Tree planting cost-effectiveness
Disclaimer: this is not a project from SoGive.
1. Summary
In this analysis, the mean, and 5th and 95th percentiles of the cost-effectiveness of tree planting in t/£ were estimated to be[1]:
For the UK, 0.07 (-0.04 to 0.3).
For The Woodland Trust (TWT), 0.04 (-0.02 to 0.2).
For Eden Reforestation Projects’ (ERP’s) mangrove projects in Madagascar, 0.3 (-0.2 to 1).
At best[2], 1 (-0.8 to 4).
These values are lower than or similar to the bar of 1 t/$ Founders Pledge (FP) arguably considers for the grants of its Climate Change Fund (CCF). In addition, the 5th and 95th percentiles of the cost-effectiveness of CCF in t/$ were guessed to be 10^-1 and 10^4 by Johannes Ackva[3] (although these should be intended as informed guesses, not resilient estimates; see here). As a result, donating to CCF is expected to be at least 1 order of magnitude more effective than supporting the best tree planting interventions[4].
Author contributions
The contributions by author are as follows:
Hanzhang Ren: in-depth review.
Melissa Bedinger: background research.
Vasco Grilo: background research, cost-effectiveness modelling, and writing.
Acknowledgements
Thanks to Andrew Baker, Haley Koleszar, Ilona Coulson-Ashworth, Matt Sharp, Sanjay Joshi, and Vince van ’t Hoff.
2. Context
Motivation
This research has the goal of assessing the cost-effectiveness of tree planting in the UK, TWT, ERP, and idealised tree planting projects. These are currently not recommended by FP nor Giving Green, which consider themselves aligned with Effective Altruism[5].
Tree planting can be appealing to donors seeking carbon offsets, especially risk averse ones who perceive it as having a higher likelihood of success than the projects of organisations working on policy advocacy or innovative green technology[6]. In reality, the risks of tree planting projects are often underestimated.
Analysed projects
The cost-effectiveness of tree planting depends on the specific project. In this analysis, it was assessed for:
Typical UK projects, via:
“WCC” projects: reference Woodland Carbon Code (WCC) projects[7].
“Small” projects[8]: reference WCC projects with an area no larger than 5 ha.
“Standard” projects: reference WCC projects with an area larger than 5 ha.
“TWT” projects: reference WCC projects from TWT.
“ERPM” projects: projects from ERP involving planting mangroves in Madagascar.
ERP was rated as Silver (Tentative) under the SoGive methodology.
There were preliminary signs that mangroves could be cost-effective (see Eden Reforestation Projects).
“ITP” projects: idealised tree planting interventions.
These offer an upper bound for how good tree planting can be.
In addition, UK tree planting charities are overviewed and rated in Appendix A, and climate change organisations analysed by the Effective Altruism community in Appendix B.
3. Methodology
Overview
The contents of the following sections are:
Cost-effectiveness: description of the model and variables which affect the cost-effectiveness.
Input distributions: definition of the distributions which determine the cost-effectiveness.
Cost-effectiveness
The cost-effectiveness depends on the following[10]:
“Heat mortality benefits (life/ha)” (HMB): heat effects on human mortality.
“Ecosystem services benefits (life/ha)” (ESB): economic and health effects of the ecosystem services, excluding the heat effects mentioned just above.
“Effect on insects (life/ha)” (EI): welfare effects on non-nematode and non-earthworm invertebrates on land (i.e. terrestrial arthropods).
“Existential risk mitigation (life/ha)” (ERM): impact on the reduction of existential risk.
“Cost (£/ha)” (C): total cost of the forestation intervention, which includes the cost to develop, plant, grow and maintain the trees.
The methodology to estimate the first 4 variables is described in the following sections. Other effects are discussed here.
The cost-effectiveness metrics analysed in this report are as follows:
“Cooling cost-effectiveness (life/£)” = CCE = HMB / C.
“Ecosystem services cost-effectiveness (life/£)” = ESCE = ESB / C.
“Insects cost-effectiveness (life/£)” = ICE = EI / C.
“Existential risk cost-effectiveness (life/£)” = ERCE = ERM / C.
“Cost-effectiveness (life/£)” = CE = CCE + ESCE + ICE + ERCE.
Heat mortality benefits
The “heat mortality benefits (life/ha)” were estimated from:
HMB = ANRE * MCC * HMBDF * RF.
“Adjusted net removal of CO2e emissions (t/ha/year)” (ANRE): total net cooling effect in terms of removal of CO2e emissions. It was estimated from[11]:
ANRE = NRE—EAC / DER.
“Net removal of CO2e emissions (t/ha/year)” (NRE): difference between the factual and counterfactual removal of CO2e emissions per unit area per unit time (biogeochemical effect).
“Effect of albedo change (t/ha)” (EAC): impact of changing the surface reflectivity on the radiative forcing (i.e. change in energy flux) expressed as the production of additional CO2e emissions per unit area (biogeophysical effect).
“Duration of emissions removal (year)” (DER): time between afforestation/reforestation and the moment after which the net removal of CO2e emissions is assumed to be null.
“Mortality cost of carbon (life/t)” (MCC): mortality caused by additional CO2e emissions in 2020.
“Heat mortality benefits duration factor (year)” (HMBDF): time between afforestation/reforestation and the end of the project, adjusted for the change of the MCC, which influences the future mortality benefits. It was estimated from[12]:
HMBDF =[13] ((1 + MCCR)^DER − 1) / MCCR.
“Mortality cost of carbon annual variation rate” (MCCR): rate at which the MCC varies.
“Risk factor” (RF): likelihood that the factual will not differ from the counterfactual. It was estimated from:
RF = (1 - TPIR)(1 - ER).
“Tree planting intervention risk” (TPIR): risk related to the tree planting project.
“Existential risk” (ER): “an existential risk is a risk that threatens the destruction of humanity’s longterm potential” (The Precipice).
Ecosystem services benefits
The “ecosystem services benefits (life/ha)” were estimated from:
ESB = (FES—CES) / GDPC * VHL / USL * ABDF * RF.
“Factual initial ecosystem services ($/ha/year)” (FES): factual ecosystem services besides climate regulation.
“Counterfactual initial additional benefits ($/ha/year)” (CES): counterfactual ecosystem services besides climate regulation.
“Real GDP per capita (int-$)” (GDPC).
“Value of fully healthy life in terms of multiples of real GDP per capita (1/QALY)” (VHL): equivalence between health and wealth.
“Utility of saving a life (QALY/life)” (USL).
“Additional benefits duration factor (year)” (ABDF): time between afforestation/reforestation and the end of the project, adjusted for the change of the GDP per capita, which influences the future additional benefits. This was estimated from[14]:
ABDF =[15] (1 - (1 + GDPCR)^-DER) / (1 - (1 + GDPCR)^-1).
As GDPCR tends to zero, ABDF converges to DER.
“Real GDP per capita annual growth rate” (GDPCR): rate at which the GDP per capita increases.
Effect on insects
The “effect on insects (life/ha)” was estimated from:
EI = (FI—CI) * WI * IW / USL * DER * RF.
“Factual number of insects per unit area as a fraction of the worldwide mean” (FI).
“Counterfactual number of insects per unit area as a fraction of the worldwide mean” (CI).
“Worldwide number of insects per unit area (insect/ha)” (WI). This was estimated from:
WI = I / LA.
“Worldwide number of insects (insect)” (I).
“Worldwide land area (ha)” (LA).
“Insects welfare (-QALY/insect/year)” (IW): utility of insect life.
Existential risk mitigation
The “existential risk mitigation (life/ha)” was estimated from:
ERM =[16] ANRE * DER * (1 - TPIR) * LTMCC.
“Longterm mortality cost of carbon (life/t)” (LTMCC):
LTMCC = ERCC * FV / CGHG.
“Existential risk due to climate change” (ERCC).
“Humanity’s future value (life)” (FV):
FV = FPS * FD * FU / USL.
“Humanity’s future population size (person)” (FPS).
“Humanity’s future duration (year)” (FD).
“Humanity’s future utility (QALY/person/year)” (FU).
“Cumulative greenhouse gas emissions between 2020 and 2100 assuming current climate policies (t)”[17] (CGHG).
Normalised cost-effectiveness
To facilitate the interpretation of the results, normalised cost-effectiveness metrics were obtained by multiplying the cost-effectiveness metrics by the SoGive life-saving Gold Standard benchmark of 5 k£/life:
“Normalised cost-effectiveness metric” = “cost-effectiveness metric (life/£)” * “5 k$/life”.
The normalised cost-effectiveness has the following interpretation:
A value of x implies that the intervention is x times as effective as the SoGive Gold Standard.
As the SoGive Gold Standard benchmark for the removal of CO2e emissions is 1 £/t, the cost-effectiveness is roughly[18] x t/£.
A negative value implies that the intervention is harmful, whereas a positive value implies that it is beneficial (as the cost-benefit ratio).
A null value implies that the intervention is neutral (unlike the cost-benefit ratio, for which a null value implies that the intervention is either infinitely harmful or beneficial).
Input distributions
Distribution types
In order to determine a range for the cost-effectiveness, the variables which affect it were modelled as:
Constant:
Mortality cost of carbon annual variation rate.
GDP per capita.
Value of fully healthy life in terms of multiples of real GDP per capita.
Utility of saving a life.
Worldwide land area.
Net removal of CO2e emissions.
Effect of albedo change.
Mortality cost of carbon.
Value of a statistical life annual growth rate.
Duration of emissions removal.
Value of a statistical life.
Factual initial ecosystem services.
Counterfactual initial ecosystem services.
Factual number of insects per unit area as a fraction of the worldwide mean.
Counterfactual number of insects per unit area as a fraction of the worldwide mean.
Worldwide number of insects.
Humanity’s future duration.
Humanity’s future population size.
Humanity’s future utility.
Cumulative greenhouse gas emissions between 2020 and 2100 assuming current climate policies.
Cost.
Insects welfare[19].
Beta:
Tree planting intervention risk.
Existential risk.
Existential risk due to climate change.
The reasons are as follows:
Constant:
Variables for which either information about quantiles could not be easily found or uncertainty is negligible.
Normal and truncated normal:
Distributions whose mean is similar to the median.
The variables with negative values were defined as normal, and those without as truncated normal.
Lognormal and loguniform:
Distributions which respect multiplicative processes, and therefore whose mean is larger than the median.
The variables with a seemingly longer tail were defined as lognormal distributions, and those with a seemingly shorter tail as loguniform distributions.
Beta:
Distributions with values between 0 and 1.
To facilitate the calculations, all distributions were assumed to be independent. Their parameters are defined in the following sections.
Distribution parameters
The type and parameters of each of the variables are presented in the table below for the various analysed projects[20]. The estimation of the distribution parameters is described in the following section.
| Variable | Type of distribution | Parameters |
|---|---|---|
| Net removal of CO2e emissions (t/ha/year) | Normal | Mean (small): 6.90. Mean (standard): 4.82. Mean (WCC): 4.84. Mean (TWT): 5.60. Mean (ERPM): 14.0. Mean (ITP): 40.7. Standard deviation (standard): 2.16. Standard deviation (WCC): 2.17. Standard deviation (TWT): 0.680. Standard deviation (ERPM): 1.70. Standard deviation (ITP): 4.94. |
| Effect of albedo change (t/ha) | Normal/Constant | Mean (small/standard/WCC/TWT): 65.1. Mean (ERPM): 322. Standard deviation (small/standard/WCC/TWT): 44.2. Standard deviation (ERPM): 221. Value (ITP): 0. |
| Duration of emissions removal (year) | Truncated normal | Minimum: 1. Maximum: 100. Mean (small): 76.5. Mean (standard): 81.0. Mean (WCC): 80.9. Mean (TWT/ERPM/ITP): 100. Standard deviation (small): 23.9. Standard deviation (standard): 25.3. Standard deviation (WCC): 25.3. Standard deviation (TWT/ERPM/ITP): 31.2. |
| Mortality cost of carbon (life/t) | Normal | Mean: 2.54*10^-4. Standard deviation: 2.58*10^-4. |
| Mortality cost of carbon annual variation rate | Constant | Value: 0. |
| Tree planting intervention risk | Beta/Constant | Alpha (small/standard/WCC/TWT/ERPM): 2.80. Value (ITP): 0. |
| Existential risk | Beta | Alpha: 1.50. Beta: 8.00. |
| Factual initial ecosystem services ($/ha/year) | Lognormal/Constant | Mean (median) (UK temperate forests): 6.01 (408). Mean (median) (ERP mangroves): 4.32 (78.8). Standard deviation (ERP mangroves): 3.09. Value (ITP): 0. |
| Counterfactual initial ecosystem services ($/ha/year) | Lognormal/Constant | Mean (median) (UK temperate grasslands): 2.26 (9.62). Mean (median) (UK cultivated areas): 4.88 (132). Mean (median) (ERP cultivated areas): 7.07 (1.18 k). Standard deviation (UK cultivated areas): 2.33. Standard deviation (ERP cultivated areas): 1.10. Value (ITP): 0. |
| Real GDP per capita (int-$) | Constant | Value (small/standard/WCC/TWT): 49.7 k. Value (ERPM/ITP): 1.64 k. |
| Value of fully healthy life in terms of multiples of real GDP per capita (1/QALY) | Constant | Value: 2.00. |
| Utility of saving a life (QALY/life) | Constant | Value: 51.0. |
| Real GDP per capita annual growth rate | Normal | Mean (small/standard/WCC/TWT): 0.900 %. Mean (ERPM/ITP): 2.14 %. Standard deviation (ERPM/ITP): 0.841 %. |
| Factual number of insects per unit area as a fraction of the worldwide mean | Lognormal/Constant | Mean (median) (small/standard/WCC/TWT/ERPM): 0.833 (2.30). Standard deviation (small/standard/WCC/TWT/ERPM): 0.414. Value (ITP): 0. |
| Counterfactual number of insects per unit area as a fraction of the worldwide mean | Lognormal/Constant | Mean (median) (small/standard/WCC/TWT/ERPM): 0.405 (1.50). Standard deviation (small/standard/WCC/TWT/ERPM): 0.391. Value (ITP): 0. |
| Worldwide number of insects (insect) | Lognormal | Mean (median): 41.4 (1.00 E). Standard deviation: 1.40. |
| Worldwide land area (ha) | Constant | Value: 1.30*10^10. |
| Insects welfare (-QALY/insect/year) | Loguniform | Minimum: 0.00400. Maximum: 0.0630. |
| Existential risk due to climate change | Beta | Alpha: 1.73. Beta: 17.3 k. |
| Humanity’s future population size (person) | Lognormal | Mean (median): 0 (1.00). Standard deviation: 47.7. |
Humanity’s future duration (year) | Lognormal | Mean (median): 15.1 (3.48 M). Standard deviation: 10.8. |
| Humanity’s future utility (QALY/person/year) | Lognormal | Mean (median): 0 (1.00). Standard deviation: 3.01. |
| Cumulative greenhouse gas emissions between 2020 and 2100 assuming current climate policies (t) | Lognormal | Mean (median): 28.9 (3.71 T). Standard deviation: 0.160. |
| Cost (k£) | Lognormal | Mean (median) (small/standard/WCC): 8.15 (3.46 k). Mean (median) (TWT): 9.13 (9.23 k). Standard deviation (small/standard/WCC): 0.717. Standard deviation (TWT): 0.804. Standard deviation (ERPM/ITP): 0.200. |
Net removal of CO2e emissions
The parameters for the net removal of CO2e emissions were estimated as follows:
For “small”, “standard”, “WCC” and “TWT” projects:
The mean and standard deviation were calculated based on the area weighted mean and standard deviation of the net removal of CO2e emissions per unit area per unit time[21].
The net removal of CO2e emissions per unit area per unit time was estimated based on the net removal of CO2e emissions, area and project duration of the respective reference WCC projects.
The results were summarised in the tab “WCC projects”.
An overview of the WCC methodology is available in Appendix C.
For “ERPM” projects:
The mean was estimated from the product between the annual net removal of CO2e emissions and the time mangroves take from planting until the net removal of CO2e emissions is no longer significant[22].
The annual net removal of CO2e emissions was calculated from the product between:
The annual net removal of CO2e emissions for global established mangroves determined in Cameron 2019[23] (Table 1, second to last row and column).
The ratio between the carbon stock density presented in Jones 2014 (section 3.4) for Madagascar closed-canopy mangroves[24] and the one in Kauffman 2017 (section “Global comparisons”) for global mangroves.
The standard deviation was estimated such that the coefficient of variation equals that of “TWT” projects.
For “ITP” projects:
The mean was set to the maximum net removal of emissions per unit area per unit time of 40.7 t/ha/year identified by One Tree Planted here based on this carbon storage calculator from Winrock International. The value respects a mean over the first 20 years after planting.
The standard deviation was estimated such that the coefficient of variation equals that of “TWT” projects.
The parameters calculation is in cells C2:D7 of tab “CEAs”.
Effect of albedo change
Except for “ITP” projects, the parameters for the effect of albedo change were estimated as follows:
The mean and standard deviation were estimated from the 5th and 95th percentiles of the radiative forcing, based on Eq. (6) of Bright 2020[25].
The 5th and 95th percentiles of the radiative forcing were estimated from the emissions equivalent of shortwave forcing (EESF) of an optimistic and pessimistic scenario for the radiative forcing caused by the albedo change.
For “small”, “standard”, “WCC” and “TWT” projects:
The 5th percentile was estimated from the mean of optimistic scenarios concerning changes from:
Dark soil arable cropland to light soil dense coniferous forest.
Long grass to light soil dense coniferous forest.
Dark soil arable cropland to light deciduous forest.
Long grass to light deciduous forest.
The 95th percentile was estimated from the mean of pessimistic scenarios concerning changes from:
Light soil arable cropland to dark soil dense coniferous forest.
Short grass to dark soil dense coniferous forest.
Light soil arable cropland to dark deciduous forest.
Short grass to dark deciduous forest.
For “ERPM” projects:
The 5th percentile was estimated from the mean of optimistic scenarios concerning changes from:
Dark soil arable cropland to mangrove forest (Alagoas, Brazil).
Long grass to mangrove forest (Alagoas, Brazil).
Water surface to mangrove forest (Alagoas, Brazil).
Light wet sand to mangrove forest (Alagoas, Brazil).
The 95th percentile was estimated from the mean of pessimistic scenarios concerning changes from:
Light soil arable cropland to mangrove forest (Southeastern Brazil).
Short grass to mangrove forest (Southeastern Brazil).
Dirty water surface to mangrove forest (Southeastern Brazil).
Dark dry sand to mangrove forest (Southeastern Brazil).
The parameters calculation is in cells C8:D9 of tab “CEAs”.
For “ITP” projects:
The mean was set to 0, which overestimates the cost-effectiveness since trees absorb light.
The standard deviation was set to 0, as it tends to be somewhat directly proportional to the mean.
Duration of emissions removal
The parameters for the duration of emissions removal were estimated as follows:
The minimum and maximum in years were set to 1 and 100, which is the longest project duration of WCC projects.
For “small”, “standard” and “WCC” projects:
The non-truncated mean and standard deviation were set to the area weighted mean and standard deviation of the project duration of the respective reference WCC projects.
For “TWT” projects:
The non-truncated mean was set to the area weighted mean of the project duration of the respective reference WCC projects.
The non-truncated standard deviation was set such that the coefficient of variation equals that of “WCC” projects.
For “ERPM” and “ITP” projects:
The non-truncated mean was set to 100 year, which is the longest project duration of WCC projects.
The non-truncated standard deviation was set such that the coefficient of variation equals that of “WCC” projects.
The calculation of the parameters is in cells C10:C14 of tab “CEAs”.
Mortality cost of carbon
The parameters for the mortality cost of carbon were estimated as follows:
The mean and standard deviation were determined from the 5th and 95th percentiles[26].
The 5th and 95th percentiles were set to the lower (“<10th percentile”) and upper (“>90th percentile”) bound of the 2020 MCC estimated in Bressler 2021, which accounts for temperature-related human mortality impacts.
Bressler’s MCC only accounts for the mortality impacts between 2020 and 2100 (see equation (3)).
However, the 2020 MCC estimate was assumed here to be representative of the “true” MCC, which includes the mortality impacts past 2100.
The 95th percentile was set to the upper (“>90th percentile”) bound estimated in Bressler 2021 for the 2020 MCC.
The parameters calculation is in cells C15:D15 of tab “CEAs”.
Mortality cost of carbon annual variation rate
The mortality cost of carbon annual variation rate was defined as follows:
The MCC tends to decrease over time. To illustrate this point:
The 2020 MCC includes the mortality impacts from 2020 on (in agreement with what was assumed just above).
The 2080 MCC includes the mortality impacts from 2080 on.
Consequently, the 2080 MCC excludes the mortality impacts from 2020 to 2080 (which are included in the 2020 MCC).
However, for simplicity, the MCC variation rate was set to zero (instead of a negative value), which overestimates the cost-effectiveness.
Tree planting intervention risk
Except for “ITP” projects, the parameters for the tree planting intervention risk were estimated as follows:
Alpha and beta were estimated from the mean and mode[27].
The mean was set to the risk factor of 20 % used by the WCC methodology[28].
The mode was set to 17.5 %, which is the mean between the minimum risk factor of 15 % used in versions of the WCC methodology before Version 2.0[29] and the mean defined just above.
The calculation of the parameters is in cells M2:N2 of tab “CEAs”.
The above does not cover all the risks which result from not following the guiding principles for CO2e removal certification proposed by Clean Air Task Force[30] (CATF). As a result, the tree planting intervention risk is likely to be underestimated.
For “ITP” projects, the tree planting intervention risk was set to 0, in order to overestimate the cost-effectiveness.
Existential risk
The parameters for the existential risk were set to those defined in Denkenberger 2022 (see section 2.3). “The resulting mean is 16%”, which agrees with Toby Ord’s best estimate of 1⁄6 for the total existential risk within the next 100 years[31].
Factual and counterfactual initial ecosystem services
Except for “ITP” projects, the parameters for the factual and counterfactual initial ecosystem services were estimated as follows:
Except for “ITP” projects, the means and standard deviations of the logarithm were estimated from the area weighted mean and standard deviation of the logarithm of the ecosystem service values on the Ecosystem Service Valuation Database[32] (ESVD) as of 1 April 2022 (extracted from here).
Data points without a value for the ecosystem services (in Int$/ha/year) were excluded.
Data points without a value for the area were supposed to have an area equal to the mean area of the study locations with values for the area.
To obtain the ecosystem service values, the data points were grouped by study location, as the same study location often corresponds to multiple data points (each one respecting a different type of service).
Study locations with a null value for the ecosystem services were excluded.
For “small”, “standard”, “WCC” and “TWT” projects:
The biome “7 Temperate forests” was considered for the factual scenario, and the mean between “9.3 Temperate grasslands” and “14 Cultivated areas” for the counterfactual.
Only study sites in the UK were considered.
For “ERPM” projects:
The biome “3.4 Mangroves” was considered for the factual scenario, and “14 Cultivated areas” for the counterfactual.
Study sites not only in Madagascar, but also in Indonesia and Mozambique were considered[33], to include a meaningful number of data points. On the ESVD:
For “3.4 Mangroves”, there were 0 studies for Madagascar, 2 for Mozambique, and 10 for Indonesia.
For “14 Cultivated areas”, there were 0 studies for Madagascar and Mozambique on the ESVD, but 2 for Indonesia.
The calculation of the parameters is in cells F6:F7 of tabs “UK temperate forests ecosystem services”, “UK temperate grasslands ecosystem services”, “UK cultivated areas ecosystem services”, “ERP mangroves ecosystem services”, and “ERP cultivated areas ecosystem services”.
It is worth noting that the mangroves ecosystem services could have been overestimated by fully including the benefits of fisheries and wood, which account for about 80% of the total value, but could be partly obtained elsewhere.
For “ITP” projects, for simplicity, the mean and standard deviation were to 0, since ecosystem services are not a major driver of cost-effectiveness (see Key conclusions).
Real GDP per capita
The values for the real GDP per capita were set to:
For “small”, “standard”, “WCC” and “TWT” projects:
The UK 2021 real GDP per capita (from the World Bank).
For “ERPM” and “ITP” projects:
The Madagascar 2021 real GDP per capita (from the World Bank).
Value of fully healthy life in terms of multiples of real GDP per capita
The value of fully healthy life in terms of multiples of real GDP per capita was set to 2/QALY, based on the equivalence between 2 doublings of consumption and 1 QALY considered by Open Philanthropy.
Utility of saving a life
The utility of saving a life was set to the value of 51 QALY/life implied by GiveWell’s moral weights[34].
Real GDP per capita annual growth rate
The parameters for the real GDP per capita annual growth rate were estimated as follows:
The minimum and maximum were assumed to be 50% lower and higher than the median.
For “small”, “standard”, “WCC” and “TWT” projects:
The median was set to the real GDP per capita growth projected by OECD 2021 (Table 1) for the UK from 2030 to 2060.
For “ERPM” and “ITP” projects:
The median was set to Madagascar’s annualised real GDP per capita growth from 1960 to 2020, the first and latest years for which data is available in the World Bank.
The calculation of the parameters is in cells C16:D17 of tab “CEAs”.
Factual and counterfactual number of insects per unit area as a fraction of the worldwide mean
Except for “ITP” projects, the parameters for the factual and counterfactual number of insects per unit area as a fraction of the worldwide mean were estimated as follows:
The mean and standard deviation of the logarithm were estimated from the median, and standard deviation of the decimal logarithm.
Rainforest was considered as a proxy for the factual (temperate forests or mangroves), and Cerrado for the counterfactual (temperate grasslands or cultivated areas).
The median, and standard deviation of the decimal logarithm for rainforest and Cerrado were set to the values defined here by Brian Tomasik.
The calculation of the parameters is in cells C18:D19 of tab “CEAs”.
For “ITP” projects, for simplicity, the mean and standard deviation were to 0, since insects are not a major driver of cost-effectiveness (see Key conclusions).
Worldwide number of insects
The parameters for the worldwide number of insects were estimated as follows:
The mean and standard deviation of the logarithm were estimated from the 5th and 95th percentiles.
The 5th and 95th percentiles were set to the lower and upper bound presented here by Brian Tomasik.
The calculation of the parameters is in cells C20:D20 of tab “CEAs”.
Worldwide land area
The worldwide land area was set to the value estimated by The World Bank for 2021.
Insects welfare
The parameters for the insects welfare were estimated as follows:
The minimum and maximum were estimated from the product between the probability of being sentient, and the minimum and maximum welfare assuming sentience.
The probability of being sentient was set to the likelihood of the wild bug feeling pain according to the Weighted Animal Welfare Index of Charity Entrepreneurship (WAWI).
The minimum and maximum welfare assuming sentience were set to the upper and lower bound provided in WAWI for the range of the “total welfare score (with evidence)”, divided by 100.
The “total welfare score (with evidence)” is limited to 100, and therefore considering 1 QALY is equivalent to experiencing such a score for 1 year seems reasonable.
Existential risk due to climate change
The parameters of the existential risk due to climate change were set to those estimated here.
Humanity’s future population size
Humanity’s future population size was set to 10^10 person. This is the median global population predicted for 2056, and therefore likely underestimates the cost-effectiveness.
The parameters of humanity’s future population size were estimated as follows:
The mean and standard deviation of the logarithm were estimated from the median and 97.5th percentiles.
The median was set to 1, given Toby Ord’s guess in The Precipice of 50 % for the total existential risk.
The 97.5th percentile was set to the upper bound of 4*10^40, calculated in Sandberg 2021 from the ratio between the mass of the Milky Way and the mean mass of a human.
The calculation of the parameters is in cells C21:D21 of tab “CEAs”.
Humanity’s future duration
The parameters of humanity’s future duration were estimated as follows:
The mean and standard deviation of the logarithm were estimated from the quantile 1⁄6 (similar to the 17th percentile) and 99.99th percentile.
The quantile 1⁄6 was set to 100 year, given the estimate of ⅙ for the existential risk in the next 100 years given by Toby Ord in The Precipice.
The 99.99th percentile was set to 100 Tyear, as guessed in section 3.1.2 of Beckstead 2013 (search for “100 trillion years”).
The calculation of the parameters is in cells C22:D22 of tab “CEAs”.
Humanity’s future utility
The parameters of humanity’s future utility were estimated as follows:
The mean and standard deviation of the logarithm were estimated from the 5th and 95th percentiles.
The 5th and 95th percentiles in QALY/person/year were set to 0.1 and 10.
The calculation of the parameters is in cells C23:D23 of tab “CEAs”.
Cumulative greenhouse gas emissions between 2020 and 2100 assuming current climate policies
The parameters for the cumulative greenhouse gas emissions between 2020 and 2100 assuming current climate policies were estimated as follows:
The mean and standard deviation of the logarithm were estimated from the 25th and 75th percentiles.
The 25th and 75th percentiles were set to the low and high estimates of CAT (as of 29 July 2022).
The calculation of the parameters is in cells C24:D24 of tab “CEAs”.
Cost
The parameters for the cost were estimated as follows:
For “small”, “standard” and “WCC” projects:
The mean and standard deviation of the logarithm were estimated from the 5th and 95th percentiles.
The 5th and 95th percentiles were set to the lower and upper bound given by Andrew Baker (Woodland Carbon Markets Advisor) in an interview we conducted in February 2022.
For “TWT” projects:
The mean of the logarithm was calculated from the logarithm of the cost of TWT program Create Woodland, including overhead proportionally to the weight of this program on the total expenditure on charitable activities.
The standard deviation of the logarithm was determined such that the coefficient of variation of the logarithm equals that of “WCC” projects.
For “ERPM” and “ITP” projects:
The mean and standard deviation of the logarithm were estimated from the 5th and 95th percentiles.
The 5th and 95th percentiles were calculated from the product between the cost per tree and the minimum and maximum planting density provided by ERP[35].
The calculation of the parameters is in cells C25:D27 of tab “CEAs”.
4. Results
The results were obtained via Monte Carlo simulations in this Google Colab program, generating 10^7 samples for each distribution[36].
The mean, and 5th and 95th percentiles of the non-normalised and normalised cost-effectiveness metrics are presented in the tables below. The full results are in cells A30:G96 of tab “CEAs”. Additionally, cumulative distribution functions and histograms of the cooling cost-effectiveness in t/£ are available here.
Cooling cost-effectiveness
| Projects | Cooling cost-effectiveness (t/£) | ||
|---|---|---|---|
Mean | 5th percentile | 95th percentile | |
| Small | 1.07E-01 | -5.37E-02 | 4.41E-01 |
| Standard | 7.19E-02 | -4.00E-02 | 3.13E-01 |
| WCC | 7.22E-02 | -4.02E-02 | 3.15E-01 |
| TWT | 3.72E-02 | -1.83E-02 | 1.52E-01 |
| ERPM | 3.04E-01 | -2.00E-01 | 1.05E+00 |
| ITP | 1.25E+00 | -7.96E-01 | 3.86E+00 |
| Projects | Cooling cost-effectiveness (life/£) | ||
|---|---|---|---|
Mean | 5th percentile | 95th percentile | |
| Small | 2.71E-05 | -1.36E-05 | 1.12E-04 |
| Standard | 1.82E-05 | -1.02E-05 | 7.95E-05 |
| WCC | 1.83E-05 | -1.02E-05 | 7.98E-05 |
| TWT | 9.43E-06 | -4.64E-06 | 3.86E-05 |
| ERPM | 7.71E-05 | -5.06E-05 | 2.66E-04 |
| ITP | 3.16E-04 | -2.02E-04 | 9.80E-04 |
| Projects | Normalised cooling cost-effectiveness | ||
|---|---|---|---|
Mean | 5th percentile | 95th percentile | |
| Small | 1.36E-01 | -6.81E-02 | 5.58E-01 |
| Standard | 9.11E-02 | -5.08E-02 | 3.97E-01 |
| WCC | 9.15E-02 | -5.10E-02 | 3.99E-01 |
| TWT | 4.71E-02 | -2.32E-02 | 1.93E-01 |
| ERPM | 3.86E-01 | -2.53E-01 | 1.33E+00 |
| ITP | 1.58E+00 | -1.01E+00 | 4.90E+00 |
Ecosystem services cost-effectiveness
| Projects | Ecosystem services cost-effectiveness (life/£) | ||
Mean | 5th percentile | 95th percentile | |
| Small | 5.41E-07 | -2.09E-05 | 3.76E-05 |
| Standard | 5.50E-07 | -2.13E-05 | 3.83E-05 |
| WCC | 5.50E-07 | -2.13E-05 | 3.83E-05 |
| TWT | 2.31E-07 | -8.58E-06 | 1.60E-05 |
| ERPM | 2.47E-03 | -2.61E-03 | 3.86E-03 |
| ITP | 0.00E+00 | 0.00E+00 | 0.00E+00 |
| Projects | Normalised ecosystem services cost-effectiveness | ||
|---|---|---|---|
Mean | 5th percentile | 95th percentile | |
| Small | 2.70E-03 | -1.04E-01 | 1.88E-01 |
| Standard | 2.75E-03 | -1.06E-01 | 1.91E-01 |
| WCC | 2.75E-03 | -1.06E-01 | 1.91E-01 |
| TWT | 1.15E-03 | -4.29E-02 | 7.99E-02 |
| ERPM | 1.24E+01 | -1.31E+01 | 1.93E+01 |
| ITP | 0.00E+00 | 0.00E+00 | 0.00E+00 |
Insects cost-effectiveness
| Projects | Insects cost-effectiveness (life/£) | ||
|---|---|---|---|
Mean | 5th percentile | 95th percentile | |
| Small | -1.37E+03 | -6.13E+03 | 6.23E+02 |
| Standard | -1.40E+03 | -6.29E+03 | 6.40E+02 |
| WCC | -1.40E+03 | -6.29E+03 | 6.39E+02 |
| TWT | -5.99E+02 | -2.63E+03 | 2.61E+02 |
| ERPM | -2.40E+03 | -1.13E+04 | 1.29E+03 |
| ITP | 0.00E+00 | 0.00E+00 | 0.00E+00 |
| Projects | Normalised insects cost-effectiveness | ||
|---|---|---|---|
Mean | 5th percentile | 95th percentile | |
| Small | -6.85E+06 | -3.07E+07 | 3.12E+06 |
| Standard | -7.02E+06 | -3.14E+07 | 3.20E+06 |
| WCC | -7.02E+06 | -3.14E+07 | 3.20E+06 |
| TWT | -2.99E+06 | -1.32E+07 | 1.30E+06 |
| ERPM | -1.20E+07 | -5.66E+07 | 6.45E+06 |
| ITP | 0.00E+00 | 0.00E+00 | 0.00E+00 |
Existential risk cost-effectiveness
| Projects | Existential risk cost-effectiveness (life/£) | ||
|---|---|---|---|
Mean | 5th percentile | 95th percentile | |
| Small | 1.76E+92 | 5.76E-51 | 8.90E+21 |
| Standard | 1.15E+92 | -6.46E-47 | 2.02E+21 |
| WCC | 1.16E+92 | -9.26E-48 | 2.04E+21 |
| TWT | 5.81E+91 | 3.62E-50 | 3.48E+21 |
| ERPM | 3.23E+96 | 3.67E-56 | 2.20E+22 |
| ITP | 3.31E+104 | 1.70E-47 | 1.88E23 |
| Projects | Normalised existential risk cost-effectiveness | ||
Mean | 5th percentile | 95th percentile | |
| Small | 8.80E+95 | 2.88E-47 | 4.45E+25 |
| Standard | 5.75E+95 | -3.23E-43 | 1.01E+25 |
| WCC | 5.78E+95 | -4.63E-44 | 1.02E+25 |
| TWT | 2.91E+95 | 1.81E-46 | 1.74E+25 |
| ERPM | 1.62E+100 | 1.83E-52 | 1.10E+26 |
| ITP | 1.65E+108 | 8.50E-44 | 9.41E+26 |
Cost-effectiveness
| Projects | Cost-effectiveness (life/£) | ||
|---|---|---|---|
Mean | 5th percentile | 95th percentile | |
| Small | 1.76E+92 | -4.95E+03 | 8.90E+21 |
| Standard | 1.15E+92 | -6.27E+03 | 2.02E+21 |
| WCC | 1.16E+92 | -6.26E+03 | 2.04E+21 |
| TWT | 5.81E+91 | -2.07E+03 | 3.48E+21 |
| ERPM | 3.23E+96 | -1.01E+04 | 2.20E+22 |
| ITP | 3.31E+104 | -1.26E-04 | 1.88E+23 |
| Projects | Normalised cost-effectiveness | ||
|---|---|---|---|
Mean | 5th percentile | 95th percentile | |
| Small | 8.80E+95 | -2.47E+07 | 4.45E+25 |
| Standard | 5.75E+95 | -3.14E+07 | 1.01E+25 |
| WCC | 5.78E+95 | -3.13E+07 | 1.02E+25 |
| TWT | 2.91E+95 | -1.03E+07 | 1.74E+25 |
| ERPM | 1.62E+100 | -5.03E+07 | 1.10E+26 |
| ITP | 1.65E+108 | -6.32E-01 | 9.41E+26 |
5. Discussion
Key conclusions
The cost-effectiveness of tree planting is driven by the existential risk cost-effectiveness, which is perfectly correlated with the cooling cost-effectiveness, as both are directly proportional to the adjusted net removal of CO2e emissions (t/ha/year). Consequently, one could assess the overall cost-effectiveness of tree planting based on the cooling cost-effectiveness, which is expected to be:
For the UK (“small”, “standard” and “WCC”), 0.07 t/£.
For TWT (“TWT”), 0.04 t/£[37].
For ERP’s mangrove projects in Madagascar (“ERPM”), 0.3 t/£.
At best (“ITP”), 1 t/£.
These values are lower than or similar to the bar of 1 t/$ FP arguably considers for the grants of its Climate Change Fund[38]. As a result, donating to it is expected to be more effective than supporting tree planting interventions. Moreover, the bar of 7.94 kt/$ suggested here is about 7 k times as high as the mean cost-effectiveness estimated for the idealised tree planting interventions.
In addition, my SoGive rating[39] for ERP is Not Recommended (Firm). Its mangrove projects in Madagascar, which accounted for 80 % of the trees planted by ERP in 2020[40], are expected to be significantly less cost-effective than the SoGive’s Gold Standard of 1 t/£. Moreover, such projects are arguably less effective than the marginal donation to ERP.
Other effects
The mean heat mortality benefits, ecosystem services benefits, effect on insects, and existential risk mitigation for “ERPM” projects, as well as guesses for the mean of some non-modelled effects, are summarised in the table below. The following sections contain further details.
The existential risk reduction is the major driver for the cost-effectiveness, which suggests the key conclusions are robust.
| Effect | Impact (life/ha) |
|---|---|
| Heat mortality benefits | 10^-1 |
| Ecosystem services benefits | 10^1 |
| Effect on insects | -10^7 |
| Existential risk reduction[41] | 10^(10^2) |
| Wild animals besides insects | -10^7 to 0 |
| Civilisational resilience | 0 to 10^0 |
| Tree planting workers | 0 to 10^-1 |
| Non-modelled cooling | -10^-1 to 10^-1 |
| Air pollution | -10^-1 to 10^-1 |
| Pandemics | -10^2 to 10^2 |
| Other longterm effects | -10^(10^2) to 10^(10^2) |
Wild animals besides insects
According to Brian Tomasik, “insects probably dominate in total sentience, unless you care drastically more about bigger animals”. Therefore it appears reasonable to assume that the strength of this effect is not larger than that of the effect on insects. Moreover, including other wild animals would tend to further decrease the cost-effectiveness, “suffering plausibly dominates happiness in nature” based on Tomasik 2015.
Civilisational resilience
Wood can be useful as a source of energy for civilisational recovery after a severe global catastrophe (for example, involving the death of 99.9 % of the global population). However, given the current global forest area of 4 Gha, the marginal value of additional woodland area is arguably negligible. Strategically creating woodland in certain places could be positive, but the location of these is unclear. Consequently, the strength of this effect is arguably not larger than that of the heat mortality benefits by more that one order of magnitude.
Tree planting workers
An upper bound for the benefits to the tree planting workers could be estimated by assuming that all the costs represent cash benefits to tree planting workers with an annual income of 236 £[42]. For “ERPM” projects, the mean cost is 1.69 k£/ha, which means the cash benefits to tree planting workers are limited to 7.14 (= 1.69 k / 236) doublings of consumption per hectare. Based on the equivalence between 2 doublings of consumption and 1 QALY considered by Open Philanthropy, and the utility of saving a life of 51 QALY/life, the maximum benefits amount to 0.07 life/ha (= 7.14 / 2 / 51). The order of magnitude of this is similar to that of the heat mortality benefits.
Non-modelled cooling
The effect on the adjusted net removal of CO2e emissions caused by tree planting changing surface temperatures and fluxes of latent and sensible heat (see, and tab “Cloud forcing”) was implicitly modelled as null. According to Lague 2016 (see Fig. 4), the magnitude of such cloud forcing effect is smaller than 5 W/m^2, which translates into CO2e emissions of 100 t/ha[43]. Multiplying this by the MCC results in 0.02 life/ha, which is smaller than the estimated heat mortality benefits.
Air pollution
Trees could improve air quality in cities (see this video from Vox), or increase air pollution if burned. On the other hand, wood could also replace other materials which cause GHG emissions (e.g. cement). Nevertheless, according to the WRI, “when burned, trees generate more CO2 emissions per unit of energy generated than fossil fuels”.
In any case, the order of magnitude of the effect on air pollution is arguably smaller or equal to that of the heat mortality benefits.
Pandemics
Deforestation can be a major driver for the emergence of zoonotic pandemics, but it is not clear whether afforestation/reforestation has the opposite effect (see Appendix D). As a result, the strength of this effect is arguably not larger than that of ecosystem service benefits by more than one order of magnitude.
Other longterm effects
The impact after reaching net zero emissions is implicitly modelled as null, but neglecting longterm effects does not appear to be critical[44]. The impact multipliers of tree planting do not seem promising (see Appendix E), and therefore it is not expected to lead to trajectory changes. As a consequence, the strength of this effect would hardly be larger than that of the existential risk reduction. Furthermore, since its sign is unclear, the expected cost-effectiveness should remain overwhelmingly positive.
Appendix A. UK tree planting charities
The register of charities of the Charity Commision for England and Wales (Charity Commision) was used to find tree planting charities in the UK. The following criteria were defined in the advanced search:
Income: no smaller than 0.5 M£.
Keywords: “tree”.
A brief overview of the 16 charities satisfying these criteria is presented in the following sections. The description of the charity activities was taken from the Charity Commision website on 14/11/2021, and the information about their regions of intervention and expenditure on charitable activities from their last annual reports as of 6/11/2021. The charities were divided into 3 categories (the first 2 respect my SoGive ratings):
More Information Needed.
Not Recommended (Firm).
Non-tree-planting charities.
My SoGive ratings[39] resulting from the analysis are presented in the table below.
| Charity | SoGive rating |
|---|---|
| Tree Aid | More Information Needed |
| International Tree Foundation | More Information Needed |
| The Woodland Trust | Not Recommended (Firm) |
| Heart of England Forest | Not Recommended (Firm) |
| Trees for Cities | Not Recommended (Firm) |
| Community Forest Trust | Not Recommended (Firm) |
| Tree Sisters | Not Recommended (Firm) |
| The Carbon Community | Not Recommended (Firm) |
| Japan Matsuri | Not Recommended (Firm) |
| Avon Needs Trees | Not Recommended (Firm) |
None of the 16 charities were on YouGov’s list for The Most Famous Charities & Organisations (Q3 2021). Information about the 16 charities is presented in the tab “Tree planting charities”.
More Information Needed
Tree Aid
Activities: “Tree Aid works with poor families, especially women, in the African drylands to unlock the potential of trees to reduce poverty & protect the environment. We provide education, training, policy & technical advice on tree based development initiatives that support poor communities build incomes, improve their management of & secure access to natural resources, & achieve nutritional security”.
Regions: Burkina Faso, Ethiopia, Ghana, Mali and Niger.
Expenditure on charitable activities: £4,517,771.
Assessment: No information about the expenditure on planting trees nor the removal of CO2e emissions. Although 2 M trees were planted in the year ended on 31 March 2020, assessing the cost-effectiveness of tree planting would not be enough to rate Tree Aid. The impact of the charity work on development does not seem negligible.
International Tree Foundation
Activities: “Protecting, promoting and planting trees to benefit the natural environment and the wellbeing and livelihoods of communities”.
Regions: UK, Kenya and another 8 african countries.
Expenditure on charitable activities: £508,569.
Assessment: No information about the removal of CO2e emissions. Some of the trees were planted in the tropics, and the ratio “expenditure on planting trees”/”number of trees planted” was 14% that of The Woodland Trust (£0.24/£3.88 = 14%). The impact of the charity work on development does not seem negligible.
Not Recommended (Firm)
The following charities are ranked as Not Recommended (Firm). Information about the removal of CO2e emissions could only be found for The Woodland Trust. However, the cost-effectiveness of UK tree planting projects is expected to be sufficiently low (0.07 t/£) for tree planting in the UK to be non-effective (see Discussion).
The Woodland Trust
Activities: “Our Objects are to conserve, restore & re-establish trees & in particular broad leaved trees, plants and all forms of wildlife & thereby to secure & enhance the enjoyment by the public of the natural environment. Our vision is a UK rich in native woods & trees for people & wildlife. Life’s better with trees strengthening the role of woods & trees in our landscapes & rekindling our love of them”.
Regions: UK.
Expenditure on charitable activities: £42,948,000.
Assessment: Initially, it was preliminarily concluded that the charity was roughly half as effective as the SoGive Gold Standard. However, the charity is ultimately rated as Not Recommended (Firm), as it is expected to be non-effective (0.04 t/£; see Discussion).
Heart of England Forest
Activities: “To establish, maintain and preserve, ideally a contiguous forest or a forest connected by corridor for the benefit of the public in the ancient borders of the Forest of Arden to the Vale of Evesham (The Heart of England). To engage in the education of the general public and for research relating to the promulgation of trees, woodlands,and wildlife and their habitats in the Heart of England”.
Regions: UK.
Expenditure on charitable activities: £1,217,000.
Assessment: Information about the removal of CO2e emissions could not be found.
Trees for Cities
Activities: “Trees for Cities is an independent charity working with local communities on tree planting projects. Our aim is to tackle global warming, create social cohesion and beautify our cities through tree planting, community, education and training initiatives in urban areas of greatest need”.
Regions: UK (86% of planted trees), Peru, Kenya, Nepal, Tanzania, Ethiopia and more.
Expenditure on charitable activities: £3,022,435.
Assessment: Information about the removal of CO2e emissions could not be found. Moreover, the mean cost to plant a tree is 2.5 times that of The Woodland Trust (the ratio between the cells D9 and D4 of tab “Tree planting charities costs” is 12.03/4.82 = 2.5).
Community Forest Trust
Activities: “The Community Forest Trust works to support the development of community forestry initiatives including City of Trees, The Mersey Forest, and White Rose Forest. Projects include mitigating climate change through the provision of high quality green infrastructure”.
Regions: UK.
Expenditure on charitable activities: £2,485,383.
Assessment: Information about the removal of CO2e emissions could not be found.
Tree Sisters
Activities: “Tree Sisters is building a global network of women to help crowd-fund tropical reforestation. We are an education based charity, developing methodologies, programs and our own organisational development out of the intelligence of living systems. We aim to inspire a wide diversity of women to take shared responsibility in normalizing conscious eco-system regeneration by funding tree planting monthly”.
Regions: Madagascar (27% of planted trees), Mozambique (17%), Nepal (7%), West Papua (6%), Cameroon (4%), Kenya (3%), Brazil (2%), India (2%) and more.
Expenditure on charitable activities: £1,550,079.
Assessment: Initially, there were signs that this charity could be effective. 57% of the trees consisted of dry deciduous forest and mangroves, and were planted via Eden Reforestation Projects, which was previously rated as Silver (Tentative) by SoGive. In addition, most of the trees were planted in the tropics, and the ratio “expenditure on planting trees”/”number of trees planted” was 6% that of The Woodland Trust (£0.24/£3.88 = 6.1%). However, the charity is ultimately rated as Not Recommended (Firm), as it is expected to be non-effective (0.3 t/£; see Discussion).
The Carbon Community
Activities: “The Carbon Community is a science led charity dedicated to carbon capture via creation of new community forests in the UK. Our tree planting will capture millions of tonnes of CO2 and make a material difference to global heating. Through our research with the world’s leading scientists we aim that every forest captures twice the amount of CO2 than traditional woodland”.
Regions: UK.
Expenditure on charitable activities: £19,267.
Assessment: Information about the removal of CO2e emissions could not be found.
Japan Matsuri
Activities: “To hold Japan Matsuri (Japan Festival) at Trafalgar Square, London in September or October. About 50,000 people attend the event. To engage in planting of cherry trees at locations throughout the UK”.
Regions: UK.
Expenditure on charitable activities: £277,992.
Assessment: Information about the removal of CO2e emissions could not be found.
Avon Needs Trees
Activities: “Avon Needs Trees is buying land in the Bristol-Avon Catchment Area to create new, permanent forest through reforesting and rewilding. Our objectives are to lock up carbon, improve biodiversity, create natural flood defences and to provide public amenity space for the local community. Our first purchase was 34 acres at Hazeland near Calne, Wiltshire, where we have planted over 10,000 trees”.
Regions: UK.
Expenditure on charitable activities: £5,087.
Assessment: Information about the removal of CO2e emissions could not be found.
Non-tree-planting charities
The following charities do not plant trees, or do not provide information about neither the expenditure on planting trees nor the number of trees planted.
JNF Charitable Trust
Activities: “JNF UK is an organisation raising funds for environmental and humanitarian causes in Israel”.
Regions: Israel.
Expenditure on charitable activities: £8,129,000.
Assessment: No trees were planted in 2020, and the expenditure on planting trees was only 0.3% of the expenditure on charitable activities in 2019. According to the 2020 annual report, “over 250 million trees have been planted by JNF member organisations on Israeli soil over the past century”. Now, however, the “focus is shifting more towards sustainably and environmentally social and economic challenges”.
The Tree Council
Activities: “The Tree Council works towards making trees matter to people; more trees, of the right kind, in the right places; better care for all trees of all ages and inspiring effective action for trees. An umbrella body and a forum for tackling issues relating to trees and woods, it promotes the improvement of the environment by the planting and conservation of trees and woods throughout the UK”.
Regions: UK.
Expenditure on charitable activities: £410,112.
Assessment: No trees were planted in the year ended on 31 March 2020, and the grants supporting “tree and hedge planting” (given to “schools, community groups and Good Gifts Guide”) only represented 8% of the expenditure on charitable activities.
Majlis Khuddamul Ahmadiyya (UK)
Activities: “To name just a few we have been collecting blood, planting trees nationwide, visiting several hospitals to meet and greet the sick, providing volunteers to the registered British charities, serving free meals to the needy, making award payments to deserving students and constantly carrying out educational and sports sessions for the members throughout the country.And a lot more!”.
Regions: UK.
Expenditure on charitable activities: £476,786.
Assessment: No trees were planted in the year ended on 31 October 2020.
Will Woodlands
Activities: “The Charity’s principal activity is the conservation, preservation and establishment of trees, plants and all forms of wildlife in the United Kingdom and to secure and enhance the enjoyment by the public of the natural environment of those territories. The Charity has continued its policy of acquiring land and establishing woodlands”.
Regions: UK.
Expenditure on charitable activities: £636,544.
Assessment: No information about the expenditure on planting trees nor the number of trees planted in the year ended on 31 March 2020. Moreover, tree planting does not seem to be significant: “a new hedge was planted in the north west corner for the Estate adjacent to the highway”.
The Albright Wood Norton Estate Charitable Trust
Activities: “The charity owns and operates woodland at Wood Norton Evesham. The objects of the charity are: To promote for the benefit of the public the conservation,protection and improvement of the physical and natural environment by re-establishing trees,plants and wildlife. The woodland is available to voluntary groups to benefit from the woodland environment with the permission of the Trustees”.
Regions: UK.
Expenditure on charitable activities: £26,049.
Assessment: No information about the expenditure on planting trees nor the number of trees planted during the period extending between 15 March 2019 and 30 June 2020.
Chase Africa (Community Health And Sustainable Environment)
Activities: “The relief of poverty and the protection of the environment overseas through training, awareness-raising and education, the provision of family planning and basic healthcare information and services, the planting and management of trees, the protection and restoration of forests, and the sustainable management of natural resources”.
Regions: Kenya and Uganda.
Expenditure on charitable activities: £384,377.
Assessment: No information about the expenditure on planting trees nor the number of trees planted in 2020. Moreover, the mission of the charity concerns global health and development: “to give women and men choice over the timing, number and spacing of their children, enable access to primary healthcare and support communities to protect their natural environment”.
Appendix B. Climate change organisations analysed by the Effective Altruism community
FP has awarded grants, in the context of The Climate Change Fund from October 2020 until November 2021, to the following organisations:
CATF.
Future Cleantech Architects (FCA).
TerraPraxis (TP).
The Economics of Energy Innovation and System Transition (EEIST) Consortium.
Carbon180.
Giving Green recommended the following organisations as of 9 March 2022 (but see the EA Forum post Why I’m concerned about Giving Green):
US policy change (“for individuals & foundations”):
Evergreen Collaborative.
CATF.
Carbon180.
Australia policy change (“for individuals & foundations”):
Beyond Zero Emissions.
Farmers for Climate Action.
Original Power.
Carbon offsets & removals (“for businesses’ net zero commitments”):
BURN.
Charm Industrial.
Climeworks.
Tradewater.
Other organisations mentioned in the EA Forum post Climate change donation recommendations:
Information Technology and Innovation Foundation (ITIF).
Coalition for Rainforest Nations (CfRN).
Eden Reforestation Projects.
Cool Earth Action.
A brief overview of the aforementioned organisations is presented below in the following order:
Organisations analysed by SoGive: from the highest to the lowest SoGive rating.
Other organisations: alphabetical.
Organisations analysed by SoGive
Clean Air Task Force
Activities: “CATF is a US-based NGO which conducts research and advocacy. Having originally been founded to reduce coal-power related air pollution via policy campaigning, its work is now much broader. CATF’s current foci include next-generation nuclear technology, zero carbon liquid fuels, and the reduction of super pollutant emissions”.
Regions: US, Europe and Africa.
Expenditure on charitable activities: $4,857,656 (FY2019).
Assessment: Rated as Gold (Tentative) under the SoGive methodology (more here) “FP (...) identified CATF as one of the two most promising charities in the climate change sphere [see FP climate change report and Climate Change Fund prospectus], and (...) estimated the cost-effectiveness of CATF’s future projects at $0.29 ($5.50 - $0.03) per tonne of CO2e emissions averted”.
Information Technology and Innovation Foundation
Activities: “ITIF is a think-tank focusing on issues at the intersection of technological innovation and public policy, one of which is clean energy. An organisation called Let’s Fund has been producing a series of write-ups of high risk high reward donation opportunities, one of which is donating to ITIF’s clean energy R&D work”.
Regions: US.
Expenditure on charitable activities: $3,626,153 (FY2020).
Assessment: Rated as Gold (Tentative) under the SoGive methodology (more here). “The Cost Effectiveness Analysis (CEA) produced by Let’s Fund [available here] estimated that, in their “realistic” scenario, $1 spent funding ITIF would result in an additional $28 of Clean energy R&D research spending. Their pessimistic and optimistic scenarios estimated $0.40 and $375 respectively”.
Coalition for Rainforest Nations
Activities: “CfRN promotes a system called REDD+, which is a mechanism by which developing countries are financially rewarded for reducing their rates of deforestation and forest degradation. The financial rewards are provided both by developed countries and also by the sale of carbon credits. CfRN’s work specifically has focused on persuading governments or corporates in developed countries to provide some of these financial rewards”.
Regions: Central America, Caribbean, South America, Africa, South Asia, East Asia and Southeast Asia (see this map).
Expenditure on charitable activities: $2,133,338 (FY2019).
Assessment: Rated as Silver (Tentative) under the SoGive methodology (more here). “FP listed CfRN as one of the two most cost-effective charities they had identified in the Climate Change sphere [see FP climate change report, but CfRN is not one of the 3 organisations mentioned in its Climate Change Fund prospectus]. This was due to a combination of their identification of deforestation as a particularly promising problem to work on, and due to the exceptional leverage they estimated CfRN had achieved in terms of encouraging governments to spend on REDD+. Founders’ Pledge estimated the cost to avert 1T of CO2e at $0.12 ($0.02 to $0.72)”.
Eden Reforestation Projects
Activities: “Eden Projects hire locals in Nepal, Madagascar, Haiti, Indonesia, Mozambique and Kenya to re-plant local forests, many but not all of which are mangroves”.
Regions: Madagascar (83% of planted trees), Indonesia (6%), Mozambique (4%), Ethiopia (3%), Nepal (1%), Kenya (1%), Haiti (0.4%), Nicaragua (0.03%) and Honduras (0.01%), according to the 2020 annual report.
Expenditure on charitable activities: $9,391,332 (2020 financial statements).
Assessment: Rated as Silver (Tentative) under the SoGive methodology (more here). “ImpactMatters (...) found Eden projects to be by far the most cost-effective tree planting organisation when measured in terms of carbon sequestered per dollar”, with a cost per tonne of CO2eq sequestered of $0.36. However, the charity is ultimately rated as Not Recommended (Firm), as it is expected to be non-effective (0.3 t/£; see Discussion).
Cool Earth Action
Activities: “Cool Earth works with communities of people living in rainforest regions to develop sustainable agreements. The specific nature of the agreements varies from community to community but the purpose is to improve their lives to the point where they can opt to resist pressure from logging companies to sell their land”.
Regions: Peru, Cameroon, Democratic Republic of the Congo, Mozambique, Cambodia and Papua New Guinea.
Expenditure on charitable activities: £1,741,087 (2020/2021 annual report).
Assessment: Rated as More Information Needed under the SoGive methodology (more here) for the following reasons. Concerns that logging was being displaced rather than prevented. The logging prevention effect might fade once a community’s relationship with Cool Earth ends. The impact likely varied by project/geography. GWWC estimated that “Cool Earth reduces greenhouse gas emissions through direct protection of forests at a rate of $1.34 per tonne of CO2-equivalent”. Despite the uncertainty, GWWC estimated that “this figure will be no higher than $1.87/tCO2eq and no lower than $0.65/tCO2eq”.
Other organisations
Beyond Zero Emissions
Activities: “Beyond Zero Emissions (BZE) is an independent think-tank whose research shows the technological and financial feasibility of achieving zero emissions in Australia in the short-term, while maintaining employment for people in regional communities that have traditionally depended on fossil fuels. Its research identifies net zero pathways to unlock economic opportunities for emissions-intensive industries and regional communities” (Giving Green analysis).
Regions: Australia (Giving Green analysis).
Expenditure on charitable activities: $1,022,364 (BZE 2019/2020 annual report).
Assessment: Giving Green recommends BZE to companies for the following reasons: “BZE’s board, staff and volunteers have strong connections with the Government, industry and regional communities most affected by the transition”; “BZE’s research, stakeholder engagement and advocacy is helping to reframe the narrative about decarbonisation in Australia: a key barrier to progress on climate policy”; “BZE’s ideas are already gaining traction with emissions-intensive industries and regional communities traditionally dependent on fossil fuels”; “Federal and state governments on both sides of politics are already implementing BZE’s policy and investment recommendations”; “Additional marginal investment could help BZE develop more detailed plans to achieve deep emissions cuts this decade, which is essential to avoid catastrophic temperature rise” (Giving Green analysis).
Burn
Activities: “Burn sells high efficiency charcoal and wood burning cookstoves, which reduce emissions by reducing fuel use compared to traditional cooking stoves. While they sell carbon credits, there is no direct link between buying such credits and removing emissions from the atmosphere; instead, they use “additional revenue provided by the emissions market [to] conduct additional marketing and R+D””.
Regions: Kenya.
Expenditure on charitable activities: Not applicable since Burn is not a charity.
Assessment: Rated as Bronze (Tentative) under the SoGive methodology (more here). Giving Green estimates that “$3.61 USD in offsets avoids 1 ton of CO2e” (they “do not believe that an offset purchase can be viably linked to removing a specific amount of CO2”, but they “believe that purchasing an offset enables BURN to distribute more stoves and directionally leads to fewer emissions”).
Carbon180
Activities: “Carbon180 is an insider policy advocacy organization that focuses on accelerating the development of carbon removal technologies and practices, which would remove carbon dioxide from the atmosphere and lock it away for at least hundreds of years. Its four initiatives include (1) building and enacting federal policy to scale up carbon removal solutions, (2) accelerating the adoption of soil carbon sequestration practices, (3) encouraging community engagement between carbon removal researchers, and (4) stimulating innovation. Its tactics include research, policy advocacy, and ecosystem building. Carbon180 also centers equity and justice in its work to ensure that carbon removal can be scaled up in a way that is sustainable with equitably-distributed benefits” (Giving Green analysis).
Regions: US (Giving Green analysis).
Expenditure on charitable activities: $1,180,450 (FY2019).
Assessment: “We [Giving Green] estimated that Carbon180’s work on federal policy can remove CO2 from the atmosphere at a cost of $0.66 per metric ton (in expectation), which compares favorably to other high-performing organizations that we have analyzed. Because our CEA model only includes short term effects of Carbon180’s work on federal legislation, it seems likely that we may even have underestimated Carbon180’s impact and cost-effectiveness. Our results should be viewed as rough, indicative estimates given the uncertainty in our different model inputs” (Giving Green analysis).
Charm Industrial
Activities: “Charm Industrial is a US-based company that converts agriculture residues into bio-oil through a process known as fast pyrolysis. Charm collects agriculture residues (such as corn stover and wheat straw) from farmers in Kansas and feeds them into a pyrolysis unit, heating them at very high temperatures (> 500° C) in the absence of oxygen for a matter of seconds, creating bio-oil. Instead of slowly decomposing and releasing greenhouse gases, the bio-oil locks up the carbon from the original biomass and is injected into EPA-regulated wells, sinking to the bottom of the geological formation where it remains for thousands or millions of years” (Giving Green analysis).
Regions: US (Giving Green analysis).
Expenditure on charitable activities: Not applicable since Charm Industrial is not a charity.
Assessment: “Charm has developed and is beginning to scale up a process that can measurably and permanently keep CO2 from agriculture and forest residues from entering the atmosphere. It avoids some of the permanence and additionality concerns associated with biochar providers, has made the purchase process streamlined for offset buyers, and is currently working on a public dashboard to further improve transparency. Their process is measurable, additional, and permanent. The only drawbacks are its current price, and the need for more data on the solidification of bio-oil in injection wells. We have determined that supporting Charm Industrial at this juncture can have a significant effect on scaling up their climate mitigation solution by reducing costs and helping improve their technology. More broadly, this can help shape a relatively new climate mitigation pathway” (Giving Green analysis).
Climeworks
Activities: “Climeworks is a Switzerland-based company that does Direct Air Capture (DAC) and either sequesters collected CO2 underground or sells it for commercial purposes including to a Swiss greenhouse and to a bottler for Coca-Cola Switzerland. Climeworks has a policy against working with fossil fuel production companies, which are a common partner for DAC companies as certain types of fossil fuel extraction can be made more efficient and productive by injecting CO2 underground in a process called Enhanced Oil Recovery (EOR). The climate impact of these projects is debated, which is why Climeworks’ focus on other types of projects is compelling” (Giving Green analysis).
Regions: Switzerland and Iceland (Giving Green analysis).
Expenditure on charitable activities: $81,088,601 (FY2019).
Assessment: Giving Green recommends companies to purchase CDR (carbon dioxide removal) credits from Climeworks for the following reasons: they “are very sure that their project is permanently removing CO2 from the atmosphere”; they “believe that each dollar invested in their work will additionally remove carbon”; they “are convinced that investments in their work right now are entirely being used to remove carbon as opposed to simply maximizing profit”; they “see long-term potential in their DACS [direct air capture and storage] technology”; “credits purchased now contribute to potentially drastic long-term cost reductions for Climeworks”. However, “the main drawback of Climeworks is that it is currently very expensive (around $1000/ton) to remove carbon relative to other options. This may be justified by the fact that supporting Climeworks will hopefully go toward reducing the cost of their frontier carbon-removal technology” (Giving Green analysis).
Evergreen Collaborative
Activities: “Evergreen Collaborative is a policy advocacy group that was founded by former staffers of Washington State Governor Jay Inslee’s 2020 presidential campaign. It designs and advocates for policy proposals while working alongside like-minded environmental organizations such as the Sunrise Movement, Rewiring America, RMI, and the Big Greens (e.g., large and well-funded environmental groups such as the Environmental Defense Fund and Sierra Club). By working on policy advocacy and acting as connective tissue between other environmental organizations, Evergreen Collaborative seeks to influence Congress, the Executive Branch, and federal agencies” (Giving Green analysis).
Regions: US (Giving Green analysis).
Expenditure on charitable activities: Not found.
Assessment: “According to a cost-effectiveness analysis model that we developed, Evergreen Collaborative’s work on federal legislation is highly cost-effective in reducing greenhouse gas emissions. Based on how much CO2-equivalent (CO2e) it could potentially reduce between 2022 and 2030, Evergreen is predicted (in expectation) to reduce emissions at a cost of about $0.54 per metric ton of CO2e under our Realistic scenario. These results should be viewed as rough, indicative estimates given the uncertainty in our different model inputs” (Giving Green analysis).
Farmers for Climate Action
Activities: “Farmers for Climate Action (FCA) is a movement of farmers, agricultural leaders and rural Australians working to ensure they are part of the solution to climate change. It is the only farmer-led organisation focused specifically on adapting to and mitigating climate change in the Australian agricultural sector[1]. FCA is a registered charity[2], and now represents more than 6,500 farmers and has over 45,000 total supporters[3]. Its work includes farmer education and training, industry and government advocacy, research and partnerships, and building farmer networks and communities of practice” (Giving Green analysis).
Regions: Australia (Giving Green analysis).
Expenditure on charitable activities: “FCA’s operating budget for the 2021-2022 financial year is $1.7 million” (Giving Green analysis).
Assessment: Giving Green recommends FCA to companies for the following reasons: “FCA’s board, staff and members have strong connections with the Nationals Party, industry leaders and regional communities most affected by climate change”; “FCA’s research and advocacy is helping to reframe the narrative about agriculture and climate change at the industry and government level”; “FCA’s education and training on climate smart agriculture is influencing action by farmers”; “Federal and state governments on both sides of politics are already implementing FCA’s policy recommendations”; “Additional marginal investment could help FCA mobilise farmers as a key voter group to shift the Nationals’ climate policies” (Giving Green analysis).
Original Power
Activities: “Original Power (OP) is working to ensure Australia’s First Nations communities benefit from the renewables boom. It uses a collective-action model to resource and support Aboriginal and Torres Strait Islander communities to self-determine what happens on their country. This work is critical because, as Australia’s traditional owners, First Nations people have unique rights over 50 per cent of Australia’s land, making them critical stakeholders in the transition from a fossil fuel-based economy to one powered by clean, renewable energy. OP supports communities in their efforts to protect cultural heritage, challenge fossil fuel developments (if this is what communities decide), and create a just transition to renewables. OP’s work can support the rapid roll out of large-scale renewables as an alternative to fossil fuel projects, in turn reducing Australia’s emissions” (Giving Green analysis).
Regions: Australia (Giving Green analysis).
Expenditure on charitable activities: “FCA’s operating budget for the 2021-2022 financial year is $1.7 million” (Giving Green analysis).
Assessment: Giving Green recommends OP to companies for the following reasons: “Original Power is helping to reduce Australia’s emissions by paving the way for renewables as a superior alternative to fossil fuels in providing jobs, economic opportunities and energy security for First Nations communities”; “Original Power is supporting First Nations communities to exercise their rights to self-determine what happens on their country”; “Original Power has a strong team with connections to First Nations communities and clean energy industry leaders and policy makers”; “Additional marginal investment could help Original Power ensure First Nations people benefit from the renewable energy revolution, drive community-owned clean energy projects and secure equitable arrangements for large-scale renewable projects on their lands” (Giving Green analysis).
TerraPraxis
Activities: “TerraPraxis is a small and brand new organisation with a special focus on speeding up progress on crucial but neglected clean energy technologies”. “TerraPraxis researches, and advocates for the expansion of, energy innovation, with a particular focus on advanced nuclear power—a form of nuclear power that is much safer and potentially cheaper than existing nuclear plants. They combine high-quality technical analysis with advocacy in energy policy circles in international fora (such as the Clean Energy Ministerial and the IEA) and in Europe and North America” (FP report).
Regions: UK, continental Europe and North America (FP report).
Expenditure on charitable activities: Not applicable since Tradewater is not a charity.
Assessment: FP recommends TerraPraxis (see FP Climate Change Fund prospectus) for the following reasons: “focus on energy innovation, especially in advanced nuclear power, a highly neglected but important energy technology”; “led by Kirsty Gogan, one of the most articulate advocates for nuclear power”; “funding is highly likely to be additional, as they find it hard to fundraise and would otherwise fund their non-profit work via paid consultancy work”; “providing them with seed funding will both produce direct impact and is a great learning opportunity to see what they can do with additional philanthropic funds” (FP report).
Tradewater
Activities: “Tradewater is an organization that finds and destroys ozone-depleting substances (ODS) and sells offsets to fund this process. In fact, it is the only such organization that we found selling these offsets to the public. Tradewater’s primary mission is to find and destroy refrigerants and other gases with especially high warming potential. They work worldwide to find these gases, purchase them, and subsequently destroy them. Tradewater’s revenue comes completely from the carbon offset market. It sells offsets directly to consumers on its website. Tradewater also sells larger batches of offsets directly and works with offset brokers as needed” (Giving Green analysis).
Regions: US, Latin America and Ghana.
Expenditure on charitable activities: Not applicable since Tradewater is not a charity.
Assessment: Overall Giving Green believes that “the listed cost of offsets sold by Tradewater ($15/ton) is a good representation of the true cost of carbon removal”. Their “main concern is that Tradewater is a for-profit company, and therefore could claim offset revenue as profit instead of reinvesting it in further ODS [ozone-depleting substances] removal projects”. Consequently, Giving Green “urges Tradewater to make their financials public in order to reassure offset buyers” (Giving Green analysis).
Appendix C. WCC
The WCC is a UK standard which provides reassurance about the carbon sequestration of tree planting projects. The procedure it follows to determine the net removal of CO2e emissions is described in section 3 of version 2.1 of the code (the updated online version of the Standard & Guidance is here). It is also briefly outlined in the following sections.
Carbon baseline
The WCC establishes a baseline for the carbon sequestration over time (counterfactual), against which the impact of the project can be measured. The baseline is based on the continuation of the land use prior to the forestation project. It includes four carbon pools:
Tree biomass above and below ground biomass.
Litter and deadwood.
Non-tree biomass above and below ground biomass.
Soil.
Carbon leakage
The WCC defines leakage as “land use change/intensification outside the project boundary but within the UK”. If it is proposed by the land manager, the project shall carry out an assessment to determine whether it will result in additional CO2e emissions. Otherwise, it is assumed to be null.
Underreporting of leakage would likely result in overestimating the net removal of emissions.
Project carbon sequestration
The net removal of emissions is calculated with the WCC Carbon Calculation Spreadsheet, whose version 2.4 is available here. The calculator includes the following:
Emissions from establishment activities, ongoing management and clearfell.
Emissions from soil disturbance.
Sequestration in tree biomass, litter and deadwood (and in a limited number of scenarios, soil).
Net carbon sequestration
The net removal of CO2e emissions (or “net carbon sequestration”) accounts for the difference between the factual and counterfactual carbon sequestration, as well as leakage. At Year 5, the net removal of CO2e emissions is based on the projected carbon sequestration. From Year 15 onwards, it is based on field survey measurements.
Appendix D. Effect of tree planting on the emergence of pandemics
According to IPBES 2020, “land-use change is a globally significant driver of pandemics and caused the emergence of more than 30% of new diseases reported since 1960”. However, to the extent the driver for this is the contact between humans and farmed animals with wildlife, the case for deforestation causing pandemics seems stronger than the case for forestation preventing pandemics:
Deforestation is often caused by urbanisation and expansion of animal farming. Consequently, preventing it tends to minimise contact with wildlife.
On the other hand, forestation would only prevent contact with wildlife if it prevented deforestation (driven by the expansion of animal farming and urbanisation). However, no evidence was found for this.
In addition, forestation could arguably contribute to causing pandemics, if the degree of contact with wildlife required to establish forests is not negligible.
A quick google search seemingly supports this. ”Deforestation could cause pandemics. but that does necessarily imply that forestation prevents pandemics. I only found evidence supporting the former. In fact, forestation could increase the risk of pandemics, as it should require more contact with wildlife.
Appendix E. Tree planting impact multipliers
In order to establish a prior view, tree planting was assessed based on the impact multipliers mentioned in the most recent climate report published by FP. These are “reasons to expect that a given funding allocation will have an above-average impact”, and do not replace the analysis of specific opportunities. “Rather, they serve as an orientation for where to search and as a prior of how surprised we should be to find high-impact opportunities in different parts of the climate funding space” (see section “Impact Multipliers” of Founder Pledge’s report for more details).
Tree planting is not expected to be effective based on FP’s impact multipliers, which are described in the following sections.
Neglectedness
Tree planting is not neglected, and therefore it does not benefit from this impact multiplier. For example, according to Google Trends, the average interest in tree planting over the last 5 years is much larger than the interest in technologies promoted by organisations which have received grants from FP. This is demonstrated in the table below, where an interest value of x represents x% of the maximum interest in tree planting during the last 5 years.
| Term | Organisation | Interest over the last 5 years |
|---|---|---|
| Tree planting | Not defined | 52 |
| Small modular reactors | TerraPraxis | 1 |
| Synthetic fuels | TerraPraxis | 1 |
| Carbon capture and storage | Carbon180 | 3 |
| Direct air capture | Carbon180 | 1 |
Risk neutrality
“A typical behavioral bias in climate giving is “risk aversion”, preferring a (seemingly) certain emissions reductions—say, from a carbon offset—to a much higher expected reduction through a more uncertain means, say, advocacy-focused philanthropy”.
Common tree planting misconceptions lead to it being perceived as a low risk intervention to decrease global warming (despite the risks being often underestimated). As a consequence, it does not benefit from this impact multiplier.
Patience
“A lot of climate philanthropy is driven by a desire to quickly reduce emissions, e.g. by marginally accelerating transitions already in progress, such as coal phase-outs in Europe or a quicker adoption of electric vehicles (when this transition is already poised to happen)”.
One of the common tree planting misconceptions is that, “once the trees are planted, the job is done”. This suggests tree planting tends to be perceived as an intervention whose benefits are produced in a short timescale (despite tree growing being a long process). As a result, it does not benefit from this impact multiplier.
Advocacy
“We think that funding organizations that are trying to influence how societal resources (incl. attention) are allocated provides a strong impact multiplier compared to funding more direct interventions”.
Tree planting is an intervention with direct effects on global warming, and therefore it does not benefit from this impact multiplier.
Trajectory changes
“We think it is likely that engaging around trajectory changes—in situations where decisions will have consequences for long time-spans due to self-reinforcing dynamics and/or stickiness provide another impact multiplier, although we think this requires pairing with assessments of neglectedness”.
“Dynamics which have this trajectory-shaping character”:
“Virtuous cycles around technological change: Cost reductions, increased demand, further cost reductions, as seen with solar, wind, electric cars, and batteries”.
“Vicious cycles around carbon lock-in: Infrastructure and regulatory choices that favour particular technologies, or capital-intensive investments with long lifetimes, etc.”.
“Virtuous cycles around social movements and shifting social norms: E.g. the emergence of a national-level climate movement, once that movement reaches national name recognition it can profit from lots of attention and funding”.
“Spreading of policy ideas: The adoption of policy ideas across the world based on one or a few leading examples, as observed with carbon taxes, emissions trading systems, binding climate laws, net-zero commitments etc”.
Tree planting does not seem to robustly benefit from any of these dynamics.
Catalytic growth
“We believe that “overhead”, unrestricted funding or funding directly targeted at improving the workings of an organization—such as operations, fundraising, but also communications and strategy—is often underprovided for small-to-medium-sized organizations”.
The largest organisation analysed (income of 60 M£) will most likely not benefit from this impact multiplier, but the smallest ones (income lower than 1 M£) could.
Global diffusion of technological change
“Outsized changes in the global energy system with emissions consequences far beyond can be induced” via:
“Research & Development & Demonstration (RD&D), i.e. early-stage innovation support (“technology push”)”.
“Demand-policies for early deployment, such as deployment subsidies, public procurement, standards requiring new technologies (“demand-pull”)”.
Tree planting is an intervention with direct effects on global warming, and therefore it does not benefit from this impact multiplier.
Policy additionality
“Being additional is not only about being additional to other funders but also about being additional to existing policy targets”.
Tree planting in the UK, “where there is a binding climate law”, does not seem to benefit from this multiplier. The UK government has the target of planting trees at a rate of 7.5 kha/year by 2024-2025.
Tree planting in low-income countries could be more additional. However, the president of Madagascar (where Eden Reforestation Projects, the international tree planting organisation analysed here, plants 80% of its trees) aims to plant trees at a rate of 40 kha/year between 2019 and 2024, although “conservation experts point to shortcomings in the plan”.
- ^
1 t stands for 10^3 kg of CO2e.
- ^
The calculation of the best mean of 1 t/£ can be illustrated by considering a net removal of emissions of 39.7 t/ha/year for 50.7 year, and a cost of 1.62 k£/ha.
- ^
Johannes is the first author of FP’s latest climate report.
- ^
Fitting a lognormal distribution to Johannes’ guesses leads to a mean cost-effectiveness for CCF of 2 kt/$.
- ^
However, ImpactMatters has published this methodology to assess tree planting, and Giving Green this report on forestry interventions.
- ^
Note risk aversion for altruistic benefits makes less sense than for personal benefits. For example, donating 1 k£ to an organisation whose annual spending is 1 M£ is about 10 times as impactful as donating 100 £, but earning 1 M£/year does not lead to 10 as much happiness as earning 100 k£/year.
- ^
- ^
The methodology used by the WCC to assess the carbon sequestration depends on the size of the forestation area (see section 3 of the WCC Standard and Guidance).
- ^
By default, the tabs mentioned hereafter refer to this same Sheet.
- ^
1 life/ha means that planting 1 ha of trees saves 1 life.
- ^
The equivalence between emissions and radiative forcing is discussed in Bright 2020.
- ^
The formula for HMBDF assumes the adjusted net removal of CO2e emissions is the same for all the years before DER, and that the MCC grows exponentially (and discretely). The HMBDF was set to DER, as the MCCR was assumed to be null (see Mortality cost of carbon annual variation rate).
- ^
Sum of the 1st DER terms of a geometric progression of common ratio 1 + MCCR.
- ^
The formula for ABDF assumes the GDPC grows exponentially (and discretely).
- ^
Sum of the 1st DER terms of a geometric progression of common ratio 1/(1 + GDPCR).
- ^
Here, the existential risk mitigation is assumed to be directly proportional to the decrease in temperature caused by the intervention, which in turn is directly proportional to the total net removal of emissions (as temperature increases roughly linearly with cumulative emissions). Note the timing of the emissions removal does not matter if the existential risk mitigation only depends on the decrease in the maximum global temperature caused by the intervention, which arguably determines the risk of crossing an existential tipping point.
- ^
The assumption of current climate policies is consistent with the Metaculus’ community prediction of 2.55 ºC (as of 2 October 2022) for “how much greater (in ˚C) will the average global temperature in 2100 be than the average global temperature in 1880”. According to the Climate Action Tracker (CAT), current climate policies are predicted to result in a global warming between 2.5 ºC and 2.9 ºC (interval which contains 2.55 ºC). The modelling of the cumulative GHG emissions would ideally consider more scenarios. The exclusion of the emissions after 2100 until net zero, after which the existential risk due to climate change is arguably negligible, tends to overestimate the existential risk mitigation.
- ^
“Roughly” because the SoGive Gold Standard benchmarks for saving lives and removing CO2e are only “roughly” consistent, since the calculation of the latter was rounded.
- ^
The mean moral weight for uniform, normal and logistic distributions is 30 % higher (see results for “fruit flies” here).
- ^
The words “mean” and “standard deviation” are often in italic because:
- For the lognormal distributions, they respect the mean and standard deviation of the logarithm of the variables (which differs from the mean and standard deviation of the variables).
- For the truncated normal distribution, they respect the mean and standard deviation of the respective non-truncated normal distribution.
- ^
Formulas for the weighted mean and standard deviation are available here.
- ^
The additional benefits were assumed to continue until the end of the project duration.
- ^
Considering established mangroves tends to overestimate the cost-effectiveness, as rehabilitating mangrove forests were determined in Cameron 2019 to remove less CO2e (Table 1, first 5 rows).
- ^
Madagascar was selected as the country because ERP plants 80% of its trees there (see tab “Eden Reforestation Projects”). Note that Benson 2017 estimated a value of 455 t/ha for Madagascar closed-canopy mangroves, which is 20% lower than that of Jones 2014. In addition, assuming closed-canopy mangroves (“tall, mature stands; canopy >60% closed”) overestimates the cost-effectiveness of “ERPM” projects if ERP plants open-canopy mangroves in Madagascar. According to the data of Jones 2014 (see J25:J27 of tab “Carbon stocks”), the carbon stock density for open-canopy I mangroves (“young, short-medium trees; canopy 30%–60% closed; influenced by background soil/mud”) is 40% lower than that of closed-canopy, and for open-canopy II (“stunted short trees, very sparse; canopy ≥10% closed; dominated by background soil/mud”) is 9% lower.
- ^
According to Bright 2020, the respective method was introduced in Betts 2000, “to which almost all CO2-eq. literature for 𝛥𝛼 [albedo change] may be traced”, and whose “research objective was to compare an albedo contrast between a fully grown forest and a cropland (i.e, 𝛥𝛼)”.
- ^
The lower bound of the 2020 MCC estimated in Bressler 2021 is negative, thus the minimum and maximum were estimated from the lower and upper bounds.
- ^
Based on the formulas for the mean and mode available in Wikipedia, it could be concluded that:
- “Alpha” = (1 − 2 * “mode”) / (1 - “mode” / “mean”).
- “Beta” = (1 / “mean” − 1) * “alpha”.
- ^
See section “Contributing to the buffer” of this page. The WCC methodology considers the following risks (see Table 1 of this document): legal/social; related to project management; financial; related to natural disturbances (fires, storms, pests and diseases, droughts, amongst others). The selected mean is a good estimate of that of the “small”, “standard”, “WCC” and “TWT” projects, but is likely to be an underestimate of that of the “ERPM” projects. Although it is mentioned here and here that the survival rate of trees planted by ERP is over 80% (which would be compatible with a mean tree planting intervention risk lower than 20 %), in the literature, mangrove restoration projects report a much lower project success and tree survival rate. For example:
- Kodikara 2017 “investigated the effectiveness of mangrove planting initiatives in Sri Lanka”, and concluded that:
-- “Nine out of 23 project sites (i.e. 36⁄67 planting efforts) showed no surviving plants”.
-- “The level of survival of the restoration project sites ranged from 0 to 78% and only three sites, that is, Kalpitiya, Pambala, and Negombo, showed a level of survival higher than 50%”.
- Wodehouse 2019 “examines village-level rehabilitation planting carried out in 13 villages (119 rehabilitation attempts at 74 sites) across two countries in southeast Asia”, and concluded that:
-- “Mean propagule survival across all rehabilitation attempts was 20% with a median of 10%”.
-- “Sixty six percent of attempts had a survival rate of less than 20%”.
- ^
In such versions, the risk factor ranged from 15 % to 40 % (see section “Contributing to the buffer” of this page). The difference between the upper bound and the mean (20 % = 40 % − 20 %) being larger than the difference between the mean and the lower bound (5 % = 20 % − 15 %) suggests a right-skewed distribution. To preserve this property, the mode was set to the lower bound.
- ^
Some of the principles are outlined in this article from Vox.
- ^
See Table 6.1 of The Precipice (also available here).
- ^
ESVD considers 23 types of services: food (1), water (2), raw materials (3), genetic resources (4), medicinal resources (5), ornamental resources (6), air quality regulation (excluding mortality impacts) (7), climate regulation (excluding mortality impacts) (8), moderation of extreme events (9), regulation of water flows (10), waste treatment (11), erosion prevention (12), maintenance of soil fertility (13), pollination (14), biological control (15), maintenance of life cycles (16), maintenance of genetic diversity (17), aesthetic information (18), opportunities for recreation and tourism (19), inspiration for culture, art and design (20), spiritual experience (21), information for cognitive development (22), and existence, bequest values (23).
- ^
Indonesia and Mozambique are the countries besides Madagascar where ERP planted the most trees. In 2020, ERP planted 83% of the trees in Madagascar, 6% in Indonesia, and 4% in Mozambique (see tab “Eden Reforestation Projects”).
- ^
- ^
It has been noted that the implied mean cost is 2 times the one estimated from the product between the cost per tree and tree planting density mentioned by the CEO of ERP in this interview.
- ^
The running time per simulation was about 20 s.
- ^
In SoGive’s shallow analysis of TWT, the cost-benefit ratio was conservatively estimated to be 2.14 £/t, which corresponds to a cooling cost-effectiveness of 0.467 t/£ (= 1⁄2.14). This is 12.6 (=0.467/0.0372) times the value estimated in the present analysis.
- ^
1 t/$ is similar to SoGive’s bar of 1 £/t.
- ^
By “my SoGive rating”, I (Vasco) mean the ratings are not necessarily endorsed by SoGive.
- ^
See C3 of tab “Eden Reforestation Projects”.
- ^
The order of magnitude of the mean existential risk reduction is not stable, but its decimal logarithm is of the order of 10^2.
- ^
Determined from the annual income of GiveDirectly’s recipients of 286 $ * 0.827 £/$.
- ^
The conversion factor from radiative forcing to CO2e emissions is in cell C22 of tab “Constants”.
- ^
For example, with respect to wild animal welfare in the far future, Saulius Simcikas writes that:
- “Digital minds can be much more efficient, thrive in environments where biological beings can’t, use many more resources, etc. Hence, I believe that they dominate in terms of importance in the far future. I think that spreading biological life is among the most important far-future considerations only if one believes with a very high credence that digital minds can’t or won’t be sentient” (for more, see here).
Thanks for this, Vasco, Hanzhang, Melissa!
A couple of thoughts:
It seems to me that on reasonable quantifications of the ways in which direct tree planting efforts in the UK do not optimize for climate impact (utter non-neglecteness, lack of advocacy, trajectory change, low policy additionality, the factors you mention in the appendix) one would have a prior where tree planting is several orders of magnitude less cost-effective than strategies that seek to optimize for impact. As such, I find a posterior of a 1-order-of-magnitude difference within reach quite surprising.
While I am generally quite pessimistic on forestry interventions because of the utter non-neglectedness and the difficulty to get to credible additionality and permanence, it seems direct tree planting in the UK is kind of close to the worst thing one can do from a climate angle. So for donors that cannot be moved from tree planting it could be interesting to see what the best charities in this space might look like, e.g. advocacy to improve REDD+ or peatland protection etc.
Re Google Trends for neglectedness: A great datasource for climate philanthropy are the Climate Works reports—those show clearly not only that forests are very well funded compared to other areas, but also that funding is growing strongly (it has been an early focus of Bezos Earth Fund, the largest climate philanthropist). In addition, IIRC, conservation philanthropy more broadly, of which a significant share focuses on forests, is several X larger than climate philanthropy.
Hi Johannes,
Thanks for sharing your thoughts!
1 t/£ is estimated to be about 1 (= log10(1/0.0722)) order of magnitude (OOM) higher than the cost-effectiveness of tree planting in the UK. However, the difference to the projects funded by CCF may be quite larger. Fitting a lognormal distribution with 2.5th and 97.5th percentiles equal to the lower and upper bound of the 95 % confidence interval you guessed (with the disclaimer that it should not be intended a resilient estimate) leads to a mean of 2.34 kt/$. This is about 5 (= log10(2.34 k / 0.0722)) orders of magnitude higher than the cost-effectiveness of tree planting for the UK.
Thanks! Yeah, a five OOM difference sounds more like what I would expect, indeed.
Well, this is very interesting! Thank you for posting it.
My first impression, reading it, is that I wonder where most of your hard numbers come from. While the references to organizations or research sources are scattered, did I miss a summary of what organizations provided numbers or data on topics like:
projected anthropogenic GHG’s
total carbon uptake by trees (for example, respiration increases, tree types)
threats to aforestation vs reforestation efforts
policies of native vs non-native reforestation
mitigation of climate impacts on forests
ecosystem management to avoid increasing niche species losses
I also wonder how estimates were done of effects on trees of climate impacts like:
drought
tree diseases
fires
invasive pests
development priorities
increasing tree respiration
These effects impact reforestation and aforestation efforts.
For me it’s a warning that your analysis must be broad but shallow because an analysis of, for example, ecosystem services will necessarily be very complicated with respect to aforestation vs reforestation and choice of plant species and impacts on insect species, and related problems with tree-destroying pests. That research is still ongoing. In your write-up here, I don’t see direct references to IPCC data or projections about value and impacts of land management. Is that represented here, earlier in your chain of research, or in your sources? They’re one of several options for research summaries in this area.
I did take a look at the www.evsd.info database, briefly, and it reads to me as economics thinking applied to ecology, and not very well. For example, their homepage lists one of the primary “ecosystem services” of mangroves as tourism. That might be true subjectively and in the moment, where there is some choice about whether the mangrove forests exist and how important tourism should be to an economy, but it fails to recognize mangrove forest role in contexts where earnings from tourism are irrelevant. Fundamentally, this looks like more dissonance between what motivates policy action and what deserves policy action.
With nature, you can’t typically change your mind. Larger systems can be hard to restore. Conditions won’t allow growth of species in the right quantities or varieties or at all. Animal and plant species, once extinct, tend to stay gone. You can’t remove toxins from soil or water or air in polluted areas. You can’t control the temperature or the weather,. You can’t grow food where you once could, or eat it when you once could, because its no longer safe. Without the economic incentive, and narrow interests define those, survival value of ecosystems are mismanaged. Once the land is depleted, it typically goes ignored. Until then, it’s dollar value is relative to current economic activity that is falsely projected to continue into the future, something like tourism. What if tourism has to stop to create enough benefit to save some niche ecosystem feature?
On my reading, this idea of ecosystem services given dollar amounts treats the value of ecosystems as one of aesthetics if the ecosystem has not otherwise been exploited for profit in some easily recognizable business process (and plausibly a non-reversible one that degrades the ecosystem). However, the real value of these systems is not aesthetic. They are survival-supporting machines that we don’t understand well and interact with poorly. If we thought of them like that, the real question here, of how to protect whole ecosystems from our misdoings, would be the focus.
In your existential risk mitigation footnote, you mention that you make existential risk directly proportional to GAST reduction, but it is not. In particular, some temperature effects can happen with small additional increases in temperature, thus why they are called tipping points, because their effects proceed with just a small additional forcing. Furthermore, some tipping element effects involve a hysteresis relationship, so that just reversing the GAST increase that caused them will not reverse their change (for example, sea level rise). Finally, some tipping elements can change in a reversible way, but the reversal over the change is longer than human time frames.
Underlying your approach could be an assumption that, having defeated climate change, the natural world is saved, so it makes sense to compare tree planting as an approach to reducing GAST rise with some other approach. However, the sharp losses to all tree populations due to existing climate change at some intermediate point in the future (for example, tipping of the Amazon within 20 years) is a near term threat to our ecosystems even if GHG’s were in principle reducible to pre-industrial levels by century’s end, other things equal. Not to mention, other things are not equal. In particular, with increase in GAST past 1.5C virtually guaranteed, we can expect tipping points to pass this century.
Ecosystem losses are covered somewhat in the IPCC report on land management, but not to the reducible form where you can assume incremental changes lead to incremental effects. More broadly, the problem is one of assuming that ecosystem management is like running a factory, that is, we are deciding how to allocate resources and make products (in a broad sense), and can always change our minds, reallocate resources, bring in engineers, install new stuff, and go back to make the older products. But we can’t do that, really. Once a system is too far gone, it won’t make sense to try and recover. I don’t think that danger or cost or timing of such outcomes is represented in your analysis.
So, yes, tree planting is not justifiable as a single solution for climate change, but in fact tree populations are threatened by climate change. To model reforestation or ecosystem protection efforts as a question of cost-effectiveness in reducing GHG’s reverses the causality. The bigger concern is whether we can protect forests by reducing anthropogenic GHG’s, real-estate development, cattle-raising, and other pressures on forest health. Forests are important in their own right, but that importance should not be reduced to economic value in a way that ignores most ecosystem survival benefits and limits.
I welcome your thoughts. Thanks.
Hi Noah,
Thanks for the comment, I strongly upvoted it!
See Cumulative greenhouse gas emissions between 2020 and 2100 assuming current climate policies.
See Net removal of CO2e emissions.
See Tree planting intervention risk (including this footnote).
These are implicitly taken into account in the net removal of CO2e emissions.
See Ecosystem services benefits and Factual and counterfactual initial ecosystem services.
I agree the analysis is broad and shallow. I commented here that:
I agree ecosystem services are quite complex, but, based on the results I got:
Regarding the EVSD, you mentioned:
Tourism is not the driver of the ecosystem services based on the data I used to model those of the “ERPM” projects:
I tend to agree that:
However, it might still be the best approach we have now. I suppose the relationship between ecosystem services and existential risk could be explored further, but I guess the main driver for the cost-effectiveness will be the net removal of CO2e emissions.
I agree temperature can increase superlinearly past a certain level of emissions. However, I expect the linear regime to continue for a while. From note (6.93) of What We Owe to the Future (WWOF) (see here; emphasis below added by me):
In principle, I agree that:
However, for the reasons described in the note above, I think we are very far from that situation.
There is an important distinction between crossing a climate tipping point and a temperature tipping point. The article you link to focusses on climate tipping points of the cryosphere, ocean and biosphere. Their most extreme (maximum for positive, or minimum for negative) impact on the global temperature is descrived inTable S3 (see Supplementary Materials):
Greenland Ice Sheet [collapse]: 0.13.
West Antarctic Ice Sheet [collapse]: 0.05.
Labrador-Irminger Sea / SPG Convection [collapse]: −0.5.
East Antarctic Subglacial Basins [collapse]: 0.05.
Amazon Rainforest [dieback]: 0.15 (= (0.1 + 0.2)/2).
Boreal Permafrost [collapse]: 0.3 (= (0.2 + 0.4)/2).
Atlantic Meriodional Overturning Circulation [collapse]: −0.5.
Arctic Winter Sea Ice [collapse]: 0.6.
East Antarctic Ice Sheet [collapse]: 0.6.
Lowlatitude Coral Reefs [die-off]: not defined (ND).
Boreal Permafrost [abrupt thaw]: ND.
Barents Sea Ice [abrupt loss]: ND.
Mountain Glaciers [loss]: 0.08.
Sahel & West African Monsoon [greening]: ND.
Boreal Forest [southern dieback]: −0.18.
Boreal Forest [northern expansion]: 0.14.
Boreal Permafrost [gradual thaw]: 0.7.
Arctic Summer Sea Ice [loss]: 0.25.
Global Land Carbon Sink [weaken]: ND.
Ocean Biological Pump [weaken]: ND.
Marine Methane Hydrates [dissociation]: 0.5.
Indian Summer Monsoon [shift]: ND.
South. Ocean Sea Ice [Ind. increase]: ND.
South. Ocean Sea Ice [Pac./Atl. loss]: ND.
South. Ocean Sea Ice [bimodality]: ND.
Equatorial Stratocumulus Clouds [breakup]: 8.
Antarctic Bottom Water [collapse]: ND.
Indian Ocean Upwelling [abrupt increase]: ND.
Tibetan Plateau Snow [abrupt loss]: ND.
Ocean Deoxygen -ation [global anoxia]: ND.
Arctic Ozone Hole [abrupt expansion]: ND.
El Nino Southern Oscillation [permanent / extreme]: ND.
Northern Polar Jet Stream [instability]: ND.
Adding up all of these, I get a maximum global warming of 10.37 ºC, which is mostly driven by the 8 ºC which could result from the breakup of the equatorial stratocumulus clouds. This:
If this tipping point exists (it is still unclear), we would have to burn all fossil fuels to cross it, which seems quite unlikely. Moreover, even if we crossed it, we would take quite some time to do so (it would take time to get to a CO2 concentration which is 4 times as high), so I guess there would be some time to adapt. (7 + 8.9)/2 = 8 ºC of warming until the tipping, plus 8 ºC due to crossing the tipping point would lead to 16 ºC of warming, but this is a very worst case scenario (which should be downweighted proportionally to its low likelihood). In addition, it is not totally clear (1400 + 2200)/8*16 = 7200 ppm would lead to a moisture greenhouse effect. However, the note from WWOF says the tipping point may be somewhere between 4,480 to 8,960 ppm, so it would certainly be quite troubling!
I agree, and would add that supporting CCF is much better than tree planting (as I clarified here).
The ESVD considers many ecosystem services (see this footnote), although there is room for improving their measurement.
Thanks again for commenting!
Thanks for the reply. If you don’t mind my adding a few more thoughts:
and then there’s some estimates of temperature effects of all the tipping elements you listed … tipped. But temperature effects on temperature, not the effects of the changed systems on the land, oceans, or atmosphere. Looking at those numbers, I have some serious doubts. For example, if you shut off the AMOC, you change ocean circulation such that warming occurs off continental shelves, triggering release of hydrates on those shelves. Globally, there’s more than is needed to raise GAST by 0.5C. There’s more accessible and releasable from just the East Siberian Arctic Shelf, 1C-2C GAST from there alone, if it’s an abrupt release of mostly methane.
If the Greenland Ice Sheet were to collapse, the accompanying sea level rise would be more than 7 meters. Off the top of my head, that would flood some countries, make floods so drastic in others that farming ceases because of saltwater inundation, and flood out most coastal cities around the world. Global farming productivity would take a hit because of loss of low-lying countries. Saltwater would infiltrate groundwater supplies in more places, and reduce the fertility of land near rivers, reducing farming productivity further. That’s one tipping point considered in isolation. EDIt: Those effects would begin with strength around 1-2 meters of rise, I estimate.
If you’re writing about the collapse of the West Antarctic Ice Sheet, that’s near-term, and 10+ feet of sea level rise. I don’t know how you feel about the civilizational impacts of an additional 10+ feet of sea level rise, but you can’t feel that good about it, and at that point, the Antarctic would not stop melting. It holds 50 meters plus of sea level rise in it.
And that’s just one problem. Jet stream instability? That’s connected heat domes across the Northern Hemisphere, extreme weather events like what happened in the UK this year, but much worse, for example, enough to trigger a multi-breadbasket failure. Shifting the monsoons? Well, that could dry out land over Asia and South America, reducing farming productivity, killing forests,desertifying land, collapsing land biological pumps, which currently absorb some 30% of anthropogenic CO2 production. Those effects would be irreversible in the short term. Carbon dumped from existing land suffering fires or drought would go into the atmosphere, adding to anthropogenic sources. The ocean would be what’s left to absorb CO2, some 25%+ of our CO2 production (EDIT: or more).
But then the problems become too big. The AMOC has stopped, land sinks of carbon are gone, methane hydrates globally are melting out, permafrost is melting out abruptly, ocean PH in the productive zones has dropped because of changed circulation, ocean heating and freshening and CO2 absorption has collapsed all fisheries, starting in the Arctic, and some phytoplankton species are dead, with others present but in smaller amounts, though that’s a guess. The ocean’s biological pump is collapsed, and the ocean is dumping more carbon into the atmosphere. In parallel, the 6th great extinction over and ours is threatening.
I just don’t think that the impacts of those tipping elements falling was thought through very well by the WWOTF research. Their analysis reverses the relevant importance of GAST, apparently. It’s not that higher GAST is nonlinearly bad in amplifying itself (though I think if all those tipping points fall, you could be looking at 20C GAST rise or more). What is bad about higher GAST is that it effects those tipping elements as it does, nonlinearly, and at fairly low temps, 2C-5C GAST rise, mostly. The consequences of the tipping elements falling are disastrous. Most people die. Most species go extinct. Most land is uninhabitable. That’s what the fall of those tipping elements means.
I believe it takes patience and bottom-up thinking to consider building a town on a melted permafrost landscape still oozing methane and developing sink holes with constant simmering fire appearing on any vegetation, or in a desert with 55C-60C temps a lot of the year or rocky areas with no groundwater, no meltwater, no rivers, and little or no rain, or on a coast subject to frequent extreme weather, toxic algae blooms, and pollutant-contaminated jellyfish die-offs, when they’re around at all.
I’ve read before that geologic time shows that species survived high temperature changes, and so there won’t be a 6th great extinction, but it is already taking place, our activities outside global warming contribute to it, and in any case relevant temperature changes from the geologic past occurred over much longer than a few hundred years, giving species time to adapt.
The lesson in climate science and given in public discussion and media coverage of climate change is, that we don’t want to go past 1.5C GAST rise. Here’s we’re looking at upwards of 3C GAST rise this century. The articles I referenced are exploring the outcomes of that rise, and its knock-on effects into the next century. The significance is the systemic effects of that GAST rise, not that the numbers are just “high” per se.
Thanks for sharing!
Unfortunately, I do not know much about the effects of the tipping points. I believe you know way more than I do! That being said, I can say I agree with Kemp 2022 that research on the effects of extreme climate change (> 3 ºC) has been neglected relative to non-extreme one.
You’re welcome, of course, and thank you for your courtesy in replying.
I don’t have time, money, training or resources to explore climate in more detail, but if I could, I would start with developing more sophisticated computer models that incorporate more physical processes deciding changes to tipping elements (for example, meltwater drainage on Greenland). Existing atmospheric models use a mesh that is too coarse, making predictions translate from climate to weather poorly. Ocean modeling is hopeless at current atmospheric mesh size, we need 1km meshes or better for climate models, not the much larger meshes in use for the atmosphere. Dedication of more and faster computer resources to climate modeling would be helpful.
However, different methodologies applied to develop less sophisticated models can still produce general foresight that limits or suggests available policy options to prevent climate-related crises. This runs into political interests ultimately concerned with the welfare of a small percentage of the global population, whose false sense of security and selfishness is actually in the way of preserving civilization.
Regardless, the survival of humanity, if that is what this fiasco comes to, could depend on terraforming this planet back to an 1800′s state. Well before then, we need very fine mesh computer models, as well as better understanding of ecosystems and practical means to create and quickly evolve new species. An alternative would be local-scale weather and ecosystem control, allowing a small area of land to have mild weather and favorable living conditions, allowing one city or small country to flourish, though that combination of advanced technology and a small human population might not work on its own. The trouble with humans and technology is that we are not born with the learnings of our parents, so a small population supported by technology that is magic to us, and whose ecosystem also depends on that technology, won’t last, unless we’ve locked in some AI or AL running it that understands it. That issue makes me think we’re better off with a planet that is habitable in many areas, not one with a small green zone run by some effectively alien technology. If that whole discussion seems implausible, that is because it absolutely is implausible.
While as a longtermist, I can see the value of having a small human population reached equitably through family planning over a few centuries, that should be achieved without large losses of existing global population. That means protecting against climate change catastrophe and systemic effects. Implying, for this century, degrowth and energy conservation and the end of consumerism, and less emphasis on democratization of advanced technologies. But all that has to happen soon to preserve most of the global population from early death.
I have watched Kemp’s lecture on foreseeing the end of the world. The discussion is about how to layer forecasting and foreseeing methods to improve prediction of existential crises, something still in the exploratory stages. Don’t associate my own conclusions with his, my weird opinions are my own, but I recommend his work.
You folks do a lot of forecasting, but foreseeing is worth exploring as well. I see the same forks in the road, but only foresee the negative ones as least surprising. Climate destruction is not a hopeless situation though, but not hopeless for whom living how? That’s the bigger question, and a worrisome one. There’s no solutions without compromises, particularly for utilitarians , that is, absent the longtermist “far future trillions of people” take, where current population is a just a blip anyway. And that’s why I’ve said before, only future people that will exist have moral status. Present people do have moral status. Present people matter, including those in the womb. Future people? Only if they will actually be alive at some point.
Climate destruction, as assessed by MacAskill’s book and Halstead’s analysis, is difficult to take seriously. You EA folks could use something better, considering your interest in preventing existential dangers. In general, EA’s massively underestimate the climate change threat to civilization, and appear to prefer techno-optimist solutions regardless, which multiplies environmental and resource management concerns for utilitarians when turned into policy.
Anyway, I envy your position as someone actively involved in foresight and forecasting, as well as your efforts to suggest policy. Congratulations on that opportunity! I have not looked into the climate nonprofit that you recommend, but I hope their work is good.
I was really glad to see you take into account albedo changes-I think this is really important for the boreal forest.
For the existential and overall cost effectivenesses, I think you’re getting implausible values because the standard deviation of your future population size and duration of humanity are too large. For instance, it appears that your 5th percentile of future population of humanity is 10^-34 people, and duration is <1 yr. I guess it’s possible to have a mean greater than the 95th percentile, but I don’t think the means are plausible. Anders Sandberg estimates the upper bound of number of human lives is around a Googol (10^100, note, Google is a misspelling of Googol). So it is not reasonable that 1 £ is going to save 10^90 lives.
Hi David,
Thanks for the comment!
You may well be right. I did not try to get the most accurate distributions. The selected ones are essentially for illustration that the longterm effect of tree planting dominates the analysis.
In the table of this section, I mention that such effect is of the order of 10^(10^2) in terms of lives saved per hectare. This was also to indicate I am only confident of getting the order of magnitude of the order of magnitude right. So, for the cost-effectiveness, I believe log10(log10(“existential/overall cost-effectiveness (life/£)”)) = 2.
The virtually null 5th percentile of the future population size makes sense to me. Given Toby Ord’s guess if 1⁄2 for the total existential risk, I guess the median population size is close to 0 (I set it to 1). The 5th percentile of the humanity’s lifespan being shorter than 1 year does not make sense, but it does not affect the mean of the distribution, which is driven by the right tail. In general, truncated distributions would be more accurate.
Okay-if you’re just trying to get the order of magnitude of the exponent correct, then this could be consistent with 10^10^1.5 ~10^30 lives/£, which is plausible. As for the median population size, I guess you are trying to get an average over the time horizon which could extend 100 trillion years, so that’s why you’re saying it’s around zero with 50% as existential risk, which makes sense.
By the way, I was confused when I read this:
I think you mean population size instead of duration.
Yes, that was it, thanks for clarifying!
Good catch, corrected!
Thanks for the report—I’ll keep my initial estimate of tree planting as “not worth spending much time on that”. This works feels very complete.
Just a suggestion: this post is kind of hard to read as it is now, so I think it could benefit from something like an executive summary that allows to grasp in a clear and simple way why tree planting is not as effective as one could expect (there are many interesting things you took into account, like albedo and impact on insects but it feels drowned in an ocean of numbers).
Hi Corentin,
Thanks for the comment!
That makes sense to me.
I would say it is very broad, as it covers many of the effects of tree planting. However, I think it is also quite shallow, as it does not go into much depth into any of them. I think it is not necessary to go into greater depth given given the resilience of the mean cost-effectiveness being at best around 1 t/£.
I very much agree the post is not very reader-friendly.
This footnote illustrates how one could obtain the upper bound for the mean cost-effectiveness of 1 t/£:
As for:
In essence:
So the net removal of CO2e emissions (t/ha/year), persistence of the intervention (year), and cost (£/ha) seem to be the factors which really matter.