Post Conference Report and Assessment by R. A. MacMillan
LandMapper Environmental Solutions Inc., Edmonton, Alberta, Canada.

The theme of the 3rd Global Workshop on Digital Soil Mapping, held in Logan, Utah from September 30 to October 3, 2008 was “Bridging Research, Production and Environmental Applications”.  Janis Boettinger and her team of local organizers, assisted by NRCS personnel from California (David Howell) and West Virginia (Amanda Moore), are commended for organizing and hosting a most interesting and successful event.

The venue at Utah State University in Logan was excellent with a comfortable and well equipped auditorium, convenient and affordable accommodations and healthy and efficient refreshment breaks and lunches. The program was well designed to highlight recent advances in methods and concepts while simultaneously addressing how these research methods can be rapidly adopted for operational mapping and applications.

A field trip was scheduled for the first day of the workshop (September 30) in order to bring registrants together to renew old acquaintances or forge new ones before the start of the formal program. The trip exposed participants to a wide diversity of landscapes and soils in a traverse from the pluvial, alluvial and deltaic sediments of Lake Bonneville in the eastern portion of the Basin and Range physiographic province, through the beautiful fall scenery of the Logan Canyon, with its residual, colluvial, lacustrine and (potentially) glacial sediments to the overthrust sedimentary rocks of the Bear River Range of the middle Rocky Mountains. Janis displayed her passion for the area and its environment by interpreting the evolution of the landscape and soils in the area in great detail and with great enthusiasm. Her explanation started with the geological evolution of the physiography of the Basin and Range and Rocky Mountain provinces, continued with elaboration of the sequence of flooding, sedimentation and retreat of pluvial Lake Bonneville, investigation of the effects of salinity on soils within the valley sediments and concluded with a review of the effects of aspect on climate and soil development in steeper terrain. The trip had 10 stops and 9 soil pits. Janis picked some strikingly beautiful locations to demonstrate how the soils in her area responded to and reflected the environmental controls under which they formed. While there was much interest in the soil profiles exposed in the pits, there was also considerable interest in looking up and out from the pits to interpret how the observed soils related to the landscapes in which they developed. This recognition of the intimate relationship between soil classes or properties at a point and controls exerted by the larger landscape setting was actually an on-going, if unstated, theme of many of the conference presentations that followed.

One major theme, repeated in many presentations, was the importance of context and hierarchies in pretty well all efforts to model the spatial distribution of soils or soil properties, especially over larger areas. Clearly, the scorpan model is based on acknowledging that environmental conditions, operating over multiple spatial scales, control the development and spatial distribution of soils and soil properties. Still, even now, there is a tendency by soil modelers to not recognize the importance of environmental hierarchies explicitly and formally. The idea of establishing environmental context that appeared in many presentations was perhaps best summed up by the homosoil concept as presented by Alex McBratney. It appears that digital soil mapping methods are beginning to relearn the importance of establishing formal hierarchies to capture spatial environmental controls operating at different scales so as to better define domains or regions of like conditions within which modeled outcomes are more likely to be predictable and model rules are more likely to be successful.

A second major theme was increased recognition of the importance of ancillary data sets, not obtained from analysis of digital elevation data (DEMs), as inputs to a wide variety of digital soil mapping procedures. Of these, the most widely reported, was increased use of remotely sensed imagery to improve predictive success. However, several other sources of ancillary input data were also described, including airborne and ground measured electrical conductivity and near infrared spectroscopy. There is an overlap here with the previous point, in that much of the use of image data sets was directed at establishing context in terms of vegetation, geology, land cover, mineralogy, age or hydrology to improve predictive models and rules within regions or domains.

We seem to be entering a second cycle of digital soil mapping in which we appear to have learned something from the problems and deficiencies that were encountered during the first cycle and are beginning to develop strategies to address these problems. While many DSM applications continue to make prominent use of similar basic terrain derivatives (e.g. slope, aspect, profile and plan curvature, and wetness index), many others are beginning to recognize, and address, the deficiencies in these initial basic measures. A more critical evaluation of the meaningfulness and utility of local measures of surface shape and orientation computed within local 3×3 windows was evident in several presentations. More attention was given to computing and assessing the utility of these various terrain metrics within larger windows that capture shape, form and context over a wider range of scales. Other efforts aimed at defining domains or regions within which these local measures were more meaningful were noted. Among inference methods, decision trees, and in particular random forests, appear to have achieved strong acceptance.

At the previous DSM workshop in Rio in 2006, Alex McBratney indicated that one of his criteria for assessing the success of a meeting was whether he had been exposed to any really exciting or interesting new ideas. We had a similar discussion during lunch on the final day of this workshop and asked the people around our table to pick out the one new idea that most captured their interest. It was indicative of the vigor and success of the workshop that so many different ideas were identified as exciting by even the small group at this one table.

A vote was held amongst those still in the audience at the end of the workshop and Alex McBratney’s idea of trying to model the parameters of functions that predict the variation in any soil property or attribute with depth across x, y space was picked as the most exciting new idea for the largest segment of the remaining audience. There was also considerable support for several other new ideas, including the conmap concept presented by Thorsten Behrens. Other new ideas that excited attendees included that of trying to explicitly map the age (a) or time factor in the scorpan equation as presented by Jay Noler and of trying to expand recognition of the spatial position factor (n) from simply recognizing x, y spatial position to providing information on spatial continuity or of spatial context with respect to any measure of interest.

It is a measure of how far and fast DSM has come that the great idea represented by use of proximal infrared remote sensing spectroscopy to rapidly infer many soil properties previously only obtainable by laboratory analysis was no longer viewed as sufficiently new to attract the majority of votes as the best new idea at the meeting. Similarly, the concept of a global digital map of soil properties (GlobalSoilMap.net), while undoubtedly ambitious and audacious, was no longer deemed “new enough” to attract the majority of votes. So the meeting was not short of new and exciting ideas and there appeared to be something exciting there for everyone.

So how well did the meeting do at addressing its proposed theme of “Bridging Research, Production and Environmental Applications”?  In my personal view, the vast majority of presentations still fell within the domain of research trials with far fewer presentations that dealt explicitly with how to move the research into production or end use. However, we can consider the global digital soil map, the Mojave DSM initiative and TEUI as examples of projects that are intended to move digital soil mapping forward from a research focus to an operational and applications focus. In my view, it is critically important that these initial “production” DSM projects succeed in producing useful and reliable products that will demonstrate that digital soil mapping methods have real validity for operational use. The discipline of DSM needs to have several clear and demonstrable successes, in projects that have the potential for significant national and international level impact, if we are to maintain on-going enthusiasm and support for DSM. We risk slipping into the realm of an irrelevant curiosity if we are not able to convincingly demonstrate useful applications of our methods.

I hope that everyone who attended the workshop enjoyed and profited from it as much as I did. It was a pleasure to see and meet many new attendees and to reconnect with many old friends and colleagues. I was encouraged at the skills, enthusiasm and geographic diversity of the many new participants. I encourage both existing and new DSM practitioners to attend these workshops and to present their work. These small, focused DSM meetings provide a great opportunity for people interested in digital soil mapping to make valuable personal contacts and to get exposed to new ideas and methods at their formative stages.

R. A. MacMillan
LandMapper Environmental Solutions Inc., Edmonton, Alberta, Canada