Research ArticleCLIMATE CHANGE

Climate legacies drive global soil carbon stocks in terrestrial ecosystems

See allHide authors and affiliations

Science Advances  12 Apr 2017:
Vol. 3, no. 4, e1602008
DOI: 10.1126/sciadv.1602008
  • Fig. 1 A theoretical framework explaining the effects of climatic legacies on soil C stocks in natural and agricultural areas.

    Higher color intensity in soil represents more soil carbon. In the example, a grassland under a current dry climate, which was previously a forest ecosystem (site A) and developed under a wetter paleoclimate now, has a greater amount of soil C than expected based on its current climate or compared to a contemporary arid grassland subjected to arid paleoclimate (site B). Shifts in land use from natural systems to agriculture have been shown to markedly reduce the amount of soil C as a result of rapid C degradation and soil erosion linked to land clearing and cultivation.

  • Fig. 2 Relative contribution of paleo- (mid-Holocene and Last Glacial Maximum) and current climate as drivers of soil carbon stocks.

    Results from variation partitioning modeling aiming to identity the percentage of variance of soil carbon explained by past and current climate variables for the Global-WoSIS (A), Global-Drylands (B), and Australia (C) data sets are shown. Shared effects of these variable groups are indicated by the overlap of circles. (D to F) Results from random forest analyses aiming to identify the top five significant (P < 0.05) bioclimatic variables regulating soil carbon for the three data sets used. Increase in the percentage of MSE is equal to the increase in the mean square error. Acronyms are available in Table 1.

  • Fig. 3 Relative contribution of paleo- (mid-Holocene and Last Glacial Maximum) and current climate as drivers of soil carbon in agricultural (n = 1167) and natural (n = 814) systems from the Global-WoSIS data set.

    (A and B) Variation partitioning modeling aiming to identity the percentage of variance of soil carbon explained by past and current climate variables for the identified agricultural and natural systems from the Global-WoSIS. Shared effects of these variable groups are indicated by the overlap of circles. (C and D) Results from the random forest analyses aiming to identify the top five bioclimatic variables regulating soil carbon for the three data sets used. Increase in the percentage of MSE is equal to the increase in the mean square error. Acronyms are available in Table 1.

  • Fig. 4 Soil carbon stocks for agricultural (n = 814) and natural (n = 1167) ecosystems from the Global-WoSIS data set.

    Analyses of variance (ANOVAs) were used to test for differences between natural and agricultural systems.

  • Table 1 Bioclimatic variables included in this study.
    Bioclimatic variableAcronym
    Annual mean temperatureAMT
    Mean diurnal rangeMDR
    IsothermalityISO
    Temperature seasonalityTSEA
    Maximum temperature of warmest monthMAXTWM
    Minimum temperature of coldest monthMINTCM
    Temperature annual rangeTRANGE
    Mean temperature of wettest quarterTWETQ
    Mean temperature of driest quarterTDQ
    Mean temperature of warmest quarterTWARQ
    Mean temperature of coldest quarterTCQ
    Annual precipitationAP
    Precipitation of wettest monthPWETM
    Precipitation of driest monthPDM
    Precipitation seasonalityPSEA
    Precipitation of wettest quarterPWETQ
    Precipitation of driest quarterPDQ
    Precipitation of warmest quarterPWARQ
    Precipitation of coldest quarterPCQ

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/3/4/e1602008/DC1

    table S1. Correlation (Pearson’s) among bioclimatic variables across different time periods.

    table S2. Correlations (Spearman ρ) among soil C stocks estimated at 0 to 10, 10 to 20, 20 to 50, 50 to 100, 0 to 20, 0 to 50, and 0 to 100 cm of soil depth for the Global-WoSIS data set.

    table S3. Results from random forest analyses aiming to identify the most important bioclimatic variables regulating soil C stocks for the three data sets used.

    table S4. Correlations (Spearman ρ) among bioclimatic variables across different time periods (current climate, mid-Holocene, and Last Glacial Maximum) and soil C contents for the Global-WoSIS, Global-Drylands, and Australia data sets.

    table S5. Direct effects of current and past climate on soil C stocks and correlations among exogenous variables (that is, climate from different periods) extending results of the structural equation models shown in fig. S6.

    fig. S1. Location of the sites included in the Global-WoSIS (n = 4381), Global-Drylands (n = 224), and Australia (n = 450) data sets.

    fig. S2. Relative contribution of paleo- (mid-Holocene and Last Glacial Maximum) and current climate of the residuals of soil C stocks (from a multilinear regression with latitude and longitude as predictors of soil C stocks).

    fig. S3. Relative contribution of paleo- versus current climate in driving soil C in tropical (n = 1354), temperate (n = 1566), continental (n = 655), and arid (n = 775) ecosystems from the Global-WoSIS data set.

    fig. S4. Relative contribution of paleoclimate, current climate, and other factors including space (latitude, longitude, and altitude), soil properties (soil pH, electrical conductivity, and sand content), and biotic features (total plant cover and species richness) in driving soil C stocks in the Global-Drylands data set.

    fig. S5. Relative contribution of paleo- versus current climate in driving soil C across different soil depths: 10 to 20 cm (all sites, n = 4234; agricultural sites, n = 1134; and natural sites, n = 790), 20 to 50 cm (all sites, n = 3797; agricultural sites, n = 1046; and natural sites, n = 670), and 50 to 100 cm (all sites, n = 2400; agricultural sites, n = 610; and natural sites, n = 448) for all sites available and also for the identified agricultural and natural systems from the Global-WoSIS.

    fig. S6. Relative contribution of paleo- versus current climate in driving soil C across different soil depths: 0 to 20 cm (all sites, n = 4234; agricultural sites, n = 1134; and natural sites, n = 790), 0 to 50 cm (all sites, n = 3786; agricultural sites, n = 1046; and natural sites, n = 674), and 0 to 100 cm (all sites, n = 2349; agricultural sites, n = 604; and natural sites, n = 435) for all sites available and also for the identified agricultural and natural systems from the Global-WoSIS.

    fig. S7. Relative contribution of paleo- versus current climate in driving soil C stocks in middle latitudes (n = 2080) and tropics (n = 2301) for the Global-WoSIS data set.

    fig. S8. Structural equation modeling aiming to identify the relative influence of the main bioclimatic variables from current, mid-Holocene, and land maximum climate (as identified by random forest analyses) on soil C stocks.

    fig. S9. Relative contribution of paleo (mid-Holocene and Last Glacial Maximum) and current climate as drivers of the residuals of soil C stocks (from a multilinear regression with latitude and longitude as predictors of soil C stocks) in agricultural (n = 1167) and natural (n = 814) systems from the Global-WoSIS data set.

  • Supplementary Materials

    This PDF file includes:

    • table S1. Correlation (Pearson’s) among bioclimatic variables across different time periods.
    • table S2. Correlations (Spearman ρ) among soil C stocks estimated at 0 to 10, 10 to 20, 20 to 50, 50 to 100, 0 to 20, 0 to 50, and 0 to 100 cm of soil depth for the Global-WoSIS data set.
    • table S3. Results from random forest analyses aiming to identify the most important bioclimatic variables regulating soil C stocks for the three data sets used.
    • table S4. Correlations (Spearman ρ) among bioclimatic variables across different time periods (current climate, mid-Holocene, and Last Glacial Maximum) and soil C contents for the Global-WoSIS, Global-Drylands, and Australia data sets.
    • table S5. Direct effects of current and past climate on soil C stocks and correlations among exogenous variables (that is, climate from different periods) extending results of the structural equation models shown in fig. S6.
    • fig. S1. Location of the sites included in the Global-WoSIS (n = 4381), Global-Drylands (n = 224), and Australia (n = 450) data sets.
    • fig. S2. Relative contribution of paleo- (mid-Holocene and Last Glacial Maximum) and current climate of the residuals of soil C stocks (from a multilinear regression with latitude and longitude as predictors of soil C stocks).
    • fig. S3. Relative contribution of paleo- versus current climate in driving soil C in tropical (n = 1354), temperate (n = 1566), continental (n = 655), and arid (n = 775) ecosystems from the Global-WoSIS data set.
    • fig. S4. Relative contribution of paleoclimate, current climate, and other factors including space (latitude, longitude, and altitude), soil properties (soil pH, electrical conductivity, and sand content), and biotic features (total plant cover and species richness) in driving soil C stocks in the Global-Drylands data set.
    • fig. S5. Relative contribution of paleo- versus current climate in driving soil C across different soil depths: 10 to 20 cm (all sites, n = 4234; agricultural sites, n = 1134; and natural sites, n = 790), 20 to 50 cm (all sites, n = 3797; agricultural sites, n = 1046; and natural sites, n = 670), and 50 to 100 cm (all sites, n = 2400; agricultural sites, n = 610; and natural sites, n = 448) for all sites available and also for the identified agricultural and natural systems from the Global-WoSIS.
    • fig. S6. Relative contribution of paleo- versus current climate in driving soil C across different soil depths: 0 to 20 cm (all sites, n = 4234; agricultural sites, n = 1134; and natural sites, n = 790), 0 to 50 cm (all sites, n = 3786; agricultural sites, n = 1046; and natural sites, n = 674), and 0 to 100 cm (all sites, n = 2349; agricultural sites, n = 604; and natural sites, n = 435) for all sites available and also for the identified agricultural and natural systems from the Global-WoSIS.
    • fig. S7. Relative contribution of paleo- versus current climate in driving soil C stocks in middle latitudes (n = 2080) and tropics (n = 2301) for the Global-WoSIS data set.
    • fig. S8. Structural equation modeling aiming to identify the relative influence of the main bioclimatic variables from current, mid-Holocene, and land maximum climate (as identified by random forest analyses) on soil C stocks.
    • fig. S9. Relative contribution of paleo (mid-Holocene and Last Glacial Maximum) and current climate as drivers of the residuals of soil C stocks (from a multilinear regression with latitude and longitude as predictors of soil C stocks) in agricultural (n = 1167) and natural (n = 814) systems from the Global-WoSIS data set.

    Download PDF

    Other Supplementary Material for this manuscript includes the following:

    • table S1 (Microsoft Excel format). Correlation (Pearson’s) among bioclimatic variables across different time periods.

    Files in this Data Supplement:

Navigate This Article