Remoteness, although a subjective concept, has indispensable consequences. It can support decision-makers in quantifying risks and feasibilities of developing newly discovered mineral deposits, resilience planning and evaluation of accessibility challenges of remote communities, or support agency budget allocations for much-needed services. Developing a remoteness index typically involves merging spatial and temporal data from a variety of incompatible sources such as topographical, census, and travel cost and duration. This paper presents a novel method for generating a measure of remoteness for any geographical location based on the nighttime satellite imagery. This continuous measure, herein referred to as Nighttime Remoteness Index (NIRI), is generated using machine learning-based models that link the intensity and statistical features of nighttime lights to remoteness; the predictive model is trained and validated using the nighttime satellite imagery and the Accessibility Remoteness Index for Australia (ARIA). This method does not require local data; hence, it is not limited by political jurisdictions or geographic boundaries. The NIRI is developed by using multivariate adaptive regression splines, and support vector machines regressions, after examining several other machine learning techniques. The NIRI maps of Australia and North America are developed based on the validated models.