French crop yield, area and production data for ten staple crops from 1900 to 2018 at county resolution

Bernhard Schauberger, Hiromi Kato, Tomomichi Kato, Daiki Watanabe, Philippe CiaisScientific 2022 · DOI

TL;DR

This paper introduces a comprehensive, county-level dataset of French crop yields, area, and production for ten staple crops spanning 1900-2018. This long-term, high-resolution data is crucial for understanding agricultural performance drivers, projecting future yields under climate change, and developing adaptation strategies.

What Problem Does This Paper Address?

Ensuring global food security in the face of climate change, a growing population, and evolving dietary patterns is a significant challenge. To effectively address this, researchers need detailed historical data on agricultural performance to understand how environmental conditions, management practices, and economic factors influence crop yields. However, such data, particularly with high spatial and long-term temporal resolution, has been scarce. This paper tackles this gap by providing a meticulously compiled dataset for France, a major European agricultural producer, covering over a century of crop data at the county (département) level.

Methods

The core of this work involved the painstaking manual digitization of historical agricultural statistics from French statistical yearbooks for the period 1900-1988. For the years 1989-2018, data was sourced from digital archives. A key methodological decision was to calculate yields (in tonnes dry mass per hectare) from production and area data, as printed yield figures in older records were often found to be erroneous. The researchers also implemented a rigorous outlier filtering process, employing multiple statistical criteria (thresholds, top percentile removal, variance-based filtering) to ensure data quality and consistency.

Datasets

The primary dataset introduced in this paper is the "French crop yield, area and production data for ten staple crops from 1900 to 2018 at county resolution." This dataset comprises:

  • Crops: Ten staple crops including barley, maize, oats, potatoes, rapeseed, sugarbeet, sunflower, durum wheat, soft wheat, and wine. Some crops are further categorized by seasonal type (spring/winter), resulting in 18 crop-cultivar types.
  • Spatial Resolution: French départements (counties, NUTS3 level).
  • Temporal Resolution: 1900-2018 (1900-2016 for wine).
  • Data Points: A total of 375,264 raw data points, reduced to 120,942 yield, 127,344 area, and 126,978 production entries after filtering.

The dataset was validated against national data from the Food and Agriculture Organization of the United Nations (FAO) for the period 1961-2018, showing very high correlations (Pearson's r > 0.95, often > 0.99) for area, production, and yield.

Key Insights and Ideas

  • The Power of Historical Data: This work underscores the immense value of long-term, high-resolution historical data for understanding complex agricultural systems and their responses to various drivers.
  • Data Curation is Crucial: The manual digitization and rigorous filtering process highlight the critical role of meticulous data curation in producing reliable scientific datasets, especially from historical sources.
  • County-Level Detail Matters: Providing data at the county level allows for a more nuanced analysis of regional agricultural performance and the identification of localized trends and vulnerabilities, which might be masked in national-level aggregations.
  • Yield Calculation from Production/Area: The decision to recalculate yields from production and area data, rather than relying on potentially erroneous published yield figures, is a significant methodological improvement for historical agricultural data.

Key Results

The paper presents a comprehensive dataset containing over 120,000 filtered data points for yields, and similar numbers for area and production, across ten crops and 96 French départements from 1900 to 2018. Technical validation against FAO data demonstrated high consistency, with Pearson's r correlations for aggregated national time series exceeding 0.95 for all metrics and most crops. The dataset includes both raw and filtered versions, with the filtered data having undergone extensive outlier removal.

Limitations and Future Work

  • Irrigation Data: The dataset does not distinguish between rainfed and irrigated yields, as information on irrigation area or usage was not consistently recorded in the historical handbooks. This could be a limitation for analyses specifically focused on water management impacts.
  • Data Availability: While extensive, the data for wine only runs until 2016.
  • Future Work: The authors suggest that this dataset can contribute to enhanced understanding of agricultural performance, enabling better detection of hazards, more accurate projections of future yields under climate change, and the derivation of effective adaptation options. The availability of the data and associated R code encourages further research into French agricultural history and its drivers.

Summary updated

Abstract

Agricultural performance is influenced by environmental conditions, management decisions and economic circumstances. It is important to quantify their respective contribution to allow for detecting major hazards to production, projecting future yields under climate change and deriving adaptation options. For this purpose, time series of agricultural yields with high spatial and long-term temporal resolution are a primary requisite. Here we present a data set of crop performance in France, one of Europe's major crop producers. The data set comprises ten crops (barley, maize, oats, potatoes, rapeseed, sugarbeet, sunflower, durum wheat, soft wheat and wine) and covers the years 1900 to 2018. It contains harvested area, production and yield data for all 96 French départements (i.e. counties or NUTS3 level) with a total number of 375,264 data points. Entries until 1988 have been digitized manually from statistical yearbooks. The technical validation indicates a high consistency of the data set within itself and with external resources. The data set may contribute to an enhanced understanding of the manifold influences on agricultural performance.
BibTeX
@inproceedings{schauberger2022french,
  author = {Bernhard Schauberger and Hiromi Kato and Tomomichi Kato and Daiki Watanabe and Philippe Ciais},
  title = {{French crop yield, area and production data for ten staple crops from 1900 to 2018 at county resolution}},
  booktitle = {Scientific},
  year = {2022},
  doi = {10.1038/s41597-022-01145-4},
}
Figure