Main Contributors: Wolfgang Britz and Gunter Wriedt
GIS processing of MARS yields: Renate Koble and Giulio Marchi

The crop yields are estimated based on estimates of water limited and non-water limited yields, courtesy of the JRC's MARS project, and per grid cell reported as the average of the water limited yield times unity minus the estimated irrigation share for that crop in the pixel cluster plus the non-water limited yields times the estimated irrigation share. A statistical estimator ensures that the average yields as reported in regional agricultural statistics from Eurostat are recovered.

The irrigation shares and yields per crop and 1×1 km cell cluster are statistically estimated based on a Highest Posterior Density estimator, which ensures the following conditions:

• The total irrigated area at NUTSIII/NUTSII region reported in the 1999 Farm Structure Survey is recovered
• The area weighted average yield per 1×1 km grid cell and crop is identical to the yield found in regional statistics
• Average yield per 1×1 km grid cell and crop are a weighted average of the non-irrigated and irrigated potential yields from MARS, multiplied with an identical scaling factor

Given the fact that there are more variables to estimate (irrigation shares per crop and HSMU, scaling factors for yields per region), the statistical estimator picks the most probable combination based on minimizing deviations from given information:

• The reported or average irrigation shares at NUTS II level for selected crops. Where no shares had been reported, those had been estimated based on regression models based the reported average irrigation shares, and using climate data as explanatory variables.
• The average irrigation shares per HSMU, defined from the global irrigation map of FAO
• A preferred scaling factor of unity for the crop yields

Britz W. and Wriedt G.:
Pan-European Estimation of crop specific yields and irrigation shares at 1x1 km grid, JRC, 2007 (pdf, foils)

Last Updated:Tuesday, October 28, 2008