Optimal Sampling Scheme for Estimation Landscape Mapping of Vegetation Cover

Abstract

An optimization of a sampling design aims at decreasing costs without losing necessary spatial information and desired precision for estimation and mapping of vegetation cover. This study concentrates on investigating optimal solutions for sampling design, considering both plot and sample size in terms of cost and variance estimated for global estimation and local landscape mapping of overall vegetation cover used in the management of soil erosion. A geostatistical method was developed based on regionalized variable theory and compared to a classical random sampling method for a case study in which optimal sampling was designed for estimating and mapping vegetation cover. Cost is introduced into the sampling design in terms of measurement time. This method has made it possible to seek optimal solutions for determining plot and sample sizes given a desired precision and allowable survey cost budget for both local and global estimation. The results show that the geostatistical method is more cost-efficient than the classical designs because it accounts for spatial dependence of variables in the sampling design. Moreover, plot size affects kriging standard error of the local estimate more significantly than sample size, while sample size has more effect on precision of the global estimate than does plot size.