Bayesian Network Adaptor

Summary

The Bayesian Network (BN) adaptor sets up a Bayesian network based decision support system (DSS) for coastal hazards and impacts in hotspot areas, according to the framework developed in task 3.3 of the RISC-Kit project (www.risckit.eu). 

The BN adaptor uses

1. .nc files of storm simulations
2. .csv files indicating receptor locations 
3. a .json file specifying the structure of the BN
4. a .json file specifying the vulnerability relationships for the hotspot area

to create

1. a .dsl / .dne file that can be opened with a Bayesian network software
2. a folder with all files that are relevant for the webviewer

Required Folder Structure in FEWS

Users have to create a personal folder, i.e. BN_<hotspotname>, in the FEWS directory Config/ModuleDataSetFiles.zip/ as well as the following subfolders and files therein:

1. input
2. model
   • <receptor1>.csv
   • <receptor2>.csv
   • ...
   • <receptorN>.csv
   • BNstructure.json
   • vulnerabilities.json
   • areas.shp (& related GIS files)
3. output
4. trainingData

The folders input, output and trainingData remain empty. They are filled and purged through the BN adaptor (BNA) and the general adaptor (GA) only.

Data Process in FEWS

The data process described in this section takes place after every single simulation with the model train. 

First, the GA writes the (near-shore) process model output (e.g. XBeach, Telemac) as

1. hazardbc_<measure>_<bin>.nc
2. hazard.nc

to Config/ModuleDataSetFiles.zip/BN_<hotspotname>/input. The first file should contain the max. value (scalar) of the relevant forcing variables, such as water level or significant wave height. <measure>_<bin> is an addition to the file name that only has to be used when the simulation includes a type I measure, i.e. measures affect the hazard pathway and require a different setup of the hydrodynamic model, such as beach nourishments, dikes etc. The second file contains the gridded maximum values of the relevant hazard variables, such as inundation depth, flow velocity or erosion. 

Subsequently, the GA calls the BNA.  

The BNA  

1. copies the two files from input to trainingData/<yyyymmdd>_<measure>_<bin>
2. reads all files in model and trainingData 
3. creates the file <hotspotname>.dsl and the folder webviewer with all relevant files and writes them to output

The user can open <hotspotname>.dsl with a Bayesian network software. Finally, the files in webviewer have to be copied/moved manually to a shared ftp (details will follow), where they can be picked up by the webviewer for display.

Required File Structures

hazard_<measure>_<bin>.nc

(to be added soon)

hazard.nc

(to be added soon)

BNstructure.json

vulnerabilities.json

(to be added soon)

receptorn.csv

(to be added soon)

Example Configuration

(to be added soon)