Measuring soil movement from space

“In this recent blog post, Matthew North gives some information on the work he is doing in his PhD on soil movement”

In the UK we rely on infrastructure for transport, trade and commerce. According to the UK government, there was approximately 400,000 kilometres of road in the UK in 2013, and approximately 16,000 kilometres of railway. As you can imagine, over such vast distances there are many environmental risks posed upon them. Potentially, one of the biggest and most variable risks to infrastructure networks is soil related ground movement. Soil ground movement includes processes such as subsidence, shrink and swell, soil erosion and washout. Because infrastructure typically spans across many soil types, the risks posed to infrastructure networks are often varied and non-linear. Several other environmental conditions also influence the amount of ground movement potential, such as the prevailing weather conditions and climate, drainage, slope, altitude and trees. These factors all have the potential to cause substantial damage to infrastructure. Even the smallest (millimetre scale) soil movements can act to weaken infrastructure, and over a long time these small movements can lead to asset failure. So this is great, but how do you measure soil related impacts to 400,000 kilometres of road network and 16,000 kilometres of railway? Well one possible answer is of course remote sensing!

There are a range of satellites and techniques available for monitoring soil movement, all with varying spatial and temporal resolutions, archives of data, and associated costs. For the work discussed in this blog, freely-available Sentinel 1 data (distributed by the European Space Agency) was used to measure and monitor soil related surface deformation of minor roads, major roads and railways in six different UK sites. The technique chosen is called the Persistent Scatterers Interferometry (PSI) with all processing being undertaken in a software package called ENVI SARscape. It’s all a bit of a mouthful, but essentially the combination of Sentinel 1 and PSI can measure millimetre movements of infrastructure caused by soil movement from 693 km’s above Earth. Pretty neat, right?

Figure 1 (below) gives a simplified overview of the methods used for this investigation.


For this study, a total 2,397 km2 of land was investigated, which included analysis of over 6,084 km of roads and 343 km of railways. In these six sites (Bristol, Bath, Bournemouth, Grantham, Kings Lynn and Peterborough), a range of land covers ranging from mixed rural to dense urban were included. A wide range of different soil types were investigated, and analysis of 7 major soil groups were included, all with varying soil movement potentials. As this is a soil blog, a description of the major soils are provided below, and there expected ground movements:

Table 1: A description of the major soil groups investigated in this study

The image processing technique is relatively complex, and I want to keep this blog as simple and easy to understand as possible. For a full account of the methods take a look to the paper which I link at the end of the blog. But put simply, the PSI technique uses the reflectance of man-made structures (like roads and railways) in a stack of 20+ satellite images. As soil moves infrastructure the brightness of infrastructure, as shown in the satellite image, changes ever so slightly. These very small changes can be quantified and converted into measurements using basic SAR principals.
These surface deformation measurements can then be analysed in a Geographic Information System where additional datasets, such as soil maps, geology, and weather and climate data are analysed with the PSI measurements to reveal relationships between soil-related infrastructure movement and environmental conditions. Statistical analysis of many hundreds of thousands of measurements can then be undertaken to reveal patterns across all in the infrastructure types in relation to the soil map.
Never before have I got so excited about millimetre movements! But the results showed infrastructure which overlaid 4 out of 7 of the major soil groups exhibited a seasonal deformation pattern in response to soil shrink and swell. The patterns of deformation were clear, a marked period of shrink for the summer drying period and heave for the winter wetting. These soils (Ground-water Gley soils, Surface-water Gley soils, Pelosols and Brown soils) are commonly found in the UK. These results are hopefully useful for infrastructure and utility managers wishing to know greater details about the risks posed upon infrastructure networks. Interestingly, very negligible differences were observed between minor and major roads, which is potentially an unexpected result. Some of the highest subsidence was found for Railways over peat soils, in particularly the Holme Fen nature reserve, East Anglia, UK (see Figure 2). Linear subsidence of up to -7.5mm per year was observed for this railway, see Figure 3. Subsidence of this magnitude can eventually lead to warping of the railway tracks, and also the un-safe operation of rail infrastructure, which is obviously a cause for concern.

Figure 2: The railway investigated at Holme Fen, East Anglia, UK. The railway and embankment are subsiding at 7.5mm per year due to the drainage of peat soils in the nature reserve. (Photo credit: A life spent chasing train’s blog.)

Figure 3: Time series analysis of the major soil groups and infrastructure types analysed in this paper. Note the linear subsidence of railways in Peterborough, showing a -7.5mm linear trend of subsidence

So, drawing to an end, this blog has hopefully introduced a new technique to monitor soil movements and infrastructure in a way that you might not had previously thought about. The potential for a nation-wide scale investigation of ground movement is still a little way off, mainly due to the vast amounts of data needed and computational resource. It is not to say that it isn’t possible though.
I’m happy to receive any comments or feedback for my first Dirt Doctors post. The full paper is available here!

Author: Matthew North

Twitter: @mattrobertnorth