The effect of spatial autocorrelation on the sampling design for thematic map accuracy assessment.

摘要

The sampling design, including sampling methods, size, and location, is one of the main concerns in the accuracy assessment of thematic maps. Spatial autocorrelation is an important factor impacting on sampling designs and needs careful consideration in accuracy assessment. The objective of this study is to investigate the effect of spatial autocorrelation on sampling designs for accuracy assessment.; In this thesis, four thematic maps (500 X 500 pixels) with combinations of two spatial autocorrelation levels (high and low) and two class proportion differences (90/10 and 60/40) were generated to study the effect of spatial autocorrelation and class proportion on sampling designs. A series of eleven sample sizes (from a minimum of 25 to a maximum of 1296) were simulated using three popular sampling designs, including simple random sampling (SRS), systematic sampling (SYS), and stratified random sampling (StrRS) on the four simulated maps. The conventional error matrix and related measures were calculated for each simulation, and precision of estimating different measures was compared among the three sampling designs.; The simulation study showed that recommending the use of a particular sampling design depends on the spatial autocorrelation level, class proportion difference, and the accuracy index that a given application requires. In general, the class proportion difference has more impact on the performance of different sampling than the spatial autocorrelation on a map. For estimating the accuracy of individual classes, StrRS achieved better precision than SRS and SYS in most cases, especially for estimating the small class. For estimating the overall accuracy, different sampling designs achieved very similar precision. If a better estimate of the kappa coefficient is required, StrRS is recommended on maps with high class proportion difference, while SRS is preferred for maps with low spatial autocorrelation and low class proportion difference.