Each year from late May to early June, the Economics, Statistics, and Cooperatives Service (ESCS) of the U.S. Department of Agriculture conducts a nationwide June Enumerative Survey (JES), consisting of interviews with farm operators in randomly selected area-sample units called segments, which are typically one square mile in size. Later in the growing season, crop yields are estimated from biological measurements conducted in small plots of land only a few square meters in size, which have been selected by two-stage subsampling from JES segments. The first stage of the subsampling is for fields planted to the crop of interest; second-stage subsampling is for plots within first-stage selected fields.

In this paper a regression-like estimator is investigated as a method to use LANDSAT data to improve ESCS yield estimates for corn and soybeans. The estimator's primary variable, which is required to be known only for sampled fields, is estimated field-level yield computed from observed plot data. The estimator's auxiliary variables are fieldlevel means of MSS radiometric values and/ or various MSS vegetative indices. By definition auxiliary variables must be known over the entire population, which in this case is all land planted to the crop of interest within some target area. Since a pixel's population membership is not known however, for pixels exterior to JES segments, the set of all pixels classified to the crop of interest is used to define a pseudo-population for the estimator. This creates an estimator bias, which is estimated from labeled LANDSAT data coinciding with JES segments.

Evaluation of the developed estimator with 1978 unitemporal Iowa data produced mixed results in sub-state analysis areas. In some areas of Iowa, no yield estimation improvements from LANDSAT were indicated. In other parts of Iowa, yield estimation improvements were moderate for soybeans and marginal for corn. Haze correction was used to develop entire-state estimators. Entire-state estimation improvements were modest for both corn and soybeans.

Date of this Version