The diurnal cycle of precipitation according to multiple decades of global satellite observations, three CMIP6 models, and the ECMWF reanalysis

NASA Precipitation Measurement Mission observations are used to evaluate the diurnal cycle of precipitation from three CMIP6 models (NCAR-CESM2, CNRM-CM6.1, CNRM-ESM2.1) and the ERA5 reanalysis. NASA's global-gridded IMERG product, which combines spaceborne microwave radiometer, infrared sensor, and ground-based gauge measurements, provides high-spatiotemporal-resolution (0.18 and half-hourly) estimates that are suitable for evaluating the diurnal cycle in models, as determined against the ground-based radar network over the conterminous United States. IMERG estimates are coarsened to the spatial and hourly resolution of the state-of-the-art CMIP6 and ERA5 products, and their diurnal cycles are compared across multiple decades of June-August in the 60°N-60°S domain (IMERG and ERA5: 2000-19; NCAR and CNRM: 1979-2008). Low-precipitation regions (and weak-amplitude regions when analyzing the diurnal phase) are excluded from analyses so as to assess only robust diurnal signals. Observations identify greater diurnal amplitudes over land (26%-134% of the precipitation mean; 5th-95th percentile) than over ocean (14%- 66%). ERA5, NCAR, and CNRM underestimate amplitudes over ocean, and ERA5 overestimates over land. IMERG observes a distinct diurnal cycle only in certain regions, with precipitation peaking broadly between 1400 and 2100 LST over land (2100-0600 LST over mountainous and varying-terrain regions) and 0000 and 1200 LST over ocean. The simulated diurnal cycle is unrealistically early when compared with observations, particularly over land (NCAR-CESM2 AMIP:21 h; ERA5: 22 h; CNRM-CM6.1 AMIP: 24 h on average) with nocturnal maxima not well represented over mountainous regions. Furthermore, ERA5's representation of the diurnal cycle is too simplified, with less interannual variability in the time of maximum relative to observations over many regions.