The world’s soils are an important resource. Mapping them enables scientists and food producers as well as land managers with the opportunity to maximize production while ensuring sustainbility and meeting conservation goals. A practical first step to achieving this goal involves an inventory of the global soil resources. V1 Magazine editor Jeff Thurston interviewed Bob Macmillan, scientific advisor to GlobalSoilMap, an international initiative involving many organisations and individuals around the world who are attempting to complete this work.
V1 Magazine: What is GlobalSoilMap.net?
Bob Macmillan: GlobalSoilMap.net is a bold new initiative to produce a global digital map of important soil properties at a grid resolution of 90 m.
We will predict values for the key soil properties of organic carbon, percent clay (texture)and pH. Having predictions of these three properties, it is possible to estimate bulk density, volumetric carbon density and available water holding capacity (AWC). Additional soil properties that we are considering mapping include cations plus exchangeable acidity (ECEC)and electrical conductivity (EC). Each property will be predicted as a continuously varying function of depth, so that the value of any property can be retrieved for any depth, or range in depth, in the soil at any location.
The essence of the project is that we are trying to collate, standardize and add value to the uneven and disparate assemblage of existing soil maps and point data that have been produced up to now but that tend to be underused and not universally available.
We have been referring to these existing data sets as legacy soils data. We want to maximize access to, and usability of, these data so as to realize maximum benefit from all the efforts (and expenditures) that went into collecting and producing this information initially.
V1 Magazine: Different organisations and people have collected soil information using different methods over time. You mentioned prediction values and uncertainty, what do you mean by that?
Bob Macmillan:We believe that the way to make all of these legacy maps and point data sets most easily assessable and effective is to use them to produce a global set of predictions of key soil property values by soil depth. Each predicted value will also have an estimate of uncertainty attached to it.
By reinterpreting and reformatting these existing data we can work to address issues related to differences in terminology, mapping concepts and methods used to compute and report soil property values in existing databases in different parts of the world. For example, it is not uncommon to find that different maps of adjacent areas may report very different values for common soil properties such as organic carbon or pH along their boundaries. We often discover that these differences result from differences in how soils were sampled, analyzed and described at different times and by different agencies. Such differences make for incompatibilities that reduce the effectiveness of current soil maps and databases. In Europe, this challenge is referred to as harmonization and it affects many trans-border data sets, not just soils. Our efforts will help produce globally harmonized estimates of soil property values.
V1 Magazine: Who is involved in GlobalSoilMap.net? Do different participants have different roles?
Bob Macmillan: GlobalSoilMap.net has been structured as a consortium of cooperating agencies, institutions and individuals. The consortium has been organized into nodes, with each node accepting responsibility for all activities within a given geographic region. We have defined seven nodes for Europe, North America, South and Latin America, Australia and Oceania, sub-Saharan Africa, Asia and West Asia – North Africa. Each node is responsible for planning, organizing, funding and implementing all activities within its area of geographic influence.
Coordination of the project is being undertaken by ISRIC – World Soil Information, located in Wageningen in the Netherlands. The project structure includes a scientific authority committee chaired by Prof. Alex McBratney from the University of Sydney in Australia.
Typically, the lead agency for each node represents the most relevant national or trans-national government agency within the geographic extent of the node that has a mandate to collect and disseminate soil information. This is logical given that these agencies have access to the largest pools of data and also have the most resources in terms of trained personnel and infrastructure. Consequently the various nodes are being led by the Joint Research Centre (JRC) of the European Commission in Europe, the USDA Natural Resources Conservation Service (NRCS) in North America, EMBRAPA-SOLOS in Brazil for South and Latin America, CSIRO in Oceania, the Tropical Soil Biology and Fertility Institute (CIAT-TSBF) in Africa and the Institute of Soil Science of the Chinese Academy of Sciences in Asia. We are still finalizing the node leadership for West Asia – North Africa and for South Asia. These node leaders provide organization and impetus to the project, lead efforts to secure necessary funding, provide access to data and provide or locate personnel and partners to implement the actual mapping.
The project is also making use of expertise in digital soil mapping that resides mainly in universities and research laboratories. These individuals have often led the efforts to develop the new technologies being used in this project. The project looks to them to assist in development and documentation of appropriate methods, training of new personnel to apply the methods and development and application of scientific standards and quality control for the project. Many of these researchers have volunteered their time and expertise to the project up to now, working within the context of existing research interests and research projects.
As far as roles go, we have identified four main functions that need to be carried out at each node, as well as within the central coordinating agency (ISRIC). These are project management, scientific coordination, collection and standardization of legacy data and fund raising. Individuals are in place in each of these roles at ISRIC.
V1 Magazine: You are speaking about a huge amount of data, much of it buried away and non-digital. What will be the challenge?
Bob Macmillan: We don’t see this as a project with a huge emphasis on collecting new data but rather on collating, reformatting and standardizing existing data. We hope to define and populate a framework that permits easy access to, and easy use of, a standardized set of soil property values.
There are some formidable challenges, both technical and organizational, to overcome and we will not solve them all immediately. Clearly, one great challenge will be to just locate and obtain all presently existing and potentially relevant maps and point data sets. Once located and obtained, there is another challenge in merging these disparate maps and data sets, in identifying where inconsistencies and errors exist and in figuring out how to resolve, or at least minimize, the adverse affects of inconsistencies and errors. This is the main responsibility of the legacy data officer in each node or location.
A second challenge will be how to deal with significant differences in the type and amount of data available for different parts of the world. In some areas the problem may be one of trying to cope with overwhelming volumes of data while in others it will be trying to deal with lack of data or poor data quality. We have devised different strategies for making predictions given different quantity and quality of existing data and maps for any given geographic area. Previous projects of a similar nature, particularly ones carried out in Australia, identified methods by which relatively sparse and scattered data sets could be mined to help extrapolate soil class and soil property maps from areas of available data to areas lacking substantial data.
The science that underlies most of our prediction efforts relies on the fact that we can model soil variation across the landscape with some success. Soil variation can be thought of as being a function of environmental and soil forming processes that are at least partially explained by readily obtainable environmental covariates. For example, a significant amount of the observed variation in soil organic carbon or pH can be related to soil forming processes that are controlled by climate or vegetation. We can obtain information on climate and vegetation at every point of interest and use this to infer, or model, the most likely value of organic carbon, or pH.
The conceptual model that we use has been termed the scorpan model to highlight the fact that soil properties can be modeled as a function of available spatial information for each of the scorpan components. These are s = soil, c = climate, o = organisms, r = relief or topography, p = parent material, a = age and n = spatial position or context. We use our existing soil maps and point data bases to help us establish rules for relationships between observed soil classes or soil properties and values of these scorpan variables at known locations. We then apply the rules to the spatially exhaustive data sets of covariates to predict values of soil properties everywhere. This is a key part of the methodology for standardizing and harmonizing the very different legacy data sets we have to work with.
We know that our predictions will be far from perfect in many locations but it is important to compare what we will produce with what is currently available on a global or continental scale. Relative to that standard, we are confident that what we can produce will represent a substantial improvement. Relative to truth, or even to very fine resolution, field scale maps of individual farms, we are not likely to fare as well.
Our efforts may be spotty and uneven at first in terms of the data present for any given area but we would hope that eventually we will identify areas that most need improvement and effect that improvement. The project will be assigning an estimate of the uncertainty of our soil property predictions for each location and soil property. Areas with the highest uncertainty will be targeted for improvement in subsequent updates of the initial database.
V1 Magazine: Ultimately then, does this work lead toward global soil modeling and greater understanding of food production strategies? Is there a connection to GEOSS, example?
Bob Macmillan: Clearly, we are trying to produce these data to serve pressing societal needs. The needs we feel we can service have to do with assessing and improving global security in the production of food and fiber, to efforts to model and mitigate climate change and assess carbon stocks, to efforts to model and assess water quality and quantity globally and regionally and to efforts to respond to crises in food production, flooding, drought, environmental degradation and so on.
We strongly believe that our approach of providing standardized and harmonized estimates of continuous variation in specific soil properties with depth is the most effective way to structure and deliver global soil information for use by modelers, strategists and decision makers working at regional to global scales. In the longer term, we can see the database evolving to support more local applications related to management at the farm or field scale. The framework is designed to accommodate such use but the actual data populating the framework initially is not likely to be of sufficient resolution or accuracy to support such intensive uses.
And yes, there is a direct connection to GEOSS. We will be familiarizing ourselves with all other relevant global initiatives and making every effort to complement these efforts. We will seek to make use of all existing related efforts. Our project has representatives who are actively involved in the GEOSS initiatives and we will work to make our products compatible with, and interoperable with, existing and emerging standards for presentation and exchange of spatial environmental data.
V1 Magazine: So where does the work begin?
Bob Macmillan: The short answer is that the work begins in Africa. Every journey starts with one first step and our first step will most likely take place in Africa.
The Africa node got started a bit ahead of most of the others because it received significant funding from the Bill and Melinda Gates Foundation and the Alliance for a Green Revolution in Africa (AGRA). So they need to select and implement prediction methods sooner than the other areas and are anxiously seeking whatever help we can provide them to do this task right now. They have already made commitments to produce initial maps and deliver initial products.
Most of the other nodes are still in initial organizational stages. If they are doing anything at all it is mainly preliminary data preparation and methods evaluation.
Technically, our first iteration of maps will represent a snapshot of soil property values that were obtained over range of times and management practices. Many of these properties are dynamic and change in time, largely in reflection of management. We will design the framework so that it can accommodate temporal change. This should permit us to model or record changes in soil properties through time in response to things like changes in management or climate.
The ability to monitor changes in soil properties and soil quality through time is of particular importance for the Africa node. This node has an additional mandate to monitor how changes in climate and soils can first be identified and then mitigated by promoting adoption of more effective integrated soil fertility management practices
The ability to monitor changes in soil properties through time is, however, a longer term vision. First we just have to fill in the grid with a first cut of our best estimates or predictions of these soil property values at specific locations.
V1 Magazine: How is the project assembling participants and how is it being financed?
Bob Macmillan: We are still trying to assemble the teams and the financial, data and computational resources that we need to implement this project operationally.
So far, we have relied on voluntary contributions, mostly provided by employees of government agencies, research centers and universities. We are also invoking our node model, wherein we look to the node leaders to identify and recruit partners and participants. Most of these partners will come from other, related, government agencies or universities but some may come from non-governmental institutes or even the private sector. As each node is officially “launched”, its efforts to grow by identifying additional partners and finding local funding become more vigorous.
Funding obtained from the Bill and Malinda Gates Foundation and from the Alliance for a Green Revolution in Africa (AGRA) has really provided the impetus to get the project started. The majority of this funding went specifically to support activities in the Africa node. However, some funding has been made available to ISRIC to support initial efforts to develop and coordinate the science, begin identifying and assembling the largest and most conspicuous legacy data sets and initiate efforts to apply for and obtain additional funding. This funding has been key to letting us build and promote the concept and identify major cooperators to take on a leadership role in each node.
Most participants to date have joined the consortium because they are excited by this bold idea and by the potential it has to put soil science in the forefront of efforts to respond to global environmental challenges. Our responsibility now is to produce concrete examples of our products and their main applications so that others can see their tremendous potential and decide they also want to contribute in some way.
We will be initiating formal fundraising activities in the very near future. We will also be working with the node leaders to develop strategies and targets for fund raising specific to each node. It may require as much as US $300 (€ 200) to complete operational mapping of the entire world. We will be targeting large scientific funding organizations as well as private sector companies with an interest in supporting projects such as ours that deliver environmental and societal benefits.
V1 Magazine: Do you see tools like geographic information mapping (GIS) and global positioning systems (GPS) and remote sensing playing a role? What would they be useful for and how can they contribute? Is this information going online and available to the public?
Bob Macmillan: Clearly, this project represents an effort to apply leading edge technologies to compile and deliver global soil information.
We will be using GIS technologies to assemble and collate all spatial data sets. We will be using remote sensing data and technologies to collect and process information from airborne and satellite sensors. The Africa node, in particular, has developed an approach that involves relating spectral properties of soil samples collected in the field with spectral signatures extracted from satellite imagery.
Many of the key environmental covariates that we will use to model the distribution of soil properties are extracted from processing of digital elevation models (DEM). The 90 m grid size adopted for this project reflects the fact that the best global data set of elevation data available at the time the project was conceived and launched was the 3 arc-second (effectively 90 m) SRTM DEM collected by the shuttle space mission. Since that time, a new 30 m ASTER GDEM has been released and several nodes have been able to obtain access to the 1 arc-second (effectively 30 m) SRTM DEM data. Selection of a 90 m grid size reflected recognition of maximum data volumes that we felt we could manage. So, while some nodes may acquire and make use of 30 m DEM data to make predictions, the final predictions are still likely to be formatted for storage and delivery at 90 m.
While many soil properties exhibit predictable trends in spatial variation associated with topography, topographic information by itself is not sufficient to support effective modeling of soil variation on a global scale. Imagery can supply a great deal of information that will be helpful in predicting soil properties. For example, image data can be used to model the lithology of surface materials in many environments and to separate areas of wetlands or organic soils from mineral uplands in other environments. Multi-date imagery can be used to model patterns of land cover and land use, which are scorpan (o) factors that exercise a strong influence over the soil properties we propose to predict. Airborne radiometric data is available for some large portions of the world and it has proven to be highly useful for modeling variation in soil parent material properties. So we will use what we have and what we can get to make the best predictions we can in each geographic area and landscape.
GPS is not likely to come much into play, except for supporting efforts to collect any new field data, such the field program being undertaken by the Africa node. While we are essentially an exercise in using legacy data, there may still be some requirements to collect new data to, for example, calibrate and standardize our legacy data sets or to assemble independent samples to assess the accuracy or uncertainty of our predictions. All new data collected will involve use of GPS to geo-reference it.
The intent of this project is to provide all information on-line, for free, using the latest and most interactive web-based delivery tools. We are still in our initial stages and have much to learn about effective on-line delivery. However, that also means that everything is still possible. We hope to store not only the current predictions but also the data and models used to generate them. This should enable rapid and regular update and provide an opportunity for interactivity. We are committed to adopting open standards and are looking to benefit from advantages in web-based delivery of spatial data.
We believe we have a good idea for making information about variation in soil properties with depth, as well as horizontally, easier to manage and interpret. Instead of storing our predictions by soil horizon or by fixed depth intervals, we propose to store the coefficients of a spline function that will describe the continuous variation of each of our predicted soil properties with depth. This offers users tremendous flexibility in the way they can query the database. They might ask for the depth at which the total carbon first exceeds some threshold of they could ask for the total carbon to a specified depth or over a depth interval of their choosing. Similarly, they might pose a query about which areas can store some specified amount of water in the top 40 cm of soil and so on. This approach opens up possibilities that would be difficult, if not impossible, to extract from conventional soil profile databases and soil maps.
V1 Magazine: While you express the immediate need to collate global information, is the project leaving the door open to add to this work? In other words, can future soil related work tie into GlobalSoilMap.net and build on it – continually growing and expanding?
Bob Macmillan: On the one hand we need to keep our focus on the immediate task at hand and on meeting our committed deliverables. On the other hand, we have here the opportunity to establish a new framework, and perhaps a new paradigm, for the production, delivery and use of soils information globally.
The project is not just about producing new global soil information, it is also about developing new global standards for how this information is produced, stored, accessed and used. This means that we have to be aware of, and make use of, advanced techniques for explicitly capturing and organizing information about our work flows, data models and end user applications. We have an opportunity to adopt new technologies that will ensure the interoperability of our data and its use by adhering to existing and emerging standards for open exchange of data and interoperability between databases.
If we consider just the soil data, we plan to structure the database so that each piece of soil data contains information about the date and method used to produce it, its spatial resolution and estimated uncertainty. This will provide a capability for future work to update the database as and where necessary. It will facilitate recording of change in soil property values through time at any given location. The ability to monitor and report on changes in soil properties will be critical to addressing questions about how soils are changing in relation to changes in climate, land use, land management practices and so on.
If we look beyond just the soil data itself, opportunities exist to tie other data sets and applications into the GlobalSoilMap.net initiative in the future. The African node has a strong emphasis on linking the soils data to information on best management practices so as to be able to offer advice on how to maximize the agricultural productivity of African soils. In many regions, linking of the soils database with databases of climate and hydrological information will provide opportunities to address issues of drought, degradation, water quality and water supply. Explicit linkage of the soil database with databases of land use and management may provide opportunities to monitor globally significant activities such as the sequestration of carbon in soils in response to adoption of improved management practices.
We would hope that GlobalSoilMap.net will provide a foundation for most future soils activities. We hope to define new standards for how soils information is produced, archived and interfaced to applications. We hope that the result will be that soils information will be consistently recognized and used as an essential component of all future decision making activities that pertain to the natural environment.
V1 Magazine: Will new policies and legislation be an outcome of this work? Do you anticipate educational materials being made available?
Bob Macmillan: Certainly the intent is to provide support for analyses that could lead to development of new policies or legislation. We envisage crop modelers making use of the new soils data to run improved versions of their models to identify where gains could be made in the production of food or fiber or where deterioration of soils may be leading to falling production. We are clearly interested in offering our data for use in models for estimating carbon sequestration and global carbon stocks. We would be happy to see our database eventually provide a spatially explicit platform for recording the change in soil attributes (such as soil organic carbon) over time in response to changes in climate or land management. We will be working directly and explicitly with downstream process modelers to ensure that the data we produce and record is suitable for input into global and regional simulation models and decision support systems.
Similar projects at a continental scale in Australia and America have produced materials that are useful for education and public information. We anticipate linking our maps with other global scale maps and databases to illustrate where and how soils vary in response to differences in climate, vegetation, land use, topography, parent materials and so on. Projects such as this have inevitably resulted in the production of improved understanding of the factors that influence the variation in soil properties in space. This improved understanding has been first reported in the scientific literature but then it has found its way into educational and general reference materials.
V1 Magazine: What are the next steps at the present time?
Bob Macmillan: We have five main technical priorities at the present time. First we are completing the process of developing and approving specifications for our anticipated output products.
Second, we are reviewing similar projects and databases as the initial stage of a data modeling exercise that will let us define a flexible and effective data model for organizing and delivering all of our spatial and non spatial data.
Thirdly, we have initiated steps to apply and evaluate various different approaches to making predictions that vary according to the amount, type and quality of data available to support the predictions. The approaches will investigate the effects of using different environmental predictors (covariates), different types and amounts of legacy (or new) soils data (evidence) and different prediction methods (statistical or heuristic models).
Fourthly, we have started to look at approaches that we can use to assess the uncertainty or accuracy of our predictions under any given combination of prediction method, available data and landscape conditions. We need to have reliable, objective criteria for making informed decisions about which methods and data work best under any given set of conditions.
Finally, we need to begin to apply selected methods to one or more large areas to produce some examples of what we propose to produce. These examples will serve to identify issues and problems that remain to be addressed. More importantly, they will provide concrete evidence that it is possible to make predictive maps for large areas efficiently in an operational sense. We anticipate that having examples of the proposed output to examine will stimulate interest, discussion, certainly criticism and ultimately confidence that we can do the job and that the products will be of acceptable quality and utility.
We propose to produce these initial examples for parts of Africa because that is the node that has expectations for early deliverables.
On the organizational side, we will continue to build the consortium through a staged series of official node launches. The Asia node will have its launch in late October in Seoul, Korea at EAFS, the East and Southeast Asia Federation of Soil Science meetings. The Latin American Node is being launched officially in Costa Rica in November at CLACS, the meetings of the Congresso Latin Americano de la Ciencia del Suelo. These launches indicate the official establishment of a node and precipitate a need to begin formal organization and activity. Additional launches are scheduled for Europe in Rome in May, 2010 and for Australia and Oceania in Brisbane in August, 2010.
Progress at the nodes will ultimately depend upon our success in obtaining funds and marshalling resources
V1 Magazine: Where do you see this work leading toward? Will GlobalSoilMap.net potentially result in increased food production? What is the relationship of this work toward assisting to support world food supply and how is that connection made? What is the process involved?
Macmillan: This work is leading towards worldwide standardization of existing soils data and free, open on-line access to the data.
Initially, we expect the data will find its greatest use amongst modelers and decision makers working at regional to global scales. We will work diligently to ensure that our products are suitable for input into the most commonly used global scale models for soil carbon storage, crop growth, climate change, hydrological runoff and degradation. We will try to structure the data model so that our data can be easily accessed and used by common models with little or no need for reformatting or reinterpretation. The data ought to be suitable for supporting regional level policy analysis and decision making as well.
We would hope to link our data to agencies and services that promote best management practices and that offer advice for increasing agricultural productivity using sustainable methods, particularly in parts of the world where productivity is not yet optimized. Initially, it is a bit optimistic to believe that an individual small land holder will be able to make effective use of the data to make farm level management decisions. In the longer term, it is entirely feasible that local farmers and land managers might download and review information for a specific piece of land to help decide how to best manage it. This will depend upon being able to update the initial data in the framework with more accurate and finer resolution data as it becomes feasible to produce it.
————————————————————————————-
Information:
Dr. R. A. (Bob) MacMillan
Science Coordinator
GlobalSoilMap.net
ISRIC – World Soil Information
PO box 353, 6700 AJ, Wageningen, The Netherlands
Email: [email protected] ; [email protected]
Web: www.isric.org, www.globalsoilmap.net, www.africasoils.net
About Bob MacMillan: He has 35 years of experience as a pedologist, soil surveyor and digital soil mapper. Bob has a B.Sc. (Honours) in Geology from Carleton University, an M.Sc. in Soil Science from the University of Alberta and a Ph.D. in GIS and Environmental Modelling (hydrology & soils) from the University of Edinburgh. Bob completed many conventional soil survey projects in Alberta, B.C., Ontario, the Maritimes and East Africa as a member of provincial and federal soil survey units in Alberta, Ottawa and Nova Scotia from 1975 to 1994. In 1995 Bob started a private consultancy offering services in environmental modelling, custom GIS application development and landscape analysis. This consultancy specialized in spatial analysis incorporating vector and raster GIS, remote sensing, digital elevation modelling, spatial statistics, database programming and simulation modelling. Bob’s major work since 1995 focused on developing and applying new methods for automated predictive ecological and soil mapping. These methods were largely based on analyzing DEM and ancillary data sets to automatically classify portions of the landscape into defined ecological, soil or landform classes. In one major project, Bob successfully produced predictive ecosystem maps (PEM) for more than 8 million hectares in BC. Bob also developed his own toolkit of custom programs (LandMapR) to support this work in automated predictive mapping. In September, 2009 Bob began work as the Science Coordinator for the GlobalSoilMap.net project whose aim is to produce a digital map of soil properties at a 90 m resolution for the entire world.
