Geostatistics as a tool to reduce the sampling effort in forest inventories

Geostatistics is one of the tools applied to investigate the spatial variability of forests to reduce costs and recognize the best productivity areas for planning. This study aimed to test the performance of geostatistical techniques in reducing the sampling effort in forest inventories. For this purpose, we used the height of dominant trees as a discriminator of the homogeneous strata to obtain a better representation of the productivity within the forest stands. We carried out the study in Pinus taeda L. stands in the Center-South of Paraná, Brazil, by using plots from a forest inventory allocated with the systematic process. Then, we tested three models to determine the site curves (Schumacher, Chapman-Richards 2, and 3 coefficients) with the thirty-seventh year being the reference age. To model the spatial patterns of the dominant height, we used the ordinary kriging, and, after that, we generated the thematic maps of the site classes. Similarly, we used the indicator kriging which allowed obtaining the probabilities of high, medium, and low productivity sites. The processing of the stratified sampling, with the support of the visual interpretation of the images, allowed us to define five strata according to productivity. Results showed that ordinary kriging is effective in defining the productivity classes. Along with geostatistical techniques, it produces more homogeneous strata and reduces the errors of the forest inventory. Moreover, the best-selected model was the Chapman-Richards (3 coefficients) for the site curves. The exponential model was the best model to identify the best areas of the probability of occurrence of sites with higher productivity. The efficiency of indicative kriging generated thematic maps to delimit the likely locations of the most promising sites. Overall, geostatistics proved to be efficient concerning error when compared to simple random sampling.