Carbon Potential

We estimate the carbon potential of all Biomes and their Ecoregions using published global datasets

This is a collaborative project. Please contact us if you wish to contribute.

Joel Fiddes and Christopher Philipson

The aim is to estimate both the soil and plant carbon potential for all global terrestrial Biomes and Ecoregions.

Example Biomes

For areas of low human impact, we explore the relationship between Soil Oganic Carbon (SOC), Plant Carbon and Canopy Cover for all Biomes accounting for variation in Ecoregions. Here we present four example Biomes: Fig 2. The relationship between carbon density and canopy cover for ‘low impact areas’. Data are presented as a heat scatter plot with separate columns for Aboveground Carbon Density (first column), Soil Organic Carbon (second column) and Total Carbon (third column). We present four example Biomes: In the first row, tropical moist forests, in the second row, tropical coniferous forests, in the third row tropical grasslands, savannahs & schrublands, and in the forth row mangroves. All biomes are presented in Fig S1. Frequency density plots adjacent to the axis highlight the distribution of each variable (carbon density and canopy cover) and therefor the highlight most common canopy cover % and most common carbon density independently of their relationship.

Methods

Low impact areas mask

In order to estimate the carbon potential of undisturbed ecosystems, we only sample Low Impact Areas (LIA) as defined in Jacobson et al. (2019). The reason for this is that we aim to estimate the carbon potential in ecosystems close to an undisturbed state as possible.

Input datasets

We use the following datasets. We are open to using other datasets if available (dependent on funding for computational time).

Dataset	Resolution	Unit	Time ref	Reference	Shortname	URL
Global Forest Change	30m	%	2000	Hansen et al. (2013)	HANSEN	https://earthenginepartners.appspot.com/science-2013-global-forest
Low Impact Areas	1km	NA	2006-2015	Jacobson et al. (2019)	LIA	https://www.nature.com/articles/s41598-019-50558-6
Global Forest Watch Above ground biomass	30m	Mg/Ha	2000	Woods Hole Research Center; See also Baccini et al.(2012)	GFW	https://data.globalforestwatch.org/datasets/aboveground-live-woody-biomass-density?geometry=-117.070%2C-21.549%2C33.047%2C70.520
Hengl Soil organic carbon	250m	Mg/ha	2000-2017	Hengl et al.(2017)	HENGL	https://doi.org/10.1371/journal.pone.0169748 https://soilgrids.org/#!/?layer=ORCDRC_M_sl2_250m&vector=1 https://files.isric.org

Woods Hole Research Center. Unpublished data. Accessed through Global Forest Watch Climate March 2020. climate.globalforestwatch.org

Hengl T, Mendes de Jesus J, Heuvelink GBM, Ruiperez Gonzalez M, Kilibarda M, Blagotić A, et al. (2017) SoilGrids250m: Global gridded soil information based on machine learning. PLoS ONE 12(2): e0169748.

Computational steps:

We follow these computational steps to generate datasets of soil organic carbon (SOC), aboveground biomass (AGB) and canopy cover (CANOPY) for all Ecoregions

1. Retrieve all global data

To retrieve all global data tiles (AGB, SOC, CANOPY), we loop the algorithm over all tiles that constitute the entire global land surface. The AGB tiles define the computation domain.

2. Resample

Resample SOC and LIA tiles to a common 30m grid

3. Mask Low Impact Areas

Mask all data layers by LIA to extract data only from low impact regions for further analysis.

4. Random Sample

Extract a 10000 random-point sample of values of each component (SOC, AGB and CANOPY) for each ecoregion. This results in a number of results tables per tile (AGB, SOC, CANOPY, 3 cols x 10000 rows) equal to number of ecoregions within a tile.

5. Aggregate tiles

Finally, we aggregate all values by Ecoregion resulting in a global dataset of Ecoregions each with SOC, AGB, CANOPY.

6. Predictions and graphs

Results are plotted as xy density scatter plots per Biome, due to large size of datasets. We model the realtionship between carbon (each SOC and AGB) and canopy cover for each Biome using linear mixed effects models with a random effect for Ecoregion. We estimate the carbon potential by predicting the SOC and AGB at the mean canopy cover for each biome. We will do this at the ecoregion level.

Example Ecoregions

Five example ecoregions from Biome 1, 'Tropical & Subtropical Moist Broadleaf Forests'

Table 2. Carbon Potential aggregated at the Biome level.

This is the average of all Ecoregions.

Biome #	Canopy Cover (%)	Plant Carbon (Mg/ha)	SOC (Mg/ha)	Biome name	Bastin 2019 Area (Mha)	Current SOC Stock (Gt)	Plant Carbon Potential (Gt)	Soil Carbon Potential (Gt)
1	83	217	501	Tropical & Subtropical Moist Broadleaf Forests	82	6.3	18	41
2	60	135	267	Tropical & Subtropical Dry Broadleaf Forests	19	0.5	3	5
3	60	149	372	Tropical & Subtropical Coniferous Forests	6.8	0.1	1	3
4	57	118	450	Temperate Broadleaf & Mixed Forests	123.5	5.8	15	56
5	40	130	419	Temperate Conifer Forests	39	1.9	5	16
6	41	49	826	Boreal Forests/Taiga	284.7	23.9	14	235
7	22	69	208	Tropical & Subtropical Grasslands, Savannas & Shrublands	166.2	2.3	11	35
8	12	58	302	Temperate Grasslands, Savannas & Shrublands	92.4	3.7	5	28
9	18	39	363	Flooded Grasslands & Savannas	8.3	0.3	0	3
10	13	61	370	Montane Grasslands & Shrublands	18.4	1.4	1	7
11	5	21	889	Tundra	110.9	16.5	2	99
12	21	68	206	Mediterranean Forests, Woodlands & Scrub	18.5	0.5	1	4
13	3	30	141	Deserts & Xeric Shrublands	73.7	2.1	2	10
14	87	248	905	Mangroves	2.1	0.2	1	2
					Total	65.5 (Gt)	79 (Gt)	544 (Gt)