Home Global warming A global database on land management, land use change and the effects of climate change on soil organic carbon

A global database on land management, land use change and the effects of climate change on soil organic carbon


Data gathering

The documentary search was carried out on January 09, 2020 (Fig. 1). The following search equation was used: (“meta*analysis” OR “systematic review”) AND (“soil organic carbon” OR SOC OR “soil organic matter” OR SOM OR “soil carbon”) in the “subject words”, that is to say titles, abstracts and keywords from the following databases:

  • Web of Science, New York, USA, http://apps.webofknowledge.com, encompassing 12,000 journals and 160,000 conference proceedings.

  • Scopus, USA, https://www.scopus.com/search. The Scopus database includes more than 41,000 referenced journals.

  • OVIDE. Publisher, USA. https://www.ovid.com. The Ovid Database includes over 10,000 scientific journal titles, books and proceedings (for Cab Abstracts on Ovid).

  • Google Scholar. https://scholar.google.com/. Publisher: Google. It contains both peer-reviewed multidisciplinary literature and gray literature. We have selected the top 150 search results, organized by relevance, as this engine is very accurate for the first pages of results displayed23.24.

Fig. 1

Methodological framework used to identify and characterize the data included in the database. The study selection criteria are (1) duplicates are removed; (2) only studies published in English (non-English studies: n = 7) with an available text (studies with an unavailable text: n = 7) are considered; (3) studies not dealing with SOC are excluded; (4) only meta-analyses are included; (5) the meta-analysis exhibits at least one effect size, i.e. a quantitative measure on COS (or a described effect size for different levels of COS content, i.e. say the COS as a covariate). The hexagons represent the different characteristics analyzed in the meta-analyses or in the primary studies.

Follow the Recommended Gold Standard of Systematic Review25.26, we used these different databases with different journal coverage to achieve a comprehensive literature search and avoid potential bias. The search was supported by a librarian to further help reduce the possibility of bias and improve the overall quality of the search strategy27.

No restriction on the year of publication was applied. All climatic zones and all countries have been taken into account. Sensitivity was favored over specificity. Sensitivity implies that the emphasis in the search procedure is on collecting the largest selection of potentially relevant studies at the risk of also obtaining a high number of irrelevant studies (thereby increasing the duration of the step). Selection). In addition to the database searches, a number of other potentially relevant meta-analyses were added by the authors of this study.

The literature search identified 1535 studies (of which 1008 were unique). These studies were compiled into a database and then screened to identify relevant studies based on the following inclusion and exclusion criteria: (i) only studies published in English with full text available were considered ; (ii) the study presents a quantitative and formal analysis of several previous empirical studies, that is to say a meta-analysis (we did not consider studies with vote-counting methods and narrative reviews were also excluded); (iii) the meta-analysis exhibits at least one effect size, that is to say a quantitative measure of the magnitude of a SOC variable, either as the main variable or as a covariate.

Studies were first sorted based on title and abstract, and if necessary, the entire manuscript was read. Each study was reviewed by two authors of this article. Rejected studies were compiled into a database exclusion sheet, along with the reasons for exclusion. Finally, 217 meta-analyses met our inclusion/exclusion criteria, including 18 with SOC as a co-variable.

Characterization of meta-analyses

Our database reports metadata (author names and affiliations, journal name, keywords, date of publication and countries covered in the meta-analysis) for the meta-analyses considered (Fig. 1). The transparency and reproducibility of each of the 217 meta-analyses were assessed based on criteria related to literature search, data extraction, data analysis and interpretations. These criteria are an adaptation of those proposed in several other studies covering various fields of research.28,29,30,31. When it is satisfied, a criterion is scored 1, and 0 otherwise. An overall quality score was assigned by calculating the proportion of criteria met.

We have also categorized each meta-analysis according to its scope, indicating whether it addresses the dimensions ‘climate change mitigation’, ‘climate change adaptation’ and ‘food security’. Dimension definitions are based on the IPCC Glossary32and the objectives of the 4p1000 Initiative15 (https://www.4p1000.org/). Keywords related to each dimension (Table 1) were defined for the classification, which was carried out manually by two different authors of this study. The title and abstract were screened, and the full text was studied if necessary. Inter-reviewer consistency was checked on a sample of 30 studies. The final database includes respectively 199, 28 and 54 meta-analyses that analyze the dimensions of mitigation, adaptation and food security.

Table 1 Keywords related to each dimension 4 per 1000.

Characterization of effect sizes

We extracted all the quantitative data related to the effects reported in the retrieved meta-analyses: size of the effect, indicator of dispersion (confidence interval, and/or standard deviation or quantiles), significance (P-value) and number of data on which the effects were calculated (Fig. 2). The database contains both effect sizes which quantify the direct effect of an intervention on SOC, and effect sizes which assess the effect on other outcomes but with SOC as co- variable (indirect effects). Data were collected from tables or figures using WebPlot Digitizer software (www.automeris.io/WebPlotDigitizer/). We also described the type of metric associated with each effect size (for example mean difference, ratio, coverage D).

Figure 2
Figure 2

Number of effect sizes (left) and number of meta-analyses (right) available in the database by type of intervention and land use. The total number of effect sizes reported in the database is shown, along with the number of effect sizes on SOC (dark shaded) or other outcomes (light shaded). For the land use change intervention, the initial and final land uses are considered in this plot. A meta-analysis can include different interventions or land uses.

We did not extract data from meta-regressions, correlations and their associated characteristics due to the difficulty of synthesizing this type of results across different studies. However, when subgroup analyzes were performed (for example, by soil characteristics or climatic zones – Table 2), we extracted the effects of moderators in order to analyze and understand the variability of SOC values.

Table 2 Co-variables reported in meta-analyses.

Interventions related to effect sizes have been grouped into main categories: land management, land use change and global change (Fig. 2, 3). We considered land use types as defined in the IPCC Guidelines for National Greenhouse Gas Inventories33such as cropland, forest land, grassland, wetlands and other land. We considered land use change as the conversion of one of the above land uses to another. We considered land management to be any intervention carried out on any of the aforementioned land uses (for example logging, restoration of wetlands, mineral fertilization). We defined global change as changes on a planetary scale other than land-use change (for example climate change). The database presents the main categories of interventions, but also the more detailed interventions reported in each study (Fig. 3). The number of effect sizes and meta-analyses is largely dominated by land management studies, particularly for cropland. Next come interventions of the land use change and global change type, for which the distribution of land uses is more balanced.

Figure 3
picture 3

Categories used for characterization of interventions for global change, land use change and management.

Non-SOC results were grouped into seven broad categories: soil chemistry, plant productivity, soil physics, soil biology, greenhouse gases, water quality, and others. The first three categories accounted for almost 20-30% of the reported effect sizes. Each category was further refined into 2 to 11 subcategories (Fig. 4). The soil nutrient and aboveground biomass subcategories alone accounted for nearly 15-20% of the effect sizes.

Figure 4
number 4

Main categories and subcategories of other effect sizes retrieved from the 217 meta-analyses and studied together with SOC. The areas are proportional to the number of effect sizes in the database.

Characterization of the primary studies used in the 217 meta-analyses

We retrieved all available references (primary studies) used by the 217 meta-analyses, searching the available list of references, additional documents and databases associated with each meta-analysis.

Primary studies were characterized by their metadata (for example DOI, authors, publication date, journal). Based on the title and abstract, if necessary, we also manually extracted the type of interventions and outcomes associated with the primary studies. Manual classification into the same intervention and outcome categories as previously described was facilitated by automatic classification based on keywords (Supplementary Table 1). The final database includes 13,632 unique primary studies (Fig. 5). 9,130 ​​primary studies were used in several meta-analyses. The geographical distribution showed the largest number of such primary studies in the United States and China, followed by Brazil and Canada, then Australia, India and some European countries (United Kingdom, Germany, Spain and Italy). The regional distribution within the five countries with the highest number of studies showed large regional disparities. Africa was the least studied continent; no primary studies have been conducted in several African countries.

Figure 5
number 5

Geographic distribution of the 13,632 primary studies included in the 217 meta-analyses (a), with details of the provincial/state breakdown for (b) UNITED STATES, (vs) China, (D) Brazil, (and) Australia and (F) India.