Updating Peatland Condition Mapping

• Soils underpin terrestrial life and play a major role in regulation of greenhouse gases.
• Peat can act as a carbon sink when in good condition, or be a major source of emissions when in poor condition.
• Management of Scottish peatland including restoration of damaged peat has a major role to play in meeting The Scottish Government's commitment to reduction in overall greenhouse gas (GHG) emissions of 80% by 2050.
• Mapping of peatlands and understanding their carbon stock is essential to implementing and improving soil carbon management practice.
• Mapping also contributes to estimates of spatial distribution of emissions and links with other land management practices and plans.

The potential emission of carbon from peatlands is linked to both the land cover/use above the peat, and the condition of the peat itself. Pristine, undamaged and unmodified peat has been shown to act as a sink for carbon whereas degraded or modified peat emits carbon based on the associated land cover or use. Peat Condition categories were defined by a team led by the UK Centre for Ecology and Hydrology (UKCEH) in 2017 ( Evans et al. 2017 ) and used by the UK Government as part of the  Greenhouse Gas (GHG) Emissions Inventory calculations . These categories have since been further developed by  Hannah Clilverd (UKCEH)  and  Rebekka Artz (James Hutton Institute) . See below for a table of categories and emission factors as used in this work.

Before we can determine peatland condition class and associated emissions, we first need to determine the area of peatland in Scotland, then identify areas where erosion is present and/or the peat has been drained - these are major factors in classifying peat as 'degraded'.

1.  Underlying datasets to determine peatland extent: 
   1. Bulk Density
   2. Carbon Concentration
   3. Soil Profile Depth
2.  Identification of Degradation Features: 
   1. Drainage
   2. Erosion
3.  Determination of condition class using land use and land cover datasets: 
   1.  Land Cover Map 2021 (LCM)  (UKCEH)
   2.  Scottish Integrated Administration and Control System (SIACS)  (SGRPID)
   3.  Land Cover Map 1988 (LCS88)  (MLURI/James Hutton)
   4.  National Vegetation Classification (NVC)  (JNCC)
   5.  National Forest Inventory (NFI)  (Forest Research)
   6.  National Geographic Database (NGD)  (Ordnance Survey)

• Extensive soil datasets exist in the National Soil Profile Database of Scotland, held at the James Hutton Institute. But these are for discrete sites based on field observations and coverage is irregular across the country.
• Environmental co-variates such as rainfall, temperature, terrain, geology and vegetation are used to predict soil properties from field-based data.
• A Machine Learning process was used to carry out this prediction, including use of the  Crop Diversity High Performance Computing  (HPC) resource hosted at the James Hutton Institute.
• The process involves 3 major steps:
  • building training data by combining field samples and environmental covariates.
  • model training using this training data to teach the model the signatures of soil properties.
  • applying the model to mapping of the environmental covariates to produce output maps of the soil properties.

• Peatland Condition categories were calculated by integration of multiple national scale land use and land cover datasets alongside the degradation features calculated  above .
• Some datasets are more useful for certain classes due to their original purpose and classification scheme.
• The maps below do not show the original classification schemes of the datasets, but rather their reclassification into peatland condition classes before being integrated together into a final dataset. See  results  for more info.

• The Final peatland condition class is determined by a cascade across the different land use and land cover datasets listed above, with the logic being listed in detail in the final report.
• For those classes that are separated into drained and undrained variations or eroded variations, they are split using the drainage or erosion dataset.
• These steps are codified into a python script allowing rapid development and rerunning of the analysis when improved inputs are available.
• The resulting peatland condition map is shown below.

• This peatland condition map should not be used as the sole source of information when deciding on areas for restoration, or for planning any other activity.
• It is strongly recommended that this map is used as a guide only and that the user carries out field-based verification of the peatland condition on any site of interest before planning work.