Soil frost and snow metamorphism simulations for the BALTEX-region with a complex hydro-thermodynamic soil-vegetation scheme

EURAD, Förderverein des Rheinischen Institutes für Umweltforschung an der Universität zu Köln e.V., Aachener Straße 201-209, 50931 Köln, Tel. +49 221 400 2220 Fax +49 221 400 2320, email: ,

Geophysical Institute of the University of Alaska at Fairbanks (UAF), 903 Koyukuk Drive, Fairbanks, AK 99775-7320, United States of America, phone: + 1 907 474 7910, fax: + 1 907 474 7290, e-mail: nicole.molders@sgi.alaska.edu,
BOBA page at UAF

The inclusion of soil frost and snow metamorphism processes is an urgent need to adequately determine the water and energy cycles of the BALTEX region in long-term studies and regional climate research. In addition, to calculate the hydrological cycle methods must be devised which allow to initialize the soil moisture and temperature fields adequately.
Building on largely extended surface and space based observational data and enhanced computational resources, BOBA seeks to achieve this goal by a combination of two novel approaches: (i) the available SVAT scheme HTSVS is to be enlarged by parameterizations to consider the processes of soil frost and snow metamorphism, and (ii) advanced data assimilation techniques are to be developed for HTSVS to allow an optimized choice of the vertical resolution of the soil and to initialize soil moisture and soil temperatures. The specific project aim is to improve the calculation of the energy and water fluxes of BALTEX-region and to improve a transferability of the developed model package for use in other regions of the earth.



Project Objectives

Perma-frost exists on a third of the earth's continents. Furthermore, large parts of the continents are regularly frozen during winter. Soil frost leads to the freezing of soil-water and restricts the mobility of soil water. Capillarity, infiltration as well as percolation are only slightly effective. The thermic stability, low air temperatures, and the consequently low saturation pressure of water vapor lead to low evaporation. Thus, moisture will be stored in frozen soil and may increase spring peak flood events. In addition, in winter, of course, transpiration plays a minor role because deciduous forests already lost their leaves. Moreover, stomatal con ductivity of coniferous forests is low in winter.
Obviously, if the freezing processes of soils are not considered in numerical modeling, too high water vapor fluxes will be predicted as there is seemingly still liquid water available that, moreover, requires less energy for evaporation than ice. Since soil frost hinders infiltration of water into the soil, rain falling onto frozen soil or melting of a snow package laying over frozen soil contribute to runoff. Therefore, a prediction of soil frost is an urgent need for cal culating runoff adequately.
The boundary between an unfrozen upper soil layer and a frozen deeper soil layer, for instance, may vary within the diurnal course. The determination of surface water and energy fluxes is extremely difficult when the exact depth of the freezing line is unknown. Unfortunately, in regional climate models as well as in numerical weather prediction, soils cannot be resolved so fine as to be able to exactly determine the depth of soil frost if acceptable simulation times are desired. Therefore, it is an urgent need to examine how coarse the vertical resolution within soil may be without achieving too large errors in the accuracy of predicted water and energy fluxes.
Another important process to be considered in simulating climate, runoff, water and energy fluxes is snowmelt and previous snow accumulation. Thus, a physically adequate formulation of snow metamorphism processes (accumulation, ablation) is required to guarantee the transferability of the model to other regions. Snow has an insulating effect. Thus, the cooling of snow-covered soils will be predicted too high if snow coverage is neglected. In addition to the water budget, snow and snowmelt also affect the energy budget, among other things, due to the change in albedo.
The energy and water fluxes, infiltration, runoff, and near surface meteorological conditions strongly depend on the distributions of soil moisture and soil temperatures that have to be initialized reasonably. Unfortunately, no high resolution network of these quantities exists so that these quantities could be initialized without special methods. Therefore, it is required to develop reasonable methods to provide three- dimensional initial distributions on the basis of only a few measurements.

a. Development of a soil frost and snow metamorphosis scheme
The contribution of University of Alaska
A comprehensive extension of the existing hydro-thermodynamic soil-vegetation model HTSVS is under way at the University of Fairbanks. Key issues are the inclusion of a soil frost module and a snow metamorphism module. (visit the UAF-BOBA page for detailled description).

b. Spatio-temporal inversion as optimization technique
The contribution of RIU at the University of Cologne
Spatio-temporal data assimilation problems are part of inverse modeling issues. The mathematical background is provided by both estimation or filtering theory and control theory. In either case, a prognostic model is a key component of the algorithm. In the case of sensible causal links between observed quantities and specific model control parameters, the solution of parameter optimization problems are admissible by advanced data assimilation algorithm. In meteorological practice, the initial values of the prognostic model variables are of prime interest. It is the four-dimensional variational data assimilation algorithm (4D-var) and the Kalman filter, which are theoretically well suited to address the optimization problem. Talagrand and Courtier (1987) and Courtier and Talagrand (1987) provided highly influential studies, which impressively demonstrated the power of the variational calculus. Numerous follow up studies and research efforts performed at leading weather centers show the widespread interest in this method. Due to its extraordinarily high computational demands, the Kalman filter application has been mostly restricted to reduced experimental methods. The general 4D-variational scheme is outline in the Figure.
Given suitable observations, control of feedback processes between a land surface model and an atmospheric model can be treated as an optimization problem in the framework of advanced space-time data assimilation methods. According to the nature of the problem, initial values of soil temperature and soil moisture can be inferred or parameter identification of unknown local parameter values can be addressed. The impact was shown to be dependent on the influence of large-scale forcing. Despite the large scope of the variational calculus, studies on the optimization of initial values of soil moisture and soil temperatures are limited.

The working plan is orientated at the phases (i) model development and (ii) optimized budget calculations. The important components for quality assurance and quality control are the development and application of the soil frost and snow metamorphism modules and the objective (numerical) analysis of soil temperatures and soil moisture to improve the determination of fluxes by means of variational boundary layer and soil data assimilation. Further, the influence of different quantities is to be determined by calculation of backward and forward sensitivities.
During the time of development the work to be performed includes the development of tangential-linear and adjoint components as well as the implementation of the complete assimilation algorithm.
Soil frost processes (freezing, thawing) will be treated similarly as described in Cherkauer and Lettenmaier (1999) to avoid inconsistency to the recent state of HTSVS. The parameterization of snow metamorphism will consider accumulation, ablation, the change in snow density and albedo. The snow mass balance and snow fraction are to be treated similarly to Yang et al. (1999).
The relationships and interactions between the simulated processes, initial values of soil moisture and soil temperatures as well as the vertical resolution of the grid within soil are important for the quality of the simulated energy and water fluxes at the interface "earth-atmosphere". Since it is impossible to resolve the soil infinitely fine and to achieve acceptable simulation times at the same time, there is an urgent need to investigate how coarse a resolution can be without obtaining unacceptably great errors. Such investigations are to be carried out in a similar way as done by Elbern and Schmidt (1999). The sensitivity analysis preformed with the adjoint model will provide valuable clues on which parameters play an important role for various frost depth. Thus, an optimal configuration of the vertical grid within the soil may be determined that, in combination with the mesoscale meteorological model MM5, allows a good quality of the simulation in an acceptable simulation time.
In the optimization procedure, the respective measured quantity is to be the starting point. As pointed out above, adjoint and tangent-linear versions of MM5 exist (Zou et al. 1997, 1998). MM5 and its tangent-linear and adjoint components serve as a driver and a test platform. Herein, the planned tangent-linear and adjoint versions of the HTSVS enlarged by soil frost and snow metamorphism processes are to be applied. Observations from NOPEX and LITFASS, snow coverage, surface moisture and surface temperature determined from satellite data (e.g., from ENVISAT, NOAA-AVHRR) and - if already available - data from the BRIDGE campaign will be used.

The planned project will bring together state-of-the-art land- surface-, tropospheric and hydrologic modeling, data assimilation, parameter optimization as well as scientific computing. The proposed work will contribute to important objectives of BALTEX, namely, the progress of regional process studies on water and energy fluxes of heterogeneous terrain and the determination of the water budget of the Baltic watershed. Furthermore, there are important relations to DEKLIM-BALTEX- groups (e.g., EVA_GRIPS, BALTIMOS). The suggested project allows to examine the annual variability of the energy and water fluxes by developing tools to adequately simulate snow metamorphism, freezing and thawing in soils. Thus, it contributes to improve land-surface modeling that is an essential need in understanding the climate system (GEWEX). The suggested work will use operational data, data of recent field campaigns (e.g., NOPEX, LITFASS), ENVISAT-data, data measured by ZAMF3, and the data that are to be measured in BRIDGE for either data assimilation or evaluation. It will also lead to improve data assimilation techniques.
The intended further-development of HTSVS by inclusion of soil frost and snow metamorphism processes is an important step forward for the calculation of ground water recharge and subsurface runoff (Mölders et al. 2000a) because it is based on a physically closed concept.