From one- to two-phase sampling to reduce costs of remote sensing-based estimation of land-cover and land-use proportions and their changes

Abstract

Abstract The estimation of the proportion of land-cover and land-use classes is considered by exploiting remote sensing-based imagery. A pure-panel survey based on point sampling is adopted. An initial sample of points is selected by means of tessellation stratified sampling (TSS). The sample points are classified based on the imagery available for the years of interest to estimate land-cover or land-use proportions and their changes. To reduce the sampling effort, the initial selection of points is viewed as the first phase of sampling and a subsample of these points is selected in a second phase by means of one-per-stratum stratified sampling (OPSS). Land-use estimation at any subsequent year is then based on the classifications performed on the points of the second-phase sample. Two-phase estimators of proportions and of their changes are suggested, and their theoretical properties are derived. Presumably conservative estimators of their variances are proposed. A check of the precision lost involved when changing from one- to two-phase sampling is determined from the assessment of land use in Italy as a case study and from an artificial population generated to resemble the current land-use situation in Italy.