Multiple regression models for time series of ozone column measurements have been a standard tool of atmospheric sciences for decades. We propose a new model, which is based on a process-oriented approach and captures all major processes relevant to the change of stratospheric ozone column. These are assumed to be short- and long-term changes by tropospheric and stratospheric pressure systems, the residual circulation, homogeneous and heterogeneous halogen chemistry, aerosols, the quasi-biennial oscillation (QBO), and the solar cycle. Daily ozone data are taken from 49 stations of the Dobson photometer network covering polar, middle, and equatorial latitudes. A focus is placed on dynamical changes in midlatitudes. Results show that the trend in ozone column between 1970 and 2003 is dominated by halogen chemistry in midlatitudes (up to - 1.4 ± 0.5 DU/year or -4.1%/decade in winter) but that there are also significant trends of other origin in some regions of the world. These trends are mainly caused by changes in horizontal advection and convergence or divergence of mass, which are both related to trends in tropospheric and lower stratospheric pressure systems like the Icelandic low (represented by a proxy based on a vertically integrated equivalent latitude profile). This dynamical contribution to the trend depends both on month and longitude and varies between -0.7 ± 0.2 DU/year (-1.90%/decade) and 0.4 ± 0.2 DU/year (1.1%/decade). Another influence on trends in midlatitudes is dilution of polar vortex air (described by a proxy based on the volume of polar stratospheric clouds). The trend in the dilution proxy reaches up to -0.2 ± 0.15 DU/year (-0.60%/decade) in all midlatitude regions during late spring. Copyright 2007 by the American Geophysical Union.