Skip to main content
. 2019 Jul 19;19:28. doi: 10.1186/s12898-019-0243-y

Fig. 3.

Fig. 3

Variable importance measure (mean decrease accuracy) from random forest classification used for predicting the start of the explosive breeding events. Variables with higher values were more important for the classification. A total of 48 environmental variables were evaluated based on the combination of eight measurements and six derived variables. The variables measured were: photoperiod (phper), atmospheric pressure (atm), atmospheric pressure variance (atm var), rainfall (rain), solar radiation (solrad), relative humidity (rh), temperature (temp) and temperature variance (temp var). The derived variables were based on their delay (Δ) and persistence (μ) along the time (0, 24, 48 and 72 h)