This thesis transforms satellite precipitation estimation through the integration of a multi-sensor, multi-channel approach to current precipitation estimation algorithms, and provides more accurate readings of precipitation data from space. Using satellite data to estimate precipitation from space overcomes the limitation of ground-based observations in terms of availability over remote areas and oceans as well as spatial coverage. However, the accuracy of satellite-based estimates still need to be improved. The approach introduced in this thesis takes advantage of the recent NASA satellites in observing clouds and precipitation. In addition, machine-learning techniques are also employed to make the best use of remotely-sensed "big data." The results provide a significant improvement in detecting non-precipitating areas and reducing false identification of precipitation.

This thesis transforms satellite precipitation estimation through the integration of a multi-sensor, multi-channel approach to current precipitation estimation algorithms, and provides more accurate readings of precipitation data from space. Using satellite data to estimate precipitation from space overcomes the limitation of ground-based observations in terms of availability over remote areas and oceans as well as spatial coverage. However, the accuracy of satellite-based estimates still need to be improved. The approach introduced in this thesis takes advantage of the recent NASA satellites in observing clouds and precipitation. In addition, machine-learning techniques are also employed to make the best use of remotely-sensed "big data." The results provide a significant improvement in detecting non-precipitating areas and reducing false identification of precipitation.