High-uncertainty regions for ECe and ESP mapping

To identify regions of high uncertainty, we first calculated the standard deviation of predictions from all individual trees in the trained Random Forest models for the year 2005. We then defined the 90th percentile of these values as the uncertainty threshold. Next, we computed the distance from each pixel to the nearest sample point and applied a threshold of 100 km. The final high-uncertainty areas were determined by combining the results of these two criteria.