Coastal areas have historically been centers of economic activity due to their advantages in maritime trade and access to larger markets. However, the ongoing process of economic globalization raises questions about whether this concentration will persist or if economic activity will increasingly shift towards interior regions. While some researchers maintain that coastal areas will remain dominant, others suggest a potential shift inland. This study aims to address this question by analyzing the spatial distribution and temporal evolution of global economic activity, focusing on the interplay between coastal and inland regions. The research utilizes nighttime light data, a proxy for economic activity, in conjunction with advanced statistical techniques to provide a comprehensive assessment of this dynamic relationship. This study is crucial for informed policymaking and management of both coastal and interior areas, particularly within the context of sustainable and coordinated land and marine development.

The introduction extensively cites existing literature supporting the historical and continuing economic significance of coastal areas, referencing works by Shi and Singh (2003), Barbier et al. (2008), and others. It also acknowledges contrasting viewpoints suggesting a potential inland shift, referencing Yimin and Haochun (2018), Kotkin (2006), and Wang et al. (2020). The supplemental information contains a more detailed literature review on three aspects: the definition of coastal areas, the agglomeration of economic and social elements in coastal zones, and the application of nighttime light data in economic estimations.

The study employs a three-step methodology. First, it validates the correlation between nighttime light data and GDP using a Spearman correlation test, confirming the reliability of nighttime light as a proxy for economic activity. Second, a random forest algorithm is employed to estimate the aggregate GDP within 100 km of coastlines (termed "near regions") globally. This algorithm is chosen for its robustness to missing and unbalanced data and its ability to handle high-dimensional datasets without dimensionality reduction. The Bootstrap sampling method is utilized to generate multiple sub-training sets, improving the model's diversity and robustness. Decision trees are built using the algorithm's node splitting rules, randomly selecting features to create an ensemble model. Finally, the study analyzes the temporal change in the proportion of global and intercontinental GDP located in "near regions" and "far regions" (areas beyond 100 km of coastlines). The regional economic proportion (ei) is calculated as the ratio of a region's economic aggregate (Ei) to the total economic aggregate of all regions (ΣEi), expressed as a percentage. This approach enables a quantitative assessment of the relative economic importance of coastal versus inland areas and their changes over time.

The analysis revealed a significant shift in the global distribution of economic activity. While the total GDP of both "near regions" and "far regions" increased between 2000 and 2018, the proportion of global GDP located in "near regions" decreased from 67.25% to 63.02%, while the proportion in "far regions" increased from 32.75% to 36.98%. This signifies a "coastal remoteness" pattern. The GDP density (GDP per sq km) in "near regions" remained significantly higher than in "far regions" but the gap narrowed from 9.3 times to 7.7 times over the period. Intercontinental analysis revealed heterogeneity. Europe and North America demonstrated a "coastal remoteness" pattern, but with less dramatic shifts compared to Asia, South America, and Oceania, which showed more pronounced shifts inland. Africa exhibited a "coastal proximity" pattern, with a slight increase in the economic proportion of its near regions. Figure 3 visually displays the land area ratio and GDP density differences between near and far regions across continents and globally. Figure 4 illustrates the changing economic proportions of near and far regions globally from 2000 to 2018. Figure 5 presents similar data for various continents, highlighting intercontinental variations and including trendline equations for each. Table 1 provides a detailed breakdown of the economic proportions of near and far regions for each continent and globally over the study period.

The observed "coastal remoteness" pattern contradicts the common assumption that economic globalization would intensify coastal agglomeration. The study attributes this shift to several factors: (1) Improved inland transportation infrastructure reduced transportation costs, allowing for economic activity to shift further inland; (2) The rising importance of high-tech industries and service sectors lessened the reliance on coastal locations for maritime trade; and (3) Diseconomies of scale and negative externalities (e.g., congestion, high land prices, pollution) in heavily concentrated coastal areas pushed economic activity inland. The findings align with Friedman's theory of regional spatial structure evolution, suggesting a transition from polar core development to a diffusion development stage, with variations across continents depending on their stage of industrialization. The historical evolution of human civilizations, from riverine settlements to port cities, and now potentially towards a more balanced inland-coastal distribution, further supports this interpretation.

The study provides novel insights into the spatial evolution of global economic activity, revealing a "coastal remoteness" trend that contrasts with traditional assumptions. The combined use of nighttime light data and random forest algorithms offers a robust and scalable approach for assessing large-scale economic dynamics. The findings highlight the need for policy adjustments to reflect the changing economic landscape, emphasizing the importance of balanced development strategies for both coastal and inland regions. Future research could integrate additional data sources (e.g., remote sensing, sensor networks, social network data) to deepen the understanding of the underlying mechanisms and drivers of this spatial shift.

The study relies on nighttime light data as a proxy for economic activity, which may not perfectly capture all aspects of economic output. The focus on GDP as the primary indicator could neglect other important factors influencing economic development. Additionally, the analysis does not directly address the causal mechanisms behind the observed patterns in detail. More in-depth investigation using diverse data sources and advanced statistical methodologies is necessary for a more comprehensive understanding.