Data assimilation in the early phase: Kalman filtering RIMPUFF

In the framework of the DAONEM project (Data Assimilation for Off-site Nuclear Emergency Management), a data assimilation module, ADUM (Atmospheric Dispersion Updating Module), for the mesoscale atmospheric dispersion program RIMPUFF (Risø Mesoscale Puffmodel) – part of the early-phase programs of RODOS (Realtime Online DecisiOn Support system for nuclear emergencies) – has been developed.

It is built on the Kalman filtering algorithm and it assimilates 10-minute averaged gamma dose rates measured atground level stations. Since the gamma rates are non-linear functions of the state vector variables, the applied Kalman filter is the so-called Extended Kalman filter. In more ways the implementation is non standard: 1) the number of state vectorvariables varies with time, and 2) the state vector variables are prediction updated with 1-minute time steps but only Kalman filtered every 10 minutes, and this based on time averaged measurements.

Given reasonable conditions, i.e. a spatially densedistribution of gamma monitors and a realistic wind field, the developed ADUM module is found to be able to enhance the prediction of the gamma dose field. Based on some of the Kalman filtering parameters, another module, ToDeMM, has been developed forproviding the late-phase DeMM (Deposition Monitoring Module) of RODOS with an ensemble of fields of ground level air concentrations and wet deposited material.

This accounts for the uncertainty estimation of this kind of quantities as calculated byRIMPUFF for use by DeMM.