To estimate the amount of fertilizer applied to each crop in each
country, the fertilizer rates for each crop were multiplied by the
respective crop areas. The expected values and standard errors for
these variables were calculated using national survey data (Defra,
2010a,b; DARDNI, 2010). Where the standard errors were small
compared to the mean (less than 25%) we assumed the uncertainty
was normally distributed, otherwise we assumed a lognormal
distribution. This is because when standard errors become larger,
there is a greater chance of sampling negative values for the variables
(which would not make sense).