A method is provided for predicting a risk for rollover of a working machine for load transportation. The method includes: obtaining ground topographic data of a geographical area located close to the working machine from a ground topographic detection system; extracting a ground gradient from the ground topographic data; obtaining weight information of the load being currently transported by means of an on-board load weighting system or by receiving load information originating from the device that loaded the load being currently transported; determining a current maximal allowed ground gradient for the working machine based on the weight information; and predicting a risk for working machine rollover if the working machine approaches a geographical area including a ground gradient exceeding or being close to the current maximal allowed ground gradient for the working machine.