Global Workshop on Digital Soil Mapping

14-17 September 2004, Montpellier (F)

One of the hottest topics in soil science today is so-called “digital” or “predictive” soil mapping (abbreviation DSM). The idea is to use as much secondary digital data as possible (digital elevation models, remote sensing images, existing soil surveys and point observations) to optimise soil survey: sampling plans, further fieldwork, map and database production, with the aim of producing a digital product which can be directly used in a GIS and interpeted on-demand, replacing the traditional static soil map. The excitement comes from the convergence of three trends: (1) the ever-increasing free or cheap digital data at increasingly-finer resolutions (e.g. SRTM elevation data), (2) increasingly-sophisticated analytical techniques (much of which is the subject of the Pedometrics IUSS Comission 1.5) and related software, and (3) successful applications in diverse environments.

To capture this excitement and keep the revolution on-track, an ad-hoc group headed by Philippe Lagacherie and Marc Voltz (INRA Montpellier) and Alex.

McBratney

(University of Sydney) decided it was high time for a workshop “to review, discuss and help develop new, rapid and economic methods for digitally mapping soil classes and attributes (and their uncertainties) principally at resolutions of 20 metres to 500 metres and extents from 5 km² to 50 000 km².” The workshop was sponsored by the IUSS, the French Soil Science Society, and hosted by LISAH (Laboratoire d'étude des Interactions Sol - Agrosystème - Hydrosystème) at INRA (l'Institut national de la recherche agronomique), Montpellier. The organisers designed a format with targeted keynote addresses by leading practicioners (9) and very brief presentations by active workers (52) organised into themes, each with extensive but structured discussion led by a discussant and summarized by a reporter. This correspondent has never been to any scientific event where more of the time was used productively, where more of the focus was on making progress and identifying the way forward rather than on posturing and preening. This was due to the ideal size (81), to the format, to the excellent hospitality (including lunches in the INRA cafeteria which outdid many a conference dinner), and above all to the cooperative spirit of almost all the participants.

Themes included (1) progress, examples and economics, (2) sampling methods, (3) representation and visualisation, (4) quality assessment, (5) new covariates, especially from continuous-field near and remote sensors, and (6) quantitative modelling to predict soil classes and attributes. Operational examples of general-purpose maps came from Australia and France, with pilot projects in the USA, Finland, England & Wales among others. Other work was aimed at specific applications (e.g. organic C stocks, desertification risks, ecological functions). New sensors which have proven successful in defined circumstances include gamma-ray spectrometry, electromagnetic induction, and hyperspectral imaging along with old favourites such as aerial photography and digital elevation models.

A running debate was between those who use “data mining” techniques to search for predictors and those who look first for understanding of the soil-landscape relations either to select predictors or to model these relations explicitly. The first group point to the huge supporting datasets and practical results, while the second group point to the classical success of Jenny-like conceptual models in understanding soilscapes. McBratney, Mendoça and Minasny (Geoderma 117:3) have tried to bridge the gap with the so-called ‘scorpan’ approach; this neologism was used freely in the workshop and provided a useful framework for discussion. This empirical approach is implicitly based on pedogenic knowledge. In addition to the well-known cl, o, r, p, t (renamed c, o, r, p, a = ‘age’) this include s = known properties of the ‘soil’ at a point to be predicted and n = spatial position (‘neighbourhood’).

The proceedings of the workshop, including keynote talks, other talks and results of the discussion sessions will be published as a book in the Elsevier Developments in Soil Science Series (editors : Lagacherie, McBratney and Voltz). Each contribution will have two referees’ reports, continuing the interactive spirit of the workshop. This book should provide plenty of good ideas to anyone considering research in this area. Already there is a second workshop being organised by Maria de Lourdes (‘Lou’) Mendonça at Embrapa Solos in Rio de Janeiro for 02-04 July 2006, where emphasis will be on application of these techniques to large areas of the world without proper soil maps, as well as on increasingly-sophisticated approaches (especially knowledge-based) and full exploitation of new sensors and data sources

D.G. Rossiter

ITC Enschede

E-mail rossiter@itc.nl

Close Window