The accuracy of the operational models can be improved by using observational data to shift the model state in a process called data assimilation. Here, a data assimilation approach using the temperature similarity to control the extent of extrapolation of point-like phenological observations is explored. A degree-day model is used to describe the spring phenology of the bird cherry Padus racemosa in the Baltic region in 2014. The model results are compared to phenological observations that are expressed on a continuous scale based on the BBCH code. The air temperature data are derived from a numerical weather prediction (NWP) model. It is assumed that the phenology at two points with a similar temperature pattern should be similar. The root mean squared difference (RMSD) between the time series of hourly temperature data over a selected time interval are used to measure the temperature similarity of any two points. A sigmoidal function is used to scale the RMSD into a weight factor that determines how the modelled and observed phenophases are combined in the data assimilation. The parameter space for determining the weight of observations is explored. It is found that data assimilation improved the accuracy of the phenological model and that the value of the point-like observations can be increased through using a weighting function based on environmental parameters, such as temperature.
Keywords: Baltic region; Data assimilation; Padus racemosa; Phenological model.