24 comparisons using peer-reviewed life-cycle data — greenhouse gases, land, water and protein, per kilogram of food or per 100 g of protein.
Beef is the single most emission-intensive food in common use: ruminant methane, feed production and pasture expansion stack on top of each other. Lentils fix their own nitrogen, need no feed conversion step, and match beef almost gram-for-gram on protein.
Feeding soy to cattle wastes most of its protein and calories. Eating the soy directly as tofu removes that conversion loss, which is why tofu's footprint is a rounding error next to beef's.
Every plant milk beats dairy on all three environmental axes. Oat is the strongest all-rounder: low water use, temperate growing regions and no almond-style irrigation pressure.
Cheese is the third most emission-intensive food after beef and lamb, because it takes around ten litres of milk to make a kilogram. Nut-based cheeses skip the animal entirely, though tree nuts remain water-thirsty.
Chicken is often promoted as the low-carbon meat, and it is lower than beef — but it still emits six times more than chickpeas per kilogram, and it costs far more animal lives per kilogram of food than any large animal.
Fermented whole soybeans keep the fibre and protein of the bean while adding B vitamins. Pigs must be fed several kilograms of crops per kilogram of meat, which is where most of the footprint originates.
Eggs sit mid-table on emissions but at the extreme end for dietary cholesterol. Tofu turned into scramble with turmeric and black salt delivers a similar plate with more protein and no cholesterol.
Swapping butter for olive oil is one of the highest-evidence single dietary changes for cardiovascular risk, and it also halves the climate cost of the fat in your cooking.
Lamb is second only to beef on emissions and first on land use per kilogram in many systems. Beans occupy the opposite end of every axis, and they are the cheapest protein per gram on the shelf.
Salmon do not make EPA and DHA — they accumulate it from algae. Taking algae oil directly removes the feed-fish demand, the sea-lice and escape problems of net pens, and the contaminant load.
Prawn farming is among the most carbon-intensive seafood because of mangrove clearance, which releases stored coastal carbon. Mushroom stems, seared hard, reproduce the bite closely.
Both are concentrated sugars, so nutrition is close to a wash. The difference is ethical: honey is a bee's winter food store, replaced commercially with sugar syrup.
Leather is not simply a by-product: hides carry a share of cattle's enormous footprint, and chrome tanning is a serious water-pollution source. Cactus, cork, mycelium and pineapple-leaf alternatives now perform for bags and shoes.
Wool carries the methane and land footprint of the sheep that grew it. For warmth, recycled synthetic fill, hemp and lyocell blends match performance at a fraction of the impact.
The burger is the clearest single-meal illustration of the gap. A black-bean or lentil patty costs about 3% of the beef version's climate impact while adding fibre no meat contains.
Randomised trials comparing pea and whey supplementation during resistance training find equivalent gains in strength and lean mass when protein intake is matched. Pea has less leucine per gram, so a slightly larger serving closes the gap.
Gelatin is hydrolysed collagen from pig skin and cattle hide. Agar, extracted from red seaweed, sets more firmly and does not melt in a warm room — useful for travel desserts and hot climates.
Fermented soy carries the same live cultures as dairy yoghurt, and unsweetened versions have slightly more protein with much less saturated fat.
Wheat gluten is the densest common plant protein and takes marinade like meat. It is not coeliac-safe, but for everyone else it is the closest textural match to braised beef.
Most dark chocolate above 70% cocoa contains no milk at all. It carries several times the flavanol content and five times the iron of milk chocolate.
Ground flaxseed is the cheapest reliable ALA source, but conversion to EPA and DHA is limited. The complete plant-based answer is flax for ALA plus a small daily algae-oil capsule.
Fortified soy milk is the closest nutritional analogue to cow's milk and is the plant milk used as the reference in most dietary guidelines for children over one year.
Mycoprotein is grown in fermenters rather than fields, so its land requirement barely registers. Check the label: some formulations still bind with egg white.
Young green jackfruit shreds exactly like slow-cooked pork and absorbs barbecue sauce well. It is a carbohydrate, not a protein, so serve it with beans or a bean-based side.
Beef, lamb and farmed shrimp have by far the highest greenhouse gas emissions per kilogram and per 100 g of protein; cheese and other ruminant dairy follow. Pulses, grains, tofu and most vegetables sit an order of magnitude lower.
The figures are drawn from Poore & Nemecek's 2018 Science meta-analysis of ~38,700 farms across 119 countries, the most comprehensive life-cycle dataset available, plus the source cited on each comparison page.
Very little. Transport is typically a small share of food emissions, and grass-fed cattle usually emit more methane per kilogram and need more land than intensively raised cattle. The food you choose matters far more than how far it travelled.
How the comparisons work
Comparison pages are useful and easy to abuse. The figures here follow a single methodology, and the assumptions are stated so you can disagree with them precisely rather than vaguely.
The environmental figures draw primarily on Poore & Nemecek's 2018 synthesis in Science, which consolidated life-cycle assessments from roughly 38,700 farms in 119 countries — still the largest dataset of its kind. It covers emissions, land use, freshwater withdrawals and eutrophication from farm through retail. Where a product is not in that dataset, national life-cycle inventories or peer-reviewed product LCAs are used and labelled as such.
A mean hides an enormous distribution. The highest-impact 25% of beef producers emit roughly twelve times more per kilogram than the lowest 25%. That variation is real and matters for policy. It also does not rescue the comparison: the lowest-impact beef still exceeds the highest-impact plant protein on both emissions and land use, which is why product substitution outperforms producer optimisation as an individual lever.
Choice of functional unit changes the appearance of a comparison more than any other decision. Per kilogram flatters watery foods; per calorie flatters fats and oils; per gram of protein is the fairest unit for foods eaten mainly as protein. Where these pages compare protein sources, protein is the unit, and where they compare drinks, volume is. Any table that silently switches units between rows should be treated with suspicion.
For nearly all foods, production dominates and transport is around 6% of total food-system emissions. The exception is air-freighted perishables, which is a narrow category. Locally produced beef is not lower-impact than imported legumes; the intuition that distance is the main variable is the single most common error in casual food-footprint reasoning.
Environmental comparisons rest on physical measurement; health comparisons rest on epidemiology plus a smaller number of trials. Where these pages state a health difference, they distinguish between mechanistic plausibility, observational association and randomised evidence — and note where the effect size is small enough that it should not drive a decision on its own.
Global means, cradle-to-retail, per kilogram of product unless stated. Individual producers vary by an order of magnitude.
| Food | kg CO₂e per kg | Land m² per kg | Freshwater L per kg |
|---|---|---|---|
| Beef (beef herd) | ~60 | ~164 | ~15,400 |
| Lamb | ~24 | ~185 | ~10,400 |
| Cheese | ~21 | ~88 | ~5,600 |
| Farmed prawns | ~12 | ~2 | ~3,500 |
| Pork | ~7 | ~11 | ~6,000 |
| Poultry | ~6 | ~7 | ~4,300 |
| Eggs | ~4.5 | ~6 | ~3,300 |
| Tofu | ~3 | ~2 | ~2,000 |
| Milk (dairy) | ~3 | ~9 | ~630 |
| Legumes | ~0.8 | ~3 | ~1,250 |
| Soy milk | ~1 | ~0.7 | ~28 |
Poore & Nemecek (2018), Science · Water Footprint Network · Our World in Data
Five questions that expose most misleading charts in under a minute.
Per kilogram, per calorie or per gram of protein — and is it consistent across every row?
Farm gate, retail or plate? Excluding processing and refrigeration flatters some products substantially.
Marketing frequently cites the lowest-impact decile of animal production against the average for plants.
Producer-funded LCAs are not automatically wrong, but they choose boundaries and allocations that tend to favour the sponsor.
Allocation between meat, leather, dairy and by-products can shift a headline number by a factor of two.
On some measures, such as local biodiversity and soil carbon on degraded land, it can be. On emissions per kilogram it is generally worse, because slower growth means more methane per unit of meat, and it uses substantially more land.
Almonds are water-intensive per nut, but per litre of milk almond still uses far less water than dairy. The comparison that circulates online usually compares whole almonds to a glass of milk.
Around 77% of global soy is fed to animals. Soy for direct human consumption — tofu, tempeh, soy milk — is a small minority of production and is often grown outside the frontier regions.
Directionally, yes; precisely, no. National figures vary with feed, energy mix and land history. The ranking of food categories is stable across almost every region studied.
It takes roughly eight to ten litres of milk to make a kilogram of hard cheese, so the dairy herd's emissions are concentrated into a small mass of product.
The underlying synthesis is from 2018 with subsequent refinements; where newer national inventories change a figure materially, the page notes it.