How to interpret it
ARIMA modeling often uses transformations or differencing to obtain a stationary representation of the series.
ForecastingDefinition
Stationarity
A time-series property in which key distributional features remain stable over time under the chosen definition.