Agricultural atlasYIELDS · ENSO · REGIONAL RECORDS

Global crop yields

Explore the geography of crop yields, then open a country and its reporting regions.

Archived release · Nine Asian countries. Open the current merged-data atlas →Release 2026-09-26
Loading coverage catalog…
Observed average yield
Average yield · t/ha
No mapped valueActual record lengths vary.

Click a country to explore its records. Scroll to zoom; drag to move.

How to read this map

Average yield is the arithmetic mean of eligible annual observations for one crop product and one reported season. We keep planted-area and harvested-area yields distinct in the location descriptions. Rice combines both source forms on a paddy-equivalent basis: reported milled yield and production are divided by 0.67 (about ×1.49), while paddy values and areas are unchanged. The conversion is approximate; original values and source forms remain in downloads. Where both forms cover the same region, season, and area basis, we prefer an entire usable paddy series, avoiding double-counting and changes of reporting form within a series. We do not add seasonal components or average district yields to invent national values.

Available record uses each series’ full record; dates can differ between regions. A fixed period retains the same requested years everywhere. The coverage filter uses that full requested interval, or the span between the first and last valid observations in available-record mode. Gray polygons have no mapped value under the current selection; gray does not mean zero yield or absence of a crop.

Country-specific defaults prefer Annual, Calendar Year, or All (Season); wheat then prefers Rabi. Otherwise we use the season with the most eligible observations. Each country’s selected season is listed. Area-basis defaults select the denominator with the most eligible observations in that country and season. The area-basis selector keeps planted and harvested records separate, including where the source changes definition over time. Choose an explicit season to restrict the map. We never pool seasons.

ENSO response fits log yield using a linear time trend and your chosen linear, quadratic, or separate warm/cold ENSO slopes. We refit the exact selected season, area basis, and period, requiring 20 positive-yield years with complete ENSO windows and at least five warm and five cold seasons. Zero yields remain in the observation plots but cannot enter the log model. The 10-year periods cannot meet this fitting requirement. Gray always means unavailable or filtered, never a zero response.

ENSO colors use a fixed −50% to +50% scale; end colors include values beyond the limits, with exact values in the regional panel and downloads. A 95% interval uses coefficient covariance from 400 resamples of three-calendar-year blocks, transformed from log response to percent. These pointwise intervals have no field multiplicity adjustment and do not measure future yield variability. Positive holdout skill means smaller squared log-yield errors than a trend-only model under both five-year block layouts, with adjacent years excluded from training. Neither filter establishes causality.

Sources, geography, and downloads

We use HarvestStat Asia v0.1 and v0.2. Reporting units are stable historical groups; a group can contain several present-day districts or provinces. Boundaries are simplified for display. We preserve the original units and observations in the source files. The combined rice view converts weights for comparison. A constant conversion changes the log-yield intercept but leaves ENSO percentage responses, their intervals, and validation scores unchanged. No national-only source records are included in this release.

Eligible records require complete production and area reporting, positive area, and finite nonnegative yield consistent with production divided by the stated area. We retain valid zeros, apply no minimum record length for a descriptive average, and exclude source-coded missing values. The verified East Java 2017 correction is applied to paddy rice. Vietnamese Summer Rice labels are joined to Autumn Rice as documented in the source analysis.

The 0.67 milled/paddy factor is a fixed harmonization assumption, used in FAO’s 2013 rice analysis. IRRI describes variation in recovery across milling systems; we do not infer a country-specific or time-varying factor. Conversion metadata.

India: Foley et al., StableBound; boundaries from Geolocet, redistributed with attribution under the supplied permission. Other statistical boundaries: FAO GAUL, CC BY 4.0 per the supplied packages. World outlines: Natural Earth, public domain. Map rendering: MapLibre GL JS 5.20.0, BSD-3-Clause.

v0.1 source notes · v0.2 source notes · Coverage catalog · Release checksums · Data contract & reproduction