Forest inventory of small areas combining the calibration estimator and a spatial model

The use of satellite images is considered to compute improved weights for field plots when estimating totals of forest variables over a region by weighted sums of plot measurements. It is intuitively appealing, and necessary in growth projections for management planning, that the weight of each plot can be interpreted as the total area of similar forest in the region. This way we can get a sound description for the whole region so that the relations and distributions of all predicted forest variables resemble the true population. Area interpretation is possible if the weights are positive and the same for all target variables. A calibration estimator provides such weights. If we need estimates for small subregions (counties in this study), we should utilize plots outside the current subregion. A spatial variogram model is suggested for computing the variance of the proposed small-area calibration estimator. Kriging provides optimal weights under such model, but the area interpretation for weights would not be possible. Estimation of county results using data from the Finnish National Forest Inventory and Landsat TM satellite showed that (i) outside plots should be utilized from a constant area around the county, i.e., the inclusion range should decrease when the size of the county increases, (ii) the combined use of neighboring plots and satellite data may lead to a large reduction in the error variance for small counties. In its current form, the method does not produce predictions for individual pixels.