In the aftermath of natural disasters, quick search-and-rescue operations, and efficient distribution of relief efforts and goods to affected areas are a top priority. However, it can be challenging to continuously process real-time information of differing reliability. That means, using it to estimate and predict type and extent of the demand for relief efforts. This paper proposes an approach for estimating the state of affected areas and predict associated relief demands and delivery times. The purpose is to account for data uncertainty typically arising under disaster circumstances, and to enable prompt adjustment of relief efforts in accordance with updated real-time information. The method's output can be used by decision makers to continuously optimize relief efforts. A numerical example based on the large-scale earthquake that occurred on September 21,1999 in Taiwan is presented.