ECN system state tagging

Clustering
Changepoint (WSPEED)

Pop plot

Choose site:

Date range (Training):

Dataset:

moth

butterfly

Number of clusters:

Choose system state variables for clustering

Date range (Output):

Note: the data is scaled before clustering.

Don't use k-means on data where you don't understand the importance(s) of the variables. Because k-means is very sensitive to scaling, and by choosing inappropriate scaling - or including inappropriate variables - the k-means result can suffer substantially. So you should rather rely on other approaches of feature selection.

Choose Meteological variable to show:

Likelihood for the next observation to be in a certain state:

row: state at t, column: state at t+1

Choose ECN statistics to show:

Prediction intervals options:

68% (1 s.d.) 95% (2 s.d.) 99.5% (3 s.d.)

crop negative prediction intervals

Divide counts across previous days with no data