Mapping and Estimation of Carbon Dioxide Storage in Forest Plantations. The Contribution of the Sentinel-2 Time Series in Increasing Estimates Precision

Forest restoration activities and tree plantations play an important role in combating global warming. On the other hand, quantifying their carbon storage is a challenging task due to very short rotations and the effort and costs required for field analysis, often in remote and hardly accessible regions. In this context, remote sensing combined with new cloud computing platforms offers unprecedented opportunities for monitoring tree plantations globally. In this study, we implemented and demonstrated over a 20-ha tree plantation in Guatemala an approach that exploits Sentinel-2 imagery time series derived metrics and cloud-free composites for mapping carbon storage. Ground data were collected over 20 plots (10-m radius) to train and validate our model, which performance resulted in high (R2 = 0.69, RMSE = 35%). Plus, we estimated the amount of carbon stored in the study area and the relative confidence intervals. Using exclusively the ground data, we estimated the average net equivalent CO2 as 4.95 Mg ha−1 ± 0.9 Mg ha−1, with a confidence interval of 95%. Nevertheless, exploiting the herein presented model and statistical procedure, the estimate was much more precise and the ratio between the variances of the design-based and the model-assisted estimates was 7.1, meaning that, by using remote sensing data, it is possible to reduce the ground sample size by a factor of 7.1 while obtaining estimates with the same precision of those do not exploiting remote sensing data. This is a crucial point for meaningful reducing the effort and the cost required for collecting data on tree plantations while still obtaining statistically rigorous estimates.