Conclusions

• Although conceptually simple, tropical edge definitions based on absolute subjective threshold values have several potential drawbacks: 1. Widening trends will be induced if large-scale trends exist in the physical quantity of interest (e.g., tropopause height, OLR). 2. Similarly, absolute edge latitude is a function of the global mean value, so different datasets may yield different latitudes, even when the shapes are very similar (e.g., OLR). • “Mean ” metrics based on the 1st moment of the distribution contain less interannual variability because they incorporate more information into the latitude calculation, and this leads to better ability to detect trends.- Better noise characteristics may yield more robust detection over shorter periods of time in both model experiments and data • For a given metric, agreement among the reanalyes varies: 1. CFSR tropopause widening>> NCEP/NCAR and JRA 2. Large disagreement in Hadley cell metrics 3. JRA much different in P-E trends, but primarily due to jump in SH in 198