It is thought that satellite thermal infrared (IR) images can aid to the detection of precipitation, an interesting possibility due to the existence of geostationary satellites with thermal IR sensors which would enable a good spatial and tem- poral tracking of rain and storms. In this letter, we explore the application of multiscale/multifractal techniques in the design of new methods for the assessment and tracking of pluviometry. We first identify the main streamlines by a singularity analysis of the wavelet projections of the IR record. From the streamlines, we derive a proxy scalar image that represents the result of pure horizontal advection. From the comparison of original and proxy we localize the places at which horizontal advection fails, which we identify with convection places. We illustrate our methodology with thermal IR images from Metosat acquired during heavy tropical rainfall, and compare the results with some data from the Tropical Rainfall Measuring Mission satellite.