We quantify climate-driven changes in agricultural production potential on a global 10×10 km grid. For each grid cell, we take the cropping patterns observed at 2020—which crops were grown and how much area each occupied—and hold them fixed across all time periods and scenarios. We then calculate how the attainable yield of these same crops changes under different climate conditions: a historical baseline (1981–2000), the recent observed climate (2001–2020, from AgERA5 reanalysis), and future climate projections under the SSP5-8.5 scenario (from CMIP6 climate model ensembles). By keeping the crop mix constant while only climate conditions vary, we isolate the direct impact of climate change on agricultural productivity.
Agricultural output in any location depends on three factors: (1) climate conditions (temperature, precipitation, growing season characteristics), (2) cropping patterns (which crops are grown and how much area each occupies), and (3) agronomic management (input intensity, irrigation, technology).
Our approach isolates the climate component by holding the other two factors constant:
With this design, any differences in production across time periods or scenarios reflect pure climate-driven changes in yield potential—not adaptation through crop switching, irrigation expansion, or technological change.
Think of this as answering: "What would happen to agricultural productivity if farmers kept growing exactly the same crops in the same proportions as observed at 2020, but faced the climate conditions of 2030, 2050, or 2080?"
This "no-adaptation" scenario establishes a baseline for measuring the adaptation challenge communities face.
Attainable yields: We use GAEZ v5 gridded attainable yield estimates under historical and future climate scenarios. Attainable yield represents the maximum productivity achievable given the climate and soil conditions of each grid cell, assuming a specified management level. It accounts for climate factors (temperature, precipitation, solar radiation, growing season), soil characteristics (nutrients, texture, depth, drainage), terrain constraints (slope, elevation), and crop physiology and growth requirements.
Attainable yield is adjusted downward from theoretical "potential yield" to reflect realistic site-specific limitations, making it a more credible benchmark than pure agro-climatic potential.
Cropping patterns: We use the observed distribution of crops and harvested areas at 2020 as our fixed reference. This captures which crops were actually grown in each location and how much area each crop occupied. These cropping patterns are held constant across all time periods and climate scenarios, so that any differences in production potential reflect only the effect of changing climate.
All spatial layers are harmonised to a common 10×10 km grid resolution and standardised crop definitions before calculation.
Let \(A_{i,c}\) denote the harvested area (ha) of crop \(c\) in cell \(i\), taken from the baseline year and held fixed. Let \(Y_{i,c}(s,t)\) denote the GAEZ yield estimate (e.g., tonnes/ha) for scenario \(s\) and time period \(t\) under the chosen management setting.
We compute total (mass) output in cell \(i\) as:
$$\Theta_i(s,t) = \sum_{c} A_{i,c} \, Y_{i,c}(s,t).$$
To aggregate across crops in a common unit, we convert crop output into calories using a crop-specific calorie conversion factor \(\kappa_c\) (kcal per tonne):
$$C_i(s,t) = \sum_{c} A_{i,c} \, Y_{i,c}(s,t) \, \kappa_c.$$
We then convert annual calories to People Fed Yearly (PFY) using a reference daily requirement \(c_{\min}\) (kcal/person/day):
$$\text{PFY}_i(s,t) = \frac{C_i(s,t)}{c_{\min} \cdot 365}.$$
In practice we use \(c_{\min}=2000\) kcal/person/day as a standard reference value.
Let \((s_0,t_0)\) denote the baseline scenario and baseline period. We report:
(and analogously for \(C_i\) and \(\text{PFY}_i\)).