods inserted half-way into the ground. Spatial distribution and variability of soli¬ fluction rates at the site were also observed. using at least 90 markers, showing persistent movement pat¬ terns at the site during subsequent years (Jaesche 1999 and 120 cm below the surface. and registered wilh a field data logger (DeltaT DL-2). Water content was registered at 6 depths between 10 and 90 cm using an automated time domain reflectometry (TDR) System (IMKOTRIME-MUX6 with P2Z-probes). During the first winter, water potential was also measured using tensiometers with prcssure-lransduccrs. As tempera¬ ture gradients tend to be minimal during winter and frost depths difficult to determine by the interpolation of temperature data.TDR measurements were chosen to indicate the transition of the soil Status from unfrozen to frozen. A comparison of TDR and thermistor data showed a generally good correlation concerning the onset of soil freezing. Daily weather data was delivered from the Sonnblick Observalory (3105 m). situated only 10 km east of the site. across the Moll valley. Information about the snow cover was obtained from periodical field observations and from calculations described below.
2.2 Model description and applications TTie model SOIL (Jansson 1998) decribes the coupled heat and water transport between atmosphere. snow cover and the soil in the vertical dimension. calculations being based on diurnal preeipitation and temper¬ ature data. SOIL explicitly considers soil freezing and subsequent changes in thermal and hydraulic charac¬ teristics of the soil. It aecounts for evaporation. sur¬ face runoff and subsurface downslope water flow in the saturated soil. as well as Infiltration into the frozen soil through previously air-filled pores. Simulated heat flow takes conduetion. convection and energy conser¬ vation into consideration. Unsaturated water trans¬ port is simulated in terms of the Richards equation. A detailed technical description is found in Jansson (1994. 1998) and. like the model itself. may also be found on the internet (http:// bgfserver.mv.slu.se/ bgfl soil.htm). Recent model applications are given by Sta¬ dler, Flühler & Jansson (1997) or Stähli, Jansson Preeipitation was assumed to occur as rain above an air temperature of 0°C , as pure snow below -1°C, and as a variable mixture at air temperatures between.The snowpack is generally treated as a homogeneous layer, with the thermal conduetivity being a function of its density (cf. Langham 1981). Snowpack density itself is modified according to the length of time since the last snowfall. the total mass of the snowpack and the liquid water content (Jansson 1998). Where a snow¬ pack is present, soil surface temperature is calculated assuming steady-state heat flow according to the ther¬ mal gradient and thermal conduetivity of the snow¬ pack and of the topsoil. If the snow liquid water con¬ tent exceeds a certain threshold value, soil surface temperature is set at 0°C as snowmelt conditions are assumed. The surface temperature of both the snow¬ pack and the snow-free soil were assumed to equal air temperature, although the model also includes the possibility of calculating surface temperatures using energy balance approaches. Snow melt was simulated applying a pure degree-day method, using a daily melt rate of 6.1 mm d ' K ', a value taken from independent studies (Jaesche, Bienert & Huwe 1998).
The hydraulic characteristics of the soils were described by an adapted Brooks and Corey parameterisation suitable for soil core measurement. Soil freezing char¬ acteristics were deduced from TDR field data. For sim-ulations, a 5 m deep soil profile was divided randomly into 11 depth increments between 7.5 to 20 cm thick in the upper metre, the thicknesses of the sections pro¬ gressive^increasing in the lower 4 m. A constant tem¬ perature of 2°C was chosen as the lower boundary temperature.
Model calculations were fed with daily temperature and preeipitation data from the Sonnblick meteoro¬ logical Station, the data adjusted to take the 450 m elevation difference into consideration by applying values of regional elevation gradients found in litera¬ ture. Temperature was modified using a constant lapse rate of 6.5-10'°C m1 (cf. Buchenauer 1990). Daily preeipitation rates were reduced by 30%, Coming close to the mean annual altitudinal lapse rate of about 100 mm per 100 m (Auer 1992a For each temperature level. daily preeipitation was additionally scaled by factors of 0.5. 0.75. 1. 1.25 and 1.5 to obtain drier or wetter scenarios. This resulted in 25 scenarios characterised by mean preeipitation. tem¬ perature and resulting frost depth. the extremes comprising a very warm and wet climate on the one hand, a rather cold and dry climate on the other hand. This approach neglects possible shifts in the seasonal distri¬ bution of Single variables, but it allows the examina¬ tion of the effects of climate change with regards to the annual variability of weather data, as can be seen in the 13 annual weather scenario data sets. 3 Results and discussion 3.1 Annual variability of solifluction rates and external control The 13-year record of mean solifluction movements shows a strong interannual variability (Fig. 2) of high movement rates, including the maximum value in winter 1988/89, is followed by several lower rates and almost complete inactivity in the 1992/93 period. There is no indication of site disturbance due to marker installation, as was observed in a review of other stud¬ ies on solifluction (Smith 1992). The 7 surface mark¬ ers for long-term observations are in relatively stable regions, while markers used during the two intensive monitoring periods (1985-1989 and 1994-1998) were preferentially positioned in mobile regions. Different inter-annual changes occurred within the two marker groups, e.g. a solifluction decrease was observed between 1996/97 and the next season in the one group, whereas an increase was observed during the same period in the other group. The differing mechanisms of solifluction that might act within these two groups will be discussed in detail elsewhere. Both results indi¬ cate the inherent uncertainties present when dealing with mean solifluction rates, for example in compari¬ son with climatic variables as shown below.
Analogous to observations made by Veit et al. (1995) with a shorter (7-year) data set, the greatest move¬ ments. like those in the 1988/89 period, followed a rather warm winter, while the smallest movements occurred after a heavy, yet not unusually cold winter (Fig. 3). Regression analysis of the 13-year data (mean and median values) with various climate indices such as mean monthly or seasonal (e.g. winter) tempera¬ tures, generally confirmed the results of Veit et al. (1995): There is no significant relation between sea¬ sonal or annual indices and movement; rather, seasonal weather characteristics, especially during autumn, influ¬ ence solifluction. Accordingly, a significant correlation of annual movement rates with preeipitation totals in October is observed (r=-0.8, p<0.01). Only a weak correlation (p<0.05) is observed between preeipita¬ tion totals during October and November (r=-0.6) and mean temperature in October (r=0.6:mean movement only). At first glance, these results seem to oppose soil freezing. But they indicate a stable weather Situation The correlation analysis further shows that cold Sum¬ mers affect slope movements, as low June, June/July and June through August mean temperatures correlate negatively (r=-0.6, p<0.05) with solifluction rates. This is seen as an indication of prolonged snowmelt periods supporting solifluction by lateral water supply (Jaesche 1999), rather than the hindrance of ground ice melt as proposed by Veit et al. (1995). snow depth (as reconstructed by model simulations) do not correlate with solifluction intensity.
3.2 Annual snow cover and frost depth: measurements and model calibration As is to be expected when observing solifluction var¬ iability and taking initial temperature measurements (Veit et al. 1995), snow and soil heat regimes were highly variable during the 1995-98 study period (Fig.   4). A surprising Observation is that ground freezing continues or even commences despite snow depths of one meter or more. This is probably caused by the complete shading of the site from November through the middle of February and strong radiation cooling at the snow surface.
The importance of snow depth, especially at the begin¬ ning of cold winter conditions, is confirmed by model run-throughs on snowfall intensity: A fourfold increase of daily preeipitation during deposition of the first 50 cm of snow (completed within 2 to 10 days) led to a decrease in annual frost depth of up to 50% (Jaesche 1999). Therefore, in order to obtain reasonable values for frost depth and solifluction, it is necessary to con¬ sider the precise seasonal course of air temperature, preeipitation and snow cover, as will be demonstrated below. Total amounts of winter snowfall or maximum Snow and frost regimes were simulated using the SOIL model (Fig. 4). Calibration of the snow sub-model with field observations meant that an additional snow input of 25% was made necessary. This is explained by drifting snow accumulation into the coneave-shaped, east (leeward) exposed site. Melt rates during the 3-month winter period of complete shading had to be reduced to 10% (cf. Jaesche 1999). Observed ablation dates at the instrumented site were met within two days values. especially during winter 1996/97. density-related Parameters were modified to describe observed snow depths and soil surface temperatures. Measured saturated water conduetivity had to be increased by a factor 101 to reflect tension changes during preeipitation-free periods. In turn, upward water flow towards the freezing front was reduced. This was necessary to prevent the delay of the frost front progression and the thus reduced total frost depth. Once the calibra¬ tion had been carried through. model results of soil temperature and frost depth during all three winter periods were in good agreement with field measure¬ ments (Fig. 4).
3.3 Long-term (13-year) variability of frost depth and solifluction activity The successful calibration of the SOIL model pro¬ vides a reasonable base for ground freezing simulations during the whole 13-year period (Fig. 5).
Again. a strong annual variability of snow cover and frost depths may be observed. The calibration period 1995-98 almost covered the füll ränge of frost depths simulated. strengthening the confidence in the validity of this exercise. While the regression results of climatic variables and solifluction rates shown above could only be interpreted assuming annual variations of ground freezing. the Simulation results presented here confirm the rela¬ tion between annual frost depth and solifluction (Fig.  6). Deep freezing of the ground promotes soil move¬ ments. Unexplained variability of solifluction mainly arises from other factors influencing soil mobility. such as soil moisture content influenced by summer snow¬ melt. It must be noted that the given linear regression is valid only for the measured ränge of frost depths. It seems possible that an exponential function would better describe the relation. especially at low frost depths Fig.  6 raises the question of whether a minimum freezing intensity is necessary for solifluction. As independent observations at individual solifluction lobes have shown. no movements occur at zero seasonal frost depth.
Under present climatic conditions. characterised by a mean annual air temperature of -2.1 CC and a mean annual preeipitation of 1120 mm. the mean annual frost depth during the 13-year period reaches about 70 cm (Fig. 5). Continual but annually variable solif¬ luction at the investigated site. leads to the formation of lobate surface features very similar to the relict fea¬ tures found at lower elevations. The reason that the relict features are found up to a maximum of 300 m below present solifluction activity cannot be explained alone by minor Holocene temperature changes of 1°C . as indicated by other proxies. The model described above offers a tool for examining freezing activity and hence solifluction according to changes in mean tem¬ perature or preeipitation. taking into account their annual variability.
3.4 Frost depth under changed climate scenarios Similarly to the model results described in the preceding section. Figure 7 shows simulated snow and result¬ ing frost depths for 13 winter periods. however with the focus on a climate change scenario moving towards colder and drier conditions. This is achieved by running the model with modified input data: daily temper¬ atures were reduced by 1.3 C. daily preeipitation rates by 50%. The changes result in drastically enhanced ground freezing. often reaching a depth of two metres.  19861988199019921994: Simulated snow cover and ground freezing under cold/dry climatic conditions. based on modified [1985][1986][1987][1988][1989][1990][1991][1992][1993][1994][1995][1996][1997][1998] daily weather data (air temperature reduction AT -1.3°C. preeipitation reduction AP -50%). Present conditions are also given (cf. Fig. 5).
Draft by the authors. diagram: L. Baumann changes applied and shows the resulting 13-year mean frost depths. The heaviest freezing occurred in the cold/dry scenario presented above (Fig. 7); mean frost depth under present conditions (Fig. 5) is to be found in the centre ofthe figure. Though individual years had shown different reactions to changed input data, there is a significant increase of long-term mean frost depth with decreasing temperature or preeipitation. Con¬ trary changes in climate variables, though. may bal¬ ance each other to leave mean frost depth unchanged. This relation is approximated by a bivariate linear regression model. applied to the absolute values of mean temperature and preeipitation, as warmer than today (7"= -1.1°C ). Assuming that the solifluction process was characterised by the same medium frost depth observed today (F -74 cm), graphic illustration (Fig. 8) (Veit 1993). By the fact that the glacier equilibrium lines show minor var¬ iations (±100 m), reduced preeipitation is indicated. Furthermore, it is possible that the calculated 325 mm preeipitation reduction for the Holocene period may be partially explained by the effects of the vertical preeipitation gradient.

Conclusions
Field observations of past and present alpine solifluc¬ tion, in combination with intensive monitoring and the Simulation of snow depth and soil heat regimes al a site affected by solifluction. represents a new and promising approach towards reconstrueting Holocene climate. Although annual movement rates are highly variable and strongly controlled by both the weather conditions during autumn and snow cover Variation, long-term solifluction activity clearly responds lo cli¬ matic change. Solifluction activity is closely linked to the seasonal frost depth. The proposed relation between mean solifluction. temperature and preeipita¬ tion at a research site in the eastern Alps may be used to explain variations between fluctuations of solifluc¬ tion and snow or timber line, or even allow quantification of preeipitation during Holocene solifluction phases. However. it is necessary to deduce mean tem¬ perature data from other sources.
Based on this general approach, future work will have to focus on the calculation of Holocene solifluction rates and the detection of altitudinal variations. The monitoring of soil heat regimes and solifluction on slopes with different exposition, as well as research of solifluction in various climatic regions is necessary to expand the applicability of the model. Wesentlichen durch die Witterungsbedingungen und die Schneebedeckung unmittelbar vor und zu Beginn der winterlichen Dauerfrostperiode gesteuert. Simulierte längerfristige Abweichungen der Bodenfrostmächtig¬ keit unter verschiedenen paläoklimatischen Szenarien werden gleichermaßen von Änderungen im Tempera¬ turals auch im Niederschlagsregime beeinflußt. Als Anwendungsbeispiel der Modellrechnungen wird die Quantifizierung holozäner Niederschläge bei maxima¬ ler Höhendepression der Solifluktion vorgestellt.