Soil organic carbon sequestration: scoping study

7 How has Scotland’s soil carbon content changed over time?

There are only a few datasets where the change in SOC concentration or stocks has been quantified. At a national level the main ones are the National Soil Inventories (1978-88 and 2007-9 resampling), the Countryside Survey, LUCAS 2009 and 2015 while the North-east experimental Farm data, ‘Woodlands’ long-term rotational experiment, Centre for Sustainable Cropping (Balruddery Farm) all provide change data for cultivated soils. Although LUCAS does include a range of land use types, there are large parts of Scotland where no samples were taken (Figure 7). Other substantial change datasets include the Afforested soils dataset for Woodlands and Birse and Robertson Survey for Woodlands, grasslands and Montane habitats.

At the national scale, the NSIS resampling in 2007-9 was specifically designed to assess change in soil properties including SOC stocks and to test sampling methods. Sampling from a soil profile pit, by 0-15 cm core and by 0-15 cm composited auger borings over a 20 m x 20 m square plot. Chapman et al. (2013) reported on the change in SOC between the two sampling campaigns (1978-88 an 2007-9) from the soil profile pits (Table 4). The original location of the samples taken in the first campaign (1978-88) were revisited and resampled. There was an average of 25 years between the original and resampled soils and stocks were calculated to 100 cm. They found no statistically significant change in SOC stocks overall nor within any of the land cover types shown in Table 4 apart from a statistically significant increase in SOC stocks for soils under woodland. They also commented that, had they only measured stocks to 15cm, the results would have shown an increase in SOC stocks overall. It is important to note that, although there was no statistically significant change SOC stocks, changes were detected (Table 4). This included a decline in SOC stocks in improved grasslands. Such differences can act as an early warning but require a larger sample size to assess if there is indeed a statistically significant change.

Chapman et al., 2013 also found a statistically significant decrease (4.03 to 3.7%) in SOC concentration in arable and grassland soils but also that the average thickness of the topsoil had increased from 29 to 32 cm. There was therefore no statistical difference in overall stocks despite the reduction in the mean values of SOC from improved grasslands. The increase in topsoil thickness was attributed to deeper ploughing between the two sampling periods by which the upper portions of the less carbon-rich subsoil were incorporated into the topsoil. This may explain the decline in SOC concentration but also demonstrates the complexities in determining change over time.

A follow up study (Lilly et al, 2016) where moorland soils had undergone afforestation confirmed the increase in SOC stocks in woodlands observed by Chapman et al. (2013).

Emmett et al, (2010) reported that the three Countryside surveys (1978, 1998 and 2007) showed no statistically significant change in SOC concentrations or stocks in the top 0-15cm in Scotland. However, they did see a significant decrease in SOC concentration between 1998 and 2007 from 3.56 % SOC to 3.23 % in the Arable and horticultural habitat but no significant change in stocks Table 5. Since the soils were sampled to a fixed depth then a decline in SOC concentration but no change in stock would suggest an increase in bulk density, that is, a more compact topsoil.

* statistically significant decrease from 1998-2007 sampling

The LUCAS dataset comprises 190 sites sampled in Scotland in 2009 and 234 in 2015 (Figure 7). Of these, 150 sites were sampled in Scotland in both years. These repeat samples encompass a number of habitats that are not always equivalent to those in IACS, NSIS or Countryside Survey. In addition, the 2009 soil dataset does not have the land cover at the time of sampling. Although Table 2 shows the mean values for the two LUCAS sampling campaigns, further data analyses is needed to evaluate this dataset for its ability to detect change over time in SOC concentrations. The time frame of the LUCAS samples means that if the positional accuracy of the samples is sufficient it may be possible to use IACS data to determine the land use history at each site and help quantify the differences in SOC concentration in arable, rotational and permanent grasslands.

In 2017, Lilly et al. (2019) resampled 37 sites in North-east Scotland that had been previously used by the Macaulay Institute for Soil Research to study crop response to P fertiliser in the 1950s up to 1980. A screw auger was used to take samples on a ’W’ pattern to a depth of 15 cm as was done previously. The land use varied from cropped fields to long-ley grassland. The original sample was retrieved from the National Soils Archive and reanalysed for SOC concentration alongside the fresh 2017 samples.

The samples were taken on average 54 years apart and the mean SOC concentrations were 4.19 % (±1.26 %) and 4.48 % (± 1.36 %) for the original samples and 2017 samples respectively. There was no statistical difference between the two set of samples. Although bulk density samples were taken in 2017, none were taken during the original soil fertility experiments. The localised nature of the study and the mixed land uses mean that care needs to be taken when interpreting this dataset alongside other national scale assessments as they are not directly comparable. The fixed depth sampling means that any possible increase in topsoil thickness was not assessed and, while topsoil thickness was measured in 2017 (providing a basis for future sampling), it was not during the original fertility experiments.

The Centre for Sustainable Cropping (CSC) at Balruddery Farm has samples from 6 arable fields collected annually since 2010 on a 18m grid pattern and at 0-20 cm (350 samples per year). This dataset provides a detailed estimation of both spatial and temporal variability in SOC concentrations.

It is clear from the list of datasets in Appendix 1 that few were collected specifically to investigate change in SOC over time, but it is possible that many could be adapted to do this. As the original sampling was done for many different reasons and designed to answer a different set of questions, care would be needed in deriving a clear set of protocols to maximise the information that could be derived. While it is better to have consistent datasets with specific sample designs, protocols and at an appropriate scale to answer specific questions on SOC, concentrations, stock and storage potential, the existing datasets could be used to give rapid, interim measures of change and also help in establishing sample densities for a monitoring scheme. Robust statistical advice would also be needed to establish if datasets could be amalgamated.