Global map of peak years

Map of countries peak-rate year and percentage increase of production at peak rate year

What is displayed?

Either for all crops combined or for a selected crop, each country is potted in a different color (yellow: there is a peak-year, orange: no peak year, gray: no sufficient data, mostly because crop not planted). Additionally, colored circles illustrate the year, when production has had a peak (color) as well as the rate of production increase for that peak year (size of circle)

What can be changed interactively?

On default the peak years of all crop combined based on their caloric value are shown. If in the panel to the left a certain crop is selected, the map changes accordingly.

When selecting a certain crop, also the info-box to the left is updated an the peak-rate year of that specific crop is updated.

Forest plot of peak years

Forest plot showing the estimated peak year and its uncertainty for each crop that is planted in the regions of interest (global or selected country)

What is displayed?

For each crop the median peak rate year of production is displayed, if identified. Uncertainty bars represent 2.5th and 97.5th percentiles derived from 5,000 bootstrap estimates (see Data & Methods). Point size illustrate the relative increase of production in percent in that peak rate year. Note, due to bootstraping the resulting distribution of peak-rate years can be skewed and the median peak rate year might not necessarily be centered in the confidence interval.

What can be changed interactively?

On default the peak years of crops of all countries (global) is shown. If a country is selected, the figure shows the crops that (a) are produced in the county and (b) that show a peak rate year.

Note, when selecting a country also the histograms in the bottom row changes and shows the total area harvested for all crops in that country (bottom left) as well as the total calories produced for each crop in that specific country (bottom right). Also the info-box to the left is updated and the peak-rate year of crop production in that country is displayd, summarized for all crops combined based on the caloric value.


Map of countries’ stability

Country-specific stability of the total caloric production or of the crop-specific caloric production. Stability is calculated as the mean of the non-time-detrended production divided by the standard deviation of time-detrended production for ten year time periods (see Data & Methods for details).

What is displayed?

Country colors indicate the stability of agricultural production estimated in mean/CV, CV denoting the coefficient of variation.

The size of the circles is relative to the total caloric production of the country.

What can be changed interactively?

Per default, stability of the total caloric production for the most recent time period is shown. In the window on the left, different crops and time periods can be selected. The map will then be updated accordingly.

Stability Histogram

This shows global or country-specific stability of the caloric production of different crops. Stability is calculated as the mean of the non-time-detrended production divided by the standard deviation of time-detrended production for ten year time periods (see Data & Methods for details).

What is displayed?

Each crop is listed on the y-axis and the x-asis shows the production stability in 1/CV, CV denoting the coefficient of variation (see Data & Methods).

What can be changed interactively?

Per default global stability for the most recent time period is shown. Time period can be changed with the time slider in the left panel. In the window on the left, the user might select different countries and time periods. The bar plot will be updated accordingly.

Note, when selecting a country also the histograms in the bottom row are updated. The total area harvested for all crops in that country (bottom left) as well as the total calories produced for each crop in that specific country (bottom right) are displayed.


Map of asynchrony between crops

Country-specific asynchrony between crops. Asynchrony between crops describes the asynchrony of the year-to-year caloric production of different crops within a given country. Asynchrony between crops was calculated based on the time-detrended total caloric production for ten year time periods (see Data & Methods for details).

What is displayed?

Country colors indicate the asynchrony between crops.

What can be changed interactively?

Per default asynchrony between crops is shown for the most recent time period. In the window on the left, different time periods can be selected. The map will then be updated accordingly.

Histogram of asynchrony within crops

Global asynchrony within crops of different countries. Asynchrony within crops describes the asynchrony of the year-to-year caloric production of the same crop at different countries. Asynchrony between crops was calculated based on the time-detrended crop and country-specific caloric production for ten year time periods (see Data & Methods for details).

What is displayed?

Asynchrony within crops (x-axis) is shown for each crop (y-axis).

What can be changed interactively?

Per default, global asynchrony within crops is shown for the most recent time period. In the window on the left, different time periods can be selected and the bar plot will be updated accordingly.


Methodology behind estimating peak rate years

Peak-rate year estimation

We use a method that is standardized, nonparametric, generalizable, and allows analysis of nonrenewable and renewable resources (Seppelt et al. 2014). In contract to the original publication, we here did not make use of the estimation of peak rate year for non-renewable resources. To estimate a peak rate year, the maximum increase of a time series must be calculated, see Figure below.

Peak Year Concept

We specifically accounted for non-stationarity of parameters of production increase. Time dependency (e.g. non-stationaritry) of production increase might due to different innovations being available to countries at a certain time. Non-parametric curve fitting offers advantages to account for that process, and does not require parametric assumptions or that a functional model be postulated (e.g. stationarity of rate of resource appropriation). This means that the different resources and drivers need not follow the same increase process (Gasser 1984).

By using a bootstrap resample to estimate the uncertainty of the peak-rate year estimate, we avoid distributional assumptions. First, we divided each time series by its maximum value to scale the values between 0 and 1. This does not affect the estimate of the year when the maximum increase of production occurs, but it allows the estimation of the relative growth rate making the different time series comparable. Then, 5,000 bootstrap samples were obtained, leaving approximately 30-50% of the data out per sample. For each bootstrap sample, we fitted a cubic smoothing spline with a smoothing coefficient selected by generalized cross-validation. This resulted in limited smoothing, so the estimate of peak-rate year is minimally biased (but with greater variance). The peak-rate year is the year of the maximum of the first derivative of the cubic spline smoother. The 50th percentile (2.5-97.5th) of the 5,000 bootstrap resamples was taken to be the peak-rate year with the uncertainty. The number of bootstrap resamples was sufficient to ensure that stable estimates were obtained. When the estimated peak-rate year was equal to the last year in the time series, we concluded that the peak- rate year had not passed. This was performed in the software R (R Core Team 2013).

Sources of uncertainty

First, uncertainty of the identified peak-rate years is due to the limited length of the time series. A shorter time series increases the variance of the estimators as the sample size is reduced. In the bootstrap resample approach taken here a random sample, smaller than the original data series, is drawn which increases further the uncertainty surrounding the peak-rate year. Time series with several minima and maxima will result in large uncertainty of the peak-rate year estimate, an adequate representation of the noise in the time series.

Second, technological and demand shifts make the time series non-stationary. Hence, we do not make model-based predictions (which would assume stationarity for statistical modeling) of rate of resource use, but detect peak-rate years in-sample for independent resources over a period where the demand for resources is not expected to decrease. New data in the future may shift the peak-rate year.

Third, there is equifinality in the interpretation of a peak-rate year. For example, a peak-rate year may be detected because of the joint effects of scarcity, structural changes such as innovations, and changes in regulation in various regions. The interaction pattern cannot be inferred directly from the data. Fourth, while this study focuses on peak-rate year at the global level, a regional specific analysis may provide a different picture and more specific information on regional action.

References

  • Gasser, T., H.-G. Müller, W. Köhler, L. Molinari, and A. Prader. 1984. Nonparametric regression analysis of growth curves. Annals of Statistics 12:210-229. 10.1214/aos/1176346402
  • R Core Team. 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria.
  • Seppelt, R., A. M. Manceur, J. Liu, E. P. Fenichel, and S. Klotz. 2014. Synchronized peak-rate years of global resources use. Ecology and Society 19:50. 10.5751/ES-07039-190450.

Methodoloy of stability and asynchrony calculation

Stability

To calculate production stability we used agricultural production data from the FAOStat database, see tab “Data Sources”. We converted crop-specific production from tons to calories using standardized nutritive factors. We only included crops, for which nutrient data could be clearly assigned. For each 10 year time interval, we only included crops for which time series were complete and where both production and harvested area were reported. For each unit and year, we then summed calorie production of all crops. To account for temporal production stability independent of long-term trends, we time-detrended annual production data by regressing annual total calorie production on year squared for each time interval and region (Renard and Tilman 2019). We calculated temporal production stability as the mean of the non-time-detrended production divided by the standard deviation of time-detrended production for each time period following Renard and Tilman (2019) and Mehrabi and Ramankutty (2019).

Asynchrony

To calculate national asynchrony between crops, we first time-detrended annual production data of each crop by regressing annual calorie production on year squared for each time interval and region (Renard and Tilman 2019). We then used this data to calculate synchrony for each time interval and region following Loreau and De Mazancourt (2008) and Mehrabi and Ramankutty (2019) with the ‘codyn’ package (version 2.0.3) in R (Hallett et al. 2016). Synchrony relates the total variance of the time-detrended national production to the sum of the respective variances within each crop. We then subtracted synchrony from 1 to receive asynchrony.

Likewise we calculated global asynchrony within crops regarding individual crops. This calculation was based on the total variance of the time-detrended global production relative to the sum of the respective variances within nations.

Further methdological details can be found in Egli et al. (2020; 2021).

References

  • Hallett, L. M. et al. 2016. codyn: An R package of community dynamics metrics. Methods in Ecology and Evolution 7: 1146–1151. 10.1111/2041-210X.12569.
  • Loreau, M., and C. De Mazancourt. 2008. Species synchrony and its drivers: Neutral and nonneutral community dynamics in fluctuating environments. American Naturalist 172: E48-E66. 10.1086/589746.
  • Mehrabi, Z., and N. Ramankutty. 2019. Synchronized failure of global crop production. Nature Ecology & Evolution 3: 780–786. 10.1038/s41559-019-0862-x.
  • Renard, D., and D. Tilman. 2019. National food production stabilized by crop diversity. Nature 571: 257–260. 10.1038/s41586-019-1316-y.
  • Egli, L., et al. 2020. Crop asynchrony stabilizes food production. Nature 588: E7–E12. 10.1038/s41586-020-2965-6.
  • Egli, L., et al. 2021. More farms, less specialized landscapes, and higher crop diversity stabilize food supplies. Environmental Research Letters 16(5): 55015. 10.1088/1748-9326/abf52.

Data Sources

Agricultural Production

All analysis base on the FAOStat database, which is maintained and regularly updated by FAO with support from the Member States.

Crop selection

To exclude crops of minor relevance, we sorted them by the average harvested area over the last ten years in descending order and included crops up to a cumulative harvested area of 90% of the total harvested area of all crops.

Country selection

To exclude countries where crop cultivation is of minor relevance, we sorted them by the average harvested area over all included crops and the last ten years in descending order and included countries up to a cumulated harvested area of 99% of the total harvested area of all countries.

Total caloric value

We used the total caloric value from FAO to estimate the variable ‘All crops combined’ by summing up the total calories produced by countries and years. See table:

Results

Availability

Results on peak rate years for each crop and country, asynchrony for decadal periods for crops and countries as well as stability for all crops combined on a decadal bases can be accessed and downloaded. Data is curated at Open Science Foundation and is updated on an annual bases shortly after FAOStat is providing its annual update. Please refer to the note on the bottom of the left column that proved the date of the last download and update of results.

Referencing Methods

For referencing methods or original analysis please cite one or all of the following references.

  • Egli, L., Schröter, M., Scherber, C., Tscharntke, T. & Seppelt, R. 2020. Crop asynchrony stabilizes food production. Nature 588: E7–E12. 10.1038/s41586-020-2965-6
  • Egli, L., et al. 2021. More farms, less specialized landscapes, and higher crop diversity stabilize food supplies. Environmental Research Letters 16(5): 55015. 10.1088/1748-9326/abf52
  • Seppelt, R., A. M. Manceur, J. Liu, E. P. Fenichel, and S. Klotz. 2014. Synchronized peak-rate years of global resources use. Ecology and Society 19: 50. 10.5751/ES-07039-190450.

Results Data

The following table provides access to the most recent versions of the dashboard’s data.

Referencing Data

All results including versions from previous years (see version tracking) is available at osf.io at the repository StabilityPeakAgricultureData.

If needed this can be referenced as Seppelt, R., & Egli, L. (2021). StabilityPeakAgricultureData. 10.17605/OSF.IO/SKD5P.

Contacts

Concept and design

Dr. Ralf Seppelt, Deptartment Computational Landscape Ecology, UFZ, Leipzig, Germany, e-mail

Dr. Lukas Egli, Deptartment Computational Landscape Ecology, UFZ, Leipzig, Germany, e-mail

Peak year analysis

Dr. Ameur Manceur, Deptartment Computational Landscape Ecology, UFZ, Leipzig, Germany, e-mail

Stability & asynchrony analysis

Dr. Lukas Egli, Deptartment Computational Landscape Ecology, UFZ, Leipzig, Germany, e-mail

RShiny implementation

Emmanuel Adeleke, M.Sc, Faculty of Biology, Chemistry and Earth Sciences, University of Bayreuth, Germany, e-mail

Copyright © 2021 UFZ, Ralf Seppelt, Lukas Egli

SOFTWARE AND DATA ARE PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Imprint

UFZ imprint applies here, see Imprint