A New Sampling Strategy for Enhancing Forest Monitoring Leveraging Remote Sensing Data

Nowadays, monitoring forests and supporting decision-makers with reliable data on biomass are more important than ever. Such data are commonly collected at the national level using spatial sample surveys. However, due to the high costs of surveying and processing, sampling intensity is often low, and the estimates’ precision may be unsatisfactory. We propose a new sampling strategy for forest inventories that takes advantage of increasingly accurate and accessible remotely sensed data made available by technological advancements. Our approach includes an assessment of the potential utility of incorporating remote sensing data into the sampling design and incorporates them only if deemed beneficial. This is achieved through a two-step sampling process. Initially, a well-spread smaller sample is drawn to examine the nature and strength of the relationship between the auxiliary and response variables. Then, depending on the observed relationship, the spatial sampling design for selecting the remainder of the sample is adjusted to (possibly) incorporate the auxiliary information. Our proposal’s performance is evaluated through Monte Carlo experiments based on simulated datasets and real-world data that motivated our analysis. Specifically, we focus on estimating the storage of terrestrial carbon in biomass and soil using satellite remote sensing optical data as auxiliary information.