July 2020

Integrating Soil and Remote Sensing Data: Monitoring Peatlands

Statisticians at the British Geological Survey have devised and tested novel statistical models that can integrate different types of data to monitor soils cost-effectively and at a scale and accuracy that was not previously possible.

Soils perform many vital roles in maintaining the environment and in providing raw materials and food. For example, peatlands are important habitats for wildlife, reduce the risk of floods and contribute to climate regulation through the storage of carbon. In common with many soil types, peatlands are under threat from human activities such as drainage, grazing, liming and afforestation. When peatlands are damaged, stored carbon is released adding to the greenhouse gases in the atmosphere. There is an increasing need to monitor the quality of peatlands in particular and soils in general to see where their quality is under threat and what protection measures are proving to be effective.

Generally, soil quality is monitored by visiting a location and by measuring properties of the soil either in situ or by taking samples to be analysed in a laboratory. The depth of peat might be measured by inserting a metal pole into the ground (Figure 1). This ground-based approach gives precise information about attributes of the soil at that location but it is time consuming, expensive and impractical to repeat across the landscape. In contrast, remote sensors mounted on airplanes or satellites can provide data that cover much larger areas. However, these data are only likely to be indirect measurements of the quality of soil. For example, airborne radiometric sensors measure the quantity of radioactive isotopes in the earth. Peat soils attenuate or reduce the radiometric signal that is emitted by rocks and a small radiometric signal might be an indication of deep peat. However, the relationship between peat depth and the radiometric data is likely to be complex since just 50-60cm of peat is required to attenuate the radiometric signal by 90% (Beamish, 2014).

Metal measuring pole in field
Figure 1: Measurement of peat depth.


It is only possible to monitor soils across large regions effectively by combining or integrating these different sources of information. This integration requires flexible models which are consistent with data sources that vary and relate to each other in complex ways. We therefore established a novel model to integrate ground-based measurements and remote sensor data which used spline functions to allow a nonlinear relationship between the two variables and for the strength of the relationship to vary. When this model was applied to 425 previously published measurements of peat depth (Young et al., 2018) in the Dartmoor National Park and radiometric potassium (K) data from the Tellus airborne geophysical survey of South West England (http://www.tellusgb.ac.uk/) it indicated that the uncertainty in the relationship between these variables was largest for the deepest peats (Figure 2). For instance, where the radiometric K signal was 0.1% the peat depth could plausibly be less than 50 cm or greater than 350 cm but for radiometric K greater than 1.5% we can be more certain that the peat depth is less than 50 cm.

Graph displaying relationship between peat depth and Radiometric K
Figure 2: The measured radiometric K % and peat depths cm (black dots), the estimated relationship between those quantities (black line) and the 90% confidence interval for that model (grey shading).


We were able to use this model and the two data sources to map peat depths across a southern portion of the national park more accurately than had previously been possible (Figure 3, left and centre). We were also able to demonstrate that the new model could be used to reduce the costs of soil monitoring by reducing the number of ground-based measurements that were required. We extended and applied a survey optimization procedure (Marchant et al., 2013) and found that the ground-based measurements could be mostly focussed where the radiometric K was low and the peat depth uncertain (Figure 3, right).

Three visual representations of the surveys
Figure 3: (left) The measured radiometric K % across a southern portion of the Dartmoor National Park and locations of existing peat depth measurements; (centre) Map of peat depth cm predicted using radiometric K % measurements and peat depth measurements; (right) Locations of an optimized 100 point survey of peat depths.

Author: Ben Marchant


Beamish, D., 2014. Peat Mapping Associations of Airborne Radiometric Survey Data. Remote Sensing, 6, 521-539.

Marchant, B.P., McBratney A.B., Lark, R.M., Minasny, B. 2013. Optimized multi-phase sampling for soil remediation surveys. Spatial Statistics, 4, 1-13.

Young DM, Parry LE, Lee D, Ray S, 2018. Spatial models with covariates improve estimates of peat depth in blanket peatlands. PLoS ONE 13 (9): e0202691.


This project was part of Ensemble’s seed-funding for soils and digital technology projects. The work of Ensemble and subsequent grants has been funded by the UK EPSRC as part of the Senior Fellowship in the Role of Digital Technology in Understanding, Mitigating and Adapting to Environmental Change grant no: EP/P002285/1.

Header photo credit:Guy Wareham, Spot height 563, peat moorland with Fur Tor just visible on the horizon.