Assessing the Ability of Climate Models to Simulate the Observed Sensitivity of Tropical Cyclone Intensity to Sea Surface Temperature

This series of studies evaluates the ability of global climate models (GCMs) to simulate the observed relationship between the upper limit of tropical cyclone (TC) intensity and sea surface temperature (SST). Previous studies addressed whether GCMs are capable of reproducing observed TC frequency and intensity distributions. This research builds upon these earlier studies by examining how well GCMs capture physically relevant relationships that are important for understanding the impacts of climate change on TC intensity. The research presented here aims to 1) quantify differences between the observed and simulated sensitivity of TC limiting intensity to SST, and 2) explore possible explanations for any differences that exist. Observed TC data are compared to simulated TCs from four different GCMs---the FSU-COAPS, GFDL-HiRAM, MRI-AGCM, and NCAR-CAM. Model horizontal grid spacing ranges from ~100 km for the FSU-COAPS to ~20 km for the MRI-AGCM. An additional comparison is made for TCs generated through a statistical-deterministic downscaling technique. This research uses a spatial tessellation approach that spatially bins North Atlantic TC and SST data into equal-area hexagon regions. For each region, the statistical upper limit of observed and simulated TC intensity (i.e., limiting intensity) is estimated using extreme value theory. For comparison with the statistical limiting intensity, reanalysis and model field data are employed to approximate observed and simulated potential intensity, respectively. Results reveal that the current suite of GCMs do not capture the observed sensitivity of TC limiting intensity to SST. While a 1° C increase in SST corresponds to a 7.9 +/- 1.19 m/s increase in observed limiting intensity, the same 1° C increase in SST is not associated with a statistically significant increase in simulated TC limiting intensity. This is found to be true both for relatively coarse resolution GCMs that do not generate TCs with intensities exceeding 50 m/s as well as for higher resolution GCMs that are capable of simulating Category 5 hurricanes. Rather than SST, it is found that simulated TC limiting intensity is highly sensitive to 700--400 hPa relative humidity. Conversely, relative humidity does not describe any of the residual variance in observed TC limiting intensity. Therefore, this research suggests that even if a model is able to resolve realistically strong TCs, those simulated TCs may not be governed by the same thermodynamic principles as those that we observe. Although GCMs do not capture the observed sensitivity of limiting intensity to SST, it is shown that the FSU-COAPS model capably reproduces the observed sensitivity of potential intensity to SST. The model generates a thermodynamic environment suitable for the development of strong TCs over the correct portions of the basin, however strong simulated TCs do not develop. This result strongly supports the notion that direct simulation of TC eyewall convection is necessary to accurately represent TC intensity and intensification processes in climate models.