Artificial intelligence and soil carbon modeling demystified: power, potentials, and perils

Abstract

The global soil carbon pool has been estimated to exceed the amount of carbon stored in the atmosphere and vegetation, though uncertainties to quantify below-ground carbon and soil carbon fluxes accurately still exist. Modeling soil carbon using artificial intelligence (AI) - machine learning (ML) and deep learning (DL) algorithms - has emerged as a powerful force in the carbon science community. These AI soil carbon models have shown improved performance to predict soil organic carbon (SOC) storage, soil respiration (R_s), and other properties of the global carbon cycle when compared to other modeling approaches. AI systems have advanced abilities to optimize fits between inputs (geospatial environmental covariates) and outputs (e.g., SOC or R_s) through advanced pattern recognition, learning algorithms, latent variables, hyperparameters, hyperplanes, weighting factors, or multiple stacked processing (e.g., convolution and pooling). These machine-oriented applications have shifted focus from knowledge discovery and understanding of ecosystem processes, carbon pools and cycling toward data-driven applications that compute digital soil carbon outputs. The purpose of this review paper is to explore the emergence, applications, and progress of AI-ML and AI-DL algorithms to model soil carbon storage and R_s at regional and global scales. A critical discussion of the power, potentials, and perils of AI soil carbon modeling is provided. The paradigm shift toward AI modeling raises questions how we study soil carbon dynamics and what conclusions we draw which impacts carbon science research and education, carbon management, carbon policies, carbon markets and economies, and soil health.

Keywords

Soil carbon / soil organic carbon / soil respiration / artificial intelligence / machine learning / deep learning / artificial neural networks

References

[1] Janzen H.Carbon cycling in earth systems—a soil science perspective.Agric Ecosyst Environ2004;104:399-417

[2] Scharlemann JP,Hiederer R.Global soil carbon: understanding and managing the largest terrestrial carbon pool.Carbon Manag2014;5:81-91

[3] Shi Z,He Y.The age distribution of global soil carbon inferred from radiocarbon measurements.Nat Geosci2020;13:555-9

[4] Grunwald S,Boettinger JL.Digital soil mapping and modeling at continental scales: finding solutions for global issues.Soil Sci Soc Am J2011;75:1201-13

[5] James J,Harrison R.Deep soil carbon: quantification and modeling in subsurface layers.Soil Sci Soc Am J2014;78:S1-S10

[6] Koarashi J,Masiello CA.Dynamics of decadally cycling carbon in subsurface soils.J Geophys Res2012;117:G03033

[7] Vasques G,Comerford N.Regional modelling of soil carbon at multiple depths within a subtropical watershed.Geoderma2010;156:326-36

[8] Rogelj J,Höhne N.Paris agreement climate proposals need a boost to keep warming well below 2 °C.Nature2016;534:631-9

[9] Paustian K,Ogle S,Robertson GP.Climate-smart soils.Nature2016;532:49-57

[10] Rumpel C,Chenu C.The 4p1000 initiative: opportunities, limitations and challenges for implementing soil organic carbon sequestration as a sustainable development strategy.Ambio2020;49:350-60 PMCID:PMC6889108

[11] Smith P,Dong H. Agriculture, forestry and other land use (AFOLU). In: Krug T, Nabuurs GJ, editors. Climate change 2014: mitigation of climate change. Cambridge: Cambridge University Press; 2014. p. 813-922.

[12] Six J,Jay breidt F,Mosier AR.The potential to mitigate global warming with no-tillage management is only realized when practised in the long term.Glob Chang Biol2004;10:155-60

[13] Sun W,Yu L.Climate drives global soil carbon sequestration and crop yield changes under conservation agriculture.Glob Chang Biol2020;26:3325-35

[14] Lal R.Soil carbon stocks under present and future climate with specific reference to European ecoregions.Nutr Cycl Agroecosyst2008;81:113-27

[15] Smith P,Cai Z.Greenhouse gas mitigation in agriculture.Philos Trans R Soc Lond B Biol Sci2008;363:789-813 PMCID:PMC2610110

[16] Stockmann U,Crawford JW.The knowns, known unknowns and unknowns of sequestration of soil organic carbon.Agric Ecosyst Environ2013;164:80-99

[17] Ogle SM,Baldock J.Climate and soil characteristics determine where no-till management can store carbon in soils and mitigate greenhouse gas emissions.Sci Rep2019;9:11665 PMCID:PMC6691111

[18] Nicoloso RS.Intensification of no-till agricultural systems: an opportunity for carbon sequestration.Soil Sci Soc Am j2021;85:1395-409

[19] Minasny B,Mcbratney AB.Soil carbon 4 per mille.Geoderma2017;292:59-86

[20] White RE,Lam SK.A critique of the paper ‘soil carbon 4 per mille’ by Minasny et al. (2017).Geoderma2018;309:115-7

[21] de Vries W.Soil carbon 4 per mille: a good initiative but let’s manage not only the soil but also the expectations.Geoderma2018;309:111-2

[22] Poulton P,Macdonald A,Powlson D.Major limitations to achieving “4 per 1000” increases in soil organic carbon stock in temperate regions: evidence from long-term experiments at Rothamsted Research, United Kingdom.Glob Chang Biol2018;24:2563-84 PMCID:PMC6001646

[23] van Groenigen JM,Hungate BA,Powlson DS.Sequestering soil organic carbon: a nitrogen dilemma.Environ Sci Technol2017;51:4738-9

[24] Baveye PC,Tessier D.The “4 per 1000” initiative: a credibility issue for the soil science community?.Geoderma2018;309:118-23

[25] Russell S. Artificial intelligence: a modern approach. Available from: https://storage.googleapis.com/pub-tools-public-publication-data/pdf/27702.pdf [Last accessed on 12 Apr 2022]

[26] LeCun Y,Hinton G.Deep learning.Nature2015;521:436-44

[27] LeCun Y. Convolutional networks for images, speech, and time series. The handbook of brain theory and neural networks. Cambridge: MIT Press; 1995.

[28] Mcbratney A,Bryce A.Pedometrics timeline.Geoderma2019;338:568-75

[29] Khaledian Y.Selecting appropriate machine learning methods for digital soil mapping.Appl Math Model2020;81:401-18

[30] Padarian J,Mcbratney AB.Using deep learning for digital soil mapping.Soil2019;5:79-89

[31] Kamilaris A.Deep learning in agriculture: a survey.Comput Electron Agric2018;147:70-90

[32] Diaz-Gonzalez FA,Correa CA,Patino D.Machine learning and remote sensing techniques applied to estimate soil indicators - review.Ecol Indic2022;135:108517

[33] Grunwald S.Grand challenges in pedometrics-AI research.Front Soil Sci2021;1:714323

[34] Ma Y,Malone BP.Pedology and digital soil mapping (DSM).Eur J Soil Sci2018;70:216-35

[35] Thompson JA,Grunwald S.Digital soil mapping: Interactions with and applications for hydropedology. In: Lin HS, Editor. Hydropedology - synergistic integration of Pedology and Hydrology. Cambridge: Academic Press; 2012. p. 665-709.

[36] Mcbratney A,Minasny B.On digital soil mapping.Geoderma2003;117:3-52

[37] Xiong X,Myers DB,Harris WG.Holistic environmental soil-landscape modeling of soil organic carbon.Environ Model Softw2014;57:202-15

[38] Breiman L. Classification and regression trees. London: Chapman & Hall; 1984.

[39] Prasad AM,Liaw A.Newer classification and regression tree techniques: bagging and random forests for ecological prediction.Ecosystems2006;9:181-99

[40] Breiman L.Bagging predictors.Mach Learn1996;24;123-40

[41] Hastie T,Friedman J. The elements of statistical learning. New York: Springer; 2009.

[42] Efron B. An introduction to the bootstrap: monographs on statistics and applied probability. London: Chapman & Hall; 1993.

[43] Freund Y. Experiments with a new boosting algorithm. 13th International Conference on Machine Learning; 1996 Jan 22; Murray Hill. 1996. p. 148-56.

[44] Friedman JH.Greedy function approximation: a gradient boosting machine.Ann Stat2001;29:1189-232

[45] Lawrence R.Classification of remotely sensed imagery using stochastic gradient boosting as a refinement of classification tree analysis.Remote Sens Environ2004;90:331-6

[46] Breiman L.Random forests.Mach Learn2001;45:5-32

[47] Cutler A,Stevens JR,Ma Y. Ensemble machine learning: methods and applications. Chicago: Media LLC; 2012.

[48] Svetnik V,Tong C,Sheridan RP.Random forest: a classification and regression tool for compound classification and QSAR modeling.J Chem Inf Comput Sci2003;43:1947-58

[49] Meinshausen N.Quantile regression forests.J Mach Learn Res2006;7:983-99

[50] Karatzoglou A,Hornik K.Support vector machines in R.J Stat Soft2006;15:1-28

[51] Smola AJ.A tutorial on support vector regression.Stat Comput2004;14:199-222

[52] Williams G. Data mining with rattle and R: the art of excavating data for knowledge discovery. New York: Springer; 2011.

[53] Somarathna P,Minasny B.Mapping soil organic carbon content over New South Wales, Australia using local regression kriging.Geoderma Reg2016;7:38-48

[54] Wold H.Soft modelling by latent variables: the non-linear iterative partial least squares (NIPALS) approach.J Appl Probab1975;12:117-42

[55] Garthwaite PH.An interpretation of partial least squares.J Am Stat Assoc1994;89:122-7

[56] Carrascal LM,Gordo O.Partial least squares regression as an alternative to current regression methods used in ecology.Oikos2009;118:681-90

[57] Tobias RD. An introduction to partial least squares regression. Proceedings of the Twentieth Annual SAS Users Group International Conference; Cary: SAS Institute Inc; 1995. p. 1250-7.

[58] Quinlan JR. Programs for machine learning. Burlington: Morgan Kaufmannn; 1993.

[59] Haykin S. Neural networks: a comprehensive foundation. Hoboken: Prentice Hall; 1999.

[60] Hecht-Nielsen R.Theory of the Backpropagation Neural Network*. In: Wechsler H, editor. Neural Networks for Perception. Elsevier; 1992. p. 65-93.

[61] Wythoff BJ.Backpropagation neural networks: a tutorial.Chemometr Intell Lab Syst1993;18:115-55

[62] Bellon-Maurel V,Palagos B,Mcbratney A.Critical review of chemometric indicators commonly used for assessing the quality of the prediction of soil attributes by NIR spectroscopy.Trends Analyt Chem2010;29:1073-81

[63] McBratney AB,Stockmann U.Pedometrics. 1st ed. New York: Springer; 2018.

[64] Williams PC. Variables affecting near-infrared reflectance spectroscopic analysis. Near-infrared technology in the agriculture and food industries. Eagan: American Association of Cereal Chemists; 1987. p. 143-67.

[65] Chang C,Mausbach MJ.Near-infrared reflectance spectroscopy-principal components regression analyses of soil properties.Soil Sci Soc Am J2001;65:480-90

[66] Adi SH.Integrative environmental modeling of soil carbon fractions based on a new latent variable model approach.Sci Total Environ2020;711:134566

[67] Peng Y,Adhikari K,Grunwald S.Modeling soil organic carbon at regional scale by combining multi-spectral images with laboratory spectra.PLoS One2015;10:e0142295 PMCID:PMC4640839

[68] Ross CW,Vogel JG.Accounting for two-billion tons of stabilized soil carbon.Sci Total Environ2020;703:134615

[69] Keskin H,Harris WG.Digital mapping of soil carbon fractions with machine learning.Geoderma2019;339:40-58

[70] Martin MP,Smith P.Spatial distribution of soil organic carbon stocks in France.Biogeosciences2011;8:1053-65

[71] Were K,Dick ØB.A comparative assessment of support vector regression, artificial neural networks, and random forests for predicting and mapping soil organic carbon stocks across an Afromontane landscape.Ecol Indic2015;52:394-403

[72] Deng X,Ma W.Baseline map of organic carbon stock in farmland topsoil in East China.Agric Ecosyst Environ2018;254:213-23

[73] Viscarra Rossel RA,Grundy MJ,Clifford D.The Australian three-dimensional soil grid: Australia’s contribution to the GlobalSoilMap project.Soil Res2015;53:845

[74] Hengl T,Heuvelink GB.SoilGrids250m: Global gridded soil information based on machine learning.PLoS One2017;12:e0169748 PMCID:PMC5313206

[75] Cao B,Russell MB.Spatial modeling of litter and soil carbon stocks on forest land in the conterminous United States.Sci Total Environ2019;654:94-106

[76] Sothe C,Arabian J.Large scale mapping of soil organic carbon concentration with 3D machine learning and satellite observations.Geoderma2022;405:115402

[77] Wang B,Orgill S.Estimating soil organic carbon stocks using different modelling techniques in the semi-arid rangelands of eastern Australia.Ecol Indic2018;88:425-38

[78] Heuvelink GBM,Poggio L.Machine learning in space and time for modelling soil organic carbon change.Eur J Soil Sci2021;72:1607-23

[79] Guevara M,Stell E.No silver bullet for digital soil mapping: country-specific soil organic carbon estimates across Latin America.Soil2018;4:173-93

[80] Poggio L,Batjes NH.SoilGrids 2.0: producing soil information for the globe with quantified spatial uncertainty.Soil2021;7:217-40

[81] Lal R.Soil carbon sequestration to mitigate climate change.Geoderma2004;123:1-22

[82] Köchy M,van der Molen MK.Global distribution of soil organic carbon, based on the Harmonized World Soil Database-Part 2: Certainty of changes related to land-use and climate.Soil Discussions2014;1:363-400

[83] Le Quéré C,Andrew RM.Global carbon budget 2014.Earth Syst Sci Data2015;7:47-85

[84] Batjes NH.Technologically achievable soil organic carbon sequestration in world croplands and grasslands.Land Degrad Dev2019;30:25-32

[85] Haaf D,Doetterl S.Global patterns of geo-ecological controls on the response of soil respiration to warming.Nat Clim Chang2021;11:623-7

[86] Kirschbaum M.The temperature dependence of organic-matter decomposition—still a topic of debate.Soil Biol Biochem2006;38:2510-8

[87] Bond-Lamberty B.A global database of soil respiration data.Biogeosciences2010;7:1915-26

[88] Bond-Lamberty B.New techniques and data for understanding the global soil respiration flux.Earth’s Future2018;6:1176-80

[89] Lu H,Ma M.Comparing machine learning-derived global estimates of soil respiration and its components with those from terrestrial ecosystem models.Environ Res Lett2021;16:054048

[90] Huang N,Song XP.Spatial and temporal variations in global soil respiration and their relationships with climate and land cover.Sci Adv2020;6:eabb8508 PMCID:PMC7541079

[91] Bond-Lamberty B.Temperature-associated increases in the global soil respiration record.Nature2010;464:579-82

[92] Hashimoto S,Ito A,Nishina K.Global spatiotemporal distribution of soil respiration modeled using a global database.Biogeosciences2015;12:4121-32

[93] Xu M.Contribution of soil respiration to the global carbon equation.J Plant Physiol2016;203:16-28

[94] Warner DL,Jian J,Vargas R.Spatial predictions and associated uncertainty of annual soil respiration at the global scale.Global Biogeochem Cycles2019;33:1733-45

[95] Stell E,Jian J,Vargas R.Spatial biases of information influence global estimates of soil respiration: how can we improve global predictions?.Glob Chang Biol2021;27:3923-38

[96] Yao Y,Viovy N.A data-driven global soil heterotrophic respiration dataset and the drivers of its inter-annual variability.Global Biogeochem Cycles2021;35

[97] Jian J,Anderson-teixeira K.A restructured and updated global soil respiration database (SRDB-V5).Earth Syst Sci Data2021;13:255-67

[98] Jian J,Thomas RQ,Hodges SC.Constraining estimates of global soil respiration by quantifying sources of variability.Glob Chang Biol2018;24:4143-59

[99] Viscarra Rossel R,Ben-dor E.A global spectral library to characterize the world’s soil.Earth Sci Rev2016;155:198-230

[100] Knox NM.Total soil carbon assessment: linking field, lab, and landscape through VNIR modelling.Landscape Ecol2018;33:2137-52

[101] Clingensmith CM,Wani SP.Evaluation of calibration subsetting and new chemometric methods on the spectral prediction of key soil properties in a data-limited environment.Eur J Soil Sci2019;70:107-26

[102] Moura-Bueno JM,Horst-Heinen TZ,ten Caten A.Environmental covariates improve the spectral predictions of organic carbon in subtropical soils in southern Brazil.Geoderma2021;393:114981

[103] Shi Z,Peng J.Development of a national VNIR soil-spectral library for soil classification and prediction of organic matter concentrations.Sci China Earth Sci2014;57:1671-80

[104] Wadoux AMJ,Poggio L.A note on knowledge discovery and machine learning in digital soil mapping.Eur J Soil Sci2020;71:133-6

[105] Liao SM. Ethics of artificial intelligence. Oxford: Oxford University Press; 2020.

[106] Guyon I,Barnhill S.Gene selection for cancer classification using support vector machines.Mach Learn2002;46:389-422

[107] Kursa MB.Feature selection with the Boruta package.J Stat Soft2010;36:1-13

[108] Nilson R,Björkegren J.Consistent feature selection for pattern recognition in polynomial time.J Mach Learn Res2007;8:589-612

[109] Xie Z.Sparse group LASSO based uncertain feature selection.2014;5:201-10

[110] Taghizadeh-Mehrjardi R,Amirian-Chakan A.Improving the spatial prediction of soil organic carbon content in two contrasting climatic regions by stacking machine learning models and rescanning covariate space.Remote Sens2020;12:1095

[111] Zhao Z,Yang Q.Model prediction of biome-specific global soil respiration from 1960 to 2012.Earth’s Future2017;5:715-29

[112] Grunwald S,Ceddia MB.The meta soil model: an integrative multi-model framework for soil security. In: Field DJ, Morgan CLS, McBratney AB, editors. Global soil security. London: Springer Nature Publ.; 2017. p. 305-18.

[113] Ng W,Montazerolghaem M.Convolutional neural network for simultaneous prediction of several soil properties using visible/near-infrared, mid-infrared, and their combined spectra.Geoderma2019;352:251-67

[114] Chalmers DJ. Reality+: virtual worlds and the problem of philosophy. New York: W.W. Norton & Company; 2022.

[115] Mitchell M. Artificial intelligence: a guide for thinking humans. New York: Farrar, Straus and Giroux; 2019.

[116] Baudrillard J. Simulacra and simulation. Ann Arbor: University of Michigan Press; 1994.

[117] Safanelli JL,Ben-Dor E.Multispectral models from bare soil composites for mapping topsoil properties over Europe.Remote Sensing2020;12:1369