Quantifying Spatial Variability Using Semivariograms : A geostatistical approach for the determination of the spatial variability of remotely sensed water color images as a function of pixel size

Natural processes occurring on the surface of the earth manifest themselves at multiple spatial scales. The amount of spatial variability revealed from remote sensing images is strongly dependent on the scale of observation. Each remote sensing image is acquired at a given spatial resolution, revealing only a limited range of the existing natural variability. The simple method presented in this book was developed for the quantification of the loss of fine-scale spatial detail in low-resolution ocean color images acquired over highly patchy coastal waters. The method can however be used for the analysis of targets other than water as well as in other types of applications like the determination of the optimum scale of observation for a field sampling campaign, the design of the spatial aspects of a new sensor, study of scale-change effects in multi-scale/multi-sensor projects and possibly in many others. The current study was accomplished in 2000 for the fulfillment of a Masters of Science degree at the Remote Sensing department of Universit de Sherbrooke under the supervision of Dr. Norm O'Neill and Jean-Claude Therriault.