OR/15/030 Methodology

Nicole Archer, Barry Rawlins, Stephen Grebby, Ben Marchant and Bridget Emmett. 2015. Identify the opportunities provided by developments in earth observation and remote sensing for national scale monitoring of soil quality. British Geological Survey Internal Report, OR/15/030.

Figure 1 shows our approach to address the aims of the project.

Figure 1 Summary of the approach used in this review study - Flow diagram of the literature review process to create a database of relevant literature and information (OUTPUT 1), which was then used to write OUTPUT 2: a review report of the most relevant findings of remote sensing and soil physical properties.

The literature review had three types of investigation to provide the relevant information and data for an up-to-date assessment. The literature review was divided into three searches: 1) Global Expert Survey, 2) Published literature search and 3) Internet search. These searches provided relevant published literature, internet data and sources, and the most recent up-to-date research from experts working in remote sensing. The data was compiled into a database to store and categorise the relevant material and is shown as Output 1, in Figure 1. The database was used to write the final review report (Output 2, in Figure 1).

Published literature search

A literature search of published information was undertaken using the Thompson Reuters Web of Science literature search engine. A systematic topic search was done, using a combination of terms listed in Appendix 1 (Table 1), which provided 43 separate search criteria. The combination of search terms was considered in four ‘fields’: the first ‘field’ always used the term ‘Soil’; the second ‘field’ was related to the terms ‘remote sensing’, ‘satellite’, or ‘airborne’ and more specific search terms relating to remotely sensed sensors, such as ‘microwave’, ‘LiDAR’ ‘hyperspectal’, ‘thermal’, ‘radar’ and ‘unmanned aerial vehicle’ and; the third field was related to a physical soil property, such as ‘organic carbon’, ‘geochemistry’, ‘bulk density’, ‘soil depth’. Each time a search was done using a combination of terms, the number of ‘hits’: i.e. references found by the search engine were recorded (and are shown in Appendix 1, Table 1). If there were too many ‘hits’ (above 600 articles), more defined search terms were used and the search was run again. The search terms used in the published literature search are shown as grey rows in Appendix 1. A total of 2,885 published citations and their abstracts were downloaded to EndNote; a reference management software package, which manages bibliographies and references. All duplicate references were discarded and the final reference list in EndNote was 2,031 published references.

This final reference list in EndNote was then divided into three EndNote reference libraries and distributed to three experts, who then read through each reference and discarded those that were not-relevant. The criteria for keeping a reference were that the remote sensing method used monitored or measured one or more of the following:

1. SOC
2. Bulk density
3. pH
4. Soil depth
5. geochemical soil indicators: N, P, K, Mg, Cu, Zn, Ni and Cd
6. peat

Once all irrelevant published references were deleted, the three resulting libraries were merged to create a final reference EndNote library. The final merged library contained 307 published references.

Internet search

Two types of Internet searches were undertaken, using the Google Chrome browser to find relevant published references, website references, sources and blog sites relating to remote sensing of soil physical properties (shown in Appendix 1, Table 2). The first search was a Google Web search and used two search criteria: 1) ‘remote sensing’ + ‘soil’ + ‘physical’ + ‘properties’ and 2) ‘remote sensing’ + ‘soil’ + ‘monitoring’. Theses searches gave an enormous amount of hits (between 2 and 4 million), we therefore only searched the hits of the first 10 pages.

As the second search criteria provided only 1 useful reference, we only used the search criteria ‘remote sensing’ + ‘soil’ + ‘physical’ + ‘properties’ for a Google Blog search, which found 4 useful hits from the first 10 pages of the Blog Search.

In total, 44 useful references were found and added to EndNote. These references were added to the final merged EndNote library, providing in total 351 references.

Global expert survey

Based on our knowledge of the soil remote sensing community, we selected and wrote to six world experts (see list in Appendix 2, Table 1) to ask whether there were new techniques which had not yet been published that may be worthy of consideration in this review. The letter sent to these experts is also shown in Appendix 2. Where the experts provided new references relevant to our review we included these in our reference database and where they provided relevant, new insights we included these in our overall interpretation. We also undertook a telephone interview with members of a project team based at the newly established company Rezatec, who are currently undertaking a project to develop a peat spotter service, where the objective is enable accurate, cost-effective and appropriate (in line with international standards) methods for identifying the location of peatlands, quantifying their intactness and assessing their carbon content.

Structuring and assimilation of information

All the new (n=15) references from the Global Expert Survey were added to the final merged EndNote library (containing 351 references). This library was again divided into three smaller libraries and given to three experts who again discarded irrelevant references which were considered to be:

1. of no relevance in a UK environment (many of the references related to arid and semi- arid environments).
2. related to the monitoring of irrelevant physical soil properties.
3. using inappropriate remotely sensed techniques; for example many references were related to using spectral techniques proximally (at ground level) or in laboratory conditions.

The remaining total number of references was 186. These final references were then categorised using a ‘tagging’ procedure in EndNote. Each reference had a Keyword field which was ‘tagged’ using keywords, relating to the remote sensor and the physical soil property being measured. These keywords were the following:

1. Hyperspectral
2. SAR
3. Microwave
4. UAV
5. Thermal
6. Radiometric
7. Spectral
8. Soil texture
9. Electrical conductivity
10. Bulk density
11. pH
12. Soil depth
13. SOC
14. Geochemistry
15. Erosion

By using tagged keywords, the references were then categorised quantitatively to understand which remote sensor was mainly used to measure different soil properties and which physical soil property had been researched the most.