Supplementary Information

Estimates of precipitation rates based on microwave emission are known to suffer from retrieval errors over land, with assumptions about surface emissivity not always being valid (Wilheit, 1986; Greenwald et al., 2007). Microwave instruments are also known to have issues separating cloud water from rain water, such that a change in cloud water may appear as a pre-cipitation change (Berg et al., 2006). Whilst active microwave instruments such as the TRMM PR also suffer from errors in their retrievals, they do not have the same difficulties distinguishing between cloud and rainwater. Both active (such as the TRMM PR) and passive microwave retrievals make assumptions about droplet size. Here we investigate changes in the TRMM PR reflectivity, which includes compara