Day 1054

Wednesday.


Proceedings of the National Academy of Sciences
Seas are rising at fastest rate in last 28 centuries
Kopp et al

The first estimate of global sea-level (GSL) change of the last ~3000 years that is based upon statistical synthesis of global database of regional sea-level reconstructions.  The 20th century rise was extremely likely faster than during any of the 27 previous centuries.  Semi empirical modeling indicates that, without global warming, GSL in the 20th centre very likely would have risen by between -3cm and +7cm, rather than the -14 cm observed.  Semi empirical 21st centre projections largely reconcile differences between intergovernmental panel on climate change projects and semi empirical models.

2/3 of US coastal flood days are estimated (based on above GSL analysis) to be climate change-driven.


1602.06947
Accelerated fitting of stellar spectra
Ting, Conroy, Rix

The number of synthetic spectra needed for a rectilinear grid grows exponentially with the label space dimensions, precluding the simultaneous and self-consistent fitting of more than a few elemental abundances.  Shortcuts such as fitting subsets of parameters separately an introduce unknown systematics and do not produce correct error covariance in the derived labels.  Present a new approach -- CHAT (Convex Hull Adaptive Tessellation) -- which includes several new ideas for inexpensively generating a sufficient stellar synthetic library, using linear algebra and the concept of an adaptive, data-driven grid.  A convex hull approximates the region where the data lie in the label space.  A variety of tests with mock datasets demonstrate that CHAT can reduce the number of required synthetic model calculations by 3 orders of magnitude in an 8D label space.  The reduction will be even larger for higher-dimensional label spaces.  In CHAT the computational effort increases only linearly with the number of labels that are fit simultaneously.  Around each of these grid points in label space and approximate synthetic spectrum can be generated through linear expansion using a set of "gradient spectra" that represent flux derivatives at every wavelength point with respect to all labels.  On this basis, these techniques provide new opportunities to fit the full stellar spectra from large surveys with 15-30 parameters simultaneously, as it reduces the spectral fitting problem to a series of simple linear regressions.