Distributions of Tropical Precipitation Cluster Power and Their Change Under Global Warming

Prior studies have indicated that observed precipitation cluster area and power distributions exhibit a power law range, with cutoffs at large cluster area and high cluster power (Peters et al, 2009, 2012). Cluster power is defined as the precipitation (here expressed in units of latent heat release) integrated over contiguous precipitating grid cells (for precipitation above a minimum threshold). Cluster power distributions are examined over the Tropics (30N - 30S) for one satellite-retrieved dataset (TRMM-3B42, May-September 1998-2008) and three global climate model (GCM) datasets (GFDL-HIRAM-C360, GFDL-HIRAM-C180, MIROC5) chosen for reasonably high resolution. Historical GCM data (May-September 1998-2008) are from the Atmospheric Model Intercomparison Project (AMIP) experiments, and future GCM data (May-September 2086-2095) are from the SST2090 experiments of the Coupled Model Intercomparison Project Phase 5 (CMIP5) database. Observed and historical GCM cluster power distributions are first compared in order to establish a baseline for comparison with future modeled cluster power data. Observed cluster power distributions have long, scale-free power law ranges between 10 - 10^5 GW, and a rapid drop off in high cluster power frequency thereafter. The phenomena leading to these clusters range from convective phenomena at the pixel scale (approximately 25 km) and mesoscale clusters through ITCZ disturbances and tropical storms. GCM simulations accurately reproduce observed distributions across the power law range. GCM cutoff values are more sensitive to rain rate threshold, due to overly widespread occurrence of low rain rates, but agree well provided the threshold is not too low. Because the cutoff affects the probability of the highest cluster power events that can be potentially very important for human impacts, changes to cluster power distributions under global warming are investigated. GFDL-HIRAM-C360 cluster power distributions from the SST2090 experiment follow the same long, scale-free distribution as GFDL-HIRAM-C360 AMIP output, but the cutoff tends to shift toward higher power. As a result the probability of large power clusters tends to increase significantly above the historical cutoff. Similar results are obtained for the other two models. A comparison to a posited Clausius-Clapeyron-scaling relationship is created by applying the same analysis to the GFDL-HIRAM-C360 AMIP historical case but with the precipitation rates rescaled by a 7% relative humidity increase per degree warming under the projected change to mean global temperature (2.16 K increase). The hypothetical Clausius-Clapeyron-scaled distribution lies between the original GFDL-HIRAM-C360-AMIP and SST2090 cluster power distributions, indicating that the change in future cluster power distributions considerably exceeds expectations based on a simple Clausius-Clapeyron scaling of precipitation intensity.