*3.2. Southern Ghana Case*

In this case, we apply the automatic registration and the warping to real precipitation data. Our goal is to gauge-adjust a satellite-based estimate with respect to the location of the rain events. We assume the gauge measurements to be more accurate, but the IMERG has a higher spatial variability that the gauges cannot reproduce because of their network's density. Thus, our goal is to keep the spatial variability of IMERG, while correcting the position mismatch with respect to the gauge. For this real case, we will only look at the warping for this case. The morphing would make the warped field similar to the gauges, even in areas without gauges.

The study area is a square domain over southern Ghana encompassing the Ghanaian cocoa region. This domain has been chosen because of the particular high density of the Trans-African Hydro-Meteorological Observatory (TAHMO) network in this area (Figure 3).

Southern Ghana has two rainy seasons. The main one extends from March to mid-July and the second one occurs during September and October. We chose a rainfall event during the main monsoon season. More precisely, we selected one hour during this event (from 18:00 to 19:00 of 22 April 2018). Given the spatial resolution of IMERG (0.1◦ lat/lon), hourly accumulation is in good agreement with the rainfall spatial and temporal variability [5]. Longer time scale would make it more difficult to identify the individual events and very few precipitation datasets are available at a sub-daily scale.

**Figure 3.** Study domain (red rectangle) and the TAHMOstations available within the domain (white dots).
