We used FAO statistics under the heading “Production”, then “Crops and animal products” which specify the quantities of plant products marketed by world agriculture in 2022. To determine the carbon capture of these plant products, the anhydrous weights (dm) are obtained by multiplying the fresh weight by 1 minus the water content (dm = fresh weight * (1 - wc)). These water contents (wc) are between less than 10% for seeds and dried fruits and more than 90% for fresh fruits and vegetables.

The consumption of food products ranges from purchase on the market to the use-by date which is generally well-informed and considers storage by refrigeration and freezing which significantly extends it. For most plants, CP is in the order of a year, while RM varies with the water content of the products between a few months and a few decades. Some seeds retain their germination power for several centuries. Appendix 1 shows the global production of the 160 agricultural commodities listed by FAO. The mean CSD weighted by anhydrous products of world agricultural crops in 2022 is 9.1 ± 0.3 years, according to formula (1).

The quantities of meat, milks and eggs consumable by humans expressed in tonnes are described in the FAO statistics “Production - Quantity” and “Primary Livestock”. These products are mainly intended to supplement human food with quality proteins and a set of micronutrients and mineral salts. They result from the conversion by animals, mainly ruminants and monogastrics, of plants from meadows, foliage and crops.

According to Mottet et al. (2017) , global livestock consumed 6 Gt/year of anhydrous fodder. The feed conversion rate, given the weight of anhydrous fodder per weight of animal protein, would be 80 for all species combined, according to these authors. A portion of fodder is included in the previous section of agricultural plant products. Still according to Mottet et al. (2017) , 14% of the tonnage consumed by global livestock is consumable by humans, A reduction of the same amount is applied to anhydrous fodder so as not to count twice this part consumable by humans.

Appendix 2 shows the 48 global livestock products listed by the FAO. The protein levels used vary between 3.2% for milk and 32% for meat products. Anhydrous fodder is calculated by applying the conversion rate of Mottet et al. (2017) . Anhydrous fodder of non-food or protein-free products such as fats, skins, waxes, honey are calculated by applying the conversion rate (dry weight of fodder/fresh weight of livestock product) obtained with protein products, i.e. 4.2.

Fodder is produced and stored for 0.8 to 3 years to await maturity and pass unproductive periods. The RM period of livestock products lasts between 40 days (eggs) and 100 years (beeswax). The mean CSD weighted by the anhydrous weights of fodder from global livestock farms is 1.2 ± 0.1 years, in accordance with formula (1), in 2022.

Statistics on the quantities of wood produced in the world are available in the “Forests” section of “Data” from FAO. These “Production - Quantity” of firewood, saw logs. pulpwood and industrial wood are expressed in cubic meters to be converted into anhydrous tonnes by applying a density varying between 0.49 and 0.70 depending on the nature of the wood.

Appendix 3 shows the 8 global forestry productions in 2022 and the durations of the CP and RM periods. Only the felling of standing trees was considered, excluding products made using the wood from these felling’s to avoid counting them twice. The plant growth period (CP) is between 25 and 50 years. The restitution by mineralization period (RM) varies between 2 years for firewood and 3 centuries for construction timber. The mean CSD weighted by anhydrous products of global forestry calculated according to formula (1) is 31.0 ± 2.7 years in 2022.

Appendix 4 shows global aquaculture and fisheries products as provided for 2021 by FAO FishStatJ data. Their mean weighted CSD calculated according to formula (1) was 1.6 ± 1.0 years. These productions are an order of 10 lower than the previous ones.

A portion of the carbon in plant products does not return to the atmosphere. It remains part of the animal biomasses, livestock and humanity that feed on it and contribute to the weight increase described in Appendix 5. The mean weighted CSD is 9.5 ± 0.7 years.

Table 2 summarizes the calculation elements depicted in Appendix 1 to 5. Considering only the commercial products, the main global contributions to atmospheric carbon capture by photosynthesis were agriculture (forages and crops) and forestry in 2022.

Storage durations (CSD) of commercial products are derived from the analysis of FAO statistics to which CP and RM durations were assigned, ranging from a few days for fresh vegetables to 300 years for construction timber. The mean CSD weighted by anhydrous commercial productions of agriculture and forestry was 9.9 ± 0.2 years in 2022. The minor captures (Appendix 4 and 5) were below the precision of the calculation and not considered further.

The figures in Table 2 do not include non-commercial plant biomass such as leaves and stems for the aerial part, and roots and exudates for the buried part. These non-commercial aerial and buried parts most often remain on site and enrich the soil with organic carbon.

Unlike the commercial parts, the above and below-ground parts of crops are not included in statistics, having no commercial value. The former part was subject of evaluation in what follows using the harvest index HI, which is the ratio of biomass of the commercial part harvested to the aerial part of the whole plant. This index is on average 0.42 ( Hay, 1995 ) for the three main crops (maize, wheat, rice) which alone provide half of the world’s agricultural production (see Appendix 1). The 157 other crops listed by the FAO have HIs which vary with the cultivar and the climate-edaphic conditions in such proportions that an average is difficult to determine. This other half of world production would merit an in-depth examination species by species. However, if we assume that the same average HI is applicable to this other half, the total capture of aerial parts of global crops would have been in this case 12.1 ± 0.3 GtCO₂/year ( Table 3 ). For the products of forestry, we took an HI of 0.8.

The below-ground parts are more difficult to assess as they are invisible and rarely quantified. According to Blume et al. (2015) , the root system represents 10% to 20% of the total biomass of a tree, 10% to 50% of the biomass of cultivated plants and 50% to 80% of grassland plants. The aerial parts of fodder are counted among the commercial products. The return to the soil in the form of excrement was neglected, minimizing the estimate. Under these conditions, Table 3 shows the distribution of CO₂ capture and storage in the two non-commercial parts of cultivated terrestrial plants, and the sum of the commercial and non-commercial parts (whole plants).

According to this calculation, capture by whole plants have reached 41.0 ± 0.6 GtCO₂ in 2022, which is twice that of commercial products alone. This figure is an understatement because it does not include the share of non-commercially exploited plant covers, whose contribution is still to specify. Unexploited continental vegetation is stable and plays a marginal role in capturing atmospheric carbon. Should unmanaged forests still capture carbon, they do so at levels much lower than most crops, which FAO mean double yield weighted by dry weights of the products is set at 19.5 ± 0.7 tCO₂/ha/y. For example, current logging regulations in Brazil adopt a standard post-harvesting recovery rate of 0.86 m³/ha/y ( Vidal et al., 2020 ), which captures 0.9 tCO₂/ha/y. It should be noted here that, contrary to popular belief, forest cutting and burning do not result in a loss of capacity to store and capture CO₂. The soil stock remains in place while spontaneous or replanted regrowth continues to capture to ensure tree growth before new cuttings 25 to 35 years later. If the forest is replaced by pastures or crops, the space it frees up captures CO₂ at much higher levels.

The capture we obtained represented 1.3% of total atmospheric CO₂, 3.2 times the continental sink of Friedlingstein et al. (2022) at 12.8 ± 3.3 GtCO₂/y, and 2.1 times that of Pan et al. (2024) at 19.8 GtCO₂/y. It also represented four times the oceanic sink estimated by Fay et al. (2023) at 10.3 ± 0.4 GtCO₂/y. Above all, it compensated the total anthropogenic emissions estimated at 41.46 GtCO₂ in 2022, including 37.15 GtCO₂ by energy and industrial combustion of fossil hydrocarbons ( Ritchie & Roser, 2024 ; data from Global Carbon Budget). The difference between these last figures is attributed to non-fossil emissions by mineralization of plant products and by consumption of fossil products other than energy, especially for fertilizers. Even if we attribute all those additional emissions (4.3 GtCO₂, Land Use Change) to plant cultivation, its net balance will remain largely positive with 36.7 GtCO₂/y removed from the atmosphere.

Our figure was significantly higher than those cited by the IPCC according to which the continental sink would be between 10 and 20 GtCO2/y ( Amthor & Baldocchi, 2001 ; Beer et al., 2010 ; Ciais et al., 2013 ; Dusenge et al., 2018 ; Friedlingstein et al., 2023 ), and we have tried to understand why such a difference. The figures cited by the IPCC are constructed from field surveys describing carbon fluxes per unit of surface area and time, then extended to the global surfaces of the different plant covers by dynamic models. These covers are forests (tropical, temperate, boreal), tundras, Mediterranean scrubs, crops, savannas, tropical pastures, temperate pastures, and deserts. The surface capture is between 3.7 (deserts) and 43.1 (tropical forest) tCO₂/ha/y ( Saugier et al., 2001 ). The share of forests was greatly overestimated, while that of crops, credited with only 10.1 tCO₂/ha/year, was significantly less than the FAO mean crop yield seen above. The capture of old-growth unexploited forests was subject to debate revealing the difficulty of its evaluation ( Luyssaert et al., 2008 ; Gundersen et al., 2021 ; Luyssaert et al., 2021 ).

The difference between our approach, relying on the chemical composition of the marketed products described by FAO statistics, and that of the authors cited by the IPCC, based on field measures of carbon fluxes extended to cover surfaces, resulted in a four times lower capture of the latter. It seems that this approach comes up against problems of representativeness of the sampling of each plant cover, which surface was defined by satellite remote sensing. In our case, the sampling is extended to all commercialized plants, to which should be added the share of unexploited plants if significant.

The soil carbon pool is made up of the non-commercialized aerial parts remaining on site and the below-ground parts after harvest. The CSD of these necromasses correspond to the average residence time of carbon before mineralization by decomposing organisms, erosion and leaching. An annual crop cycle benefits from the carbon stocks already present in the soils which are partly consumed. The additional storage of successive annual cycles makes it possible to stabilize the carbon pool towards an equilibrium. It is this equilibrium that we have tried to render by choosing the retention times from the literature. For forests, the retention times varied between 0.9 and 152 years ( Wang et al., 2017 ). They increased with latitude (x4) and decreased with temperature (x10) and precipitation (x4). Temperate regions achieve a compromise between high and low latitudes. Also, an average residence time of 40 years ( Balesdent & Recous, 1997 ; Pellerin et al., 2019 ) for the soils of crops and meadows, and of 75 years for those of exploited forests were retained ( Table 4 ). Assuming a normal distribution of mineralization over time, these durations were divided by 2, in accordance with formula (1).

Since the capture duration CP cannot be shorter for whole plants than for commercial products, we have taken the same duration as the latter in the following.

Figure 2 shows how were distributed over time the atmospheric CO₂ absorption and restitution by products and whole plants of agriculture and forestry harvested in 2022 according to Equation (2) and their argument values depicted in Table 5 .

Should all the years after 2022 be identical to it, then successive carbon captures and storages would constitute a stock of 294 GtC by biofixation of 1.078 GtCO₂ from atmosphere (or CSD times the whole plant capture of 2022 harvest), in addition to the pre-existing one, after 26.3 years (or CSD years) and which remains at this level thereafter. This is of the order of magnitude of the stocks of continental vegetation alone, without those of soils and permafrosts, of Canadell et al., 2021 , (see Table 1 ). By going back further into the last century, it would be possible to quantify the carbon stocks of whole cultivated plants that have been built up since then.