An investigation of sample survey methodologies, including the application of remote sensing, for the production of agricultural statistics in Tanzania.

One of the many applications of remote sensing data, acquired by earth-orbitting resource satellites on a regular basis, is the estimation of crop areas in agricultural statistics. The evaluation of crop area statistics for traditional farming in sub-Sahara African developing countries using the rapidly advancing remote sensing data is aptly under consideration.

Because of operational problems in the developing countries in acquiring and exploiting images of all areas of interest, an approach has been proposed that combines remote sensing data, Landsat and SPOT, and ground sampling data in a flexible manner depending on materials, equipment available, expertise and experience of personnel involved. Under the prevailing circumstances of rather limited high-tech resources in the region, the visual interpretation of imagery has been used and emphasized.

A ratio-type estimator for the mean has been proposed using information on auxiliary variables for improvement at the estimation stage. The exact as well as asymptotic bias and variances of the estimator have been obtained. Comparison on infinite and finite populations with several other ratio-type estimators is done on the properties of bias and efficiency. The simplicity and suitability of the estimators in practical situations were also taken into consideration. The performances of the estimators have been examined using both hypothetical values and the data collected in 1990 in the field work performed during the study period.

Attention is turned to various methods of attaining yield rate. A crop-cut method and farmer’s statements are discussed and tests performed on them with the Tanzanian agricultural sample survey data of 1987/88. Finally a production estimate, that could easily be extrapolated over larger area, is obtained from the product of area estimate and yield rate as a result of integrating satellite data with the field ground sampling and farmers’ interview data respectively.