Remote sensing as a tool for agricultural statistics: a case study of area frame sampling methodology in Hellas

The 10-year Monitoring Agriculture with Remote Sensing (MARS) project, launched by EU in 1989, initiated several studies with the aim of providing technical support to the European Agricultural Guidance and Guarantee Fund (EAGGF) and helping the Member States' Administrations to improve and industrialise their methods using remote sensing technology. One of the main targets of the project is to provide the framework for implementation of this technology assessing the use of high resolution satellite images as part of the measures to be taken by the national governments to improve the ground survey estimates based on an area frame rather than a holding-based approach. The area frame methodology is primary used for crop area, and yield or production estimates. The statistical units (the segments) of an area frame are directly bound to a stratified geographical region, the limits of which are known in advance. Thus, the elements of the population (the frame) are also known. Since the population is stratified no segment may be shared by two or more strata. Samples are obtained by dividing a region into blocks of equal (square) shape and repeating a pattern of elements across the region. In this context, two-dimensional systematic aligned sampling methodology, with a distance threshold in a stratified area frame on a square grid, is developed. The main purpose of this paper is to review the area frame of square segments methodology when it is applied in combination with satellite imagery (also known as supervised classification), and to provide the main features that appear when it is used in agricultural statistical surveys. Further, it is to report results of area estimates obtained from the implementation of this fast acquisition statistical data methodology, using Landsat-tm images in certain cultivated Hellenic areas, particularly productive in soft and durum wheat, maize, cereals, sugar beets, cotton, tobacco, olives trees and vines. Although the results show that there is some improvement in using the supervised classification methodology, a revised stratification methodology is proposed and a new sample is extracted for the Hellenic regions of Macedonia and Thrace, using no satellite data. The new classification is simpler, easier and less costly to implement than the one that is in current use. The developed regression model provided more accurate and viable acreage estimates than a previously applied model, and it may be extended to all cultivated Hellenic regions. Finally, the acreage, yield and production estimates obtained from the supervised classification methodology are compared with those obtained from the rapid estimates methodology which was also developed in the framework of the MARS project and is reviewed here. This comparison shows a noticeable agreement in the results obtained. (C) 1998 Elsevier Science B.V. All rights reserved.