Photo Credit: Neil Palmer / CIAT

Characterizing global production systems just got easier.

As of 2003 the canned pre-cooked meat product, Spam, was sold in 41 countries on six continents. Hormel Foods Corporation has updated its product in a variety ways, such as adding jalapeños or a hint of hickory smoke. Why are we talking processed meat? Who can mention SPAM without mentioning Spam? (If the name 'SPAM' bothers you, call it MapSpam - it’s the same frontwards and backwards). IFPRI / HarvestChoice has also updated its product, SPAM, from circa 2000 to centering on the year 2005. That’s right; SPAM 2005 is out for release, downloadable, and freely available to the public through our website (see harvestchoice.org/products/data ). That means more accurate and up-to-date spatially disaggregated crop production statistics and more of it.

For the past ten years the International Food Policy Research Institute (IFPRI) / HarvestChoice has developed and applied an approach to spatially disaggregate (downscale) crop production statistics, dubbing their product the Spatial Production Allocation Model (SPAM). SPAM generates spatially-disaggregated, crop and production system specific area, production and yield estimates using a triangulation approach across a broad set of relevant background data. The detailed spatial datasets represent a rich platform for exploring the social, economic, and environmental consequences of agricultural production at a highly disaggregated scale. The SPAM model and its outputs have become a critical input to many initiatives within and beyond the CGIAR, including several high-profile agricultural and environmental modelling initiatives (for example, GEOGLAM and IMPACT).

What’s new with SPAM two (thousand five)?

SPAM 2005 runs the model for 42 crops, whereas SPAM 2000 treated just 21 crops (including a crop category called ‘other crops’ which lumps together the rest of the crops not covered in the other 20). The total number of crops in SPAM 2000 account for at least 80% of the earth’s total crop production as reported by FAO, but in SPAM 2005 it’s closer to 100% (fodder crops are excluded). Typically national institutions publish their country’s production statistics for administrative levels immediately below national (region/province), but for some countries data is also available at finer scales - at two or more levels below national (district/county). A more detailed crop statistic guarantees a more accurate downscaling.

The 42 crops in SPAM 2005 resulted from disaggregating some SPAM 2000 crops like ‘other pulses’ and ‘other oils’, as well as by adding more cash crops, fruits and vegetables, which were considered under ‘other crops’ in SPAM 2000, and not directly modelled but calculated as a residual of the 20 crops being allocated. The 42 crops are listed below.

We are currently working on a second version of SPAM 2005 which will incorporate feedback from collaborators who have had access to a preliminary version of the data. In the meantime, since we update SPAM every five years, we are already starting to assemble subnational statistics for SPAM 2010. Stay tuned!

Should we compare SPAM 2000 and SPAM 2005 results?

No, these should not be compared nor considered time-series data for the following reasons:

  • Crop-land information for both versions is not compiled by the same method and has different base years
  • The irrigation maps are different
  • Crop suitability areas and yields have been updated in 2005 as compared to 2000
  • Sub-national statistics are more recent and the crop distribution/mix of a country may have changed
  • The allocation methodologies have been improved
  • SPAM 2005 has 42 explicitly modeled crops, whereas in SPAM 2000 there were 21

How we assimilate SPAM: A Cross Entropy Approach

After collecting national and sub-national statistics on individual crops for each country, more information is folded into the data:

  • Granular information on cropland from satellite imagery
  • Potential cropping areas and potential yields for individual crops
  • Irrigated areas
  • Knowledge regarding four distinct production systems (irrigated high inputs, rain-fed high inputs, rain-fed low inputs, and subsistence)
  • Producer prices

SPAM assesses the crop production of each crop in each production system per pixel (the 5 arc min resolution units which in this case amounts to about 9 x 9 km grids at the equator and increasingly less the further a location moves towards the poles).

To allocate production of 42 crops into each pixel, a cross entropy approach is used. The initial crop areas are broken down by production systems, and the initial crop distributions are fed into a routine of minimization of cross entropies, subject to spatial constraints taken from the granular data.

A scaling process is then applied at the end to ensure that the re-aggregated crop production stats at national level are consistent with FAOSTAT numbers for each crop, pegged to the average 2004-2006 values for area and production.

How to cite SPAM data and methodology

SPAM’s methodology is described at length in the 2014 article below. The origin of the ancillary data sources are also found there.

  • Datasets: You, L., Z. Guo, J. Koo, U. Wood-Sichra, Yue Gong. Spatial Production Allocation Model (SPAM) 2005 Version 1. http://HarvestChoice.org (Accessed _____)
  • Article: You, L., S. Wood, U. Wood-Sichra, W. Wu. 2014. Generating global crop distribution maps: From census to grid. Agricultural Systems 127 (2014) 53-60

The 42 Crops of SPAM 2005

Wheat, rice, maize, barley, pearl millet, small millets, sorghum, other-cereals, potato, sweet potato, yam, cassava, other-roots-and-tubers, dry bean, chickpea, cowpea, pigeon pea, lentil, other-pulses, soybean, groundnut, coconut, oil palm, sunflower, rapeseed, sesame seed, other-oil-crops, sugarcane, sugar beet, cotton, other-fibres, arabica coffee, robusta coffee, cocoa, tea, tobacco, banana, plantain, tropical fruit, temperate fruit, vegetables and other-crops.

Citation

HarvestChoice, 2014. "What's New with SPAM Two (Thousand Five)." International Food Policy Research Institute, Washington, DC., and University of Minnesota, St. Paul, MN. Available online at http://harvestchoice.org/node/9613.

Jul 17, 2014