Vol 437 8 September 2005 doi 10 1038 nature04038 LETTERS Carbon losses from all soils across England and Wales 1978 2003 Pat H Bellamy1 Peter J Loveland1 R Ian Bradley1 R Murray Lark2 Guy J D Kirk1 More than twice as much carbon is held in soils as in vegetation or the atmosphere1 and changes in soil carbon content can have a large effect on the global carbon budget The possibility that climate change is being reinforced by increased carbon dioxide emissions from soils owing to rising temperature is the subject of a continuing debate2 9 But evidence for the suggested feedback mechanism has to date come solely from small scale laboratory and field experiments and modelling studies2 9 Here we use data from the National Soil Inventory of England and Wales obtained between 1978 and 2003 to show that carbon was lost from soils across England and Wales over the survey period at a mean rate of 0 6 yr21 relative to the existing soil carbon content We find that the relative rate of carbon loss increased with soil carbon content and was more than 2 yr21 in soils with carbon contents greater than 100 g kg21 The relationship between rate of carbon loss and carbon content is irrespective of land use suggesting a link to climate change Our findings indicate that losses of soil carbon in England and Wales and by inference in other temperate regions are likely to have been offsetting absorption of carbon by terrestrial sinks The National Soil Inventory was made to obtain an unbiased estimate of the distribution of the soils of England and Wales and of the chemistry of the topsoil 0 15 cm depth 10 Samples were collected and soil profiles described at the intersections of an orthogonal 5 km grid over the whole area Methods This yielded about 6 000 sites of which 5 662 could be sampled for soil Figure 1a shows the distribution of soil organic carbon contents across England and Wales measured in the original sampling 1978 83 Sufficient subsets of the sites were resampled at intervals from 12 to 25 yr after the original sampling to be able to detect changes in carbon content with 95 confidence Methods This was done in three phases in 1994 95 for arable and rotational grassland sites 853 of the original 2 578 sites in 1995 96 for managed permanent grassland sites 771 of the original 1 579 and in 2003 for non agricultural sites bogs scrub rough grazing woodland and so on 555 of the original 1 505 Roughly 40 of the original sites were resampled This is the only soil inventory on such a scale anywhere in the world to have been resampled To allow for the varying time interval between samplings annual rates of change in carbon were calculated for each site by assuming that the process of change was linear over the sampling interval An analysis of known rates of change in soil carbon under different conditions showed this to be reasonable Figure 2 summarizes the results grouped by soil type and land use Some differences between soils and land uses are apparent for example peat soils lost carbon an order of magnitude faster than brown soils and man made soils and bogs and upland grass lost carbon an order of magnitude faster than lowland heath which appears to have gained carbon on average But we found no statistically significant relations between rate of change and land use rainfall class or soil textural class whether for the data as a whole or for outlying data However we found a significant negative linear correlation between rate of change and original organic carbon Figure 1 Changes in soil organic carbon contents across England and Wales between 1978 and 2003 a Carbon contents in the original samplings and b rates of change calculated from the changes over the different sampling intervals Values at sites that were not resampled were calculated from their original organic carbon contents using equation 1 The changes were negative in all but 8 of the sites 1 National Soil Resources Institute Cranfield University Silsoe MK45 4DT UK 2Rothamsted Research Harpenden AL5 2JQ UK 2005 Nature Publishing Group 245 LETTERS NATURE Vol 437 8 September 2005 Figure 2 Rates of change in soil organic carbon content grouped by soil type and land use a Soil type24 grouping b land use grouping Circles indicate mean values error bars indicate 95 confidence intervals content Corg that is the rate of loss increased with Corg Fig 3 The relation applied for the data as a whole and for the three main land use groupings separately though with different slopes and intercepts We took a random subsample of 1 000 observations from the data as a whole and obtained the following relation by linear regression specifying an exponential variogram for the residuals Methods Rate of change in Corg 0 6 2 0 0187 original Corg 1 The standard error s e of the intercept is 0 148 and the s e of the slope is 0 00081 the rate of change in Corg is in units of g kg21 yr21 and Corg is in g kg21 The residuals from this regression applied to the whole data set show some regional features notably a tendency to overestimate the rate of loss in uplands in the northwest and southwest of England But only 8 of the unexplained variation is spatially structured so a more sophisticated statistical model is unjustified Figure 1b shows the distribution of rates of change across England and Wales for all the sites with values for the sites that were not resampled obtained using equation 1 From the data in Fig 3 the relative rates of change rate of change mean Corg over sampling interval in units of yr21 were 1 43 0 14 20 69 21 84 22 71 23 01 and 22 35 for original Corg ranges in g kg21 0 20 20 30 30 50 50 100 100 200 200 300 and 300 respectively Hence the relative rates of loss also tended to increase with Corg although not above 300 g kg21 Soils with Corg 50 g kg21 did not lose significant amounts of carbon that is not detectable over 10 yr and those with Corg 20 g kg21 appear to have gained it However for soils with Corg 100 g kg21 relative rates of loss over the survey period were 2 yr21 Given that the bulk of the Figure 3 Rates of change in soil organic carbon content grouped by original carbon contents and indicated land uses a Arable rotational grass b permanent grass c non agricultural d all The ranges of carbon content g kg21 are from left to right in each panel 0 20 20 30 30 50 50 246 UK s carbon stocks are in organic soils11 this result gives cause for concern Table 1 gives estimates of the total changes in carbon in the upper 15 cm of soil for the