@article{fdi:010070122,
  title = {{M}apping soil organic carbon on a national scale : towards an improved and updated map of {M}adagascar},
  author = {{R}amifehiarivo, {N}. and {B}rossard, {M}ichel and {G}rinand, {C}. and {A}ndriamananjara, {A}. and {R}azafimbelo, {T}. and {R}asolohery, {A}. and {R}azafimahatratra, {H}. and {S}eyler, {F}r{\'e}d{\'e}rique and {R}azafindrabe, {F}. and {R}azakamanarivo, {H}.},
  editor = {},
  language = {{ENG}},
  abstract = {{A}ssessment of soil organic carbon stocks ({SOC}s) is highly relevant considering that {SOC}s is the central driver in climate change mitigation and soil fertility. {I}n {M}adagascar, a first attempt at mapping {SOC}s on a national scale was undertaken in 2009 with previous soil data. {A}dvanced research on soil carbon mapping on a global scale is required to enable better land use decisions. {T}his study aims to (i) evaluate the drivers of soil organic carbon ({SOC}) storage in the first 30 cm soil layer on a national scale from spatially explicit explanatory environmental variables and a recent soil database and (ii) update the spatial distribution of {SOC}s at this scale through digital mapping. {A} spatial model was established using random{F}orest, a decision tree algorithm and based on 10 pedoclimatic, topographic, and vegetation variables. {T}he model was developed with 1993 available soil plot data (collected from 2010 to 2015) and their environmental information ("{VALSOL}-{M}adagascar" database). {T}hese data were divided into two sets: a first set (n = 835) used for model calibration and a second set (n = 358) for external validation. {R}esults showed that mean annual temperature ({MAT}, °{C}), mean annual precipitation ({MAP}, mm), elevation (m) and {N}ormalized {D}ifference {V}egetation {I}ndex ({NDVI}) were the most important predictors of {SOC}s. {T}he retained prediction model had a {R}oot {M}ean {S}quared {E}rror ({RMSE}) equal to 25.8 {M}g{C}·ha− 1. {T}he predicted {SOC}s from fitted models ranged from 28 to 198 {M}g{C}·ha− 1 with total {SOC}s to 4137 {T}g{C}. {D}epending on soil type, {F}erralsols (35 to 165 {M}g{C}·ha− 1) and {A}ndosols (48 to 198 {M}g{C}·ha− 1) had relevant results related to the number of soil profiles (n = 856 and 171 respectively). {D}espite the need for in-depth analysis, the model and map produced in the present study has greatly improved our knowledge of the spatial distribution of {SOC}s in {M}adagascar and reduced uncertainty compared to the former map. {T}his map provides new perspectives in sustainable land management in {M}adagascar.},
  keywords = {{CARTE} {PEDOLOGIQUE} ; {CARTOGRAPHIE} {AUTOMATIQUE} ; {FERTILITE} {DU} {SOL} ; {CARBONE} {ORGANIQUE} ; {DISTRIBUTION} {SPATIALE} ; {MODELISATION} ; {BASE} {DE} {DONNEES} ; {GESTION} {DU} {SOL} ; {MADAGASCAR}},
  booktitle = {{D}igital soil mapping across the globe},
  journal = {{G}eoderma {R}egional},
  volume = {8},
  numero = {},
  pages = {29--38},
  ISSN = {2352-0094},
  year = {2017},
  DOI = {10.1016/j.geodrs.2016.12.002},
  URL = {https://www.documentation.ird.fr/hor/fdi:010070122},
}