This page explains the main terms and ideas behind the CADI maps. For a full technical description of our methodology, see our Technical Report.
Attainable yield is the biophysically attainable crop yield at a location, given climate, soil, and terrain constraints and a specified level of inputs and management, as modelled by the FAO/IIASA Global Agro-Ecological Zones (GAEZ) framework.
Attainable yield reflects:
What attainable yield is not: It is not an observation of what farmers actually harvest. It does not incorporate market access, input use decisions, policy, conflict, or other socioeconomic factors. It is designed to isolate the role of biophysical constraints and climate conditions and should not be interpreted as observed production data.
For each grid cell, we consider the crops that are grown during the reference period (approximately year 2020) and hold this crop mix fixed across all time periods and scenarios. This means the same set of crops and harvested area shares is used whether we are computing production under the 1981–2000 climate or the 2081–2100 climate. Any differences in output therefore reflect only the effect of changing climate on the yields of the crops currently grown—not adaptation through crop switching or land-use change.
The crop data comes from GAEZ v5 Module VI, which downscales national agricultural statistics onto spatial cropland to provide the distribution of crops and harvested areas. The crops we consider are:
Cash crops (cocoa, coffee, tea, cotton, tobacco, rubber) and residual categories (fodder, crops n.e.s.) are excluded and assigned zero caloric value, as our indexes focus on food production for human consumption. Fruits and nuts are also excluded due to the excessively wide range of caloric values within this group.
To aggregate the different crops that grow in a given cell into a single, intuitive measure, we convert crop-specific attainable yields (e.g., tonnes per hectare) into food-energy equivalents measured in kilocalories per hectare per year (kcal/ha/year). This caloric yield allows us to aggregate the output from multiple crops within a single grid cell into a single measure of caloric production potential.
People Fed Yearly is a way to express how much food a grid cell can produce in terms of how many people it could feed for a whole year. It converts the total caloric production potential into a simple, intuitive number:
$$\text{PFY} = \frac{\text{Total calories that can be produced per year}}{2{,}000 \;\text{kcal/day} \times 365 \;\text{days}}$$
We use 2,000 kcal per person per day as the reference—a widely used adult guideline in both EU and US nutrition labelling.
Important: The number of people that could be fed is a simplified, energy-only metric. It does not account for diet composition, animal feeding, food waste, market access, or trade. It tells you how much caloric energy a place could produce, not how many people it actually feeds.
We report changes in PFY in two complementary ways:
Both metrics matter. A 50% loss in a low-productivity region may be catastrophic for local food security even if the absolute numbers are small. Conversely, a 5% loss in a breadbasket region can affect millions of people.
Our maps display how much agricultural productivity changes—or is projected to change—in each 10 km grid cell around the world. Two views are available:
In all cases, cropping patterns—which crops are grown and how much area each occupies—are held fixed at their observed year-2020 distribution. Only climate conditions vary across time periods, so any differences represent the pure climate-driven change in production potential.
The climate data sources (AgERA5 for historical periods and CMIP6 ensembles for future projections) are built into the GAEZ v5 modelling framework and are not a choice specific to this project.
AgERA5 is a global gridded climate dataset produced by the Copernicus Climate Change Service for agricultural and agro-ecological applications. Based on the ERA5 reanalysis, it provides daily data at approximately 10 km resolution for variables such as minimum and maximum temperature, precipitation, solar radiation, vapour pressure, and wind speed.
In GAEZ v5, AgERA5 is the climate input used to estimate crop yields for the historical periods 1981–2000 and 2001–2020.
For future periods, GAEZ v5 uses the delta method rather than raw climate model output. The procedure has three steps:
In short, GAEZ v5 future projections are based on the observed 1981–2000 baseline climate plus modelled changes, not on raw climate-model values themselves.
The historical and future periods in GAEZ v5 are not constructed in exactly the same way. The 2001–2020 estimates are based directly on AgERA5, which represents the climate actually observed during those years. Future projections, by contrast, are generated with the delta method — taking the monthly climate change signal from bias-corrected CMIP6/ISIMIP3b models and applying it to the 1981–2000 AgERA5 baseline.
As a result, future maps should be interpreted as showing how the historical baseline climate would change under the modelled future signal, rather than as raw future climate-model output. This difference in construction matters when comparing observed and projected periods: part of the contrast reflects climate change itself, but part also reflects the fact that the two periods are generated differently. In some locations, this can contribute to differences in magnitude or even in the sign of change.
The Shared Socioeconomic Pathways (SSPs) are a set of scenarios developed for the IPCC Sixth Assessment Report. They describe plausible futures for global development and greenhouse gas emissions, spanning different challenges to mitigation and adaptation.
Scenario labels such as SSP1-2.6, SSP3-7.0, and SSP5-8.5 combine a socioeconomic storyline with a forcing pathway targeting an approximate radiative forcing level in 2100 (2.6, 7.0, or 8.5 W/m²). Key scenarios include:
We report results under two scenarios: SSP3-7.0 and SSP5-8.5. We present SSP3-7.0 as the primary scenario throughout the key findings, with SSP5-8.5 shown alongside for comparison.
The two scenarios produce very similar outcomes through mid-century — for example, the number of countries experiencing net agricultural losses is identical at 2041–2060 (101 out of 166 countries under both). They diverge more sharply toward the end of the century, where SSP5-8.5 produces deeper losses.
The GAEZ model generates its projections using the 1981–2000 period as a baseline. When we compare these projections to the changes actually observed during 2001–2020, the observed deterioration in agricultural productivity tracks between SSP3-7.0 and SSP5-8.5. Lower-emissions scenarios (e.g., SSP1-2.6) project changes that are smaller than what has already been observed, which would paradoxically suggest an improvement going forward — not because conditions are expected to get better, but because the damage already observed has outpaced those projections.