Current satellite retrieval techniques are capable of producing rainfall retrievals on high spatial resolutions; however, high uncertainties are present in such high-resolution products. Therefore, a quantitative representation of retrieval uncertainties is required for widespread adoption of satellite products. An elegant way to express these uncertainties is to generate a realistic ensemble of rainfall replicates that each element of that is consistent with the original satellite observation while containing a random element. Current methods for simulating rainfall include physical-based and stochastic models. However, these models are not capable of conditioning the replicates on spatial constrains. Such a capability is required if rainfall data are to be conditioned on remote sensed or in-situ measurements. In this paper we present a new conditional method for generating realistic rainfall replicates. The method is based on matching the statistics of sub-band coefficients of the rainfall retrieval and the replicate. The conditioning capability allows for limiting the replicates in two ways. First, to define a spatial support that the rain storms should be inside that; this allows for constraining the replicates to lie inside potential rainy areas, e.g. cloudy areas from GOES observation. Second, to force specific points to have pre-defined values; this allows for fusing point measurements of rainfall, e.g. in-situ measurements, into the generating procedure. Results show that this method is capable of generating realistic replicates and applying the conditions to them. We illustrate this algorithm with an example using NEXRAD-IV radar data and GOES brightness temperature observations.