Open-Earth-Monitor Global Workshop 2026

Statistical modeling and uncertainty analysis plays a critical role in evaluating climate and environmental data. Concepts such as standard error of the mean and design-based estimation seem to be increasingly used to manipulate prediction errors and tradable changes. Advanced trend estimation and change-point models are essential for accurately identifying long-term shifts in essential climatic variables such as soil organic carbon and above ground biomass. Subtracting two above-ground biomass (AGB) maps can create false data because map uncertainties propagate into the difference, compounding the errors from both individual maps and inflating apparent change signals. Rather than revealing true environmental dynamics, naive subtraction often produces an apparent "change" that is actually just statistical noise. Quantile Regression Random Forests (QRRF) offer a powerful, non-parametric approach to estimating the true distribution of errors by retaining all observations within the terminal leaf nodes of the forest, rather than just calculating the conditional mean. This allows the model to estimate the full conditional cumulative distribution function and extract specific percentiles to form prediction intervals. We demonstrate how this method can be used to determine tradable carbon sequestration without taking additional risks.