Optimal information acquisition under a geostatistical model

Studies examining the value of switching to a variable rate technology (VRT ) fertilizer program assume producers possess perfect soil nitrate informat ion. In reality, producers estimate soil nitrate levels with soil sampling. The value of switching to a VRT program depends on the quality of the esti mates and on how the estimates are used. Larger sample sizes, increased spa tial correlation, and decreased variability improve the estimates and incre ase returns. Fertilizing strictly to the estimated field map fails to accou nt for estimation risk. Returns increase if the soil sample information is used in a Bayesian fashion to update the soil nitrate beliefs in nonsampled sites.