Thank you for this thought-provoking post. I deeply appreciate seeing attention given to non-charismatic and extremely neglected animals such as soil arthropods, and the willingness to invest serious time into exploring how we might help them. It’s rare and valuable work.
Most comments here focus on questions about the sentience of these small animals, the implications of your conclusions or the appropriateness of basing moral action on highly uncertain utilitarian calculations. I’d like to contribute from a different angle: your soil arthropod abundance estimates. This is the part I understand best, and where I think there may be relatively low-hanging fruit for improving the precision of your numbers. While I work in ecology, I’m not a soil specialist, so these are informed observations rather than definitive expert judgments.
- You note that “0.1 kg of chicken meat causes 2.87 animal-days more in directly affected animals, but 6.16 million animal-years fewer in soil nematodes, mites, and springtails for feed crops replacing temperate grasslands, savannas, and shrublands.” I’d suggest also considering cropland replacing forest. In Europe, much former cropland has been abandoned as agriculture concentrated on more productive areas, and these lands often undergo natural succession toward forest. Reduced meat demand could plausibly accelerate this trend, leading to forest regrowth rather than grassland expansion. In tropical regions, the dynamics are less certain, but a significant share of meat production still comes from newly deforested land. To capture the true impact on soil arthropod communities, at least some fraction of cropland should be compared against forest densities, not only grassland or savanna.
- Your soil arthropod abundance estimates already account for regional and habitat variation, but they appear to treat pasture as a single global category. I’d expect large differences between, say, nutrient-poor Brazilian pastures and nutrient-rich Siberian steppe pastures. It seems unlikely that these support similar densities of nematodes, mites, and springtails. Regional differentiation here might significantly change the results.
- Using your numbers, pasture densities come out much lower than cropland densities, which is counterintuitive. Cropland is usually subject to tillage (which can kill many surface-dwelling nematodes) and often receives substantial insecticide applications that likely harm a broad range of soil arthropods. In contrast, pasture arthropods are primarily exposed to antiparasitics like ivermectin. While most antiparasitics take time to degrade and are likely toxic to some coprophagous invertebrates, their total mass applied per hectare and their spatial footprint (concentrated in dung pats) are far smaller than the mass and coverage of insecticides on cropland. To me, this suggests that soil arthropod densities in pastures should be higher than those in croplands within the same biogeographic region, at least outside intensively treated livestock systems.
- Most arthropod density studies in temperate and subarctic regions sample in summer, when abundance and richness peak. Winter densities can be orders of magnitude lower, especially in subarctic soils. Extrapolating summer values to the full year may overstate average densities, so a season-weighted metric could be more accurate. In tropical systems, I’d expect most sampling to occur in the wet season, when densities are also highest. Adjusting for these seasonal biases would make your cross-system comparisons more robust.
- In your product comparisons, you account for the land area needed per kilogram of product. But did you also account for the split between cropland and pasture for each product in each region? Beef and mutton in tropical rangelands or the Asian steppe, for example, are often produced primarily from pasture rather than cropland. This proportion matters if the two land uses have very different soil arthropod densities.
Overall, I suspect these factors (land-use change baselines, regional pasture differences, cropland vs. pasture toxicity profiles, seasonal sampling biases and land-use composition by product) could shift some of your estimates by an order of magnitude or more. If we want to translate your insights into actionable guidance for career choices, donations or consumption patterns, resolving these uncertainties and producing estimates specific to each product and biogeographic region would make the conclusions much stronger.
Thanks again for opening this conversation, and for taking seriously the welfare of animals most people overlook!
Thanks for the great points, David! I strongly upvoted your comment.
I presented results for cropland replacing replacing temperate grasslands, savannas, and shrublands, but I consider many other replacements in my sheet. I estimate tropical and subtropical forests have 1.76 (= 5.25*10^6/(2.99*10^6)) times as many soil nematodes, mites, and springtails per m2 as tropical and subtropical grasslands, savannas, and shrublands, and that temperate forests have 1.42 (= 8.92*10^6/(6.30*10^6)) times as many soil nematodes, mites, and springtails per m2 as temperate grasslands, savannas, and shrublands. These ratios are higher than 1, so I think replacements of cropland with forests would tend to decrease soil-animal-years more than suggested by the results I presented in my post. This would strengthen my point that effects on soil animals are much larger than those on directly affected animals. However, I very much agree with your broader point that one should consider a mix of cropland and pasture replacing a mix of biomes (instead of just cropland replacing a single biome, as I did).
I also agree that regional and seasonal differences should be taken into account. Table S4 of Rosenberg et al. (2023)has not only the mean densities I used for each biome, but also 95 %confidence intervals(CIs), and the number of sites regarding each estimate. Acari are mites, collembola are springtails, formicidae are ants, and isoptera are termites.
Overall, I am still thinking that increasing agricultural-land-years is a good heuristic to decrease soil-animal-years. However, I am only confident that the effects on soil animals are larger than those on farmed animals. It would be great to have way more research decreasing the uncertainty about the effects on soil animals, not only about their welfare per animal-year, but also about the changes in their animal-years.
However, I very much agree with your broader point that one should consider a mix of cropland and pasture replacing a mix of biomes (instead of just cropland replacing a single biome, as I did).
I have just published a post where I consider this.
Hi Vasco,
Thank you for this thought-provoking post. I deeply appreciate seeing attention given to non-charismatic and extremely neglected animals such as soil arthropods, and the willingness to invest serious time into exploring how we might help them. It’s rare and valuable work.
Most comments here focus on questions about the sentience of these small animals, the implications of your conclusions or the appropriateness of basing moral action on highly uncertain utilitarian calculations. I’d like to contribute from a different angle: your soil arthropod abundance estimates. This is the part I understand best, and where I think there may be relatively low-hanging fruit for improving the precision of your numbers. While I work in ecology, I’m not a soil specialist, so these are informed observations rather than definitive expert judgments.
- You note that “0.1 kg of chicken meat causes 2.87 animal-days more in directly affected animals, but 6.16 million animal-years fewer in soil nematodes, mites, and springtails for feed crops replacing temperate grasslands, savannas, and shrublands.” I’d suggest also considering cropland replacing forest. In Europe, much former cropland has been abandoned as agriculture concentrated on more productive areas, and these lands often undergo natural succession toward forest. Reduced meat demand could plausibly accelerate this trend, leading to forest regrowth rather than grassland expansion. In tropical regions, the dynamics are less certain, but a significant share of meat production still comes from newly deforested land. To capture the true impact on soil arthropod communities, at least some fraction of cropland should be compared against forest densities, not only grassland or savanna.
- Your soil arthropod abundance estimates already account for regional and habitat variation, but they appear to treat pasture as a single global category. I’d expect large differences between, say, nutrient-poor Brazilian pastures and nutrient-rich Siberian steppe pastures. It seems unlikely that these support similar densities of nematodes, mites, and springtails. Regional differentiation here might significantly change the results.
- Using your numbers, pasture densities come out much lower than cropland densities, which is counterintuitive. Cropland is usually subject to tillage (which can kill many surface-dwelling nematodes) and often receives substantial insecticide applications that likely harm a broad range of soil arthropods. In contrast, pasture arthropods are primarily exposed to antiparasitics like ivermectin. While most antiparasitics take time to degrade and are likely toxic to some coprophagous invertebrates, their total mass applied per hectare and their spatial footprint (concentrated in dung pats) are far smaller than the mass and coverage of insecticides on cropland. To me, this suggests that soil arthropod densities in pastures should be higher than those in croplands within the same biogeographic region, at least outside intensively treated livestock systems.
- Most arthropod density studies in temperate and subarctic regions sample in summer, when abundance and richness peak. Winter densities can be orders of magnitude lower, especially in subarctic soils. Extrapolating summer values to the full year may overstate average densities, so a season-weighted metric could be more accurate. In tropical systems, I’d expect most sampling to occur in the wet season, when densities are also highest. Adjusting for these seasonal biases would make your cross-system comparisons more robust.
- In your product comparisons, you account for the land area needed per kilogram of product. But did you also account for the split between cropland and pasture for each product in each region? Beef and mutton in tropical rangelands or the Asian steppe, for example, are often produced primarily from pasture rather than cropland. This proportion matters if the two land uses have very different soil arthropod densities.
Overall, I suspect these factors (land-use change baselines, regional pasture differences, cropland vs. pasture toxicity profiles, seasonal sampling biases and land-use composition by product) could shift some of your estimates by an order of magnitude or more. If we want to translate your insights into actionable guidance for career choices, donations or consumption patterns, resolving these uncertainties and producing estimates specific to each product and biogeographic region would make the conclusions much stronger.
Thanks again for opening this conversation, and for taking seriously the welfare of animals most people overlook!
Thanks for the great points, David! I strongly upvoted your comment.
I presented results for cropland replacing replacing temperate grasslands, savannas, and shrublands, but I consider many other replacements in my sheet. I estimate tropical and subtropical forests have 1.76 (= 5.25*10^6/(2.99*10^6)) times as many soil nematodes, mites, and springtails per m2 as tropical and subtropical grasslands, savannas, and shrublands, and that temperate forests have 1.42 (= 8.92*10^6/(6.30*10^6)) times as many soil nematodes, mites, and springtails per m2 as temperate grasslands, savannas, and shrublands. These ratios are higher than 1, so I think replacements of cropland with forests would tend to decrease soil-animal-years more than suggested by the results I presented in my post. This would strengthen my point that effects on soil animals are much larger than those on directly affected animals. However, I very much agree with your broader point that one should consider a mix of cropland and pasture replacing a mix of biomes (instead of just cropland replacing a single biome, as I did).
I also agree that regional and seasonal differences should be taken into account. Table S4 of Rosenberg et al. (2023) has not only the mean densities I used for each biome, but also 95 % confidence intervals (CIs), and the number of sites regarding each estimate. Acari are mites, collembola are springtails, formicidae are ants, and isoptera are termites.
Overall, I am still thinking that increasing agricultural-land-years is a good heuristic to decrease soil-animal-years. However, I am only confident that the effects on soil animals are larger than those on farmed animals. It would be great to have way more research decreasing the uncertainty about the effects on soil animals, not only about their welfare per animal-year, but also about the changes in their animal-years.
I have just published a post where I consider this.