Date of Award

Fall 12-2021

Level of Access Assigned by Author

Open-Access Thesis

Degree Name

Master of Science (MS)

Department

Earth Sciences

Advisor

Bradfield Lyon

Second Committee Member

Sean Birkel

Third Committee Member

Kirk Maasch

Additional Committee Members

Benjamin Lintner

Abstract

Most climate models participating in the Coupled Model Intercomparison Project, Phase 5 (CMIP5) struggle to accurately simulate all aspects of the climatological global sea surface temperature (SST) distribution, especially in regions of coastal and equatorial upwelling. Of particular interest in this thesis are coupled model SST biases in the tropics and their potential influence on environmental conditions associated with tropical cyclone (TC) genesis and intensity. To evaluate this, two atmospheric general circulation models (AGCMs), the ECHAM5 and CAM5, were employed, first to examine their ability to capture observed TC characteristics when forced with observed SSTs (control runs) and then by comparing these results with output from model runs forced with observed SSTs plus the climatological, multi-model (31) mean CMIP5 monthly SST bias (bias runs), both covering the period 1979-2004. Three TC indices were evaluated, two of which were statistically based that relate atmospheric and SST conditions to TC formation and the third was physically based which evaluates the potential intensity of a TC (maximum nearsurface windspeed) based on SST and vertical profiles of atmospheric data. All three indices were evaluated over the near global domain. The statistically based indices were the Tropical Cyclone Genesis Index (TCGI) and the Genesis Potential Index (GPI) whereas the physically based index was the Potential Intensity (PI). Results using monthly mean atmospheric conditions from the control runs of ECHAM5 and CAM5 were found to produce similar results with those obtained from using daily NCEP reanalysis data, especially in terms of the spatial patterns of the three TC indices. Results between ECHAM5 and CAM5 control runs were found to be very similar and therefore in order to avoid being repetitive, only ECHAM5 was subsequently used for the analysis of the response to coupled model SST biases. While the AGCMs were found to generally produce too many cyclones as compared to NCEP reanalysis data they were able to capture the observed pattern of TC behavior very well. Pattern correlations were calculated between ECHAM5 control run results and NCEP reanalysis data for TCGI, GPI, and PI with values of 0.98, 0.56, and 0.86 obtained, respectively. These values increased confidence that despite the coarse resolution of the AGCM it was doing a good job at capturing the observed spatial patterns of the TC indices considered. Comparing annual average results from the bias and control runs, the CMIP5 climatological SST biases were found to result in statistically significant regional changes in TCGI, PI, and GPI. In some locations, the proxies for TC frequency (TCGI, GPI) and intensity (PI) are enhanced, in others they are reduced. Generally, the regions of overestimated (underestimated) TCGI, GPI, and PI were found to be collocated with areas of positive (negative) SST biases. There were a few notable exceptions, but these were generally localized in nature. The sensitivity of the TCGI to the individual component variables from the bias runs was also examined and tested for statistical significance. Results from this sensitivity testing showed that the relative SST was the most influential term whereas absolute vorticity had the smallest overall impact. Seasonal analyses were also conducted for PI and GPI, owing to the different hurricane seasons across each ocean basin. Results generally showed a strong correlation between regions of over (underestimated) TCGI, PI, and GPI to regions of negative (positive) SST biases. The results obtained from this study indicate that ECHAM5 and CAM5 are able to accurately simulate observed TC behavior when forced with observed SSTs, especially in terms of spatial patterns of the TC indices. When comparing these results to those obtained when ECHAM5 is forced with observed SST plus the multi-model mean SST bias it becomes apparent that the SST biases are indeed having an impact on these simulations. Results between ECHAM5 control runs and observations showed that despite the coarse resolution of the coupled climate models, they can be used to realistically capture TC characteristics. When examining the influence of the coupled model SST biases on the three TC indices, it was found that the index with the greatest sensitivity to these biases was the TCGI. Regions of over/underestimated TCGI were generally found to be collocated with positive/negative SST biases despite a few exceptions. These impacts were generally the greatest in the Pacific Ocean, with lesser impacts across other ocean basins. This was particularly true across the eastern and western tropical Pacific where in some regions the over/underestimated number of TC genesis events exceeded 10 on an annual time scale.

Share