GlobalSoilMap for Soil Organic Carbon Mapping and Modeling

Dominique Arrouays, B. Minasny, A.B. McBratney, M. Grundy, N. McKenzie, J. Thompson, A. Gimona, S. Y. Hong, S. Smith, A. Hartemink, S. Chen, M. P.

Martin, V. L. Mulder, A. C. Richer-de-Forges, N. P.A. Saby, I. Odeh, J. Padarian, G. Lelyk, L. Poggio, I. Savin, V. Stolbovoy, Y. Sulaeman, D.

Nursyamsi, G. Zhang, M. H. Greve, Z. Libohova, P. Lagacherie, P. Roudier, J. G.B. Leenaars, G. B.M. Heuvelink, L. Montanarella, P. Panagos, J. Hempel

INRA InfoSol, France

The demand for information onfunctional soil properties is highand has increased over time. Thisis especially true for soil organiccarbon (SOC) due to its major rolein climate change mitigation andadaptation and in maintainingand enhancing many soilecosystem services, among whichis food production.

The GlobalSoilMap project aimsto produce a digital soil map ofthe world. The ultimate objectiveof the project is to build a freedownloadable database of acommon set of key soil propertiesat 3 arc-second resolution with adefined spatial entity, specifieddepth increments and uncertaintycalculations (Arrouays et al.,2014).

How maps are produced:Many countries have abundantlegacy soil data that includes soilmaps at a variety of scales, soilpoint data collected over decades,environmental covariateinformation and a network ofpartners that have contributed tobuilding the soil information overmany years. In addition, manyhave some additional soil sitessampled for soil carbon stock andbaseline assessment. The majorityof the data needed to produceGlobalSoilMap soil property mapswill, at least for the firstgeneration, come mainly fromarchived soil legacy data, whichcould include polygon soil mapsand point pedon data and fromavailable covariates such asclimatic data, remote sensinginformation, geological data, andother forms of environmentalinformation.

The specifications and theproductsThe GlobalSoilMap specificationsrequire the estimation of soilproperty values along with theiruncertainty at each of sixspecified depth increments (0-5,5-15, 15-30, 30-60, 60-100 and100-200 cm) for the followingsoil properties: SOC, clay, silt,sand and coarse fragmentcontents, pH, ECEC, soil depthand available depth to rooting,bulk density (whole soil and fineearth fraction), and availablewater capacity.As many methods may have beenused to measure the soilproperties included in theminimum data set they have to betranslated to a standard methodand the specifications provideguidance on how to do this.The spline function and similarmethods are used to transformhorizon data into continuousdepth functions of soil properties.

Fig. 1: Log-Log scatterplot of countries areas versus number of soil profiles for selected countries (Arrouays et al., submitted)

Fig. 3: A decision tree for digital soil mapping based on legacy data (from Minasny and McBratney, 2010)

The GlobalSoilMap specificationsdo not prescribe the methods ofprediction, because of diverse soillegacy data situations in variouscountries. However, Minasny andMcBratney (2010) provide a flowchart that outlines differentmodels that can be applied.

The estimation of uncertainties isa unique feature but also a majorchallenge of this project. Theuncertainties may determine, forexample, whether the soil mapsare sufficiently accurate for theintended use or where to conductnew surveys or additional soilsampling to obtain more accuratepredictions of soil properties.

A working Group “Global SoilMap” has been launched by theIUSS following an initiativecoming from many organizations

Functional soil property mapshave been produced using digitalsoil mapping techniques andexisting legacy information andmade available to the usercommunity for application. Inaddition, uncertainty has beenprovided as a 90% predictioninterval based on estimatedupper and lower class limits.Main scientific challenges includetime related and uncertaintyissues. Combining local andglobal predictions should be theway forward both to enhance thequality of digital soil maps andtheir use, and to map the entireworld. For this purpose both top-down and bottom-up approachesare necessary. Bottom-upproducts take full advantage oflocal data and knowledge, and oflocal soil distribution controllingfactors. However, global modelshave a big advantage in that theyavoid spatial gaps and may be auseful tool for harmonizingcountries products.

We believe that GlobalSoilMapconstitutes the best availableframework and methodology toaddress global issues about SOCmapping.

Fig. 2: The process for estimating depth functions from soil data relating to soil layers of horizons

INTRODUCTION

OBJECTIVES

METHODOLOGY

Fig. 4: Soil Organic Carbon and related uncertainties for Australia (Viscarra Rossel et al., 2014)

Fig. 5: Origin of the organizations proposing the IUSS WG on GlobalSoilMap

CONCLUSION

MAIN RESULTS