ISeeSnow - model intercomparison pilot-study

The ISeeSnow project aims to initiate an intercomparison project for avalanche simulation tools. We want to start a conversation among the modelling community with a pilot study comparing results from mainly thickness-(depth-) integrated models based on a Voellmy friction relation.

To keep it simple, we exclude any model verification tests that might require a more complex model setup. We also exclude model validation tests that potentially include model optimization.

So the focus is on standard simulations with prescribed friction parameters for two different topographies: an idealized topography and a real-world example.

We ask the participating groups to employ their default configuration with a Voellmy friction relation for these simulations. The values of the friction parameters mu, xi and the release thickness are provided alongside the topography and release area input data. As a third test case, a simulation run with purely Coulomb friction should be performed for the idealized topography. If the respective model configuration is not designed to use a friction relation with only Coulomb friction, we ask the participants to set the xi value in the Voellmy friction relation to an extremely high value, forcing the effect of the turbulent friction term to be negligible.

The AvaFrame-team will provide input data and parameter values for the three test cases and compare the gathered simulation results. The analysis is performed using the functionalities of the ana3AIMEC and ana4Stats modules of AvaFrame.

How to participate in ISeeSnow?

The requirement is to perform one simulation per test case (three in total) and provide the result datasets as described in the Section Result datasets.

The runISeeSnowAnalysis.py script provides the option to test the postprocessing analysis based on ana3AIMEC on your simulation results. In order to run this script, ensure that your simulation results (in the required format and following the prescribed naming convention described in the Section Result datasets) are located in the data/testCase/Outputs directories of the respective test case (i.e. IdealizedTopo, RealTopo, CoulombOnly).

The analysis results will be saved to data/testCase/Outputs/ana3AIMEC for each of the three test cases. There is also the option to provide outputDirectoryPath and testCase as command line arguments. If provided, the simulation result files will be fetched from the outputDirectoryPath directory (one per test case) and the required topography information from the data/testCase/Inputs directory, analysis results will also be saved to data/testCase/Outputs/ana3AIMEC. The aimec analysis further requires e.g. a thalweg (shapefile), this data is also provided alongside the model input data.

Note that running the runISeeSnowAnalysis.py script requires an AvaFrame advanced installation.

Additionally, the directory exampleWorkflowCom1DFA provides an example workflow to run the test cases - it is exemplary set up for com1DFA- but you are invited to include a call to your model in there.

In the following, information on provided model input data and required format of result datasets is given:

Model input data:

The input data comprises a digital elevation model (DEM) (regularly spaced points with a spatial resolution of 5 meters) and a release area scenario per test case. The release area is provided as shapefile OR alternatively as release thickness field .asc file. The .asc file has the same extent and spatial resolution as the DEM and provides the actual values of release thickness at each cell (values different from zero give the release thickness, whereas areas outside of the prescribed release polygon are represented by values of 0). Note on release thickness .asc file: The .asc files in the Inputs directories are created based on the DEM and the release polygon read from the corresponding shapefile. In the resulting raster, cells are set to belong to the release area as soon as there is an intersection with the release polygon. For this reason, the area of the release thickness field exceeds the area of the release polygon.

• digital elevation model as .asc file (format: https://desktop.arcgis.com/en/arcmap/10.3/manage-data/raster-and-images/esri-ascii-raster-format.htm) with a spatial resolution of 5 meters
  • testCase IdealizedTopo: DEM_IdealizedTopo.asc
  • testCase RealTopo: DEM_RealTopo.asc
  • testCase CoulombOnly: DEM_CoulombOnly.asc
• release area scenario as shapefile with release area feature (polygon), homogeneous release thickness throughout release area
  • testCase IdealizedTopo: REL/release1HS.shp
  • testCase RealTopo: REL/realWog.shp
  • testCase CoulombOnly: REL/release1HS.shp
• release area scenario as asc file with release area feature thickness, homogeneous release thickness throughout release area, covering the DEM extent and a spatial resolution of 5 meters
  • testCase IdealizedTopo: RELTH/release1HSField5m.asc
  • testCase RealTopo: RELTH/realWogField5m.asc
  • testCase CoulombOnly: RELTH/release1HSField5m.asc
• simulation parameter text file with values for mu and xi, as well as release thickness value (see description of Voellmy-type friction relation for example here: https://docs.avaframe.org/en/latest/theoryCom1DFA.html#voellmy-friction-model)

Note: The release thickness is measured normal to the surface.

IdealizedTopo

RealTopo

CoulombOnly

The corresponding files can be found in the directory data, where IdealizedTopo refers to the idealized test case, RealTopo represents the real-world topography and CoulombOnly provides the same topography and release area scenario as in the case of the idealizedTopo test case, but should be run with purely Coulomb friction. For all three test cases, the DEM (DEM_IdealizedTopo.asc, DEM_RealTopo.asc and DEM_CoulombOnly, respectively), the release area shapefile and the release area thickness field are located in the directory Inputs. The friction parameter values and the release thickness value, can be found in the corresponding simulationParameterValues_testCase.csv files. Data source info for the RealTopo test case can be found here. The idealized topographies have been generated using the in3Utils module. This topography generation is also included in the exampleWorkflowCom1DFA/runISeeSnowTest.py script.

Result datasets:

We ask all the participating groups to perform a simulation for the two test cases, and to provide the following result datasets:

• fields of peak flow velocity and peak flow thickness (measured normal to the surface) as .asc file covering the entire computational domain (DEM extent). Peak refers to the maximum flow variable value over the entire duration of the simulation. The peak fields should have a spatial resolution of 5 meters and the same extent of the DEM - hence the .asc files should have the same header as the provided dem file (if there is an issue to provide the results in this format, contact us). We require a specific naming of the peak field .asc files: releaseName_simulationID_simType_modelType_resultType.asc, from now on referred to A_B_C_D_E.asc, where:

  • A - releaseName: refers to the name of the release shapefile
  • B - simulationID: needs to be unique for each simulation - use: 01IdealizedTopo, 02RealTopo and 03CoulombOnly
  • C - simType: use null
  • D - modelType: name of your model - NOT allowed to include underscores
  • E - result type: is pft (peak flow thickness) and pfv (peak flow velocity)

  Note: underscores are not allowed except to separate the five elements of the file name and no data values should be provided as nans.

• a csv file with information on computation duration (CPU), avalanche flow time, total volume at initial time step and also final time step, spatial resolution (example: simulationResultTable.csv) - one file listing the values of all simulations

• a text file with information on model configuration, i.e. parameter values, numerical configuration, model version etc., the naming should be consistent with the peak field files: releaseName_simulationID_simType_modelType.txt. These configuration files can optionally be also provided as .ini files, to be interpreted by configparser.

Note: the spatial resolution of the model input data is 5 meters (regular grid) and the result fields are required to also have a spatial resolution of 5 meters. However, this applies to the submitted result fields, the simulations can be performed using your default numerical setup (also if using a different spatial resolution), but they need to be submitted as rasters with a cell size of 5 meters.

All the listed result files should be provided as archive file (zip, tar.gz, etc.) containing the simulationResultTable with one line per simulation, and one subdirectory per avalanche test case: IdealizedTopo, RealTopo and CoulombOnly each of them providing the respective peak fields of flow velocity and flow thickness and the model configuration file.

In addition, we ask the participating groups to provide a paragraph describing their simulation tool, e.g. mathematical model, numerical methods, configuration, code availability and a reference. This should be added as a simple .txt file in the archive file together with the result datasets.

Once you have prepared the required result datasets, you can send the archive file to us via email. In case the file size is too big, contact us and we will inform you on how to submit the data. With the submission you agree to the publication of the results in full. The deadline for submission is on November 30th.

Provide your own test case

We invite you to provide further even test cases that can be used in a potential continuation or future intercomparison projects. We will collect these and publish them as an open dataset (properly attributed, citable), hopefully building a test dataset that benefits the whole community. The minimum requirements for an event test case are:

• release area scenario
• release thickness
• documented runout line