Abstract. Greenhouse gas stabilisation in the atmosphere is one of the most pressing challenges of this century. Sequestering carbon in the soil by changing land use and management is increasingly proposed as part of climate mitigation strategies, but our understanding of this is limited in quantitative terms. Here we collate a substantial national and regional data set (15790 soil cores), and analyse it in an advanced statistical modelling framework. This produced new estimates of the effects of land use on soil carbon stocks in the UK, different in magnitude and ranking order from the previous best estimates. Soil carbon stocks were highest in woodlands, followed by rough grazing and semi-natural grasslands, then improved grasslands, and lowest in croplands. Estimates were smaller than the previous estimates, partly because of new data, but mainly because the effect is more reliably characterised using a logarithmic transformation of the data. With the very large data set analysed here, the uncertainty in the differences among land uses was small enough to identify consistent mean effects. However, the variability in these effects was large, and this was similar across all surveys. This has important implications for agri-environment schemes, seeking to sequester carbon in the soil by altering land use, because the effect of a given intervention is very hard to verify. We examined the validity of the 'space-for-time' substitution, and although the results were not unequivocal, we estimated that the effects are likely to be over-estimated by 5\u201333 %, depending upon land use.