Machine learning is applied to Earth Observation (EO) data to derive data sets thatare used to characterise, comprehend and conserve natural resources, contributingto progress towards international accords. However, the derived datasets containinherent uncertainty and need to be quantified reliably to avoid negative downstreamconsequences. In response to the increased need to report uncertainty,we bring attention to the promise of conformal prediction within the domain ofEO. Conformal prediction is an Uncertainty Quantification (UQ) method that offersstatistically valid and informative prediction regions while concurrently beingcomputationally efficient, model-agnostic, distribution-free and can be applied ina post-hoc manner without requiring access to the underlying model and trainingdataset. We assessed the current state of uncertainty quantification in the EO domainand found that only 20% of the reviewed datasets incorporated a degree ofuncertainty information, with unreliable methods prevalent. Next, we introduceGoogle Earth Engine native modules that can integrate into existing predictivemodelling workflows and demonstrate the versatility, efficiency, and scalabilityof these tools by applying them to datasets spanning continental to global scales,regression, and classification tasks, featuring both traditional and deep learningbasedworkflows. We anticipate that the increased availability of easy-to-use implementationsof conformal predictors, such as those provided here, will drivewider adoption of rigorous uncertainty quantification in EO, thereby enhancingthe reliability of uses such as operational monitoring and decision-making.