Abstract
Abstract. This paper details a new method of regression for sparsely sampled data sets for use with time-series analysis, in particular the Stratospheric Aerosol and Gas Experiment (SAGE) II ozone data set. Non-uniform spatial, temporal, and diurnal sampling present in the data set result in biased values for the long-term trend if not accounted for. This new method is performed close to the native resolution of measurements and is a simultaneous temporal and spatial analysis that accounts for potential diurnal ozone variation. Results show biases, introduced by the way data are prepared for use with traditional methods, can be as high as 10%. Derived long-term changes show declines in ozone similar to other studies but very different trends in the presumed recovery period, with differences up to 2% per decade. The regression model allows for a variable turnaround time and reveals a hemispheric asymmetry in derived trends in the middle to upper stratosphere. Similar methodology is also applied to SAGE II aerosol optical depth data to create a new volcanic proxy that covers the SAGE II mission period. Ultimately this technique may be extensible towards the inclusion of multiple data sets without the need for homogenization.
Highlights
The Stratospheric Aerosol and Gas Experiment (SAGE) II flew onboard the Earth Radiation Budget Satellite (ERBS) for over 20 years from its launch in October 1984 until its retirement in August 2005
SAGE II has aerosol measurements alongple linear term in addition to effective stratospheric chlorine (EESC) terms could be in- side ozone measurements, so in theory these data could be cluded, but we found that it results in pathological behavior used as a predictor variable
The use of two orthogonal EESC predictor variables instead of a piecewise linear trend allows for a variable turnaround time in ozone due to differing mean ages of air
Summary
The Stratospheric Aerosol and Gas Experiment (SAGE) II flew onboard the Earth Radiation Budget Satellite (ERBS) for over 20 years from its launch in October 1984 until its retirement in August 2005. The combination of precise measurements and a long data record has seen SAGE II data consistently used for long-term ozone trend analysis (e.g., WMO, 1988, 2011; SPARC, 2010) This is performed via multiple linear regression of monthly zonal mean ozone data to a set of predictor variables. EOF analysis is performed on EESC data sets (Newman ation of ozone is related to but not entirely dependent upon et al, 2007) for multiple mean ages of air SAGE II has aerosol measurements alongple linear term in addition to EESC terms could be in- side ozone measurements, so in theory these data could be cluded, but we found that it results in pathological behavior used as a predictor variable. O3 uncertainty exceeds 200 % below 30 km, and exclusion www.atmos-chem-phys.net/14/13455/2014/
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
