In this section
@article{Ball:2017:10.5194/acp-17-12269-2017,
author = {Ball, WT and Alsing, J and Mortlock, DJ and Rozanov, EV and Tummon, F and Haigh, JD},
doi = {10.5194/acp-17-12269-2017},
journal = {Atmospheric Chemistry and Physics},
pages = {12269--12302},
title = {Reconciling differences in stratospheric ozone composites},
url = {http://dx.doi.org/10.5194/acp-17-12269-2017},
volume = {17},
year = {2017}
}
TY - JOUR
AB - Observations of stratospheric ozone from multipleinstruments now span three decades; combining these intocomposite datasets allows long-term ozone trends to be estimated.Recently, several ozone composites have been published,but trends disagree by latitude and altitude, even betweencomposites built upon the same instrument data. Weconfirm that the main causes of differences in decadal trendestimates lie in (i) steps in the composite time series when theinstrument source data changes and (ii) artificial sub-decadaltrends in the underlying instrument data. These artefacts introducefeatures that can alias with regressors in multiple linearregression (MLR) analysis; both can lead to inaccuratetrend estimates. Here, we aim to remove these artefacts usingBayesian methods to infer the underlying ozone time seriesfrom a set of composites by building a joint-likelihoodfunction using a Gaussian-mixture density to model outliersintroduced by data artefacts, together with a data-driven prioron ozone variability that incorporates knowledge of problemsduring instrument operation. We apply this Bayesianself-calibration approach to stratospheric ozone in 10 bandsfrom 60 S to 60 N and from 46 to 1 hPa (∼ 21–48 km) for1985–2012. There are two main outcomes: (i) we independentlyidentify and confirm many of the data problems previouslyidentified, but which remain unaccounted for in existingcomposites; (ii) we construct an ozone composite, withuncertainties, that is free from most of these problems – wecall this the BAyeSian Integrated and Consolidated (BASIC)composite. To analyse the new BASIC composite, we usedynamical linear modelling (DLM), which provides a morerobust estimate of long-term changes through Bayesian inferencethan MLR. BASIC and DLM, together, provide astep forward in improving estimates of decadal trends. Ourresults indicate a significant recovery of ozone since 1998 inthe upper stratosphere, of both northern and southern midlatitudes,in all f
AU - Ball,WT
AU - Alsing,J
AU - Mortlock,DJ
AU - Rozanov,EV
AU - Tummon,F
AU - Haigh,JD
DO - 10.5194/acp-17-12269-2017
EP - 12302
PY - 2017///
SN - 1680-7316
SP - 12269
TI - Reconciling differences in stratospheric ozone composites
T2 - Atmospheric Chemistry and Physics
UR - http://dx.doi.org/10.5194/acp-17-12269-2017
UR - http://hdl.handle.net/10044/1/54147
VL - 17
ER -