«Snow modeling using SURFEX with the CROCUS snow scheme Dagrun Vikhamar-Schuler, Karsten Müller and Torill Engen-Skaugen report Title Date Snow ...»
Snow modeling using SURFEX with the CROCUS
Dagrun Vikhamar-Schuler, Karsten Müller and Torill Engen-Skaugen
Snow modeling using SURFEX with the CROCUS snow scheme
September 30, 2011
Section Report no.
Division for Model and Climate Analysis, R&D Department no. 7/2011 Author(s) Classiﬁcation zFree j Dagrun Vikhamar-Schuler1, Karsten Müller2 and Torill Engen-Skaugen1 Restricted Norwegian Meteorological Institute (met.no), Oslo, Norway ISSN 1503-8025 Norwegian Water Resources and Energy Directorate (NVE), Oslo, Nor- e-ISSN 1503-8025 way Client(s) Client’s reference
Norwegian Water Resources and Energy Directorate (NVE). Project number:
Also supported by the UK Met Oﬃce. 302H47 (NVE) Abstract We have used the SURFEX land surface model with the CROCUS snow scheme to model the snowpack stratigraphy at locations of weather stations in Norway. To evaluate the results we used available snow depth and snow proﬁle measurements. Forcing data were compared by evaluating the modeled snow depth. This comparison showed highest sensitivity to the temperature and precipitation datasets. Best estimates of the snow depth are obtained when the model is forced with observations of temperature and precipitation. The experiments also show that observations of the other input parameters may be replaced by NWP data (UM-4 km forecasts) without increasing the errors notably. Particularily, the postprocessed HIRLAM-8 km temperature forecasts produce interesting results, due to the high spatial resolution (0.5 km). Alternative precipitation datasets (postprocessed UM-4 km forecasts and HARMONIE forecasts) did not improve the modeled snow depth notably. These ﬁndings make it possible to run the model at locations of weather stations equipped with limited number of sensors, e.g. precipitation stations.
Modeled and measured snow/ground temperatures were compared at the Filefjell-Kyrkjestølane station, showing promising results. A thorough evaluation of the modeled snowpack layers and their physical properties should be a topic for future work, when more snow proﬁles will be available.
Keywords Snow avalanche, snow depth, snow model, CROCUS, SURFEX, NWP, observations.
Chapter 1 Introduction
1.1 The R&D project on snow avalanches The study presented in this report is carried out within the R&D project “Avalanche danger and senorge.no”.
The project is coordinated by NVE1 and has ﬁve partners (NVE, met.no2, NGI3, NPRA4 and NNRA5 ).
The project is planned for three years (2010-2012). The overall aims of the project are testing and developing methods to establish a regional avalanche forecasting system. The avalanche danger level will be set by international standards based on avalanche, weather and snow data. The R&D project is described in several documents found at this webpage: http://www.nve.no/no/Flom-og-skred/Skred/FoU— skred/FoU-prosjekt-Snoskredfare-og-senorge/. The project is organised in eight subprojects, and the snow simulations presented in this report is one of them.
1.2 Aim and motivation of the snow simulations The aim of this work is modeling of snow proﬁles which in turn will support the avalanche experts to determine the danger level for a region. The snow proﬁles show detailed information about the layer stratigraphy of the snowpack. Each layer is described by physical properties such as density, temperature, grain size, grain type and liquid water content. Snow proﬁles also show the historical development of a snowpack during a winter season, from the accumulation period to the melting period, which gives additional information about the risk for snow avalanches.
Snow models are used for avalanche forecasting in France and Switzerland, but in different ways. The two most advanced snow models developed for avalanche forecasting are CROCUS and SNOWPACK.
CROCUS is developed by MétéoFrance (Brun et al., 1992, 1989). For more information see Section 2.
SNOWPACK is developed by WSL Institute for Snow and Avalanche Research (SLF) in Switzerland (Bartelt and Lehning, 2002; Lehning et al., 1999, 2002a,b). In areas with less dense station network (snow and weather observations) there is greater need for models in general. Switzerland has a denser station network than France, while Norway has a quite sparse station network compared to these two countries.
In Switzerland snow models are used if no manual observations are available (Christine Pielmeier, SLF, Switzerland, pers. comm., 2011), otherwise snow observations are extensively used in the operational avalanche warning. Snow models are also applied for estimating fresh snow depth at the stations (Lehning et al., 1999). The models replace precipitation measurements in the high mountains by measureing new snow depth and converting it to SWE (snow water equivalents) by modelling the density and settling. Precipitation is difﬁcult to measure directly under windy and subfreezing conditions. Here snow depth measurements combind with modeled snowpack properties gives more accurate estimates.
Norwegian Water Resources and Energy Directorate Norwegian Meteorological Institute Norwegian Geotechnical Institute Norwegian Public Roads Administration Norwegian National Rail Administration In France representative snow proﬁles for areas with homogeneous climate zones are modeled by CROCUS on a daily basis. Snow proﬁles for six different expositions and different elevation ranges within each climate zone are simulated and evaluated by the avalanche forecaster. An expert module called MEPRA is used to evaluate the large amount of simulated snow proﬁles and to compile a report that the forecaster use in the daily avalanche bulletin production.
Evaluation of the snow models, CROCUS and SNOWPACK, as a tool for establishing a new operational avalanche warning system in Norway, is quite interesting. The sparse station network (snow and weather observations) in Norway makes the modeling approach more feasible compared to the Swiss observation-based approach. The results presented in the present report is one step in this validation procedure. The results presented in the present report is one step in this evaluation procedure.
Results of a sensitivity study of the forcing data by running the SURFEX land surface model using the CROCUS snow scheme is presented. A sensitivity study of the forcing data will assess the different weather data sets (observations, weather prognoses), and identify which data sets are most suited for state-of-the art snowpack modeling. The model has been run for selected locations of weather stations in Norway (1D runs). Evaluation of the modeling results is performed using observed snow depth, snow temperature data and available snow proﬁles.
The report is organized in six chapters, starting with this introduction. The SURFEX model and the CROCUS snow scheme are described in Chapter 2. The applied datasets and a summary of the modeling results are presented in Chapter 3 and Chapter 4, respectively. Discussion and outlook are included in a last chapter before the concluding remarks. Results for all stations are included in three appendices B, C and D.
Chapter 2 The SURFEX model and the CROCUS snow scheme The CROCUS model was developed by MétéoFrance for snow avalanche warning purposes in the early 1990’s (Brun et al., 1992, 1989). It has been run operationally by MeteoFrance for snow avalanche warning and it is part of the SCM chain (SAFRAN-CROCUS-MEPRA), see Figure 2.1. SAFRAN is the meteorological analysis system, which based on various data sets (snow observations, weather observations, numerical weather prediction (NWP) data), interpolates into hourly data of the input parameters needed by CROCUS (Durand et al., 1999). The output parameters from SAFRAN are computed for climatologically homogeneous zones, and not for regular grid cells. For every zone, CROCUS is run to model the evolution of the snowpack. Hence, the result represents the average snowpack of this zone.
Furthermore, the output from CROCUS are interpreted by the expert model MEPRA. MEPRA computes two stability indices (natural and accidental) and proposes a risk level for every zone on a 6 level scale.
Figure 2.1: The SAFRAN-CROCUS-MEPRA chain.
Recently the CROCUS model was included as a snow scheme within the SURFEX model (v. 5) (Brun et al., 2011; Vionnet et al., 2011). SURFEX (SURFace EXternalisée) is a land surface model, also developed by MétéoFrance (LeMoigne, 2009a,b). The integration of CROCUS into SURFEX, SURFEX itself and postprocessing routines are continuously upgraded and a stable release of SURFEX (v. 7) is planned (Eric Brun and Samuel Morin, MétéoFrance, pers. comm., April 2011 ). SURFEX can be run stand-alone (ofﬂine mode) or in a coupled system with atmospheric models e.g with HARMONIE using AROME physics (inline mode). In addition to CROCUS, two other snow schemes (ISBA-FR and ISBAES) are available in SURFEX. ISBA-FR models the snowpack as a single layer, while ISBA-ES models the snowpack with three layers. CROCUS is the most advanced snow scheme, and the snowpack is modeled with up to 50 layers. These three snow schemes were compared in Boone and Etchevers (2001).
They found in 2001 that the CROCUS snow scheme is about ﬁve times more computer demanding to run compared to the simple one layer ISBA-FR snow scheme.
Figure 2.2: Simple visualization of input and output of the SURFEX land surface model.
Included are also the snow and soil schemes used in the presented analysis.
In this report we have run SURFEX in ofﬂine mode, using the CROCUS snow scheme in combination with the DIFF soil scheme (Figure 2.2). The DIFF soil scheme included in SURFEX is a diffusive approach for modeling soil layers and soil properties. We have run SURFEX for single points (1D runs) at locations of selected weather stations. SURFEX requires a number of forcing data (Table 2.1).
Output from our setup of the SURFEX model were NetCDF ﬁles with prognostic variables from the CROCUS snow scheme and the DIFF soil scheme. The stratigraphy of the snowpack is modeled using a one-dimensional ﬁnite-element grid, and each snow layer is described by the thickness, temperature, dry density, liquid water content and grain types. Each layer represents a snowfall event. An example of model output is shown in Figure 2.3. The NetCDF output ﬁle was postprocessed with the Snowtools toolbox (Morin and Willemet, 2010). This toolbox contains a collection of Python scripts for manipulation and plotting CROCUS output. More information is also found in Section 4.3.1.
Table 2.1: Forcing data required by SURFEX (LeMoigne, 2009a).
Symbols from the KDVH met.no climate database for observations are used in the plots in this report. *Speciﬁc humidity was computed from observed relative humidity (UU) (Rogers and Yau, 1989). Snowfall and rainfall rate was computed from hourly observed precipitation (RR1) using a threshold temperature of
273.65 K (0.5 ). For short-term prognoses from UM and HARMONIE we summed convective and large-scale precipitation for each of the precipitation types, rain and snow. More technical details are described in Appendix A.
Figure 2.3: Example of model output of snow temperature from the SURFEX model run with the CROCUS snow scheme, and postprocessed with the Snowtools toolbox (Morin and Willemet, 2010).
The colour represents the snow temperature (K), where blue represents low temperatures and red represents higher temperatures. This example is from the Grotli met-station during the winter season 2009/2010. Maximum snow depth was approximately 0.70 m before the snow melt began in April.
Chapter 3 Dataset Different weather data sets are available at met.no. These are observations from weather stations, prognoses from different NWP models and postprocessed prognoses. The best available weather forecasts are broadcasted at http://www.yr.no/, which is the weather forecast portal of met.no. Our sensitivity study aims to identify the best state-of-the-art datasets for snow modeling. Therefore, we have applied a selection of these datasets.
3.1 Observations Currently met.no operates approximately 630 weather stations located within the Norwegian mainland.